DEVELOPMENT OF SYSTEM FOR ONLINE/OFFLINE QUALITY CONTROL OF NONWOVEN FABRICS/FUNCTIONAL FABRICS USING DIGITAL IMAGE PROCESSING TECHNIQUES

Transcription

1 DEVELOPMENT OF SYSTEM FOR ONLINE/OFFLINE QUALITY CONTROL OF NONWOVEN FABRICS/FUNCTIONAL FABRICS USING DIGITAL IMAGE PROCESSING TECHNIQUES A Thesis submitted to Gujarat Technological University for the Award of Doctor of Philosophy in Textile Engineering by Krishma Suresh Desai Enrollment No.: GUJARAT TECHNOLOGICAL UNIVERSITY AHMEDABAD July 2016

2 DEVELOPMENT OF SYSTEM FOR ONLINE/OFFLINE QUALITY CONTROL OF NONWOVEN FABRICS/FUNCTIONAL FABRICS USING DIGITAL IMAGE PROCESSING TECHNIQUES A Thesis submitted to Gujarat Technological University for the Award of Doctor of Philosophy in Textile Engineering by Krishma Suresh Desai Enrollment No.: under supervision of Prof. (Dr.) P. A. Khatwani GUJARAT TECHNOLOGICAL UNIVERSITY AHMEDABAD July 2016

3 Krishma Suresh Desai iii

4 DECLARATION I declare that the thesis entitled Development of System for Online/Offline Quality Control of Nonwoven Fabrics/Functional Fabrics Using Digital Image Processing Techniques submitted by me for the degree of Doctor of Philosophy is the record of research work carried out by me during the period from August 2011 to July 2016 under the supervision of Prof. (Dr.) P. A. Khatwani, Professor & Head, Dept. of Textile Technology, SCET and this has not formed the basis for the award of any degree, diploma, associateship, fellowship, titles in this or any other University or other institution of higher learning. I further declare that the material obtained from other sources has been duly acknowledged in the thesis. I shall be solely responsible for any plagiarism or other irregularities, if noticed in the thesis. Signature of the Research Scholar : Date: 29/07/2016 Name of Research Scholar: Krishma Suresh Desai Place : Surat. iv

5 CERTIFICATE I certify that the work incorporated in the thesis Development of System for Online/Offline Quality Control of Nonwoven Fabrics/Functional Fabrics Using Digital Image Processing Techniques submitted by Smt. Krishma Suresh Desai was carried out by the candidate under my supervision/guidance. To the best of my knowledge: (i) the candidate has not submitted the same research work to any other institution for any degree/diploma, Associateship, Fellowship or other similar titles (ii) the thesis submitted is a record of original research work done by the Research Scholar during the period of study under my supervision, and (iii) the thesis represents independent research work on the part of the Research Scholar. Signature of Supervisor: Date: 29/07/2016 Name of Supervisor: Prof. (Dr.) P. A. Khatwani Place: Surat v

6 Originality Report Certificate It is certified that PhD Thesis titled Development of System for Online/Offline Quality Control of Nonwoven Fabrics/Functional Fabrics Using Digital Image Processing Techniques by Smt. Krishma Suresh Desai has been examined by us. We undertake the following: a. Thesis has significant new work / knowledge as compared already published or are under consideration to be published elsewhere. No sentence, equation, diagram, table, paragraph or section has been copied verbatim from previous work unless it is placed under quotation marks and duly referenced. b. The work presented is original and own work of the author (i.e. there is no plagiarism). No ideas, processes, results or words of others have been presented as Author own work. c. There is no fabrication of data or results which have been compiled / analysed. d. There is no falsification by manipulating research materials, equipment or processes, or changing or omitting data or results such that the research is not accurately represented in the research record. e. The thesis has been checked using Plagiarism Detector (copy of originality report attached) and found within limits as per GTU Plagiarism Policy and instructions issued from time to time (i.e. permitted similarity index <=25%). Signature of the Research Scholar : Date: 29/07/2016 Name of Research Scholar: Krishma Suresh Desai Place : Surat Signature of Supervisor: Date: 29/07/2016 Name of Supervisor: Prof. (Dr.) P. A. Khatwani Place: Surat vi

7 PhD THESIS Non-Exclusive License to GUJARAT TECHNOLOGICAL UNIVERSITY In consideration of being a PhD Research Scholar at GTU and in the interests of the facilitation of research at GTU and elsewhere, I, Krishma Suresh Desai having hereby grant a non-exclusive, royalty free and perpetual license to GTU on the following terms: a) GTU is permitted to archive, reproduce and distribute my thesis, in whole or in part, and/or my abstract, in whole or in part ( referred to collectively as the Work ) anywhere in the world, for non-commercial purposes, in all forms of media; b) GTU is permitted to authorize, sub-lease, sub-contract or procure any of the acts mentioned in paragraph (a); c) GTU is authorized to submit the Work at any National / International Library, under the authority of their Thesis Non-Exclusive License ; d) The Universal Copyright Notice ( ) shall appear on all copies made under the authority of this license; e) I undertake to submit my thesis, through my University, to any Library and Archives. Any abstract submitted with the thesis will be considered to form part of the thesis. f) I represent that my thesis is my original work, does not infringe any rights of others, including privacy rights, and that I have the right to make the grant conferred by this non-exclusive license. g) If third party copyrighted material was included in my thesis for which, under the terms of the Copyright Act, written permission from the copyright owners is required, I have obtained such permission from the copyright owners to do the acts mentioned in paragraph (a) above for the full term of copyright protection. vii

8 h) I retain copyright ownership and moral rights in my thesis, and may deal with the copyright in my thesis, in any way consistent with rights granted by me to my University in this non-exclusive license. i) I further promise to inform any person to whom I may hereafter assign or license my copyright in my thesis of the rights granted by me to my University in this nonexclusive license. j) I am aware of and agree to accept the conditions and regulations of PhD including all policy matters related to authorship and plagiarism. Signature of the Research Scholar: Name of Research Scholar: Krishma Suresh Desai Date: 29/07/2016 Place: Surat Signature of Supervisor: Name of Supervisor: Prof. (Dr.) P. A. Khatwani Date: 29/07/2016 Place: Surat Seal: viii

9 Thesis Approval Form The viva-voce of the PhD Thesis submitted by Smt. Krishma Suresh Desai (Enrolment No ) entitled Development of System for Online/Offline Quality Control of Nonwoven Fabrics/Functional Fabrics Using Digital Image Processing Techniques was conducted on. (day and date) at Gujarat Technological University. (Please tick any one of the following option) The performance of the candidate was satisfactory. We recommend that he/she be awarded the PhD degree. Any further modifications in research work recommended by the panel after 3 months from the date of first viva-voce upon request of the Supervisor or request of Independent Research Scholar after which viva-voce can be re-conducted by the same panel again. (briefly specify the modifications suggested by the panel) The performance of the candidate was unsatisfactory. We recommend that he/she should not be awarded the PhD degree. (The panel must give justifications for rejecting the research work) Name and Signature of Supervisor with Seal 1) (External Examiner 1) Name and Signature ) (External Examiner 2) Name and Signature 3) (External Examiner 3) Name and Signature ix

10 ABSTRACT As a result of globalization & also increasing competition, it has become very important for any industry to develop solutions regarding the quality of products. Effective monitoring and control, better data predictions, quick response to query is necessary for effective Quality Control. The research work involves a development of system for online/offline quality control of nonwoven fabrics/functional fabrics using Digital Image Processing Techniques. The principal object is to determine the quality of fabric and frequency of different types of defects occurred in the fabric during the process of manufacturing. Human Visual Inspection of fabrics has been a criteria for Visual Assessment of fabric quality in the Textile Sector since long. It included the detection of fabric defects generally. However, this method cannot detect more than 60% of the overall defects for the fabric if it is moving at a faster rate and thus the process becomes insufficient and costly. Therefore, automatic fabric defect inspection is required to reduce the cost and time waste caused by defects. The present works aims at recording the number of defects in the fabric per unit length with the help of camera well supported with the software solution using MATLAB and displaying the defects and hence the quality of fabric on computer screen. At the same time, the present invention also displays the frequency of different types of defects occurred in fabric during the manufacturing process. A quality monitoring device has been developed as a part of the research work and the quality of the fabric can be assessed by mounting the beam containing the fabric on the developed device. The Fabric Quality Monitoring Device comprises of following major components: 1. Fabric delivery roller 2. Fabric take-up roller 3. Chain and sprocket driving arrangement 4. Electric motor x

11 5. Video camera 6. Computer loaded with specially developed software solution 7. Input Power The fabric is moved forward by the driving mechanism and finally it is wound on to the take-up roller. The quality of fabric can be checked by directing the light towards the face surface of fabric or back surface of fabric. During the forward movement of fabric, it is scanned by the camera. A series of images are taken by the camera which are then transferred to the computer loaded with the suitable software solution specially developed for analysing the fabric for various defects. The captured images are processed using MATLAB. Various parameters like mean, sd, histogram of the intensity values are studied for estimating & identifying the standard images. The images of the samples with defect are then processed for obtaining the defect statistics. The proposed algorithm will check for variability and give defect statistics and classify as per Defect Area. It will also check for no. of Defects in the Fabric Lot and give % Defects in the Fabric. On the basis of the defect statistics a fabric grading system has been developed which will classify the fabric for specific application. Thus, the developed system will guide the users by way of providing the information related to the defects and their frequency of occurrence during the manufacturing process. xi

12 Acknowledgement I express my deep sense of gratitude to my honourable guide Prof. (Dr.) P. A. Khatwani, Professor & HOD, Dept. of Textile Technology, Sarvajanik College of Engineering & Technology, Surat, for his immense interest, co-operation, enthusiastic and motivating attitude and systematic guidance and involvement throughout the work. It is for his tireless supervision that this work has developed into its present form. I am very much thankful and grateful to my Co-supervisor, Dr. Hamed SariSarraf, Professor, Electrical and Computer Engineering, Texas Tech University, TX, Doctoral Progress Committee members Dr. P. C. Patel, Professor, Faculty of Engg. & Tech., M. S. University and Mr. R. S. Backaniwala, Director, M/s. Himson Ltd. for mentoring me and providing me valuable guidance as and when required. I extend my sincere thanks to Mr. R. S. Backaniwala, Director, M/s. Himson Ltd., Mr. Pinal Dakoria, M/s. Technofab, Udhana Magdalla Road, Surat, Mr. Nimish Gajjar, M/s. N.M. Gajjar Hotels Pvt. Ltd, Village Kim, Surat Mr. Rakesh Bhai, M/s. Wovlene Tecfab India, Hazira, Surat, Mr. Arun Nag, M/s. Ginni Filaments, GIDC, Panoli for extending their full co-operation for sampling of the fabrics for the research work. My gratitude goes out to the assistance and support of Dr. Akshai Aggarwal, Ex. Vice Chancellor, Dr. Rajul Gajjar, I/c. Vice Chancellor & Dean, PhD Programme, Mr. J. C. Lilani, I/C Registrar, Ms. Mona Chaurasiya, Research Coordinator, Mr. Dhaval Gohil, Data Entry Operator and other staff members of PhD Section, GTU. I express my deep sense of gratitude to all the members of Governing Body of Sarvajanik College of Engg. & Tech. for their encouragement & full support in carrying out my experimental & testing work at the laboratories of Sarvajanik College of Engg. & Tech., Surat. xii

13 Most importantly, none of this would have been possible without the love and patience of my dear family members, my husband Dr. Pathik Naik for providing me constant support and strength, my son Tishya for his love & understanding, my mother Ms. Ranjanben Desai & Father Mr. Sureshbhai Desai, who always stood by me in my venture and to them I solemnly dedicate this thesis. Special thanks to all those who have directly or indirectly contributed for the progress and completion of my work. xiii

14 Table of Contents CHAPTER 1 INTRODUCTION Background Objectives Methodology Scope of Thesis 4 CHAPTER 2 LITERATURE REVIEW Introduction Functional Fabrics Nonwovens Need for Quality Control Manufacturing of Functional Fabrics Introduction Woven Fabrics Knitted Fabrics Nonwovens Special Fabrics Quality Control of Functional Fabrics Introduction Quality Parameters of different Functional Fabrics 10 xiv

15 2.3.3 Methods of assessment of Functional Fabrics Common Defects in Functional Fabrics Introduction Defects in Woven Fabrics Defects in Knitted Fabrics Defects in Nonwovens Fabric Inspection Visual Manual / Traditional Inspection Methods Automatic Fabric Inspection Systems Use of Image Processing in Assessing Structural Variation in Different Areas of Textiles Image Processing & Structural Variability in Fibres Image Processing & Structural Variability in Yarn Image Processing & Structural Variability in Fabrics Image Processing Introduction Human and Computer Vision System Fundamental Steps in Digital Image 26 xv

16 Processing Methods of Assessment of Image Analysis or Image Quality Image Processing & Fabric Inspection Introduction Implementation, Challenges and Difficulties Image Acquisition Components of Fabric Inspection System Algorithms for Automatic Defect Detection Introduction Structural Approaches Statistical Approaches Spectral Approaches Model-based Approaches Fabric Grading Introduction Various Approaches Summary of Literature Review 49 CHAPTER 3 PROPOSED SYSTEM & RESEARCH APPROACH Problem Description Proposed System Research Approach & Hypothesis 52 xvi

17 CHAPTER 4 SYSTEM DESIGN & DEVELOPMENT Introduction Device Development Development of manually operated device Further modification of device for image acquisition of fabric in roll form Automation of the Device Working of System using the Device Fabric Manufacturing and Defect Analysis Introduction Fabric Sampling Fabric Defects Image Acquisition for the Learning Phase Introduction Geotextiles Spunbond Fabrics Image Processing Methodology Introduction Steps involved in Processing of Images General Image Parameters of the Images 85 xvii

18 4.5.4 Methodology for Defect Detection Software Used Stages of Implementation Fabric Grading 95 CHAPTER 5 RESULTS AND DISCUSSIONS Introduction General Image Parameters of the Images Woven Geotextiles Spunbond Nonwovens Histogram Analysis Woven Geotextiles Spunbond Nonwovens Thresholding Woven Geotextiles Spunbond Nonwovens Defect Detection & Validation Defect Detection in Images of Woven Geotextiles Validation of Results of Geotextiles Defect Detection in Spunbond Images Validation of Results of Spunbond 149 xviii

19 CHAPTER 6 CONCLUSIONS & FUTURE WORK Objectives Achieved Conclusions Possible Future Scope 160 References Bibliography List of Publications Appendices xix

20 List of Figures No. Descriptions Page No. 2.1 Defects in woven fabrics Defects in knitted fabrics Defects in Nonwovens Manual Fabric Inspection Power Driven Inspection machines Automatic Fabric Inspection System Commercial Automatic Fabric Inspection System Elbit Vision 32 System 2.8 Various Illumination Configurations Influence of Illumination in Textile Materials Grading of Nonwoven Fabrics Process flow chart of the developed system Manually operated device Design of Box Base Photograph of the developed scanbox Photograph of the Modified Device Drive and main parts of the Device Different View Angles of Developed Device Top View & Illumination Arrangement Side View of Machine with Switch Board Passage of Fabric Final System Selection of Fabric Type Capture & Process Option Final Output Defects mentioned in Table Defects mentioned in Table Image of Defect Free Sample Missing End/Chira 76 xx

21 4.19 Slub (Warp) Stain (Daggi) Slub (Weft) Missing Pick/Jerky Gout Image of Defect Free Sample Drop/Bond point Fusion Pin Hole Wrinkle Hard Filament Holes Calender Cut Steps involved in processing of Images RGB Image & Grayscale converted Image of Spunbond Nonwoven Histogram of a Defect Free & Defective Region of Spunbond Fabric Grayscale Image with histogram & Contrast Adjusted Image with 88 histogram of Spunbond Fabric 4.35 Grayscale Image & Filtered Image of Spunbond nonwoven fabric Grayscale Image & Filtered Image of Woven Geotextile fabric Grayscale Image & Binary Image of spunbond nonwoven fabric Binary Image & Dilated Binary Image of Geotextile woven fabric Comparison of the Mean Intensity Values between the unprocessed 99 Images of the various Geotextiles 5.2 Comparison of the Mean Intensity Values between the enhanced 99 Images of the various Geotextiles 5.3 Comparison of the Mean Intensity Values between the unprocessed 101 Images of Defects & general Fabric in Geotextiles 5.4 Comparison of the Mean Intensity Values between the enhanced 101 Images of Defects & general Fabric in Geotextiles 5.5 Comparison of the Mean Intensity Values between the unprocessed 102 Images of the various Defects in Geotextiles 5.6 Comparison of the Mean Intensity Values between the enhanced Images of the various Defects in Geotextiles 102 xxi

22 5.7 Comparison of the Mean Intensity Values between the unprocessed 104 Images of the various Spunbond Fabrics 5.8 Comparison of the Mean Intensity Values between the enhanced 104 Images of the various Spunbond Fabrics 5.9 Comparison of the Mean Intensity Values between the unprocessed 106 Images of Defects & general Fabric in Spunbond Fabrics 5.10 Comparison of the Mean Intensity Values between the enhanced 107 Images of Defects & general Fabric in Spunbond Fabrics 5.11 Comparison of the Mean Intensity Values between the unprocessed 107 Images of the various Defects in Geotextiles Spunbond Fabrics 5.12 Comparison of the Mean Intensity Values between the enhanced 108 Images of the various Defects in Spunbond Fabrics 5.13 Histogram of Images of some samples of defect free images of 109 Geotextiles 5.14 Histogram of defect free region and Missing End (Chira) Histogram of defect free region and Slubs (Warp) Histogram of defect free region and Stain (Daggi) Histogram of defect free region and Slubs (Weft) Histogram of defect free region and Missing Pick (Jerky) Histogram of defect free region and Gout Histogram of Images of some samples of defect free images of 113 Spunbond Nonwoven 5.21 Histogram of Drop / Bond Point Fusion Histogram of defect free region and Pinhole Histogram of defect free region and Wrinkle Histogram of defect free region and Hard Filament Histogram of defect free region and Hole Histogram of defect free region and Calender cut Grayscale Image - Missing End Binary Image - Missing End Highlighted Missing End Grayscale Image Slub (Warp) Binary Image - Slub (Warp) 120 xxii

23 5.32 Highlighted Slub (Warp) Grayscale Image Stain (Daggi) Binary Image - Stain (Daggi) Highlighted Stain (Daggi) Grayscale Image Slub (Weft) Binary Image - Slub (Weft) Highlighted Slub (Weft) Grayscale Image - Missing Pick (jerky) Binary Image - Missing Pick (jerky) Highlighted Missing Pick (jerky) Grayscale Image - Gout Binary Image - Gout Highlighted Gout Comparison of Defect Area obtained from the System with those 128 obtained from Manual - visual Examination 5.46 Comparison of Length/Width of Biggest Defect obtained from the 128 System with those obtained from Manual - visual Examination 5.47 Comparison of Multiple Images of Regions with same defect Test Image Test Image Test Image Test Image Test Image Test Image Test Image Test Image Test Image Test Image Grayscale Image Drop/Bond Point Fusion Binary Image - Drop/Bond Point Fusion Highlighted Drop/Bond Point Fusion Grayscale Image Pin Hole 140 xxiii

24 5.62 Binary Image - Pin Hole Highlighted Pin Hole Grayscale Image Wrinkle Binary Image - Wrinkle Highlighted Wrinkle Grayscale Image Hard Filament Binary Image - Hard Filament Highlighted Hard Filament Grayscale Image - Hole Binary Image - Hole Highlighted Hole Grayscale Image Calender Cut Binary Image - Calender Cut Highlighted Calender Cut Comparison of Defect statistics obtained from the System with 148 those obtained from Manual - visual Examination 5.77 Comparison of Total Area between Multiple Images of Regions 151 with same defect 5.78 Comparison of Objectionable Area between Multiple Images of 151 Regions with same defect 5.79 Test Image Test Image Test Image Test Image Test Image Test Image Test Image Test Image Test Image Test Image xxiv

25 List of Tables No. Descriptions Page No. 2.1 Description of Defects in Woven Fabrics Description of Defects in Knitted Fabrics Common Defects in Non Woven Fabrics Comparison of CCD and CMOS camera Comparison of Line Scan and Area Scan Camera Penalty Points in 4-Point System Penalty Points in 10-Point System Specifications of Geotextiles Specifications of Industrial Fabrics Specifications of Coated Fabrics for Composites Specifications of Spunbond Nonwovens Specifications of Needle Punched Fabrics Specifications of Spun Laced Fabrics Identified Defects in Geotextiles Identified Defects in Spunbond Fabrics Details of the images of the defects of Geotextiles Details of the images of the defects of Spunbond Fabrics Defect Classification Sample wise average values of Image parameters for unprocessed 98 images of each type of Geotextile. 5.2 Sample wise average values of Image parameters for enhanced 99 images of each type of Geotextile. 5.3 Defect wise values of Image parameters for unprocessed images of 100 identified defects in Geotextiles. 5.4 Defect wise values of Image parameters for enhanced images of 100 identified defects in Geotextiles. 5.5 Sample wise average values of Image parameters for unprocessed 103 images of each type of Spunbond Nonwoven 5.6 Sample wise average values of Image parameters for enhanced 104 xxv

26 images of each type of Spunbond Nonwoven 5.7 Defect wise values of Image parameters for unprocessed images of 105 identified defects in Spunbond Nonwoven 5.8 Defect wise values of Image parameters for enhanced images of 106 identified defects in Spunbond Nonwoven 5.9 Defect Statistics obtained from the System Defect Statistics obtained from Manual - visual Examination Manual Grading of the Defects in Geotextile Fabrics Grading of Defects in Geotextile Fabrics using the System Comparison between the Grading of Woven Geotextile Images 130 obtained by Manual - visual Examination & System 5.14 Grading of Multiple Images of Regions with same Defect Comparison between the Grading achieved with developed System 137 as against Manual - visual grading 5.16 Defect Statistics obtained from the System Defect Statistics obtained from Manual - visual Examination Manual Grading of the Defects in Spunbond Fabrics Grading of Defects in Spunbond Fabrics using the System Comparison between the Grading of Spunbond Images obtained by 150 Manual - visual Examination & System 5.21 Grading of Multiple Images of Regions with same Defect Comparison between the Grading achieved with developed System as against Manual - visual grading 157 xxvi

27 List of Appendices Appendix A Appendix B Appendix C Details of Patent Filed Software Code Originality Report xxvii

28

29 CHAPTER 1 Introduction 1.1 Background: Today, quality is to be considered as the most important parameter in any industry and the textile industry is no exception. Each industry today aims to produce the highest quality goods in the shortest amount of time leading to requirement of an enhanced quality control system. Quality may be defined by the sum of all those attributes which can lead to the production of products acceptable to the consumer when they are combined and one such important attribute is the defects or the faults in the fabric. Thus keeping fabric defects to a minimum is of prime importance from a quality perspective. A process quality control system includes testing & inspecting of fabric, analysing the observations so made and then making the decisions to improve the performance of the system. As no manufacturing process is 100% defect-free, especially when considering the fabric manufacturing process, the success of the process is significantly highlighted by the success in detecting the objectionable fabric defects to maximum. The frequency and nature of the defects in fabrics determines the quality of the product in terms of grading as well as the price of the fabric. First, second, third quality fabrics are often the terms used in the textile industry to determine the grading of fabrics as per the frequency and nature of defects. The profit margin also varies with the grading, decreasing from first quality to the last quality. This basis of quality grading is often subjective.

30 Introduction With the improvement in the production process and also improvement in materials and technology, the quality levels have increased to a great extent. This has also lead to customer expectations of getting a minimum defect fabric. The defects at present are frequently examined by human inspectors. The major limitation here is the human perception may vary from individual to individual, high labour cost as well as the time being involved. Chances of missing out the defects by operators is common, mostly due to tiredness, boredom, inattentiveness, fatigue and lack of time and the inspection so done may not be reliable. Thus the method of inspection plays a significant role in detection of objectionable faults and hence proper grading of fabrics. In order to have maximum benefit from the inspection process, the process should have high degree of accuracy. Therefore an automatic inspection system may be highly desirable as it gives possibly the best objective and consistent evaluation. An automated inspection system consists of a computer based vision system which may be offline or online. The major components include the fabric monitoring system and the defect analysing and classifying software. A high resolution camera is used for monitoring of the fabric offline or online in most of the commercial systems available in the market along with the software module identifying the defects. The software module uses various image processing tools for enhancement of the images captured and then extracting the variability or the defects. 1.2 Objectives: The objectives of the research work are as below: To develop cost and quality effective system for targeting mainly the growing Indian Technical Textile Market. To help the user in selection of proper quality of nonwoven / functional fabrics for specific end use applications. To help the user to avoid unnecessary wastage of time and materials, which otherwise would be due to wrong selection of materials for any specific application 2

31 Objectives Very bright prospects ahead for the system to be developed considering very high market growth from 10 billion dollars in 2009 to expected 31 billion dollars in Methodology: Qualitative as well as formulative approach has been used for this research work. The structure / qualities and properties of nonwoven/functional fabrics are influenced largely by factors like type & structure of raw material, type of fabric-woven, knitted, nonwoven, special fabrics, etc. which also influences the surface texture of the fabric. The system developed is capable of identifying defects in various functional fabrics and classify according to their nature, frequency and size. Different varieties of functional fabrics-woven and nonwoven were manufactured. As mentioned earlier, no manufacturing process is 100% defect free, thus causing defects in the fabrics. A device has been developed as a part of the quality monitoring system. The important parts of the device are fabric delivery roll, illumination for capturing images of surface of the fabric, CMOS camera for capturing images of surface of fabric and a fabric take up roller. A series of images are taken by the camera which are then transferred to the computer loaded with the suitable software solution specially developed for analysing the fabric for various defects. The captured images are processed using MATLAB. The processing of images basically involves image enhancement and feature extraction to identify type of defect. The principal objective of image enhancement is to modify attributes of an image to make it more suitable for analysing it and estimating and identifying the defect free images as well as the images with defects. The features of the defects are then extracted and depending upon the statistics of the defect, they are graded, thus guiding the users by way of providing the information related to the defects and their frequency of occurrence during the manufacturing process. 3

32 Introduction 1.4 Scope of Thesis: 6 different varieties of functional fabric including woven as well as nonwoven fabrics were manufactured for the study. The defects obtained in the manufactured fabric were a result of the fabric manufacturing process and were assessed visually as well as with the software developed using the proposed algorithm using MATLAB. However, the experts from IIT had suggested to consider only one variety of fabric preferably spunbond nonwoven fabric during the Research Week held during month of April 2015 at Gujarat Technological University, Ahmedabad. They had also suggested to consider some of the major defects occurred during the manufacturing of spunbond fabrics and also to validate the results so obtained by taking multiple images of same defects. After considering the inputs from the experts of IIT, the study has been narrowed down to 2 varieties of functional fabrics i.e. Woven Geotextiles & Spunbond Nonwovens. 6 types of defects in each variety have been focused on in the study. The captured images were processed using MATLAB. Various parameters like mean, sd, histogram of the intensity values were studied for estimating & identifying the standard images. The images of the samples with defect were then processed for obtaining the defect statistics. The proposed algorithm will check for variability and give defect statistics and classify as per Defect Area. It will also check for no. of Defects in the Fabric Lot and give % Defects in the Fabric. On the basis of the defect statistics a fabric grading system has been developed which will classify the fabric for specific application. 4

33 CHAPTER 2 Literature Review 2.1 Introduction: Functional Fabrics: Functional Fabrics are those fabrics which are manufactured primarily for their technical and performance properties rather than their aesthetic or decorative characteristics (such as filters, machine clothing, conveyor belts, abrasive substrates etc.). The major areas where the functional fabrics are used are agriculture & horticulture, geotextiles, building & construction, Industrial filtration, conveying, cleaning, hygiene and medical, automobiles, shipping, railways and aerospace, packaging, personal and property protection, footwear sport and leisure. Since the functional fabrics are manufactured primarily for their technical applications, the fabrics may be produced by weaving, felting, lace making, net making, nonwoven processes and tufting or a combination of these processes.

34 Literature Review Nonwovens: Nonwovens owing to their different and specific structure from the woven and knitted fabrics are majorly used as functional fabrics. There are various definitions describing the nonwovens, but the most commonly used are those by the Association of the Nonwovens Fabrics Industry (INDA) and the European Disposables and Nonwovens Association (EDANA) which has been described below: The nonwovens definition adopted by EDANA, the European nonwovens industry association is: they are manufactured sheet, web or bat of directionally or randomly oriented fibres, bonded by friction, and /or cohesion and/or adhesion, excluding paper or products which are woven, knitted, tufted stitch bonded incorporating binding yarns or filaments, or felted by wet milling, whether or not additionally needled. The fibres may be natural or man-made origin [1]. They may be staple or continuous or be formed in situ. INDA, the North American Association has a slightly different, wider definition which has the merit of apparent simplicity: a sheet, web or batt of natural and or man-made fibres or filaments excluding paper, that have not been converted into yarns and that are bonded together by any of the following means [1] : Adding an adhesive Thermally fusing the fibres or filaments to each other or to the other meltable fibres or powders. Fusing fibres by first dissolving, and then resolidifying their surfaces. Creating physical tangles or tuft among the fibres. Stitching the fibres or filaments in place. The manufacturing of nonwovens basically started with the use of textiles on other areas than for the apparel use. With the development of Industrial textiles, the nonwovens started gaining a larger market. The oldest nonwoven i.e. felt was made up of wool (felting of wool because of its scaling structure & was used as rugs. With the invention of man-made fibres, the application of nonwovens started being used in various industrial applications. Earlier cotton & natural fibres such as hemp & jute were extensively used in production & nowadays because of advantageous properties of synthetic fibres, they are used. 6

35 Introduction Need for Quality Control: Quality is basically signifies the needs of customer and failing to maintain the required quality standard may affect the cost of the product. With the increase in number of applications of technical textiles in different areas during the days to come, and to avoid rejection of fabrics, it becomes necessary to design and develop the system to check the quality of such varieties of fabrics in much shorter time and with utmost accuracy. As mentioned earlier functional textiles are fabrics used for specific performance based applications, it becomes very necessary to maintain standards of quality required for these applications. These might become possible by implementing an efficient quality control system. Offline and online fabric inspection for woven fabrics has served as an important tool to cater to the needs of good quality control systems especially for the fabrics to be used by the garment industry [2-4]. Similar systems seem to be a necessity for maintaining the standards of the functional fabrics as well as the nonwoven fabrics. 2.2 Manufacturing of Functional Fabrics [5] : Introduction: As discussed earlier Textile fabrics are most commonly woven but may also be produced by other methods of fabric manufacturing like knitting, nonwoven processes, lacing, tufting etc. A brief overview on all these process has been described in the following sections Woven Fabrics: Woven fabrics are made on looms. They consist of two sets of yarns that are interlaced and lie at right angles to each other. The threads that run along the length of the fabric are known as warp (ends) and the threads that run from one side to the other side of the fabric, are weft (picks). In triaxial and in three-dimensional fabrics yarns are arranged differently. 7

36 Literature Review Functional fabrics are manufactured on any of the weaving machines as per their application. Their strength, thickness, extensibility, porosity and durability can be varied and depend on the weave used, the thread spacing, that is the number of threads per centimetre, and the raw materials, structure (filament or staple), linear density (or count) and twist factors of the warp and weft yarns. Higher strengths and greater stability can be obtained from woven fabrics than from any other fabric structure using interlaced yarns. Structures can also be varied to produce fabrics with widely different properties in the warp and weft directions Knitted Fabrics: Knitted fabrics are formed by interloping of yarn(s) i.e. forming yarn(s) into loops, each of which is typically only released after a succeeding loop has been formed and intermeshed with it so that a secure ground loop structure is achieved and are manufactured on knitting machines. The loops are also held together by the yarn passing from one to the next. They can be divided into weft knit fabric and warp knit fabrics. Warp Knitted: Loops formed across the width of fabric & each weft thread is fed more or less at right angles. Weft Knitted: Loops formed vertically down the length of the fabric from one thread. Weft knitted is more versatile method of fabric formation, more production & simplest method of converting yarn to fabric Nonwovens: As discussed earlier this fabrics are neither woven nor knitted but are produced when the fibres of a web/batt are intermingled or bonded by mechanical, chemical or thermal means. 8

37 Manufacturing of Functional Fabrics Special Fabrics: Lacing: It is an exquisitely detailed fabric and commonly used in lingerie, formal wear, and decorative trim. Previously the lace was handmade. Lace are now also produced on Raschel knitting machines as well as needle looms for the use of lace in areas of technical textiles like hometech and indtech. Netting: It is made by looping and knotting yarns in an open pattern, usually a characteristic geometric design of rectangular, square or diamond shapes. Originally, net was made by hand but now most contemporary versions are produced on lace machines. Depending on the size of the yarn, nets can range from very fine fabrics to heavy, coarse materials. The open pattern of the netting makes it useful for decorative accents as well as functional purposes such as fish nets. Braiding: A braid is a rope like thing, which is made by interweaving three or more strands, strips, or lengths, in a diagonally overlapping pattern. Braiding is one of the major fabrication methods for composite reinforcement structures. It is done by intertwining of yarns in any direction. From a domestic art of making laces, it evolved as a fabric made by narrow width looms. Of late, Crochet knitting machines have replaced large numbers of traditional braiding machines. Braiding can be classified as two and three-dimensional braiding. Two-dimensional braid structure can be circular or flat braid. They are formed by crossing a number of yarns diagonally so that each yarn passes alternately over and under one or more of the others. Two dimensional braids are produced through circular braiding machine and rotary machine. Three-dimensional braiding is relatively new and was developed mainly for composite structures. In it, a two dimensional array of interconnected 2-D circular braids is created on two basic types of machines- the horn gear and cartesian machines. 9

38 Literature Review 2.3 Quality Control of Functional Fabrics: Introduction: By quality it s understood that all the features and characteristic values of a product or a service fulfil the fixed and expected requirements with regard to their suitability. Since the functional fabrics and nonwovens have their specific end-use, the test method and the quality assessment is done specific to its application. Earlier, nonwovens were subjected to the same quality control testing methods used for pulp and paper industries, but the practices has now changed and separate testing methods have been developed specific to the application of the particular fabric [1]. Some of the methods to analyse the quality of these kinds of fabrics have described in this section Quality Parameters of different functional fabrics: The important parameters for characterization of technical textiles including nonwovens include various surface & structural characteristics like mass per surface unit, nonwoven thickness, elasticity and air permeability, defects, coating quality, etc. All these features and characteristic values with respect to the expected requirements with regard to their suitability accounts for assessment of quality. The presence of defects may lead to reduction in the profit margin by about 45%-65% [3]. Thus, considering the surface characteristics as an important and a common parameter in assessment of the quality performance of any functional fabrics, estimating the same by means of fabric inspection methods is highly desirable. The main objective of the visual assessment is detection of objectionable defects or fabric faults in the fabrics. Thus the different types of fabric faults can be considered as important quality parameters for different functional fabrics Methods of Assessment of functional fabrics: As mentioned earlier the test method and the quality assessment of the functional fabrics and nonwovens are done specific to its application. Some of the common methods include visual examination methods and automatic examination methods. Some of the other automatic examination methods include Image Analysis, Infrared Measurement Principles and using weight sensors for assessing uniformity. 10

39 Quality Control of Functional Fabrics Image analysis: This method is commercially used by many companies to measure the fibre orientation (fibre orientation analyzer) in web, measurement of uniformity of web, measuring porosity of fabrics (TRI/Micro Absorbmeter), etc [1]. Infrared measurement principles [1] : The light with wavelength near infra-red (NIR) and mid of infrared is selectively absorbed by moisture or organic bases coatings. The amount of absorption that occurs is related to the concentration of the absorber and therefore allows a coat weight or moisture measurement to be obtained. The NIR method gives a single point measurement after the coater or re-moisturizer is required for coating and moisture measurement. It is a non-contact measurement and the sensor operates well at a distance of 200 mm as is required for coatings. Using weight sensors for uniformity: This method primarily detects the weight per unit area and the thickness of the fabric [1]. It involves weight gauges or photoelectric sensors to determine the weight per unit area. In the online system measuring the same the delivery and feed are regulated according to the values of the weight. 2.4 Common Defects in Functional Fabrics: Introduction: A Fabric Defect is basically any abnormality in the Fabric that hinders its acceptability by the consumer. Fabric defects may be characterised by an imperfection that impairs worth or utility and spoils the utility of material [6]. The defects in the fabrics may be accounted as a result of use of defective yarn, error in process of manufacturing of the fabric or the error in the wet processing of fabrics in case of finished fabrics. As mentioned above, the defects may also be resulted due to defective manufacturing process, hence the type of manufacturing process may result in set of defects characterised by it s manufacturing process. The defect type, their principal causes and remedies of the defects found in woven, knitted and nonwoven have been briefed in this section. 11

40 Literature Review Defects in Woven Fabrics: There are many fabric defects which occur during the process of weaving. Some of these defects are described in Table 2.1 [6, 7, 8] : Table 2.1 : Description Of Defects In Woven Fabrics Sr. Fabric No. Defect 1. Missing End (Chira) 2. Slubs (Warp) 3. Stain (Daggi) Definition Principal Causes Remedy There may be one end or a group of ends missing in the fabric. Thick untwisted portion in warp yarn These stains are due to lubricants or dust. 4. Slubs Thick untwisted (Weft) portion in weft yarn 5. Missing It is a strip which Pick extends across the (Jerky) width of fabric & has the pick density lower than the required one. 6. Gout Foreign matter woven in a fabric by accident. Usually lint or waste. 7. Weft Bar It is a bar or band which extends across the width of fabric. If the broken ends are not mended immediately by the operator, these missing ends will occur in the fabric. Variation in draft during spinning. Improper material handling, bad oiling & cleaning practices Variation in draft during spinning. It is caused by faulty let - off & take - up motions; Also, if the loom is not stopped immediately in case of weft break, few picks are liable to be missed in the fabric. It is caused when the hardened fluff or foreign matter such as pieces of leather accessories, pieces of damaged pickers etc., is woven into the texture of the fabric. If the weft yarn is not regular/uniform; If there is more variation in count of weft yarn; If there is shade variation in case of dyed weft yarn; If the difference in blend composition is more in case of blended Weft yarn. This defect can be minimised by minimising missing ends in the weaver s beam and by providing an efficient warp - stop motion on a loom. Set the draft as per the requirement. By proper material handling as well as good oiling & cleaning practices, this defect can be avoided. Set the draft as per the requirement. This defect can be remedied by proper setting of let - off & take - up motions & also by using an efficient brake - motion. This defect can be remedied by preventing the foreign matter from falling onto the warp between the reed & the fell of the cloth. This defect can be remedied by better process control to get the uniform & regular weft yarn. The mixing - up of different counts, shades etc., of weft yarns can be avoided by better house - keeping. 12

41 Common Defects in Functional Fabrics TABLE 2.1 : Description of Defects in Woven Fabrics (cont.). Sr. No. Fabric Defect Definition Principal Causes Remedy 8. Float (Jala) When there is no When there is an entanglement The entanglement of proper interlacement of adjoining ends in the region ends & hence the float of warp & weft over a between the heald shafts & the can be avoided by certain area of fabric, fell of the cloth. mending the broken end a float is formed. immediately and by providing efficient warp - stop motion. 9. Patti or Crammed Picks It is also a band which extends across the width of fabric Because of improper working of take - up motion, sometimes more pick density is obtained than the required one & this higher pick density appears as a band in the fabric 10. Starting Marks It is similar to a jerky Because of mechanical faults in the loom such as loose fitting of reed, loose or worn - out crank, crank - arm or crank - shaft bearings etc, the starting marks will occur in the fabric whenever the loom is started 11. Shuttle Smash When many ends break due to shuttle trap, this defect will occur 12. Reed Mark Similar to missing end. 13. Temple Marks These are fine holes caused near the selvedges of a fabric. 14. Box Marks These are fine weftway dark or oily lines in the fabric. There are many causes for the shuttle trap like wrong timing of shedding, soft picking, unbalanced shuttle, insufficient checking of shuttle in the boxes etc. When the wires of reed are damaged or bent during running of a loom, the space between those wires & hence between the warp ends is increased. This increased space between the ends is clearly seen in the fabric as a missing end. Caused by improper use of temples. This defect will occur if the weft yarn is trapped between the box front plate & the shuttle, & becomes oily & if the balloon of weft yarn, formed during its unwinding from the pirn, touches the picker guide spindle & becomes oily. By proper setting of take - up motion this defect can be avoided. This defect can be remedied by proper maintenance of the loom. Setting of proper timing, using balanced shuttle. This defect can be avoided either by replacing or by straightening the damaged wires of the reed. This defect can be avoided by suitably selecting the temples for the fabrics to be produced. This defect can be remedied by tying a cloth piece to box front plate & keeping the shuttle box as clean as possible. 13

42 Literature Review Images of the defects in woven fabrics: Some of the common defects found in woven fabrics have been illustrated in Figure 2.1 [6,7]. FIGURE 2.1 : Defects in woven fabrics 14

43 Common Defects in Functional Fabrics Defects in Knitted Fabrics: Knitting process also caters to fabric defects and some of these defects are described in Table 2.2 [9, 10, 11] : TABLE 2.2 : Description of Defects in Knitted Fabrics Sr. No. Fabric Defect Definition Principal Causes Remedy 1. Barriness Barriness defect High Yarn Tension Ensure uniform Yarn appears in the Knitted Count Variation Tension on all the feeders. fabric in the form of Mixing of the yarn lots The average Count variation horizontal stripes of Package hardness variation in the lot should not be more uniform or variable than width. Ensure that the yarn being used for Knitting is of the same Lot. Ensure that the hardness of all the yarn packages is uniform using a hardness tester. 2. Needle Needle lines are Due to defective needle. Replacing all the defective Line prominent vertical Dirty needle slot. needles having, lines along the length Needle too tight or loose in bentlatches, hooks or stems. of the fabric which the slot. Removing the fibers are easily visible in Due to improper lubrication accumulated in, the Needle the gray as well as of needles tricks (grooves). finished fabric Replacing any bent Needles, running tight in the tricks. Checking the Needle filling sequence in the Cylinder / Dial grooves (tricks) 3. Hole Local Holes obtained Due to badly tied knot. Ensuring uniform yarn when yarn breaks Needle break due to slub. tension on all the feeders, during loop Due to high tension of yarn. with a Tension Meter. formation. Rate of yarn feed should be strictly regulated, as per the required Stitch Length. Eyelets & the Yarn Guides, should not have, any fibers, fluff & wax etc. stuck in them. The yarn being used, should have no imperfections, like; Slubs, Neps & big knots etc. 4. Oil Mark Oil lines are Due to improper lubrication. Fibers, accumulated in the prominent vertical Fibers & fluff accumulated needle tricks, cause the oil to lines which appear in the needle tricks, which seep into the Fabric. along the length of the knitted fabric tube. remain soaked with oil. Remove all the Needles & the Sinkers of the machine, The lines become periodically. permanent if the Cleaning the grooves of the needle oil used is not Cylinder & Dial of the washable & gets machine thoroughly. baked due to the heat Blowing the grooves of the during the finishing of Cylinder, Dial & Sinker the fabric. ring, with dry air after cleaning. 15

44 Literature Review TABLE 2.2 : Description of Defects in Knitted Fabrics (cont.). Sr. No. Fabric Defect 5. Foreign Materials/ Fly Definition Principal Causes Remedy Contaminations appear in the form of foreign matter such as; dyed fibers, husk, dead fibers etc. in the staple spun yarn or embedded in the knitted fabric structure. 6. Press Off Fabric press off appears as a big or small hole in the fabric caused due to the interruption of the loop forming process as a result of the yarn breakage or closed needle hooks. 7. Sinker Lines Sinker lines are prominent or feeble vertical lines appearing parallel to the Wales along the length of the knitted fabric tube. 8. Spirality Spirality appears in the form of a twisted garment after washing. The seams on both the sides of the garment displace from their position & appear on the front & back of the garment. If foreign materials knit with the yarn. When all or some of needles on circular knitting fail to function and the fabric either falls off the machine or design is completely disrupted or destroyed. Bent or Worn out Sinkers Sinkers being tight in the Sinker Ring grooves. High T.P.I. of the Hosiery Yarn Uneven Fabric tension on the Knitting machine. Unequal rate of Fabric feed on the Stenter, Calender & Compactor machines. Blowing the grooves of the Cylinder, Dial & Sinker ring, with dry air during cleaning. Needle detectors, should be set precisely, to detect the closed needles & prevent the fabric tube from completely pressing off. Proper yarn tension should be maintained, on all the feeders. Replace all the worn out or bent sinkers causing Sinker lines in the fabric. Sinker lines are very fine & feeble vertical lines appearing in the fabric. Remove the fibers clogging the Sinker tricks. Use the Hosiery yarns of the recommended TPM level for Knitting. Ensure uniform rate of feed of the dyed fabric on both the edges while feeding the fabric to the Calender, Compactor or Stenter machines. 16

45 Common Defects in Functional Fabrics Images of the defects in knitted fabrics: Some of the common defects found in knitted fabrics have been illustrated in Figure 2.2 [9, 10, 11]. FIGURE 2.2 : Defects in knitted fabrics 17

46 Literature Review Defects in Nonwovens: Some of the common defects obtained in the Nonwovens are described in Table 2.3 [1,12,13]. They are illustrated in Figure 2.3 [1,12,13]. TABLE 2.3: Common Defects in NonWoven Fabrics Name of Defect Eye brows Drops / bond point fusion Pinholes Thinspots Wrinkles Hard filaments Insects Monomer / polymer Drips Holes External Contamination / Dirt Oil contamination Melt-blown fly Broken Filaments Calendar cut Scratch Clumps Blowback Un-bonded web Rough Fabrics Streaks Bad uniformity Spills Description Stretching or folds in fabric Fused fibres on surface Very small holes in fabric Low density of fibres in a particular area Wrinkle formation Fused filaments on surface Trapping of insects in web/fabric Formation of spots on surface by droppings of monomer / polymer Holes in fabric/ web Contamination due to external factors like dust, dirt, etc. Contamination due to oil on surface of fabric Drip in melt blown fabrics Filaments are broken in web Cut marks due to calendaring Scratches in web/fabric Compact mass of fibres Fibres in opposite direction of normal orientation Loose fibres in web Surface of fabric is uneven Thin line marks in fabric Uneven fabric Coming out of fibres/lump of fibres from the surface FIGURE 2.3 : Defects in Nonwovens 18

47 Fabric Inspection 2.5 Fabric Inspection: Visual Manual / Traditional Inspection Methods: Visual Inspection of fabrics basically involves visual checking of the manufactured fabric lot for the defects produced in it during the fabric manufacturing process. As mentioned earlier a defect may be characterised by an imperfection that impairs worth or utility and spoils the utility of material and thus is majorly concerned with the texture characteristics of the surface of the fabric. Visual checking thus involves identifying and recording of the defects. The traditional method of visual inspection involves manual - visual inspection by well-trained human inspectors. The fabric inspection is performed manually by human inspectors and using off-line stations as shown in the Figure 2.4. A trained person inspects all types of fabrics and identifies all defects and then divides them into the corresponding grades and accordingly the performance of the fabric may be assessed. FIGURE 2.4: Manual Fabric Inspection 19

48 Literature Review The oldest inspection method involved pulling of fabric over a table by hand. Now-a-days power driven inspection machines as shown in Figure 2.5 are used. There are different input/output variations (roll to roll, batch to batch, fold to fold) available in the latest checking machines with a wide range of optional equipment for flexible inspection routine. FIGURE 2.5 : Power Driven Inspection machines The inspection process is done in suitable environment with proper ventilation as well as lighting. The frame through which the fabric passes is normally at degree angles to the inspector. The illumination may involve top and back lighting which is used according to the fabric to be checked. The speed of the inspection machine is normally less so as to allow the inspector to view the fabric viewing area properly. When the inspector sees any defect on the moving fabric, he records the defect name and the size of it. Also he mends all the possible repairable defects [14]. 20

49 Fabric Inspection The main drawback of this method was its accuracy with identification rate about 70%. [15] Many defects are missed, and the inspection is inconsistent, with its outcome depending on the training and the skill level of the personnel. Also the work of inspectors is very tedious and time consuming. They have to detect small details that can be located in a wide area that is moving through their visual field. In addition, the effectiveness of visual inspection decreases quickly with fatigue Automatic Fabric Inspection Systems: As discussed above, manual fabric examination is very tedious and requires highest level of concentration, which can be maintained only for about 20 to 30 minutes. Even in a wellrun operation, the reproducibility of a visual inspection will rarely be over 50 % [2]. Therefore the introduction of automated inspection was done which gives more reliable results which are free from the subjective deficiencies of visual examination methods. There have been a lot of research and development being done in the field of automated visual inspection techniques for detection of fabric defects. Image Analysis [3,4,15,16] is widely being used for detection of defects. Also a variety of advanced approaches like ultrasonic imaging system [17] & laser optical systems [18] have been proposed. Ultrasonic system uses an ultrasonic imaging technique, which uses pair of ultrasonic transducers operating efficiently in air at a predetermined frequency [17]. Each pair has a transmitter and a receiver. The signals received are processed using signal processing and image processing to assess defects in fabrics. Laser Optical system comprises of a laser unit and cylindrical lenses, which performs 1-D imaging of the weft yarn [18]. The signals received are used basically to give the weft density, weft count, fabric length and fabric speed. The variations from the mean weft density and mean weft count are used as variables to detect various fabric defects in the woven fabrics. Among the approaches being researched, the most common one is the use of image analysis that is use of computer vision system to detect the defects in the fabric [3,4,15,16,19-30]. Even the commercial systems largely follow the same approach. Commercial application of image analysis for defect detection are being done by various companies like I-Tex of 21

50 Literature Review Elbit Vision Systems ltd., Uster Fabricscan of M/s. Uster Technologies, Barco Vision s Cyclops of M/s. Barco, SCANTEX of Sam Vollenweider, RoiBox web inspection systems [russuka akerbag] etc. but the main hindrance of it s use in small scale industries is the capital cost [31, 2]. In the automatic fabric inspection the checking of the surface characteristics of the fabric is done with the help of a camera instead of human inspectors [2,3,16,31]. The images of the fabric surface are captured by a camera and then the image analysis of the captured image is done to detect the defects in the fabric with the help of any image processing software. Basically Image Analysis can be subdivided into two steps: i. Image acquisition: the images of the fabric samples are captured using CCD or a digital camera. ii. Image analysis/pattern recognition by measuring scale and shape to identify the defects. 2.6 Use of Image Processing in Assessing Structural Variation in Different Areas of Textiles: Image Processing & Structural Variability in Fibres: Microscopic evaluation of cross section of the fibres is largely done for analysis of fibres and it s characteristics since the evolution of microscopes. The sample preparation involved in the microscopic evaluation needs to be very precise and the whole process is time consuming also leading to research being done for the use of image processing for the same. Image analysis is an attractive alternative to existing systems for investigating some quantitative fiber characteristics. A lot of studies are being done in assessing the structural variation in fibres to identify the fibre structure as well as some properties like fibre fineness and maturity [32-38]. Xu and Huang (2005) [35] suggests a method for analysing cotton fibres by taking images of fibre cross section and processing the images for estimating the fineness and maturity of the fibres. Xu, Pourdeyhimi & Sobus (1993) [53] also suggests fibre cross sectional shape 22

51 Use of Image Processing in Assessing Structural Variation in Different Areas of Textile analysis using Image Processing Techniques which can be used to determine various fibre parameters and properties. Studies have also been done by using longitudinal images of cotton fibres for fibre analysis [33]. Ghith, Fayala & Abdeljelil (2011) proposes a maturity analysis of fibres by image analysis by using longitudinal images of cotton fibres [33]. The main advantage of using this approach was that it eliminated the précised sample preparation which was required cross sectional fibre approach. It was quick, reliable and unbiased technique which was used to evaluate fiber maturity and fineness Image Processing & Structural Variability in Yarn: Due to the various advantages of use of image processing in analysing fibre characteristic, a review on the use of image processing for yarn analysis becomes necessary. The commercial systems for yarn analysis and measuring the yarn parameters uses the capacitive sensors but a lot of research is being done to check the use of image processing for the same [39-44]. The major yarn parameters include the yarn diameter, yarn mass or fineness, twist and hairiness. The images of surface characteristics of the yarn are captured using cameras which are processed using any image processing software and image processing tools for estimating of the said parameters. Parameters of fancy yarn and textured yarn have also been estimated using image analysis. Pan et all (2011) suggests recognition of parameters of slub yarn parameters also using image analysis [43]. Semnani and Gholami (2009) suggests identification of defective points in false twist textured yarns [39] Image Processing & Structural Variability in Fabrics: Just like textile fibres and textile yarn, image processing has been explored to assess the various fabric characteristics [45-61]. Image processing has been proved an efficient tool for analysing fabric surface characteristics to determine various fabric parameters [45-61]. 23

52 Literature Review Pattern recognition has been vastly looked upon for analysing the woven fabric structure [48,49,50,55,59]. A paper by Salem and Nasri proposes a method for weave identification by analysing the structure of woven fabrics using image processing. They have proposed a texture analysis for recognition of the weave patterns and classified the weaves accordingly [50]. Successful determination of fabric air permeability [45,56] and porosity in fabric [46, 57] has been obtained by using the various methods of image processing. Even surface properties like pilling [47, 51, 60] has been considerably estimated using this approach. Texture analysis has offered good scope for measuring Fabric handle properties like wrinkle [61], fabric smoothness [52], drape [53] and surface roughness [54]. As discussed earlier effective methods for quality control of woven fabrics have been proposed by various researchers by incorporating computer vision and image processing [3,4,15,16,19-30]. A line of ongoing research studies show variety of approaches have been looked upon for estimating homogeneity in the fabrics majorly used for apparel purposes. Studies in this area has shown quite positive results in detecting fabric faults in woven fabrics and grading them depending on their texture characteristics [15,19,22,23]. The studies have vastly been done for woven fabrics used in apparel. The use of image processing for texture analysis of nonwoven fabrics is little explored. However, fibre orientation and web uniformity has shown positive results, but the area of defect detection in nonwovens need to be explored [62-67]. 2.7 Image Processing: Introduction: Image Processing basically refers to processing of images more specifically digital images by means of a digital computer [68]. The digital images are again obtained by means of various imaging machines which cover almost the entire electromagnetic (EM) spectrum. 24

53 Image Processing The applications of Image Processing now-a-days are numerous and include areas like agriculture, biometrics, character recognition, forensics, industrial quality inspection, face recognition, medical image analysis, security and surveillance, etc Human and Computer Vision System [68, 69] : Nixon & Aguado (2008) defines "Human vision as a sophisticated system that senses and acts on visual stimuli and has evolved for millions of years for survival" [69]. The major components of the human vision system can be categorised as the eye (physical component), a processing system (an experimental model which is not determined precisely) and analysis by the brain. They define "Computer vision system as system which processes images acquired from an electronic camera, just like human vision system where the brain processes images derived from the eyes". It can be interpreted that computer and human vision system are similar in functionality, however the computer vision system may not be able to exactly function like a human vision system as the human vision system is not complete understood yet. The concepts of the human vision system have been vastly used for developing the computer vision systems. The components of computer vision system basically include a camera, camera interface (optional) and a processing unit. Camera is the basic sensing element. The main types of cameras include vidicons (older analogue technology), charge coupled devices (CCDs), complementary metal oxide silicon (CMOS). The processing unit includes computer system and computer software for processing of images. The software includes system development softwares as well as commercial image processing softwares. Use of mathematical 25

54 Literature Review systems/mathematical tools like mathcad, mathematica, maple, matlab, scilab for processing and analysing images Fundamental Steps in Digital Image Processing [68] : Digital Image Processing refers to processing of a two-dimensional picture by a digital computer. A digital image is an array of real or complex numbers represented by a finite number of bits/elements, each of which has a particular location & intensity values. These elements are referred as picture elements, image elements or pixels. It is a basically a matrix (a two dimensional array) of pixels. The value of each pixel is proportional to the brightness/intensity value of the corresponding point in the scene. The important steps involved in image processing are: Image acquisition: Image acquisition is the first step where in the image of an object is acquired using a suitable image sensor which can be any imaging device most common one is the camera as described in the above section. Image enhancement: Image enhancement is the process in digital image processing which improves the interpretability or perception of information in the images acquired for human viewers. It bring backs the detail that is obscured or lost. It may also be used to highlight certain important features of the image. Changing the contrast of an image for better view is one of the tool for image enhancement. It may be described as a subjective pre-processing tool also. Image restoration: Image restoration is a process which improves the appearance of an image by elimination or compensation for any kind of degradations in images. The degradations in images are normally a result of motion blur, noise, camera misfocus, etc. It is an objective process, as the restoration techniques tend to be based on mathematical or probabilistic models of image degradation. Image Compression: Compression is a technique of reducing the storage required to save an image, or the bandwidth required to transmit it. It may be defined as the process of encoding data using a representation that reduces the overall size of data. It allows the use 26

55 Image Processing of the images on platforms, where storage is a limitation, thus allowing its use in wider range of applications. It is widely used in computers in the form of image file extensions, such as the jpg file extension used in the JPEG (Joint Photographic Experts Group) image compression standard. Image segmentation: Segmentation means partition of an image into its constituent parts or objects. It is the process of separating distinct regions containing each pixels with similar attributes within an image i.e. set of pixels separating one image region from other. An autonomous segmentation is considered as one of the most difficult tasks in digital image processing as a rugged segmentation procedure may provide successful solution of imaging problems to identify objects individually and on the other hand weak or erratic segmentation algorithms may be a failure. Accuracy in segmentation influences the success rate of object recognition Image Recognition: Recognition is the process that assigns a label to an object based on its descriptors like pattern matching, face or character recognition, etc. and is widely used for applications like factory automation, monitoring and security surveillance. Colour Image processing: Colour Image processing uses the variations in the colour or the intensity spectrum in an image as the basis for extracting features of interest in an image Methods of Assessment of Image Analysis or Image Quality: It is important to determine the image characteristics before and after processing of images to analyse the images and also to interpret them. The different methods used for measurement of image quality have been discussed here: Subjective Quality Assessment: It is the method which uses subjective experiments and involves human to vote for the quality of image in a controlled test environment. It may be described as a method of analysing images by measurement techniques which provide numerical values that 27

56 Literature Review quantify viewer's satisfaction. It depends on the type, size, range of images, observer's background, experimental conditions like lighting, display quality, etc. It involves ranking of images on a 5-point scale as: bad, poor, fair, good and excellent. As it involves subjective ranking, the ranking may differ from individual to individual and therefore is difficult to design any constructive method for performance improvement. Objective Quality Assessment: It can be described as a method to measure image quality algorithmically. It involves use of image metrics like Peak Signal-to-noise (PSNR) and Mean Squared Error (MSE), etc. to determine general image characteristics. The measurement techniques are easy but they may not consider human visual sensitivities. These methods are not effective in predicting distortion visibility and also for enhancing images with large luminance/intensity variations. A combination of numerical and graphical measures is considered a better means for analysing image quality. Perceptual Quality Assessment: This type of assessment is basically advancement of objective quality measurement as it is based on human visual perception like image discrimination and tasked based models, so as to characterize the variations in quality across the whole image. It suggests methods to analyse images using algorithms which does not require human judgement. Algorithms basically classified into three types based on the assessment of quality of images with respect to a perfect (reference) image : full reference, no-reference & reduced-reference. The algorithms using texture of the images has been considered vastly for automated interpretation of digital images [70-75]. These algorithms follow statistical, structural, spectral approaches or model based approaches [70-75]. The same has been summarised below: Statistical Approach: The statistical properties of images based on intensity variations in the image are used for texture analysis. The statistical properties include mean, variance, skewness, kurtosis, etc. These values are used to discriminate between the images. 28

57 Image Processing Structural Approach: It involves extraction of structural information from the visual scene to define the image. Structural distortion is used to analyse image quality. Spectral Approach: In images following any periodic patterns, spectral approach serves as a solution for image analysis. The periodic patterns are estimated using the frequency domain of the intensity of the images. The same is used to extract the texture characteristics. Differences in the frequency content are used for segmentation of images with different textures. Model Based Approach: It is effective for texture images which are stochastic in nature. The image parameters are estimated to generate an image model which can be processed as well as synthesized. It is useful for modeling natural textures. 2.8 Image Processing & Fabric Inspection: Introduction: As mentioned in the earlier section, image processing operations transform an input image to an output image from which the desirable characteristics can be extracted for interpretation and analysis. As described earlier, image processing is vastly used in variety of areas for monitoring and control. It has been described in Section about the use of image processing in fabric inspection, most commonly the woven fabrics for quality monitoring and control. The inspection process using image processing basically compromises of two main phases :- the learning phase and the detecting phase. A number of images of surface of the fabric are processed in learning phase to understand and analyse the basic image characteristics of the fabrics with/without defects. In the detecting phase, the images of the fabrics with defects are further processed to extract defect characteristics [19-30,76-79]. Automation in the fabric inspection process has been one of the most difficult tasks to be implemented practically and commercially in the textile industry. The implementation, challenges and difficulties, various approaches, components of the system and use of different algorithms have been discussed in this section. 29

58 Literature Review Implementation, Challenges and difficulties: Due to the complexity [80, 19-30, 76-79, 80-87] in the practical implementation of the fabric inspection systems especially for the functional fabrics, the research in this field is widely open. Based on the study of the literature review carried out in this area, the challenges and difficulties in the implementation of the same have been briefed as below: There are a variety of fabric faults obtained differing with respect to nature, size, frequency and severity. Variations in defect characteristics obtained in woven, knitted and nonwoven fabrics. Classification of defects. Characterization of defects in case of nonwovens because of their homogeneous fibrous structure. Grading of fabrics, especially functional fabrics as the grading varies as per the application area of the fabric. Online Image acquisition. Economical quality monitoring systems. Extremely high data flow. Distortion in images. Use of different algorithms for different nature of defects. Nature of the fabric characteristics normally imparts stretch and skew of fabric texture Image Acquisition: Image acquisition is the first and important step in developing and designing any quality monitoring system using image processing. Image acquisition is basically the process of obtaining a digital image from some hardware based source. The source usually depends upon the application area and ranges from a desktop scanner to a massive optical telescope. The image sensing mechanism can be further classified as image acquisition using a single sensor like photo diode, image acquisition using sensor strips like flatbed scanners and 30

59 Image Processing & Fabric Inspection image acquisition using sensor arrays such as CCD, CMOS camera. Trials have been done using on line-scan CCD camera, digital camera, high resolution scanners, laser scanner, optic scanner, etc. Studies related to PC based real time inspection systems have mentioned about the influence of image acquisition on the over image quality, the algorithms used and thus the effectiveness of the system as a whole[some excel sheet]. The fabric inspection usually uses a digital camera (CCD or CMOS) as a source for image acquisition and has been described in detail in the next section Components of Fabric Inspection Systems: An automatic fabric inspection system mainly comprises of 4 sections: Fabric unwinding section, image acquisition section, fabric rewinding section and monitoring and analysis section as shown in figure. The image acquisition, monitoring and analysis sections differentiate between the conventional fabric inspection systems and automatic fabric inspection systems. Figure shows commercially available fabric inspection systems from Elbit Vision System-EVS. FIGURE 2.6 : Automatic Fabric Inspection System 31

60 Literature Review FIGURE 2.7 : Commercial Automatic Fabric Inspection System Elbit Vision System The image acquisition section comprises of a source of image acquisition, source of illumination & frame grabbers. The source of image acquisition is normally by means of a CCD - charge coupled device or a CMOS - (Complementary Metal-Oxide Semiconductor) camera. The comparison between the two has been briefed out in the table below: TABLE 2.4 : Comparison of CCD and CMOS camera CCD Here many signal processing functions are performed outside the sensor. Higher power consumption than CMOS, hence has heat issues. Heat issues can increase interference. CMOS It incorporates amplifiers, A/D-converters and often circuitry for additional processing. Less power consumption than CCD, hence low temperature inside camera. They tend to suffer more from structured noise. There are mainly two types of techniques involved in image acquisition: line and area scan [88, 89]. The main difference between the two has been highlighted in the table below: 32

61 Image Processing & Fabric Inspection TABLE 2.5 : Comparison of Line Scan and Area Scan Camera Line Scan They contain a single row of pixels to capture image. The image is reconstructed in software line by line as the object moves past the camera. They are expensive and used for high end applications. Best suited for high-speed processing or fast-moving conveyor line applications. Area Scan They contain a matrix of pixels that capture an image. The image of the whole area is captured. They are more general purpose than line scan cameras, and offer easier setup and alignment Best suited for applications where the object is stationary. Illumination is a major issue in any image acquisition systems as the type of illumination effects the quality of the image captured [80]. The different illumination configurations have been illustrated in Figure 2.8 [90]. FIGURE 2.8 : Various Illumination Configurations Guruprasad & Behera (2009) [2] suggested that the basis of choice of an illumination depends on the fabric density, defect types and stage in which the inspection is carried out. 33

62 Literature Review Figure 2.9 shows the influence of illumination in a textile material. The front or top lighting is normally used for enhancing surface texture while backlighting is normally used to enhance the structure of translucent fabrics. Literature suggests use of Infra-red lighting [29], fluroscent lamp [91] and halogen lamp [78], however the influence of them is hardly seen and selection of lamp to suit the cost economy would be desirable. FIGURE 2.9 : Influence of Illumination in Textile Materials A frame grabber is an electronic device used for capturing still frames from analogue or digital video stream. Use of frame grabbers is normally used in computer vision system. Use of frame grabbers is expensive and with the use of digital cameras, they can be replaced by any kind of video multiplexer unit. The Monitoring and analysis section comprises of a PC platform with inspection software module. It is the main image processing and analysing unit and the main functions of this section are defect detection and control of image acquisition as well as the whole system. The algorithm used in the software module majorly contributes to the effectiveness of the system. The various approaches incorporated in designing the algorithms have been discussed in the next section. 34

63 Algorithms for Automatic Defect Detection 2.9 Algorithms for Automatic Defect Detection: Introduction: As mentioned in the previous section, the image processing and the defect detection algorithm contributes largely to the effectiveness of the inspection system. The various approaches and algorithms commonly implemented in the image processing have been discussed briefly in Section Since the algorithms play a crucial role in automated defect detection, the various approaches and different algorithms used specifically for automated textile inspection system has been discussed in this section Structural Approaches: Structural approaches use the idea that textures are made up of basic elements appearing in more or less regular and repetitive arrangement. It involves extraction of texture elements to interpret the texture. It is said to work well for regular texture and often to synthesize textures [72,73,74,92,93]. Structural approach using different algorithms like: studying the skeleton and background texture to identify defects [94], defect detection using a texture blobs detection algorithm has been proposed [95]. Images of plain and twill woven fabric samples were used. The algorithms needed to be specific with textures and also a lot of computation was complex. Maximum frequency difference comparison was tried out as an improved solution to blob detection algorithm giving a higher detection success rate [96]. One of the major drawback of this approach was missing out or confusion between the structure of the defect and an irregularity periodic in nature [97,98]. Also the variations in the fabric structure lead to complications in the extraction of texture primitives [3, 80]. 35

64 Literature Review Statistical Approaches: Statistical approaches propose measurement of the spatial distribution of intensity values, which are defined by pixel values in a digital image. On the basis of statistical behaviour of results, the texture analysis is done as well any deviation found from the mean values may serve a reason for identification of the defects. While following this approach, an important assumption which is made is that the statistics of the defect free region is stationary and these regions extend over significant portion of images [3, 80]. This approach normally attributes to the study of the intensity pattern within the image normally defined by the intensity distribution within the image. Brief introductions to some of the methods using statistical approaches have been described below: Thresholding Approach: It involves the study of the gray intensity values [9,100,101,102]. It is also referred to as gray-level thresholding. Simple, bi-level and multilevel thresholding are the common ways of using this approach. Simple gray-level thresholding is direct and can be used to detect a defect with high contrast. The intensity histogram of this type of images normally shows a single peak clearly defining the threshold value. The gray level higher or lower than the threshold accounts to a defect. This approach has been used in developing fabric inspection system for woven fabrics and have been able to identify defects like big holes and dark stains in woven fabrics [3,80,103]. Also the implementation of this kind approach is quite easy, but cannot be used to detect all kinds of defects. Dust particles, lint and light conditions may introduce false alarms. Studies using different methods of thresholding include simple thresholding methods, global thresholding methods, adaptive thresholding methods, otsu or automatic thresholding methods. Islam et all. suggests to use a decision tree for determining a general threshold as different weaves and structure of the fabric may lead to different threshold value [15]. Use of fast adaptive thresholding limit to detect low contrast defects in galvanized metallic strip has been studied upon [104]. Correlation: Cross correlation (between images) and autocorrelation(within image) are the normal techniques involved in this approach. 36

65 Algorithms for Automatic Defect Detection Normalized cross-correlation approach: A correlation between the database of images with defect and free of defect is derived and defined by a cross correlation function [3,80]. This function provides a measure of similarity between the images and significant variation is measured, which indicates presence of defect. Autocorrelation function (ACF) approach [99,105,106, 07] : The characteristics of the repetitive structure may be extracted as a part of pre-processing and then a correlation between the image itself and the image translated with displacement vector is measured. Therefore it can be considered quite similar to the power spectrum of fourier transform. The intensity of the maxima is supposed to be constant for a repetitive primitive for an image of defect free fabric while it will change dramatically in an image with fabric defect and therefore this approach have been studied for fault detection in woven fabrics but the main limitation is it cannot analyse a texture without a reference of primitive [108]. Hoseini et all (2013) used Otsu's approach for thresholding for patterned and plain woven fabrics [107]. Histogram based Approaches: It involves study of the histogram i.e. the intensity distribution of the image and the commonly used methods are use of statistical moments, cumulative histogram or rank order and enhancing contrast using histogram. Statistical moments approach - It involves obtaining statistical parameters of the image intensity distribution or the histogram like mean, standard deviation, skewness and kurtosis and therefore is quite simple in implementation though considerable preprocessing of the images may be required for images with non-uniform illumination conditions. Texture features are obtained directly from the gray level image by computing moments in local regions [29, 100]. It is useful in segmenting binary images containing textures with iso-second order statistics. Rank order- The rank-order function is also based on histogram analysis, the difference being the intensity values are sorted in the ascending order also can be termed as cumulative histogram. The histogram and the rank function both provide the same 37

66 Literature Review information, but the rank functions are used as rank distances can be efficiently computed. Also the median and the other rank-order filters based on the rank function have been proved useful in modifying image local properties and facilitating adaptive rank order functions [109]. These approaches have been quite successful in determining defects in ceramic tiles [110, 111]. Since the fabric texture is likely to have a varying spatial distribution, orientation, etc., the approaches based on classical histogram and cumulative histogram analysis has limited its use in fabric defect detection. Studies have been proposed by S. Hariharan, S. A. Sathyakumar, P. Ganesan on measuring of fibre orientation in nonwovens using image processing but not on detection of the faults and their classification [65]. Paper describes the application of image processing techniques for measuring the fibre orientation in nonwovens. Three types of nonwoven fabrics were studied. Spatial uniformity of fibrous structures has been described statistically by using index of dispersion. Results show the technique is capable to identify variation in geometrical dimensions of very small textile objects. Studies propose measuring of fibre orientation in nonwovens using image processing but not on detection of the faults and their classification. By elaborating the digitization algorithm along with numerical methods would give solutions for obtaining characteristics of nonwovens and thus improving the quality. Histogram properties approach [20, 99,105,112] - In this approach, histogram analysis is done i.e. a point-to-point analysis is done. The properties of the histogram may be used to enhance the images by histogram equalization. Histogram equalization operation reassigns gray level values of pixels to achieve a more uniform intensity distribution in the image thus enhancing the contrast in the image. Fractal Dimension(FD): It is basically a ratio providing a statistical index of complexity comparing how the pattern changes with the scale at which it is measured. Fractal methods have been fruitfully used to model the statistical qualities like roughness and self similarity in many natural textures [80,113,114]. Conci and Proenca (1995) suggested use of fractal 38

67 Algorithms for Automatic Defect Detection dimension with differential box counting method, which minimised computational complexity [115]. Eight types of woven fabric defects were studied and the detection accuracy obtained was 96%. However, due to poor localization of fabric defects, chances of false alarms rate were found. Also due to wide range of fabric textures, the range of FD also becomes wide limiting the use of the algorithm. Edge: The gray level transitions are determined by variations in the intensity values and so lines, spots and other spatial discontinuities can be represented using these gray level transitions and this has been used to detect fabric defects [116,117]. This approach has been found suitable for images of plain woven fabrics with low resolution [116,117]. The main limitation of this approach is that the gray level transitions are influenced by the noise generated due to the structure of the fabrics and therefore can result in false detection of faults. Lai et all. (2005) have developed a Nonwoven Defect Detecting Method (NDDM) using gradient conversion and watershed transformations for segmentation of nonwoven images into basically too thick and too thin regions [82]. Textures of defect free nonwovens and the ones with defect both have thick and thin regions and the regions which are too thick or too thin are considered to be regions with defects. The methodology suggested by them involved determining region average i.e the normal deviation from mean. The images which had deviations more than the normal deviation were identified as defected regions. The study included images of three types of nonwovens (standard as well as defective)- thermal calendared, needle punch and spun bond. The defects included tears, folds and heavy spots. The NDDM was effective in finding the too thick and too thin areas, however it lack the identification of defects. Morphological operations approach: Morphological operations basically involve extraction of boundaries and skeletons if any within the texture images by erosion and dilation [3,109,118,119]. These operations result in better smoothing, sharpening and noise removal. This approach has been proved quite well along with multi level thresholding and critically selecting the structuring elements. A simple thresholding pre processing 39

68 Literature Review operation to obtain a binary image resulted in missing out of commonly occurring defects of woven fabrics when processed with morphological operations. Use of Filters: Use of appropriate filters as a part of pre-processing an image facilitates discrimination between two textures which can be obtained by using of various filters in the learning phase. Optimal Filters, Eigenfilters or Independent Component Analysis approach have been found to be used for removal of noise in the images [80, ]. The eigen filters adapt automatically to the class of the texture to be treated and therefore has found its use over traditional approaches. They are useful in separating pair wise linear dependencies between image pixels but not very useful for higher order dependencies. Since the most fabric textures are found having higher order relationships among image pixels, a fabric defect detection using Independent Component Analysis (ICA) was tried upon. But these approaches are highly sensitive to local fabric distortions and background noise and therefore limited to its implementation for fabrics with random structures. Local linear transforms Approach [3,80,129,130] : It uses common bi-dimensional transforms like Discrete Cosine Transform (DCT), Discrete Sine Transform (DST) or Discrete Hadamard Transform (DHT) as a statistical justification for the extraction of texture properties. As mentioned earlier the important information of fabric textures is contained in higher order relationships among the pixels, obtaining this information by using a simple histogram along selected axes in the space of pixel values in a specified neighbourhood was tried upon. Also this method was tried out as a substitute for eigen filters. Co-occurrence matrix approach [3,70,80,105, ] : This approach is one of the popular statistical approach used for texture analysis and has been used for characterization of textures such as grass, wood, etc. There are normally repetitions in the gray level configurations in any texture and this is used as the basis of co-occurrence matrix, which contains information about the positions of pixels having similar pixel (gray level) value. This information is used to approximate the probability distribution function of any texture 40

69 Algorithms for Automatic Defect Detection thus defining the model of texture. Texture features such as energy, entropy, contrast, homogeneity and correlation are derived from the co-occurrence matrix and has been used to characterize various textures. Salem & Nasri (2011) used the GLCM approach for weave identification in woven fabrics found it quite useable for identification of satin, twill and plain weaves [133]. Raheja et all.(2013) proposed a defect detection system for woven fabrics recently. They implemented it on an embedded DSP platform, where they used GLCM to extract textures of the fabric [134]. The energy feature was found to be the best of use to determine the defects in the texture. Pritpal and Prabhjyo (2015) implemented this approach for studying its effectivity in woven fabrics [135]. They observed shift in the values of the GLCM parameters in the images with fabric defects compared to the one void of defects. This technique was found to be computationally expensive thus limiting it s use for online fabric inspection. Also the number of gray levels needed to be reduced to keep the size of co-occurrence matrix manageable, which might reduce the accuracy of the method. Iivarinen & Rauhamaa(1998) have developed a method for inspecting web surfaces using a second order statistical measure of gray level variration i.e. co-occurence matric [86]. The shape features, gray level histogram and texture features extracted were used for classification of the defects. However, the said study was used for images of paper web and not the textile web. Neural-Networks Approach [3,15,80,81,105, ] : Artificial neural networks are basically a family of models or network of neurons inspired by biological neural networks and are used to estimate functions depending on large number of known or unknown inputs. They are specified using their architecture (variables in the network and their topological relationships), activity rule (the change in activity of neurons in response to each other) and learning rule (understanding of the activity of neurons and its dependence on the target values). They are considered among the fastest and most flexible classifiers, especially for the fault detection owing to their non-parameteric nature and high ability to define complex decision regions. Hence they are being used for monitoring and control in the 41

70 Literature Review wide range of manufacturing industry. They have been found to be non scalable to massive datasets. However, fabric defect segmentation using artificial neural network(ann), feed forward neural networks (FFN), back propagations neural network (BPN) with fuzzy logic, pulse coupled neural network(pcnn) have been implemented. They have been found to be quite successful in automatically detecting the fabric faults. Liu et all. (2011) have done a comparision of the bayesian neural network and LVQ neural network for visual uniformity recognition of nonwovens and have found bayesian neural network more efficient for classifying nonwovens based on uniformity [81] Spectral Approaches Spectral approaches are based on spatial-frequency domain features that are less sensitive to noise and intensity variations. They are largely being implemented in the latest computer vision studies as they simulate the human vision system. They have overcome the drawbacks of many low-level statistical methods. The main limitation to this kind of approaches is they require high degree of periodicity, so have successfully implemented in defect detection of uniform textured materials like woven fabrics. In spectral-domain approaches, the texture features are normally extracted using the Fourier, Gabor and Wavelet Transform. The same has been discussed in the following sections: Fourier Transforms: It is widely used to characterize textured images in terms of frequency components. They offer good noise immunity and enhancement of the periodic features and therefore can be used to extract the power spectrum of images of defect free fabrics to determine or modelise the general fabric structure. In an image with defect, the general fabric structure is disturbed leading to change in corresponding intensity in its frequency spectrum at the specific positions of the defect. Various methods of fourier analysis like Optical Fourier Transforms (OFT), Discrete Fourier Transforms (DFT), Inverse Discrete Fourier Transforms(IDFT), Fast Fourier Transforms (FFT) and Windowed Fourier Transforms(WFT) have been studied for its implementation in detecting defects in woven fabrics. 42

71 Algorithms for Automatic Defect Detection OFT are obtained in optical domain by using lenses along with spatial filters and the defect detection in woven fabrics using them have found to be relatively easy and fast [80]. It involves modulation of the luminous intensities of the zero and first order diffraction patterns by the existence of fabric defects [118]. These techniques are suitable for global defects and to some extent local defects but not suitable for textures with difficulty in differentiating defect free and defected regions. DFT and IDFT are useful for their digital implementation as they recover images from spatial domains. The images of textures of woven fabrics are characterised by the warp and weft yarns in terms of warp and weft patterns. Each set of yarn can be modulated by its profile using 1-D. Sarraf and Goddard (1996) have used the local statistics of these 1-D parameters to administer yarn densities and offering a future scope to use the technique for defect detection in woven fabrics [139]. A method using DFT and Hough transform was implemented by Tsai and Heish(1996) [140]. 1D hough transform was used to remove the line patterns of the woven fabric textures by detecting the high energy frequency components in the fourier domain image, which were set to zero. It was back transformed into a spatial domain image. After restoration of the image, the defect free images had homogeneous line region had approximately uniform gray level, while in the defective regions it was not the case. The DFT based approaches have not been found suitable in the images of the fabric in which the frequency components of the defects and the fabric structure are mixed up in the fourier domain restricting its use in patterned fabrics. Use of 2 D Fourier spectral analysis has been used for characterizing nonwoven web structure globally. The mean image gray level at different positions along the machine direction and across the cross direction of the web was analyzed to check the uniformity of the web [64]. FFT and WFT have the ability to localize and analyse the texture features in both spatial and frequency domain thus giving scope for it s implementation against the above methods especially for local defects 43

72 Literature Review Gabor filters approach: When the window function is guassian, the WFT becomes gabor transform. It is found to achieve optimal localisation in the spatial and frequency domain [3,105, , ]. Research shows implementation of gabor filters in texture analysis and has been found quite similar to texture recognition abilities of human brains [141,142]. Junfeng & Huanhuan(2011) implemented this approach for defect detection in woven fabrics. A new method has been suggested here [126]. About 5 kinds of defects for woven fabrics have been considered here. The experimental results obtained confirmed the satisfied performance and the low computational requirement and the performance of algorithm. The method was not suitable for all textile flaws. As the method was not suitable for all textile flaws, and so gave scope for further work to be done in the area. Trials using a bank of gabor filters has been done for it s implementation in defect detection systems in woven fabrics [ ]. Filter banks are set of filters with predetermined parameters in frequency and orientation. They may turn out to be computationally intensive. Recent study has been carried out by Kang et all. (2015) for defect detection on printed woven fabrics. They have used a genetic algorithm to construct the optimal gabor filter, which can remove noise of fabric background to facilitate segmentation of defects [124]. However, the defects which were not clear with respect to fabric background were found difficult to be recognised. Optimized Finite Impulse Response (FIR) filters approach: The above spectral approaches have not been completely able to identify fabric defects that produce very subtle intensity transitions. Therefore use of Finite Impulse Response filters (FIR) has been explored in the said area. The implementation of these filters is of computational ease as they have more free parameters than gabor filters. Wavelet analysis (transform) approach: The wavelet transform is considered to be one of the standard tool for image processing for multi resolution analysis and is being explored for image compression applications. The use of this approach has been explored in the area of defect detection in textiles [127,143]. Sari-Sarraf and Goddard(1999) developed a defect detection system for woven fabrics using 2-D discrete wavelet transform. The system could detect small defects and had an overall detection rate of 89% [127]. Rallo et all. (2009) proposed an algorithm which combines gabor analysis of the sample image with a statistical analysis of the wavelet coefficients [128]. The algorithms were tried for image 44

73 Algorithms for Automatic Defect Detection samples of paper and woven fabric. However, it failed to identify the main fabric structure and so could not built the set of filters in the frequency domain. Studies by Liu et all. (2011) on comparision of neural network for visual uniformity recognition of nonwovens have used a Gaussian density (GGD) model in wavelet domain for texture analysis of nonwoven images [81]. Another approach using Laplacian pyramid has been implemented by Weickret (1999)] for online grading of nonwoven fabrics [21]. The algorithm calculates a measure of quality by analysing cloudiness (concentration of fibres) using Laplacian pyramid decomposition. The pyramid needed to be modified at boundaries to reduce errors. It was found more successful than use of wavelet and fourier approaches. The use of algorithm was restricted to assessing only the cloudiness of nonwovens Model-based approaches: The model based approach involves capturing of the process that generates texture by generating of model of texture by determining the parameters of a pre-defined model. The application of this approach has been tried when statistical and spectral approaches have failed for detection of faults in fabrics. It involves synthesizing of textures and therefore requires the image features at different levels of details matching the possible models of different image classes. Thus the task becomes very tedious and computational intensive when large number of models is considered. Some model based approach like Gauss Markov Random Field (GMRF) Model, Poisson s Model and Model-based clustering have studied for it s use in the texture analysis based defect detection [80, 101]. The study of the approaches used for defect detection in the fabrics shows that each of the approach has it s own advantage and limitations. Therefore combination of the various approaches have been explored like combination of more than one statistical approach like thresholding and morphological or neural networks [15,78,136], statistical and spectral methods [22, 70,81, , 145]. 45

74 Literature Review Studies by Liu et all. (2011) on comparision of neural network for visual uniformity recognition of nonwovens used combination of statistical and spectral methods for grading of nonwoven fabrics [81]. They have proposed the Gaussian density (GGD) model, LVQ and Bayesian neural network. The authors proposed the, LVQ and Bayesian neural network. However, its recognition rate has to be improved. The paper describes all the aspects of the proposed research in it. Introduction of neural networks increases the complexity in the study and therefore increases computation time to improve the recognition rate. The paper proposes comparison of neural networks rather than going for simple classification of the faults in nonwovens as required by industrial point of view and therefore the study is limited to validation from the point of view of experiments. The paper is mainly focused on uniformity recognitions of nonwovens. So the objectives of research should be centered around the quality of nonwovens only. But here in this article it is related to mere comparison of complex techniques to be used for finding uniformity of nonwovens rather than finding out some other simple solutions which may be useful in the industry. The paper suggests method which is valid from point of view of experiments, the problem regarding detection and grading of faults in nonwovens still remains for the industrial applications to be explored in future. Also only one variety of nonwovens has been considered over here and solutions for other types of nonwovens/functional fabrics are required to be developed Fabric Grading: Introduction: A proper system of defect analysis or defect grading becomes a necessity for the commercial viability between the sellers and consumers. The simplicity, accuracy and easy execution of the grading system are benefitted by one and all in the business. The flow based and point systems are the most common ones being used for the woven and knitted fabrics being used for apparels. These systems are basically based on size, number and frequency of the defects. The grading systems have been discussed briefly in the next section. 46

75 Fabric Grading Various Approaches [14,146,147] : Flow Based/Metric System of Measurement: The defects are basically classified as major and minor defects: Major: A defect severe enough if exposed to place an end item in seconds. Minor: An imperfection that may or may not cause a second, depending upon its location in the end item and/or its chance of being lost in fabrication. The width of the fabric is measured in cms. Substitution of 25 cms for 9 increment of defect measure is to be done. The results are calculated in points per 100 square meters. Point Systems: In the point system penalty points are given to defects depending on the size and orientation of the defect. The most common ones are 4-Point system and 10-Point system implemented by the textile industries. The Four-Point System: It assigns 1-4 penalty points according to the size and significance of the defect. Points more than 4 are not assigned for any single defect. It remains same for defect in either length or width direction. Only major defects are considered and no penalty points are assigned to minor defects. Total defect points per 100 square yards of fabric. Fabric rolls containing more than 40 points are considered "seconds" and the grading is the same for any end product. TABLE 2.6 : Penalty Points in 4-Point System Length of Defect Penalty Points Upto 3 inches inches inches 3 Over 9 inches 4 Holes and Openings (1inch or less) 2 Holes and Openings (over 1inch) 4 47

76 Literature Review The Ten-Point System: In this system, the fabrics are graded as first and second quality. It assigns 1, 3, 5 and 10 penalty points according to the size and significance of the defect. Different points are to added for defects in warp and weft directions, thus making the system difficult to use. TABLE 2.7 : Penalty Points in 10-Point System Warp Defect Weft Defect Penalty Points Upto 1 inches Upto 1 inches inches 1-5 inches inches 5 inches to the half of fabric inches Full width 10 There is no literature available for this kind of grading for nonwovens as well as functional fabric. However, the Smart View WQM automatic fabric quality system from M/s. Cognex grades the nonwovens as Poor, Usable, Good, Better, and Outstanding [148]. The system gives the grading in terms of grade value from 1 10 depending how the material looks like. A quality value of 1 is reported when the material is of poor quality; value of 5 is reported when the material is of good quality or value of 10 when the material is outstanding. FIGURE 2.10 : Grading of Nonwoven Fabrics 48

77 Summary of Literature Review 2.11 Summary of Literature Review: From the literature survey carried out, it can be concluded that need for an efficient, consistent cost effective quality control system is always a basic requirement for the textile industry. The industry especially dealing with the functional fabrics is expected to deliver first quality fabrics. It is also understood that the defects or faults in the fabric tend to affect the quality grade of fabric and often affects the pricing of the fabric. Thus accurate defect detection is a very important and traditionally visual inspection of the fabrics is very common at present in the local textile industrial sector. But due to various drawbacks of the traditional visual inspection systems like the low detection rate of defect detection, high time consumption, inefficiency of the system owing to fatigue of classer, etc., automated defect inspection system were introduced. These kind of few commercial systems are currently available but are still considered as very expensive. Therefore developing a PC based real time fabric inspection system has been an area of research interest globally. Extensive studies relating to this area show a lot of studies being done in the area of fabric inspection systems for woven and knitted fabrics used for apparels. It was found that defect analysis was done by image analysis or rather texture analysis using various image processing algorithms. The various approaches and algorithms using computer vision and image processing in the area of defect detection in various textures have been studied. It was found that each method had it s own advantage and disadvantage. The survey shows that most of the studies carried out involved a very few varieties of fabric and also were limited to executing of the algorithm for about only 2 or 3 fabric samples with different specifications of fabrics of similar nature have been considered. A study considering more fabric samples with different fabric specifications but of similar nature requires to be done. Also very limited studies have been done in the area of inspection systems for functional as well as nonwovens. Owing to the fibrous structure of nonwovens, the texture is completely 49

78 Literature Review different from woven fabric giving scope for exploring the method of implementation of various approaches for texture analysis. It was also found that most of the implemented PC based inspection systems used high quality equipment for image acquisition, and therefore a study using low cost camera could be explored. It was also found that a grading system does not exist for functional fabrics and nonwoven. The scope of the thesis thus aims at developing cost and quality effective device for a quality monitoring and analysis of functional as well nonwoven fabrics. 50

79 CHAPTER 3 Proposed System & Research Approach 3.1 Problem Description: Extensive literature survey of the papers gives an idea of the different approaches that have been considered in designing quality control systems in the area of textiles. The study has been summarized in Section The problem thus can be extracted out from the summary of the literature review and can be described as follows: A study exploring the use of low cost acquisition devices with scalable algorithm suitable for different fabric specifications of woven, knitted or nonwoven fabrics. A need for cost and quality effective device for a quality monitoring and analysis of functional as well nonwoven fabrics.

80 Proposed System & Research Approach 3.2 Proposed System: Thus a cost and quality effective device for functional/nonwoven fabrics has been intended to be developed during this research work. The developed system will basically have two modules hardware and software. The hardware components include a fabric quality monitoring device having a fabric delivery roller, fabric take-up roller, motorised driving arrangement, PC or a laptop and a cmos camera for image acquisition. The software module consists of an application for monitoring and analysing the fabric. The system will be described in detail in the next section. 3.3 Research Approach & Hypothesis: Qualitative as well as formulative approach has been used for this research work. A device was developed as a part of the quality monitoring system. The important parts of the device are fabric delivery roll, illumination for capturing images of surface of the fabric, CMOS camera for capturing images of surface of fabric and a fabric take up roller. Functional/nonwoven fabrics were manufactured. Defects in the fabrics were a result of the manufacturing process. Two categories of fabrics were selected for the study: Woven Geotextiles and Spunbond Nonwovens. Different GSM fabrics in each category had been sampled. A series of images of the selected fabrics taken by the camera with the help of the device were processed for studying the general features of the acquired images using MATLAB. The histogram trend for all the images was studied. Images of fabrics with common identified defects were acquired and processed using MATLAB for the general image parameters. The defects were detected basically using a unique algorithm developed for texture analysis. The defects were graded and compared against manual - visual grading of the same. 52

81 Research Approach & Hypothesis The algorithm was validated by processing it for multiple images of the same defective regions. The algorithm was also checked for multiple images of different fabrics for validation. The proposed system will therefore check for variability and give defect statistics and classify as per Defect Area. It will also check for no. of Defects in the Fabric Lot and give % Defects in the Fabric. On the basis of the defect statistics a fabric grading system has been developed which will classify the fabric for specific application. 53

82 CHAPTER 4 System Design & Development 4.1 Introduction: The main object of the research work is developing cost and quality effective system for functional as well nonwoven fabrics. Therefore based on the literature review done the said system designing was the first step. Efficient fabric inspection system is the prime requirement of online quality control of the fabrics. The main components of the fabric inspection system are: Fabric Unwinding Section It has a fabric delivery roller on which the fabric roll to be assessed is mounted. Image Acquisition Section It is the section where the images of the fabrics are captured using a CMOS camera. Fabric Rewinding Section After the capturing of the images the fabric is to be rewound and which is done by a fabric take-up roller. Monitoring and Analysis Section (Hardware and Software) It comprises the PC and the software module for analyzing the fabric.

83 Device Development The process flowchart followed for the development of the system has been demonstrated below: Device Development Image Acquisition Fabric Sampling Classification Image Processing Enhancing Images Extracting Features FIGURE 4.1 :- Process flowchart of the developed system 4.2 Device Development: Device development was the first and integral step to achieve our main object i.e. a cost and quality effective system for functional/nonwoven fabrics. To design a device which was cost effective as well as suitable for the said study, the development of device was done in phases. The first phase involved development of a device for optimum illumination and optimum camera distance from the fabric for getting considerable quality of images. In the second phase, the device was advanced for capturing of images in roll form, since that would be a requirement for online image acquiring system. In the third and last phase, the automation of device was done. The software module was developed using MATLAB and the code for the same has been shown in the appendix B. 55

84 System Design & Development Development of manually operated device: The device was initially designed mainly for the image acquisition of the fabric samples in general. Therefore a manually operated device was designed and developed and the same has been illustrated in Figure 4.2. As mentioned in the literature review, the configuration of illumination influenced the overall processing and the efficiency of the system, provision for illumination from bottom and top both were given. The main parts of the device included a frame / stand, scan box with provision from inside, top lights for top illumination and a CMOS camera. A slot for adjustment of camera height was provided on the side arms on the machine frame. The scan box had provision for illumination from inside. The design of the base of scan box and a photograph of the scan box has been shown in Figure 4.3 and Figure 4.4 respectively. Two 20 Watts CFL bulbs (which were then replaced with LED bulbs for better outcomes) were mounted inside the box after the box was prepared for illumination. White acrylic sheet was used for cover. Aluminium foil was provided in bottom surface of the box for getting better reflecting properties. FIGURE 4.2 : Manually operated device 56

85 Device Development FIGURE 4.3 : Design of Box Base FIGURE 4.4: Photograph of the developed scanbox. 57

86 System Design & Development The fabric samples cut of random size suitable to be laid flat on the scan box and were kept on the scan box for image acquisition. Images of fabric were taken using the CMOS camera which was connected with a laptop via a usb cable. The quality of images was checked for top and bottom illumination. Also trials were taken for checking the optimum distance of camera from the fabric surface to be monitored. Once the optimum height was obtained, the device was further modified for image acquisition of fabric in roll form Further modification of device for image acquisition of fabric in roll form: To achieve similar conditions as online monitoring, the device was further modified for the image acquisition of the fabric in the roll form. Figure 4.5 shows the photograph of the modified device. The main parts added to the device included a fabric let off/delivery roll, fabric take up roll, a handle and driving for positive control of fabric. FIGURE 4.5: Photograph of the Modified Device 58

87 Device Development A chain and sprocket arrangement was used for positive control of the fabric from one end (fabric delivery) to another (fabric take up). The fabric was cut to a width of 18cms and wound onto the spindle of the fabric let off roll. The spindle was mounted on the frame and it was positively driven by a chain and sprocket arrangement mounted on both the sides of the spindle. The open end of the fabric was wound on the take up roll. The fabric could now be moved from one end to another by means of a handle, which was mounted on the fabric take up roll spindle. A number of images of the manufactured fabric were acquired using this device to find out the general parameters of the fabric images. Also the images for the fabric regions with selected identified defects were acquired for further processing Automation of the Device: In the last phase, the automation of the device was done to offer scope for continuous monitoring of fabric in running condition. The manual-handle driven device was transformed to a motor driven device. The physical modifications in the device included motorised driving arrangement, vertical top illumination, frame for camera mounting, replacement of the fabric take up and delivery roll spindles by package holders, covers for the device along with the switches for operation of the device. The Figure 4.6 shows the main parts and the driving arrangement of the fabric delivery and take up roll. An AC motor with 20rpm, 100mA & 7 kg cm torque is mounted on the machine frame at the front. The drive from the motor is positively transferred to the fabric take-up roll by means of a chain and sprocket. The fabric delivery roll can be driven both positively and negatively. Sprockets are provided on each side of the package holder, which acts as dead weight for facilitating negative driving of the fabric delivery roll. Positive driving arrangement of the fabric take up roll was facilitated by providing side chain on the sprockets. 59

88 System Design & Development FIGURE 4.6 : Drive and main parts of the Device Package Holders were designed so as to facilitate easy change of fabric rolls. The design and the photographs of the final machine is shown in Figures from Figure 4.7 to Figure

89 Device Development FIGURE 4.7 : Different View Angles of Developed Device a) Back View b) Side view c) Front view d) Side view 61

90 System Design & Development FIGURE 4.8 : Top View & Illumination Arrangement FIGURE 4.9 : Side View of Machine with Switch Board 62

91 Device Development FIGURE 4.10 : Passage of Fabric FIGURE 4.11 : Final System 63

92 System Design & Development The switch box on the top machine cover consists of 4 main swtiches:- Switch 1 Main switch for power supply to the device. Switch 2 for starting the motor Switch 3 - for using Top Illumination Switch 4 - for using Bottom Illumination A power and an usb cable are provided for getting the power supply to the machine and inter-phasing with a PC/Laptop respectively Working of System using the Device: The stepwise sequence of operating the device has been described below: Step 1: The System (Device and PC/laptop) is made active by supplying power using the Switch 1. Step 2: The method of illumination is selected using switch 3 (top illumination), switch 4 (bottom illumination), both (top & bottom illumination simultaneously) or none (for natural illumination). The selection of illumination depends on the nature of the fabric to be used. The fabrics studied as a part of this research have been found suitable for top illumination and the same has been used for capturing of the images. Step 3: The fabric to be checked is selected using the software module via. PC/laptop. The screenshot of the same has been shown in Figure

93 Device Development FIGURE 4.12 : Selection of Fabric Type Step 4: The process of fabric monitoring is now started by pressing switch 2 which will start motor so the drive to take-up roll will be transmitted by chain and sprocket wheel and hence the fabric to be scanned will start winding on this take-up roll. Step 5: The images of the fabric are acquired using the capture option from software module via. PC/laptop. The screenshot of the same has been shown in Figure Step 6: The command for the fabric grading is given using the process option from software module via. PC/laptop. The screenshot of the same has been shown in Figure The final output of the system is shown in Figure

94 System Design & Development FIGURE 4.13 : Capture & Process Option FIGURE 4.14 : Final Output 66

95 Fabric Manufacturing and Defect Analysis 4.3 Fabric Manufacturing and Defect Analysis : Introduction: Since a quality control system for functional fabrics was intended as a part of this research work, the next step after development of the device was to produce functional fabric of different segments. 6 different varieties of functional fabric including woven as well as nonwoven fabrics were manufactured for the study. The details of the fabrics have been described in the next section. Since a quality control system was intended to be developed for different functional fabrics, it was decided to consider the defects occurring commonly in the said varieties of the fabrics. So the defects considered were the defects which were obtained in the manufactured fabric as a result of the fabric manufacturing process. Different types of defects as mentioned in Section were found in the manufactured fabrics. However, the experts from IIT had suggested to consider only one variety of fabric preferably spunbond nonwoven fabric during the Research Week held during month of April 2015 at Gujarat Technological University, Ahmedabad. They had also suggested considering some of the major defects occurred during the manufacturing of spunbond fabrics. Also, they suggested validating the results so obtained by taking multiple images of same defects. After considering the inputs from the experts of IIT, the study has been narrowed down to 2 varieties of functional fabrics i.e. Woven Geotextiles & Spunbond Nonwovens. 6 types of defects in each variety have been focused on in the study. The description of the selected defects in the said two varieties has been made in Section

96 System Design & Development Fabric Sampling: The details of the fabrics manufactured have been described in this section. The fabrics produced were of full width as they were manufactured at the local industries. The fabric manufactured was cut to obtain a fabric width of 18cms as per the requirement of the device developed. Fabrics of varying specifications had been manufactured in each category. After the cutting of the fabric width, the fabrics of varying specifications were stitched to obtain a roll. The roll so obtained was ready to be monitored and analysed by the developed device. Woven Functional Fabrics: Geotextiles Manufactured at M/s. Technofab, Udhana Magdalla Road, Surat. Machine Specifications: Sulzer Projectile Loom PU model loom 3.5 m and 5 m width Speed-230 rpm for 3.5 m & 180 rpm for 5 m TABLE 4.1 : Specifications of Geotextiles Sample Name epi x ppi warp x weft Denier GSM G1 38 x x G2 38 x x G3 34 x x G4 38 x x G5 21 x x G6 36 x x G7 34 x x

97 Fabric Manufacturing and Defect Analysis Industrial Fabrics Manufactured at M/s. N.M. Gajjar Hotels Pvt. Ltd., R.S. No. 286/1, Block No. 229, Mota Borasara, Village Kim, Surat. Machine Specifications: Water Jet Loom TABLE 4.2 : Specifications of Industrial Fabrics Sample Name epi x ppi warp x weft GSM Denier I1 48 x D x 200D 56 I2 60 x D x 200 D 78 I3 40 x D x 200D 80 I4 60 x D x 200 D 86 I5 44 x D x 390 D 204 I6 30 x D x 390 D 100 I7 32 x D x 400 D 100 I8 40 x D x 360 D 120 I9 36 x D x 234 D 66 I10 40 x D x 261 D 90 I11 36 x D x 185 D 45 I12 68 x D x 77 D 66 Coated Fabrics for Composites The fabric was manufactured at SSI Unit Vapi and coating was done at SCET, Surat. All fabric samples were coated using padding mangle. Curing of samples CC1, CC2 & CC3 was done at 150 C and curing of samples CC4, CC5 & CC6 was dome at 170 C. Machine Specifications: Power Loom Speed-130 rpm TABLE 4.3 : Specifications of Coated Fabrics for Composites Sample Name epi x ppi warp x weft GSM Ne CC1(Plain weave without coating) 76 x x CC2(resin treated 100 gpl) 76 x x CC3(resin treated 500 gpl) 76 x x CC4(2/1 twill weave without coating) 64 x 56 8 x CC5(resin treated 100 gpl) 64 x 56 8 x CC6(resin treated 500 gpl) 64 x 56 8 x

98 System Design & Development Nonwoven Fabrics: Spunbond- Manufactured at M/s. Wovlene Tecfab India, A-42/5, Ichchhapore G.I.D.C, Near GEB Substation, ONGC Road, Hazira, Surat Machine Specifications: Chinese make spunbond machine -1.6 m width Capacity : 5 tonnes/day GSM range : TABLE 4.4 : Specifications of Spunbond Nonwovens Sample Name GSM NS1 40 NS2 60 NS3 60 NS4 60 NS5 60 NS6 85 NS7 120 NS8 135 NS9 60 NS10 60 NS

99 Fabric Manufacturing and Defect Analysis Needle Punched Fabrics Manufactured at M/s. Autotech Nonwovens, Fairdeal Textile Park, NH8 Kim. Machine Specifications: Korean make needle punching machine -1.6 m width GSM range :above 150 TABLE 4.5 : Specifications of Needle Punched Fabrics Sample Name GSM NP1 270 NP2 280 NP3 200 NP4 300 NP5 400 NP6 400 NP7 400 NP8 450 Spun Laced (Hydroentangled) Fabrics Manufactured at M/s. Ginni Filaments, GIDC, Panoli. Machine Specifications: Plant Line with capacity of metric tonnes per annum : Perfojet, France GSM range :40 TABLE 4.6 : Specifications of Spun Laced Fabrics Sample Name Type GSM SL1 PV5050P SL2 PV5050A SL3 PV6535E SL4 PV3565P SL5 PV SL6 PV SL7 PV3070E SL8 PV2080P

100 System Design & Development Fabric Defects: Geotextiles: 6 types of common defects obtained while manufacturing of the geotextile fabrics were identified and considered for the study. The details of the same have been described in Table 4.7 and illustrated in Figure 4.15 TABLE 4.7 : Identified Defects in Geotextiles Sr. Fabric No. Defect 1. Missing End (Chira) 2. Slubs (Warp) 3. Stain (Daggi) 4. Slubs (Weft) 5. Missing Pick (Jerky) Definition Principal Causes Remedy There may be one end or a group of ends missing in the fabric. Thick untwisted portion in warp yarn These stains are due to lubricants or dust. Thick untwisted portion in weft yarn It is a strip which extends across the width of fabric & has the pick density lower than the required one. 6. Gout Foreign matter woven in a fabric by accident. Usually lint or waste. If the broken ends are not mended immediately by the operator, these missing ends will occur in the fabric. Variation in draft during spinning. Improper material handling, bad oiling & cleaning practices Variation in draft during spinning. It is caused by faulty let off & take up motions. Also, if the loom is not stopped immediately in case of weft break, few picks are liable to be missed in the fabric. It is caused when the hardened fluff or foreign matter such as pieces of leather accessories, pieces of damaged pickers etc., is woven into the texture of the fabric. This defect can be minimised (a) by minimising missing ends in the weaver s beam & (b) by providing an efficient warp stop motion on a loom. Set the draft as per the requirement. By proper material handling as well as good oiling & cleaning practices, this defect can be avoided. Set the draft as per the requirement. This defect can be remedied by proper setting of let off & take up motions & also by using an efficient brake motion. This defect can be remedied by preventing the foreign matter from falling onto the warp between the reed & the fell of the cloth. 72

101 Fabric Manufacturing and Defect Analysis FIGURE 4.15 : Defects mentioned in Table 4.7 Spunbond Fabrics: 6 types of common defects obtained while manufacturing of the spunbond fabrics were identified and considered for the study. The details of the same have been described in Table 4.8 and illustrated in Figure

102 System Design & Development TABLE 4.8 : Identified Defects in Spunbond Fabrics Sr. No. Fabric Defect Definition Principal Causes Remedy 1. Drops / bond Fused fibres on Breaking of bundle of Proper setting of draw ratio. point fusion surface filaments during the process. 2. Pinholes Very small Damaged surface of delivery Filing of surface of roller. holes in fabric roller. 3. Wrinkles Wrinkle Improper tension across the Maintaining uniform tension. formation width of fabric. 4. Hard Fused filaments Breaking of filaments during Proper setting of draw ratio. filaments on surface the process. 5. Hole Holes in fabric/ web 6. Calendar cut Cut marks due to calendaring Improper supply of polymeric material across the width of fabric, blockage of spinnerette holes. Rough surface of calendar roll. Maintaining proper supply of polymeric material across the width of fabric, cleaning of spinnerrrate. Polishing of surface of roller. FIGURE 4.16 : Defects mentioned in Table

103 Image Acquisition for the Learning Phase 4.4 Image Acquisition for the Learning Phase: Introduction: The first step of the Image analysis process is image acquisition of the fabric images. The texture images of the defective as well as defect free samples of geotextiles and spunbond fabrics were acquired using the device. Initially the images were captured by providing illumination or source of light from bottom or top surface of fabric as well as by adjusting the height of camera with reference to the surface of fabric for optimization of different parameters leading towards the designing of the developed device. More than 400 images were captured initially for the extraction of basic image characteristics of the textures of the defective as well as defect free fabric samples. The images of the geotextile and spunbond fabrics have been illustrated in section and respectively Geotextiles: As mentioned earlier number of images were required to acquire the basic characteristics of the geotextiles. The actual image acquired of a fabric region free of defects has been shown in Figure FIGURE 4.17 : Image of Defect Free Sample 75

104 System Design & Development The actual images of the 6 identified defects in the geotextiles have been shown in the figures from Figure 4.18 Figure 4.23 and the details of the images have been shown in Table 4.9. These images have been used for the study of the defect pattern. Multiple images of the same region were taken and used for processing during the testing phase. TABLE 4.9 : Details of the images of the defects in Geotextiles Sr. No. Defect Name Fabric Sample GSM Image No. 1. Missing End (Chira) G Slubs (Warp) G Stain (Daggi) G Slubs (Weft) G Missing Pick (Jerly) G Gout (Foreign Matter) G FIGURE 4.18 : Missing End / Chira 76

105 Image Acquisition for the Learning Phase FIGURE 4.19 : Slub (Warp) FIGURE 4.20 : Stain (Daggi) 77

106 System Design & Development FIGURE 4.21 : Slub (Weft) FIGURE 4.22 : Missing Pick / Jerky 78

107 Image Acquisition for the Learning Phase FIGURE 4.23 : Gout Spunbond Fabrics: As mentioned earlier number of images were required to acquire the basic characteristics of the spunbond fabrics. The actual image acquired of a fabric region free of defects has been shown in Figure

108 System Design & Development FIGURE 4.24 : Image of Defect Free Sample As mentioned earlier 6 types of defects had been identified for the learning phase. The details of the images of the fabrics with defect captured have been shown in the Table The actual images of the 6 identified defects in the spunbond have been shown in the figures from Figure 4.25 Figure These images have been used for the study of the defect pattern and its influence in nonwoven fabrics. Multiple images of the same region were taken and used for processing during the testing phase. TABLE 4.10 : Details of the images of defects in spunbond fabrics Sr. No. Defect Name Fabric Sample GSM Image No. 1. Drop/Bond Pt. Fusion NS Pin Hole NS Wrinkle NS Hard Filament NS Hole NS Calender Cut NS

109 Image Acquisition for the Learning Phase FIGURE 4.25 : Drop / Bond Pt. Fusion FIGURE 4.26 : Pin Hole 81

110 System Design & Development FIGURE 4.27 : Wrinkle FIGURE 4.28 : Hard Filament 82

111 Image Acquisition for the Learning Phase FIGURE 4.29 : Holes FIGURE 4.30 : Calender Cut 83

112 System Design & Development 4.5 Image Processing Methodology: Introduction: The images of the fabrics need to be processed for the extracting their characteristics. As mentioned earlier, there are limited studies carried out in the area of automated quality control system for functional fabrics, the biggest task was to decide the best possible approach and method for the processing of the images. Since the main object of this study was to develop a simple cost effective quality control system for the functional/nonwoven fabrics, a combination of various approaches described in Chapter 2, Section 2.9 was used. Two varieties of fabrics resulting due to completely different manufacturing processes needed to be studied as a part of the study. The fabric structure of the woven geotextiles and the spunbond fabrics is completely different. Therefore different algorithms needed to be used and implemented. The basic steps involved in processing of images were the same and has been described in the next section Steps involved in Processing of Images: The flowchart shown in Figure 4.31 shows the steps involved in processing of images. The algorithm for processing of the images of both the fabric samples have been designed as per the same. A brief discussion about the various steps has been described in this next section. 84

113 Image Processing Methodology FIGURE 4.31 : Steps involved in processing of Images General Image Parameters of the Images: All the images were first studied and were processed for obtaining basic characteristics using gray level conversion, contrast adjustment and studying their histogram. Gray Level Conversion: The images acquired are in true colour as can be seen in the images (Figure 4.32 a ) and they store the information of Red, Blue and Green (RGB) levels for each pixel. The computation involved with the RGB images is more and they also use more of disk space and memory. Thus the first step involved in processing is to convert these images into a gray image. A gray image has only gray level values for each pixel (0-255). In a 8-bit system the image display can show a maximum of 256 gray levels. Figure 4.32 shows a RGB image and grayscale converted image. We have used a simple graylevel conversion approach. 85

114 System Design & Development FIGURE 4.32 : RGB Image (a) & Grayscale converted Image (b) of spunbond nonwoven fabric Histogram: Histogram of an image is basically a graphical representation of the intensity distribution of the grayscale image. It plots the number of pixels for each gray level value. The tonal distribution within the image can be judged by the histogram and therefore it serves as an important tool for analysing the basic characteristics of fabric images. Figure 4.33 shows the histogram plot of defective and defect free region of a fabric. The intensity values ranges from i.e dark to light for an image in 8-bit mode. The range and intensity distribution of all the images for both the fabrics were studied from the histogram. The multiple peaks in the histogram (b) shows the presence of defects in the image. Also it indicates more values of intensity towards the lighter side and this information may be used to determine the nature of the defect. 86

115 Image Processing Methodology FIGURE 4.33 : (a) Histogram of a defect free region of spunbond fabric; (b) histogram of a defective region of spunbond fabric. Contrast Adjustment: The contrast of an image is basically a measure of its dynamic range, or the "spread" of its histogram. As seen from Figure 4.34(a), a grayscale image needs to be enhanced, especially to find any regions different from the background. Therefore contrast enhancements are needed to be done. They are typically performed as a contrast stretch followed by a total enhancement, although these could both be performed in one step. Different approaches may be used for contrast adjustment: contrast stretch and adaptive contrast stretch. A contrast stretch improves the brightness differences uniformly across the dynamic range of the image. 87

116 System Design & Development In adaptive contrast stretch the peaks are located in the histogram and then marching in both the directions until only a small number of pixel intensities are rejected. Both the methods were tried on for the images of geotextiles and spunbond fabrics. Based on the nature of histogram, adaptive contrast stretch was found suitable on spunbond nonwoven fabrics. Figure 4.34 shows the captured original image and contrast adjustment image along with their respective histogram. FIGURE 4.34 : Grayscale Image with histogram (a) & Contrast Adjusted Image with histogram (b) of spunbond nonwoven fabric Noise removal: Noise in an image is a random variation of the colour information in the images and is resulted due to the errors of limitations of the image acquisition process. It results in image pixel values that may not be the true intensity of the real scene considered. The noise in an image is reduced by using an appropriate filtering technique. A filtering technique is also an enhancement of an image by emphasizing or removing certain features. Linear, median and adaptive filtering are the common methods used. Since the filtering technique involves 88

117 Image Processing Methodology emphasizing or removing certain features, there are chances of losing important image information. Figure 4.35 shows a grayscale and filtered image of spunbond fabrics. It can be seen that the gray image has considerable noise which is removed in the processed image, but at the same time in the spunbond fabrics due to their random fibrous structure, important structural information may also be removed and therefore only contrast enhancement of the images was done. FIGURE Grayscale Image (a) & Filtered Image (b) of spunbond nonwoven fabric A lot of noise was found in the geotextile fabric (Figure 4.36a) owing to the highly lustrous surface texture of the same. So the noise removal in the fabric was an important task. We used 2 step filtering method. Firstly a 2D Gaussian filtering of images was done to remove the general noise in the image (Figure 4.36b). In the second step, another filtering operation was done (Figure 4.37c). The pixel weight for each pixel in the image was calculated based on the gradient magnitude at the pixel and was replaced by the same. The gradient magnitude is influenced by the geometric closeness and photometric similarity within the image and it reflects to edge and texture directly. Pixels in smooth regions i.e small gradient magnitude pixels have large weight while those with large gradient magnitude i.e the edge regions have small weight. Thus a high contrast is applied and therefore it has better results in differentiating the influence of noise and the influence of defect. Even the weak edges and minute details of the input image are preserved, while the actual noise may be removed. 89

118 System Design & Development FIGURE 4.36 : Grayscale Image (a) & Filtered Image (b & c) of woven geotextile fabric Methodology for Defect Detection: Histogram analysis during the learning stage showed that the presence of major defects could be easily detected, however the minor defects especially the one whose intensity levels are almost same as the structure itself or the local defects could not be identified. The defects in the fabric in terms of image representation can be basically characterised into two types depending upon it s appearance in the texture images: defects with intensities lower than the mean intensity in an image - pin hole, wrinkles, hard filaments, holes, calendar cut defects with intensities higher than the mean intensity in an image - drop/bond point fusion, stains. After the images were studied for general parameters, the images with defect were then processed. The images were processed using gray level conversion, contrast enhancement and noise was removed. The images were then converted to binary images using the 90

119 Image Processing Methodology optimum threshold for each defect. The binary images were further processed for finding the region of interest i.e. the defects using the morphological operations and finally after segmenting the region of interest the statistics of these regions were studied to classify the defects. Thresholding: Once the images are properly enhanced in a manner that the important information of the images is not eliminated, the next step is to replace the gray level intensity into binary values. Thresholding is the process of converting a gray scale image into binary image by replacing the predetermined gray level intensity values to 0 or 1. In simple words, thresholding methods replace each pixel in an image with a black pixel if the image intensity I(i,j) is less than some fixed constant T (that is, I(i,j) < T), or a white pixel if the image intensity is greater than that constant. Simple or multilevel thresholding methods are commonly used for determining the value of T. Determining of T is an important task since it will decide the information about the pixels to be further processed. The filtering and noise removal process in the images of the woven geotextiles enhanced the difference between the defective and non defective part in the images and so converting the enhanced image into a binary image by selecting an appropriate threshold was found suitable to extract the defective regions in the images. The spunbond nonwoven fabrics are characterised by thick and thin places, and also the images could not be processed for noise removal, so deciding of common threshold for defect extraction was found difficult. Therefore different thresholds were used for extracting the extreme light and extreme dark regions. Figure 4.37 shows a grayscale and binary image of a spunbond fabric after performing simple thresholding. 91

120 System Design & Development FIGURE 4.37 : Grayscale Image (a) & Binary Image (b) of spunbond nonwoven fabric Morphological Operations: The binary images obtained by simple thresholding are normally the distorted ones, the ones influenced by noised and texture as can be seen in Figure The morphological operations thus intend to correct these errors in the image so that images are more accountable. These operations can also be applied to grayscale images. They include operations like erosion and dilation in the images considering the corresponding neighbourhood pixels and work in accordance with structuring element (SE). SE is a small binary image; a small matrix of pixels and basically has the same role as convolution kernels in linear image filtering. As can be seen in the Figure 4.38b, the morphological operations allow to add or to remove pixels from the corresponding neighbourhood. FIGURE 4.38 : Binary Image (a) & Dilated Binary Image (b) of geotextile woven fabric 92

121 Image Processing Methodology Feature Extraction: A feature is basically the "interesting" part of an image or the region of interest in an image, and these features are used as a starting point for many computer vision algorithms in computer vision and image processing. The concept of feature detection refers to methods that aim at computing abstractions of image information and making local decisions at every image point whether there is an image feature of a given type at that point or not. The resulting features will be subsets of the image domain, often in the form of isolated points, continuous curves or connected regions. Edges, corners, blobs (regions of interest) and ridges are the common types of features. Once the desired binary image free from noise and background is obtained, the regions of interest (defects) if any need to be identified and studied. The techniques involved in the feature detection and extraction have been used for the same. After the detection of these features, necessary attributes, such as the edge orientation, edge detection, statistical features, shape and other parameters can be obtained and computed as per requirement. Common approach was found suitable for both the varieties of fabrics. The connected components in the binary images were found out, which are known as objects or blob. Each blob has it s own statistics which were found out. Stepwise elimination of these blobs was done to identify the defective regions. The blobs with minimum area which cannot be termed as defect were identified and eliminated in the first step. In the second step the features of the remaining blobs were obtained and the general parameters for the defects were calculated. The same were compared with the statistics obtained by manual - visual examination for validation and classification. 93

122 System Design & Development Software Used: Our proposed algorithm, technique and their all optimizations were accomplished during this study by implementing several Matlab scripts (Appendixes B) Stages of Implementation: For the final implementation of the image processing techniques and the system, the system needed to pass through the learning and the testing or validation. Learning Stage: It is known also as the training phase. It involves study or inspection of the fabric images with no defects or may also be called as the standard images. More than 200 images of both the fabric samples were captured. The main object during this phase is to calculate the important parameters of these images like mean intensity of the images, study of histogram and it s properties. These values were used to determine the image processing parameters for the detect detection. After the learning of the basic parameters of the images free from defect, images of fabrics with defects were processed for learning the pattern of each type of defect. The processing algorithm was designed by optimising various parameters at each of the above mentioned stages of algorithm like contrast enhancement, thresholding, etc. Testing & Validation Stage: The designed algorithm for the optimum defect detection was capable to identify regions with defects but it needed to be tested. The testing was done basically by three means: The area, shape and size of the defects were extracted using the algorithm which was compared with the visual examination of the defects. Based on the defect statistics, the grading of the defects was done. Multiple images of the same defect regions were captured and processed to check for any variation in the definition and grading of defects. 94

123 Fabric Grading Regions of the fabric other than those studied for defect detection were captured and processed to grade the same. 4.6 Fabric Grading: On the basis of the defect parameters obtained as result of the processing of images of the fabric lot & considering the proposed classification of defect, a fabric grading system was developed. The Defect Classification is shown in Table 4.11 TABLE 4.11 : Defect Classification Defect Name (DN) Woven Geotextile Spunbond Nonwoven Missing End (Chira) Drops / bond point fusion Slubs (Warp) Pinholes Stain (Daggi) Wrinkles Slubs (Weft) Hard filaments Missing Pick (Jerky) Holes Gout Calendar cut Defect Size (DS) Mendable- 10 % Defective Area Permissible- 30% Defective Area Critical - 60% Defective Area Rejected - 80% Defective Area Defect Frequency (DF) Frequency of occurrence of defects Defect Orientation (DO) Machine Direction/Warp Way Cross Direction/Weft Way 95

124 System Design & Development The Proposed Grading System is shown below: Grade of Fabric A (Best) B (Good) C (Poor) D (Rejected) Proposed Performance of Fabric The defect has no or very negligible influence, the fabric can thus be used for suggested applications. DN - All DS Mendeable (10% Defective Area/upto 3 length of defect) DF 1- within an image frame, 10% of the total images for the fabric roll. DO - any Substandard applications of suggested areas are possible with this grade of fabrics. DN - All DS Permissible (30% Defective Area/ 3-6 length of defect) DF 1- within an image frame, 30% of the total images for the fabric roll. DO - any Can be considered after repairing or taking preventive measures for suggested areas of applications. DN - All DS Critical (60% Defective Area/ upto 6-9 length of defect) DF more than 1 within an image frame, 60% of the total images for the fabric roll. DO any Not to be considered for any suggested applications. DN - All DS Rejected (80% Defective Area more than 9 length of defect) DF more than 1 within an image frame, 80% of the total images for the fabric roll. DO - any 96

125 CHAPTER 5 Results and Discussions 5.1 Introduction: In this chapter we present the results of the various trails conducted during the learning and the testing phase for the implementation, optimisation and validation of the designed algorithm as well as an effective quality control system for the geotextiles as well as spunbond nonwovens. For designing a suitable algorithm for defect detection, the optimization of various image processing parameters was needed to be done. The images were processed by taking trials with different values and combinations using Matlab script. Various direct and indirect parameters were studied for obtaining the optimum algorithm and the results for the significant ones have been presented here. The main parameters which were to be optimised were deciding the value of threshold, value of the structuring element. A vast number of images of both the fabrics had been studied during each phase. Since it is difficult and insignificant to display the results of all the images, only the results having significant contribution for implementation, optimisation and validation during the learning and the testing phase have been tabulated here. 97

126 Results and Discussion 5.2 General Image Parameters of the Images: This section shows results of the general image parameters of the fabric images of defective as well as defect free regions. As mentioned earlier vast number of images were acquired and processed for extracting the basic characteristics of the fabrics. The average values for unprocessed and enhanced images of each variety of geotextiles as well as spunbond fabrics were obtained and the readings of the defects of the same have been tabulated in Section and Section respectively Woven Geotextiles: The images of defect free regions of geotextiles were processed and the results have been tabulated in this section. The images of regions with defects also had been processed to obtain the basic intensity distribution for comparison with the defect free ones. TABLE 5.1 : Sample wise average values of Image parameters for unprocessed images of each type of Geotextile. Sample Name GSM Mean Intensity SD (variation in intensity) Threshold Average Maximum Intensity G G G G G G G Average Minimum Intensity 98

127 Mean Intensity Mean Intensity General Parameters of the Images TABLE 5.2 : Sample wise average values of Image parameters for enhanced images of each type of Geotextile. Sample Name GSM Mean Intensity SD (variation in intensity) Threshold Average Maximum Intensity G G G G G G G Average Minimum Intensity G1 G2 G3 G4 G5 G6 G7 Fabric Sample & GSM Mean Intensity (unprocessed Images) Mean FIGURE 5.1 : Comparison of the Mean Intensity Values between the unprocessed Images of the various Geotextiles G1 G2 G3 G4 G5 G6 G7 Fabric Sample & GSM Mean Intensity (enhanced Images) Mean FIGURE 5.2 : Comparison of the Mean Intensity Values between the enhanced Images of the various Geotextiles The mean intensity of the unprocessed images was found to be almost same for different types of fabrics. Minor variations in the mean value of intensity can be mainly accounted 99

128 Results and Discussion due to noise in the images on account of errors in image acquisition. After processing of the images, it was found that the mean intensity of sample G5(290 GSM) with maximum GSM was very less. TABLE 5.3 : Defect wise values of Image parameters for unprocessed images of identified defects in Geotextiles. Sr. No. Defect Name 1. Missing End (Chira) 2. Slubs (warp) Fabric Sample GSM Mean Intensity SD (variation in intensity) Threshold Average Maximum Intensity G G Stain G Slubs G (weft) 5. Missing Pick (jerky) G gout G Average Minimum Intensity TABLE 5.4 : Defect wise values of Image parameters for enhanced images of identified defects in Geotextiles. Sr. No. Defect Name 1. Missing End (Chira) 2. Slubs (warp) Fabric Sample GSM Mean Intensity SD (variation in intensity) Threshold Average Maximum Intensity G G Stain G Slubs G (weft) 5. Missing Pick (jerky) G Gout G Average Minimum Intensity 100

129 Mean Intesntiy Mean Intensity General Parameters of the Images G6 G1 G6 G5 G4 G2 Missing End (Chira) Slubs (warp) Stain Slubs (weft) Missing Pick (jerky) Fabric GSM & Defect Type gout Mean Intensity (Defect) Mean Intensity (General) Mean (Defect) Mean (General) FIGURE 5.3 : Comparison of the Mean Intensity Values between the unprocessed Images of Defects & general Fabric in Geotextiles G6 G1 G6 G5 G4 G2 Missing End (Chira) Slubs (warp) Stain Slubs (weft) Missing Pick (jerky) Fabric GSM & Defect Type gout Mean Intensity (Defect) Mean Intensity (General) Mean (Defect) Mean (General) FIGURE 5.4 : Comparison of the Mean Intensity Values between the enhanced Images of the Defects & general Fabric in Geotextiles 101

130 Mean Intensity Mean Intensity Results and Discussion G6 G1 G6 G5 G4 G2 Mean Intensity Mean Missing End (Chira) Slubs (warp) Stain Defect Slubs (weft) Missing Pick (jerky) gout FIGURE 5.5 : Comparison of the Mean Intensity Values between the unprocessed Images of the various Defects in Geotextiles G6 G1 G6 G5 G4 G2 Mean Intensity Mean Missing End (Chira) Slubs (warp) Stain Defect Slubs (weft) Missing Pick (jerky) gout FIGURE 5.6 : Comparison of the Mean Intensity Values between the enhanced Images of the various Defects in Geotextiles 102

131 General Parameters of the Images It can be seen from Tables , Figure 5.4 & Figure 5.5, that no significant difference was found between the basic image parameters between the images of defect free and defective regions. Also no significant information was obtained for further processing of images for defect detection Spunbond Nonwovens: The images of defect free regions of spunbond nonwovens were processed and the results have been tabulated in this section. The images of regions with defects also had been processed to obtain the basic intensity distribution for comparison with the defect free ones. TABLE 5.5 : Sample wise average values of Image parameters for unprocessed images of each type of Spunbond Nonwoven. Sample Name GSM Mean Intensity SD (variation in intensity) Threshold Average Maximum Intensity NS NS NS NS NS NS NS NS NS NS NS Average Minimum Intensity 103

132 Mean Intensity Mean Intensity Results and Discussion TABLE 5.6 : Sample wise average values of Image parameters for enhanced images of each type of Spunbond Nonwoven. Sample Name GSM Mean Intensity SD (variation in intensity) Threshold Average Maximum Intensity NS NS NS NS NS NS NS NS NS NS NS Average Minimum Intensity NS1 NS2 NS3 NS4 NS5 NS6 NS7 NS8 NS9 NS10 NS11 Fabiric Sample & GSM Mean Intensity Mean FIGURE 5.7 : Comparison of the Mean Intensity Values between the unprocessed Images of the various Spunbond Fabrics NS1 NS2 NS3 NS4 NS5 NS6 NS7 NS8 NS9 NS10 NS11 Fabric Sample & GSM Mean Intensity Mean FIGURE 5.8 : Comparison of the Mean Intensity Values between the enhanced Images of the various Spunbond Fabrics 104

133 General Parameters of the Images It can be seen from the Figures 5.7 & 5.8 that there is some difference between the values of each type of fabric, and the same might influence the thresholding value. The higher GSM fabric had maximum mean intensity while the lower GSM had minimum mean intensity, but the other fabrics exhibited almost same mean intensity value. The variation in the intensity levels in the images can be accounted due to the basic fibrous structure of the nonwovens. Again here also no significant difference found between the images for different gsm fabrics. The contrast enhancement in the images increase the range of the high and low level intensity values giving a considerable improvement in the visual appearance of the images. TABLE 5.7 : Defect wise values of Image parameters for unprocessed images of identified defects in spunbond nonwovens. Sr. No. Defect Name 1. Drop/ Bond Pt. Fusion Fabric Sample GSM Mean Intensity SD (variation in intensity) Threshold Average Maximu m Intensity NS Pin Hole NS Wrinkle NS Hard NS Filament 5. Hole NS Calender NS Cut 7. Thin Spots NS Average Minimu m Intensity 105

134 Mean Imtensity Results and Discussion TABLE 5.8 : Defect wise values of Image parameters for enhanced images of identified defects in spunbond nonwovens. Sr. No. Defect Name 1. Drop/ Bond Pt. Fusion Fabric Sample GSM Mean Intensity SD (variation in intensity) Threshold Average Maximu m Intensity NS Pin Hole NS Wrinkle NS Hard NS Filament 5. Hole NS Calender NS Cut 7. Thin Spots NS Average Minimu m Intensity NS7 NS6 NS9 NS3 NS5 NS10 NS4 Mean Intensity (Defect) Mean Intensity (General) Mean (Defect) Mean (General) Drop/ Bond Pt. Fusion Pin Hole Wrinkle Hard Filament Hole Fabric GSM & Defect Type Calender Cut Thin Spots FIGURE 5.9 : Comparison of the Mean Intensity Values between the unprocessed Images of Defects & general Fabric in Spunbond 106

135 Mean Intensity Mean Intensity General Parameters of the Images NS7 NS6 NS9 NS3 NS5 NS10 NS4 Mean Intensity (Defect) Mean Intensity (General) Mean (Defect) Mean (General) Drop/ Bond Pt. Pin Hole Wrinkle Hard Filament Hole Fusion Fabric GSM & Defect Type Calender Cut Thin Spots FIGURE 5.10 : Comparison of the Mean Intensity Values between the enhanced Images of the Defects & general Fabric in Spunbond NS7 NS6 NS9 NS3 NS5 NS10 NS4 Mean Intensity Mean Drop/ Pin Hole Wrinkle Bond Pt. Fusion Hard Filament Defect Hole Calender Cut Thin Spots FIGURE 5.11 : Comparison of the Mean Intensity Values between the unprocessed Images of the various Defects in Spunbond Fabrics 107

136 Mean Intensity Results and Discussion NS7 NS6 NS9 NS3 NS5 NS10 NS4 Mean Intensity Mean Drop/ Pin Hole Wrinkle Bond Pt. Fusion Hard Filament Defect Hole Calender Cut Thin Spots FIGURE 5.12 : Comparison of the Mean Intensity Values between the enhanced Images of the various Defects in Spunbond Fabrics Similar to the case of the woven geotextiles, no significant difference found from the basic image parameters between the images of defect free and defective regions. Difference in the Mean Intensity Values can be seen between major and minor defects. 5.3 Histogram Analysis: Study of histogram was done in both the fabrics to analyse the distribution of intensities within the fabrics. The results along with discussions for geotextiles and spunbond fabrics have been described in Section and Section respectively Woven Geotextiles: Histograms of a large amount of images for all varieties of geotextile fabrics were studied. The histograms of some of the images of fabrics void of defects have been shown in Figure The histogram of the defective fabric images were also studied and have been shown from Figure 5.14 to Figure

137 Histogram Analysis FIGURE 5.13 : Histogram of Images of some samples of defect free images of Geotextiles. The histogram of the images show a single peak and since it is a woven fabric, the distribution of the intensity is symmetric about the mean intensity. Also higher gsm fabric shows a broader peak. FIGURE 5.14 : Histogram of defect free region and Missing End (Chira) 109

138 Results and Discussion The presence of missing end accounts to the deviation of the shape of histogram from the defect free image. FIGURE 5.15 : Histogram of defect free region and Slubs (Warp) The presence of slubs accounts for more pixels at the peak. FIGURE 5.16 : Histogram of defect free region and Stain (Daggi) The presence of stain accounts to the deviation of the shape of histogram from the defect free image. The shape is similar to the histogram of missing end (Figure 5.2). This is because both the defects are in the same sample. Thus we can conclude that the histogram shape is not affected by the type of defect. 110

139 Histogram Analysis FIGURE 5.17 : Histogram of defect free region and Slubs (Weft) The shape of the histogram of the defect is higher towards the low intensity and accounts for the presence of defect. FIGURE 5.18 : Histogram of defect free region and Missing Pick (Jerky) Similar to the above figure, here also the presence of defect is reflected by more no of low intensity pixels or a higher spread of intensities of pixels. 111

140 Results and Discussion FIGURE 5.19 : Histogram of defect free region and Gout The presence of gouts just like slubs(warp) accounts for more pixels at the peak. Histogram analysis shows that presence of defects could be largely detected from the shape of histogram. The threshold value was therefore easy to be decided from the histogram Spunbond Nonwoven: Histograms of a large amount of images for all varieties of spunbond fabrics were studied like geotextiles. The histograms of some of the images of fabrics void of defects have been shown in Figure

141 Histogram Analysis FIGURE 5.20: Histogram of Images of some samples of defect free samples of Spunbond Nonwoven The images of fabrics void of defects show a single peak. Also it can be seen that the shape and pattern of intensity distribution of each sample varies. Also the histograms are not symmetric around the mean intensity values like the woven geotextiles. The defect wise histogram of the images of the fabrics with defects and with it s corresponding defect free fabric image for each defect have been shown in from Figure 5.21 to Figure

142 Results and Discussion FIGURE 5.21: Histogram of defect free region and Drop / Bond Point Fusion Variations in the histogram indicate presence of defect. FIGURE 5.22: Histogram of defect free region and Pinhole Very minute variation in the histogram indicates presence of minor defect. 114

143 Histogram Analysis FIGURE 5.23: Histogram of defect free region and Wrinkle Very minute variation in the histogram indicates presence of minor defect. FIGURE 5.24: Histogram of defect free region and Hard Filament The region having this defect falls near the higher intensity region. 115

144 Results and Discussion FIGURE 5.25: Histogram of defect free region and Hole Prominent variation in the histograms can be seen, especially the multiple peaks concludes presence of major defects. FIGURE 5.26: Histogram of defect free region and Calender cut Very minute variation in the histogram indicates presence of minor defect. Quite similar intensity distribution cane be seen concluding very minor defect. Histogram analysis shows that presence of major defects could be easily detected, however the minor defects especially the one whose intensity levels are almost same as the structure itself could not be identified. The deviation of skewness in histograms of defective images from the defect free images is useful in determining the threshold for further processing of 116

145 Histogram Analysis the image. It was seen that the defects could be broadly categorised as the ones having more skewness deviation towards dark intensity (drop/bond point fusion, stains) and the ones having more skewness deviation towards light intensity (pin hole, wrinkles, hard filaments, holes, calendar cut). However, a common threshold for all fabric images and defects could not be obtained. 5.4 Thresholding: Woven Geotextiles: From the mean intensity values of the processed images of the geotextile fabrics, it was found that G5 fabric exhibited a very less mean intensity and so a different threshold(-0.2) was required for it, while all the other fabrics a common threshold of 0.2 was found suitable Spunbond Nonwoven: Since the nonwovens are characterised by thin and thick regions, different thresholds were used to obtain regions with the extreme light and dark regions. Threshold Intensity for dark regions was 100 and for light regions was 180 used. All pixels with intensity below 100 were eliminated to obtain the extreme dark regions and all pixels with intensity above 180 were eliminated to obtain the extreme light regions. 5.5 Defect Detection & Validation: Defect Detection in Images of Woven Geotextiles: The images of the regions with defects were processed using the method as described in Chapter 4, Section The visual results of the grayscale image, binary image after thresholding and the image with highlighted defective regions for each defect have been shown in the Figures from Figure

146 Results and Discussion FIGURE 5.27 : Grayscale Image - Missing End FIGURE 5.28 : Binary Image - Missing End 118

147 Defect Detection & Validation FIGURE 5.29 : Highlighted Missing End FIGURE 5.30 : Grayscale Image-Slub (Warp) 119

148 Results and Discussion FIGURE 5.31 : Binary Image-Slub (Warp) FIGURE 5.32 : Highlighted Slub (Warp) 120

149 Defect Detection & Validation FIGURE 5.33 : Grayscale Image - Stain (Daggi) FIGURE 5.34: Binary Image - Stain (Daggi) 121

150 Results and Discussion FIGURE 5.35 : Highlighted Stain (Daggi) FIGURE 5.36 : Grayscale Image Slub (weft) 122

151 Defect Detection & Validation FIGURE 5.37: Binary Image Slub (weft) FIGURE 5.38 : Highlighted Slub (weft) 123

152 Results and Discussion FIGURE 5.39 : Grayscale Image Missing Pick (jerky) FIGURE 5.40 : Binary Image Missing Pick (jerky) 124

153 Defect Detection & Validation FIGURE 5.41 : Highlighted Missing Pick (jerky) FIGURE 5.42 : Grayscale Gout 125

154 Results and Discussion FIGURE 5.43 : Binary Image Gout FIGURE 5.44 : Highlighted Gout 126

155 Defect Detection & Validation It is evident from the processed images, that the algorithm was able to extract the defective areas from the unprocessed images. It was found that in Figure 5.38, some non defective regions also had been highlighted giving a false alarm of defect. The same had been eliminated by using the statistics of these regions while grading the image. The area, length/width and the orientation of the identified defects were calculated and the same has been tabulated in Table 5.9. The values were compared with the values obtained by manual - visual examination of the defects and the same have been tabulated in Table TABLE 5.9 : Defect Statistics obtained from the System Sr. No. Defect Name Fabric Sample GSM Defect Area (sq. inch) Length/Width of Biggest Defect (inch) Defect Orientation 1 Missing End G Warp-line (Chira) 2 Slubs (warp) G Warp-line 3 Stain G Warp-line 4 Slubs (weft) G Weft-line 5 Missing Pick G Weft-line (jerky) 6 Gout G circular TABLE 5.10 : Defect Statistics obtained from Manual - visual Examination Sr. No. Defect Name Fabric Sample GSM Defect Area (sq. inch) Length/Width of Biggest Defect (inch) Defect Orientation 1 Missing End G Warp-line (Chira) 2 Slubs (warp) G Warp-line 3 Stain G Warp-line 4 Slubs (weft) G Weft-line 5 Missing Pick G Weft-line (jerky) 6 Gout G circular 127

156 Length/Width of Biggest Defect (Inch) Defect Area (Sq. inch) Results and Discussion Manual Automatic Missing End (Chira) Slubs (warp) Stain Slubs (weft) Missing Pick (jerky) Gout Defect Type FIGURE 5.45 : Comparison of Defect Area obtained from the System with those obtained from Manual - visual Examination Manual 2 Automatic 1 0 Missing End Slubs (warp) Stain Slubs (weft) Missing Pick (Chira) (jerky) Gout Defect Type FIGURE 5.46 : Comparison of Length/Width of Biggest Defect obtained from the System with those obtained from Manual - visual Examination 128

157 Defect Detection & Validation It can be seen from Table 5.9 & Table 5.10 that the defect statistics obtained from the system are similar to the ones obtained from manual - visual examination. It can be seen from the Figures 5.45 that there is considerable difference between the area obtained in Missing Pick, however less difference can be seen between the width of the defect, which might be due to non uniformity in the intensity of the defect Validation of Results of Geotextiles: Grading of Geotextiles: Since the defects were a result of the manufacturing process, most of the images had multiple defective regions. Therefore, grading of the images was done based on the length/width & the number of these defective areas and also considering the severity of the defects as done in the manual - visual grading system for the woven fabrics. The manual - visual grading of the fabrics with identified defective regions is shown in Table The total defective area present in the image, percentage defective area, length/width of the biggest defect in the image, total number of objectionable defects & grading of the image have been calculated using the algorithm and the same has been tabulated in Table TABLE 5.11 : Manual Grading of the Defects in Geotextile Fabrics Image. No. Main Defect Present Fabric Sample GSM Manual - visual Overall Grading 1 Missing End (Chira) G6 210 Major 2 Slubs (warp) G1 160 Major 3 Stain G6 210 Major 4 Slubs (weft) G5 290 Minor 5 Missing Pick (jerky) G4 210 Major 6 Gout G2 120 Major 129

158 Results and Discussion TABLE 5.12 : Grading of Defects in Geotextile Fabrics using the System Image. No. Main Defect Present 1 Missing End (Chira) 2 Slubs (warp) Fabric Sample GSM Total Defective Area (sq inch) Total Percentage Defective Area Length / Width of Biggest Defect (inch) No. of Defects Overall Grade G C G D 3 Stain G C 4 Slubs G A (weft) 5 Missing Pick (jerky) G C 6 Gout G C TABLE 5.13 : Comparison between the Grading of Woven Geotextile Images obtained by Manual - visual Examination & System Image. No. Main Defect Present Fabric Sample GSM Manual - visual Overall Grading Overall Grade using System 1 Missing End (Chira) G6 210 Major C 2 Slubs (warp) G1 160 Major D 3 Stain G6 210 Major C 4 Slubs (weft) G5 290 Minor A 5 Missing Pick (jerky) G4 210 Major C 6 Gout G2 120 Major C Table 5.13 shows that the grading system used for classification as suggested in Chapter 4, Section 4.6 is comparable with manual - visual grading major and minor defects 130

159 Length / Width of Biggest Defect (inch) Defect Detection & Validation Grading of Multiple Images of Geotextiles with same defect: The results obtained as above were validated by capturing multiple images of same fabric regions. 5 images were taken of the same regions to check for variability in the length or width of the biggest defect and the grading obtained as per the algorithm. The results are tabulated in Table 5.13 and Figure 5.35 show the variability of the order of only 5-10 % which is considered to be negligible. TABLE 5.14 : Grading of Multiple Images of Regions with same defect Image. No. Main Defect Present 1 Missing End (Chira) Fabric Sample GSM Length / Width of Biggest Defect (inch) Grading as per System I1 I2 I3 I4 I5 I1 I2 I3 I4 I5 G C C C C C 2 Slubs (warp) G D D D D D 3 Stain G C C C C C 4 Slubs (weft) G A A A C C 5 Missing Pick G C C A C C (jerky) 6 Gout G C C C C C I Missing End (Chira) Slubs(warp) Stain Slubs (weft) Missing Pick(jerky) Defect Type gout I2 I3 I4 I5 FIGURE 5.47 : Comparison of Multiple Images of Regions with same defect 131

160 Results and Discussion It can be seen from Figure 5.47, that there is some variation in the length of the maximum defect sensed in the line defects in the warp direction. It is due to the fact that most of the times the continuity of the line defects is not maintained while thresholding and results in 2 3 breakages in line as can be seen from Figure As a result of this, the software interprets 2 or 3 defects (objects) instead of one. But by considering the number of these kind of defects (objects) along with the length/width results in obtaining the desired grading of the Defects as can be seen from Table Grading of other Geotextile Images than the Studied Ones: The developed algorithm was thus found reliable and was tested for a number of images of other fabric regions. Some of the images with highlighted defective regions and the grading so obtained have been shown in Figures 5.48 Figures FIGURE 5.48 : Test Image 1 132

161 Defect Detection & Validation FIGURE 5.49 : Test Image 2 FIGURE 5.50 : Test Image 3 133

162 Results and Discussion FIGURE 5.51 : Test Image 4 FIGURE 5.52 : Test Image 5 134

163 Defect Detection & Validation FIGURE 5.53 : Test Image 6 FIGURE 5.54 : Test Image 7 135

164 Results and Discussion FIGURE 5.55 : Test Image 8 FIGURE 5.56 : Test Image 9 136

165 Defect Detection & Validation FIGURE 5.57 : Test Image 10 The results show that the algorithm is able to detect the defective regions and the results show the variability of the order of only 5-10 % which is considered to be negligible. TABLE 5.15 : Comparison between the Grading achieved with developed System as against Manual - visual grading Test Image Grading As Per Software Manual - visual Grading 1 C Major Missing End 2 C Major Missing Ends 3 C Major Missing End & Stretch Marks 4 C Major Stretch Marks 5 A No Faults 6 C Major Missing End & Stain 7 C Major Stain & Missing Pick 8 D Major Stretch Marks 9 C Major Faulty Selvedge 10 C Major Missing End & Stretch Marks It can be seen from Table 5.15 that Test Image 1-4, 6, 7, 9 &10 have been graded as C quality which matches perfectly with the severity of the defective region. Also Test Image 5 which had no defective regions had been graded as A quality. The grading obtained for Test Image 8 was also found to be perfect considering the severity of the defect. It can be 137

166 Results and Discussion concluded that 10 out of 10 Test Images gave perfect results and therefore it can be said that the algorithm is able to detect the defective regions successfully and the results show the variability of the order of even less than the range 5-10 % as specified earlier Defect Detection in Spunbond Images: The images of the regions with defects were processed using the method as described in Chapter 4, Section The visual results of the grayscale image, binary image after thresholding and the image with highlighted defective regions for each defect have been shown in the Figures from Figure 5.58 to FIGURE 5.58 : Grayscale Image Drop/Bond Point Fusion 138

167 Defect Detection & Validation FIGURE 5.59 : Binary Image Drop/Bond Point Fusion FIGURE 5.60 : Highlighted Drop/Bond Point Fusion 139

168 Results and Discussion FIGURE 5.61 : Grayscale Image Pin Hole FIGURE 5.62 : Binary Image Pin Hole 140

169 Defect Detection & Validation FIGURE 5.63 : Highlighted Pin Hole FIGURE 5.64 : Grayscale Image Wrinkle 141

170 Results and Discussion FIGURE 5.65 : Binary Image Wrinkle FIGURE 5.66 : Highlighted Wrinkle 142

171 Defect Detection & Validation FIGURE 5.67 : Grayscale Image Hard Filament FIGURE 5.68 : Binary Image Hard Filament 143

172 Results and Discussion FIGURE 5.69 : Highlighted Hard Filament FIGURE 5.70 : Grayscale Image Hole 144

173 Defect Detection & Validation FIGURE 5.71 : Binary Image Hole FIGURE 5.72 : Highlighted Hole 145

174 Results and Discussion FIGURE 5.73 : Grayscale Image Calender Cut FIGURE 5.74 : Binary Image Calender Cut 146

175 Defect Detection & Validation FIGURE 5.75 : Highlighted Calender Cut As compared to the woven fabrics, the extraction of the defects was quite difficult due to the presence of thin - thick regions in the nonwoven fabrics. The detection rate of the local defects i.e. the one tending to merge with the basic structure was found to be low. The area of the identified defects was calculated and the same has been tabulated in Table The values were compared with the values obtained by manual - visual examination of the defects and the same have been tabulated in Table TABLE 5.16 : Defect Statistics obtained from the System Sr. Defect Name Fabric GSM Defect Area Defect Type No. Sample (sq. cm) 1 Drop/Bond Pt. NS Irregular(dark) Fusion 2 Pin Hole NS Small circle (light) 3 Wrinkle NS Irregular lines (light) 4 Hard Filament NS Irregular lines 5 Hole NS Irregular (light) 6 Calender Cut NS Very Fine line/curve 147

176 Defect Area (Sq. cms) Results and Discussion TABLE 5.17 : Defect Statistics obtained from Manual - visual Examination Sr. Defect Name Fabric GSM Defect Area Defect Type No. Sample (sq. cm) 1 Drop/Bond Pt. NS Spots Fusion 2 Pin Hole NS Small hole 3 Wrinkle NS lines 4 Hard Filament NS Patch 5 Hole NS Holes 6 Calender Cut NS Fine line/curves Manual Automatic 0 Drop/Bond Pt. Fusion Pin Hole Wrinkle Hard Filament Defect Type Hole Calender Cut FIGURE 5.76 : Comparison of Defect statistics obtained from the System with those obtained from Manual - visual Examination It can be seen from Table 5.16, Table 5.17 and Figure 5.76 that the defect statistics obtained from the system are quite similar to the ones obtained from manual - visual examination. The detection rate of Calender Cut was found to be less as compared to the other defects, the reason being very less difference between the intensity levels of the defect and defect free area. Also the defect was very fine, however some information about the defect could be gathered to mark presence of little variation. 148

177 Defect Detection & Validation Validation of Results for Spunbonds: Grading of Spunbond Images: As mentioned earlier, the fabric images had multiple defective regions and so the grading of the spunbond images was done. The grading was based on the total defective area, objectionable defective area and the number of defective areas. The manual - visual grading of the fabrics with identified defective regions is shown in Table The total defective area present in the image, percentage defective area, objectionable defective area, percentage defective area, total number of objectionable defects & grading of the image have been calculated using the algorithm and the same has been tabulated in Table TABLE 5.18 : Manual Grading of Spunbond Fabrics Image. No. Main Defect Present Fabric Sample GSM Manual - visual Grading 1 Drop/Bond Pt. Fusion NS7 120 Minor 2 Pin Hole NS6 85 Minor 3 Wrinkle NS9 60 Major 4 Hard Filament NS3 60 Minor 5 Hole NS5 60 Major 6 Calender Cut NS10 60 Minor TABLE 5.19 : Grading of Spunbond Images using the System Image. No. Main Defect Present Fabric Sample GSM Total Defect -ive Area % Total Defective Area Number of Objection -able Defects Objection -able Area % Objection -able Defective Area GRAD E OF FABRI C 1 Drop/Bond NS B Pt. Fusion 2 Pin Hole NS A 3 Wrinkle NS B 4 Hard NS B Filament 5 Hole NS D 6 Calender Cut NS A 149

178 Results and Discussion TABLE 5.20 : Comparison between the Grading of Spunbond Images obtained by Manual - visual Examination & System Image. Main Defect Present Fabric Sample GSM Manual - visual GRADE OF No. Grading FABRIC 1 Drop/Bond Pt. Fusion NS7 120 Minor B 2 Pin Hole NS6 85 Minor A 3 Wrinkle NS9 60 Major B 4 Hard Filament NS3 60 Minor B 5 Hole NS5 60 Major D 6 Calender Cut NS10 60 Minor A Grading of Multiple Images Spunbond Images with same Defect: The results obtained as above were validated by capturing multiple images of same fabric regions. The results are tabulated in Table 5.21 and show the variability of the order of only 5-10 % in the defect statistics which is considered to be negligible, while no variation was seen in the grading of the images thus giving a consistent result. TABLE 5.21 : Grading of Multiple Images of Regions with same defect Image. No. Main Defect Present Fabric Sample GSM Total Area of defect in sq. cm as per software Objectionable Area of defect in sq. cm as per software Grading as per System I1 I2 I3 I1 I2 I3 I1 I2 I3 1 Drop/Bond NS B B B Pt. Fusion 2 Pin Hole NS A A A 3 Wrinkle NS B B B 4 Hard NS B B B Filament 5 Hole NS D D D 6 Calender Cut NS A A A 150

179 Defect Detection & Validation Comparison of Total Area of defect between Multiple Images (sq. cm) Drop/Bond Pt. Fusion Pin Hole Wrinkle Hard Filament Hole Calender Cut I1 I2 I3 FIGURE 5.77 : Comparison of Total Area between Multiple Images of Regions with same defect Comparison of Objectionable Area of defect between Multiple Images (sq. cm) Drop/Bond Pt. Fusion Pin Hole Wrinkle Hard Filament Hole Calender Cut I1 I2 I3 FIGURE 5.78 : Comparison of Objectionable Area Multiple Images of Regions with same defect 151

180 Results and Discussion Grading of other Spunbond Images than the Studied Ones: The developed algorithm was thus found reliable and was tested for a number of images of other fabric regions. Some of the images with highlighted defective regions and the grading so obtained have been shown in Figures 5.79 Figures FIGURE 5.79 : Test Image 1 152

181 Defect Detection & Validation FIGURE 5.80 : Test Image 2 FIGURE 5.81 : Test Image 3 153

182 Results and Discussion FIGURE 5.82 : Test Image 4 FIGURE 5.83 : Test Image 5 154

183 Defect Detection & Validation FIGURE 5.84 : Test Image 6 FIGURE 5.85 : Test Image 7 155

184 Results and Discussion FIGURE 5.86 : Test Image 8 FIGURE 5.87 : Test Image 9 156

185 Defect Detection & Validation FIGURE 5.88 : Test Image 10 TABLE 5.22 : Comparison between the Grading achieved with developed System as against Manual - visual grading Test Image Grading As Per Software Manual - visual Grading 1 C Major 2 C Major 3 B Minor 4 B Minor 5 A No Faults 6 B Major 7 A No Faults 8 A No Faults 9 A No Faults 10 B Minor It can be seen from Table 5.22 that Test Image 1 & 2 have been graded as C quality which matches perfectly with the severity of the defective region. Also Test Image 5, 7, 8 & 9 which had no defective regions had been graded as A quality. The grading obtained for Test Image 3, 4 & 10 was also found to be perfect considering the severity of the defect. Test Image 6 had little deviation from the actual severity as it was graded of B quality by the system. It can be concluded that 9 out of 10 Test Images gave perfect results 157

186 Results and Discussion and therefore it can be said that the algorithm is able to detect the defective regions and the results show the variability of the order of only 5-10 % which is considered to be negligible. It is evident from this section that the developed system is able to detect the defective regions quite well as well grade the fabrics according to the severity of the defects for both the woven geotextiles and spunbond nonwovens. 158

187 CHAPTER 6 Conclusion & Future Work 6.1 Objectives Achieved: Successfully designed & developed prototype of device well supported with the user friendly software module to help the users: In selection of proper quality of nonwoven/functional fabrics for specific end use applications To avoid unnecessary wastage of time and materials, which otherwise would be due to wrong selection of materials for any specific application Mainly dealing with the development of functional textiles having very high growth potential during the days to come

188 Conclusion & Future Work 6.2 Conclusions: Designed & developed prototype device for monitoring the quality of nonwoven/functional textiles. Prepared algorithm for development of software module most suitable for different varieties of fabrics. Tested nonwoven fabrics for different quality parameters and validate the results so obtained by capturing multiple images of same fabric samples using image processing technique. The results show the variability of the order of only 5-10 % which is considered to be negligible. Tested other functional fabrics for different quality parameters and validate the results so obtained by capturing multiple images of same fabric samples using image processing technique. The results show the variability of the order of only 5-12 % which is considered to be negligible. 6.3 Possible Future Scope: The research work has lead to the development of a cost and quality effective solution for the woven geotextile fabrics and spunbond nonwoven fabrics. The study offers a scope for further research on using image processing techniques for more applications in the area of textiles which can be listed as below: Development of similar systems for other categories of technical textiles. Development of similar systems for other types of nonwoven fabrics. Influence of yarn faults on the fabric faults can also be explored using image processing. Classification of fabrics based on the fabric parameters. 160

189 List of References 1. P. Madhavmoorthi and G. Shetty, Nonwoven, Mahajan Publishers Pvt. Ltd., Ahmedabad. 2. Guruprasad, R., & Behera, B. K. (2009). Automatic Fabric Inspection Systems. The Indian Textile Journal, Mahajan, P., Kolhe, S., & Patil, P. (2009). A review of automatic fabric defect detection techniques. Advances in Computational Research, 1(2), Brad, R., & Brad, R. (2004). A Vision System for Textile Fabric Defect Detection. In 2nd International Istanbul Textile Congress, Istanbul, Turkey, April 22-24, Horrocks, R., & Anand, S. C. (2000). Handbook of Technical Textiles. Woodhead Publishing Limited, Cambridge, England. 6. J. B. Goldberg. (1950). Fabric Defects-Case Histories of Imperfections in Woven Cotton and Rayon Fabrics. Mc Graw-Hill Book Company, INC., USA. 7. P. A. Khatwani (2001) Fabric Defects-Causes and Remedial Measures Proceedings of the Seminar on Emerging Challenges of Globalisation for Powerloom Sector organised jointly by NCUTE, Delhi and The Southern Gujarat Chamber of Commerce & Industry, Surat, pp P. A. Khatwani (2001) Fabric defects and remedial measures Proceedings of ISTE- AICTE approved Short Term Training Programme on Modern Developments in Weaving & Processing Techniques for Cotton, Blends & Manmades, Surat. 9. Shubham Yadav(2013) Faults in the Knitted Fabrics. [Online]. Available: Mazadul Hasan, Knitted fabric faults and their remedies. [Online]. Available: Causes and Remedies of Various Knitting Faults. [Online]. Available: Conor O Neill, R. M. / I. V. (2010). Reduce Waste - Save Time and Cost Application benefits of automated inspection for roll to roll packaging converters. In 2010 Place Conference, New Mexico, USA. 13. NIS 200 brochure, Lenzing Instruments GmbH & Co. KG 14. Tti-Inspections (Pvt.). (brochure.). Fabric Inspection Using Four-Point System. Lahore , Pakistan 161

190 15. Islam, A., Akhter, S., & Mursalin, T. E. (2006). Automated Textile Defect Recognition System Using Computer Vision and Artificial Neural Networks. World Academy of Science, Engineering and Technology, Ngan, H. Y. T., Pang, G. K. H., & Yung, N. H. C. (2011). Automated fabric defect detection A review. Image and Vision Computing, 29(7), Chien, H. T., Sheen, S. H., Lawrence, W. P., Razazian, K., & Raptis, A. C. (1999). On- Loom, Real-Time, Noncont Act Detection Of Fabric Defects By Ultrasonic Imaging. Review of Progress in Quantitaitve Nondestructive Evaluation, 18, Allgood, G. O., Treece, D. A., Mee, D. K., & Mooney, L. R. (2000). Textile laseroptical system for inspecting fabric structure and form. In Machine Vision Applications in Industrial Inspection VIII and Conference 3966B: Surface Characterization for Computer Disks, Wafers, and Flat Panel Displays II Proceedings of SPIE Volume J. Zhang & X. Meng (2010). A Fabric Defect Detection System Based on Image Recognition. In Intelligent Systems and Applications (ISA), nd International Workshop (pp. 1 4). 20. R. Thilepa & M. Thanikachalam (2010). A Paper on Automatic Fabrics Fault Processing Using Image Processing Technique In MATLAB. Signal & Image Processing : An International Journal, 1(2), J. Weickert (1999). A Real-Time Algorithm for Assessing Inhomogeneities in Fabrics. Real Time Imaging, 5, M. S. Loonkar & D. Mishra (2015). A Survey-Defect Detection and Classification for Fabric Texture Defects in Textile Industry. International Journal of Computer Science and Information Security, 13(5), S.N. Niles, S. Fernando and W.D.G. Lanerolle (2015). A System for Analysis, Categorisation and Grading of Fabric Defects using Computer Vision. RJTA, 19(No.1), R. Brad & R. Brad (2004). A Vision System for Textile Fabric Defect Detection. In 2nd International Istanbul Textile Congress, Istanbul, Turkey, April 22-24, Singh, U., Moitra, T., Dubey, N., & Patil, M. V. (2015). Automated Fabric Defect Detection Using MATLAB. International Journal of Advanced Research in Computer Engineering & Science, 03(06),

191 26. Brad, R., & Brad, R. (2004). Automated Fabric Defect Inspection for Quality Assurance Systems. The 83rd Textile Institute World Conference, Shanghai, (1996), Karunamoorthy, B., Somasundareswari, D., & Sethu, S. P. (2015). Automated Patterned Fabric Fault Detection Using Image Processing Technique In MATLAB. International Journal of Advanced Research in Computer Engineering & Technology, 4(1), Islam, A., Akhter, S., & Mursalin, T. E. (2006). Automated Textile Defect Recognition System Using Computer Vision and Artificial Neural Networks. World Academy of Science, Engineering and Technology, Abouelela, A., Abbas, H. M., Eldeeb, H., Wahdan, A. a., & Nassar, S. M. (2005). Automated vision system for localizing structural defects in textile fabrics. Pattern Recognition Letters, 26(10), Fazekas, Z., Komuves, J., Renyi, I., & Surjan, L. (1999). Automatic Visual Assessment of Fabric Quality. In IEEE International Symposium on Industrial Electronics (pp ). 31. Chi-ho Chan, Hugh Liu, Thomas Kwan, Grantham Pang (1998). Automation Technology for Fabric Inspection System. [Online]. Available : Drobina, R., & Machino, M. S. (2006). Application of the Image Analysis Technique for Textile Identification. Autex Research Journal, 6(1), Adel, G., Faten, F., & Radhia, A. (2011). Assessing Cotton Fiber Maturity and Fineness by Image Analysis. Journal of Engineered Fibers and Fabrics, 6(2), Xu, B., Pourdeyhimi, B., & Sobus, J. (1993). Fiber Cross-Sectional Shape Analysis Using Image Processing Techniques. Textile Research Journal, 63(12), Xu, B., & Huang, Y. (2004). Image Analysis for Cotton Fibers Part II: Cross-Sectional Measurements. Textile Research Journal, 74(5), Zghidi, H., Walczak, M., Błachowicz, T., Domino, K., & Ehrmann, A. (2015). Image Processing and Analysis of Textile Fibers by Virtual Random Walk. In Proceedings of the Federated Conference on Computer Science and Information Systems (Vol. 5, pp ). 37. Veit, D., Homes, I., Bergmann, J., & Wulfhorst, B. (1996). Image Processing as a tool to improve machine performance and process conrol. International Journal of Clothing Science and Technology, 8(1/2),

192 38. Das, D., Ishtiaque, S. M., & Mishra, P. (2010). Studies on fibre openness using image analysis technique, 35(March), D. Semnani & A. Gholami (2009). A sharp technique for identification of defective points in false twist textured yarns. Indian Journal of Fibre and Textile Research, 34(4), Carvalho, V., Gonçalves, N., Soares, F., Belsley, M., & Rosa. (2011). An Overview Over Yarn Mass Parameterization Methods. In Sensor Devices 2011: The Second International Conference on Sensor Device Technologies and Applications (pp ). 41. Bahl, K., & Kainth, J. S. (2014). Evaluation of Yarn Quality in Fabric using Image Processing Techniques 3(3), Fabijaanska, A., & Jackowska-Strumillo, L. (2012). Image processing and analysis algorithms for yarn hairiness determination. Machine Vision and Applications, 23(3), Pan, R., Gao, W., Liu, J., & Wang, H. (2011). Recognition the Parameters of Slub-yarn Based on Image Analysis. Journal of Engineered Fibers and Fabrics, 6(1), Carvalho, V., Gonçalves, N., Soares, F., Vasconcelos, R., & Belsley, M. (2013). Yarn Parameterization and Fabrics Prediction Using Image Processing. Textiles and Light Industrial Science and Technology, 2(1), B. Wilbik-Hałgas, R. Danych, B. Wiecek & K. Kowalski (2006). Air and water vapour permeability in double-layered knitted fabrics with different raw materials. Fibres and Textiles in Eastern Europe, 14(3), Robinson, D., Ramsundar, P., & C. B. Samantaray. (2014). Analyzing Porosity in Thermal Barrier Coatings : Edge Detection of Images using MATLAB. 121st ASEE Annual Conference & Exposition. Indianapolis, IN, Paper ID # Jasinska, I. (2009). Assessment of a Fabric Surface after the Pilling Process Based on Image Analysis. Fibres & Textiles in Eastern Europe, 17(2), Liqing, L., Jia, T., & Chen, X. (2008). Automatic recognition of fabric structures based on digital image decomposition. Indian Journal of Fibre and Textile Research, 33(December), Jmali, M., Zitouni, B., Sakli, F., & Ksar, I. (2007). Automatic recognition of woven fabric patterns by extraction of the characteristics of texture. International Journal of Clothing Science and Technology, UK,

193 50. Ben Salem, Y., & Nasri, S. (2009). Automatic recognition of woven fabrics based on texture and using SVM. Signal, Image and Video Processing, 4(4), Semnani, D., & Ghayoor, H. (2009). Detecting and Measuring Fabric Pills Using Digital Image Analysis. World Academy of Science, Engineering and Technology, Huimin, C., Hongbo, G., & Weiyuan, Z. (2008). Digital analysis of fabric smoothness appearance on point-sampled model. Journal of Textile Research, 29(9), Kenkare, N., & Plumlee, T. M.-. (2005). Fabric Drape Measurement: A Modified Method Using Digital Image Processing. Computer, 4(3), Xin, W., Georganas, N. D., & Petriu, E. M. (2011). Fabric Texture Analysis Using Computer Vision Techniques. Instrumentation and Measurement, IEEE Transactions on, 60(1), Georganas, N. D., & Petriu, E. M. (2009). Fiber-level structure recognition of woven textile. In 2009 IEEE International Workshop on Haptic Audio visual Environments and Games (pp ). 56. Kanade, P., Shah, N., Agrawal, S., & Patel, D. (2012). Image Analysis Technique for Evaluation of Air Permeability of a Given Fabric. International Journal of Engineering Research and Development, 1(10), Çay, A., Vassiliadis, S., Rangoussi, M., & Tarakçıo, I. (2005). On the use of image processing techniques for the estimation of the porosity of textile fabrics, Pranut Potiyaray, Chutipak Subhakalin, U. (2010). Recognition and re-visualization of woven fabric structures. International Journal of Clothing Science and Technology, 22(2/3), Zhao, S., Jakob, W., Marschner, S., & Bala, K. (2012). Structure-aware synthesis for predictive woven fabric appearance. In SIGGRAPH 2012 Proceedings (Vol. 31, pp. 1 10). 60. Saharkhiz, S., Ph, D., & Abdorazaghi, M. (2012). The Performance of Different Clustering Methods in the Objective Assessment of Fabric Pilling. Journal of Engineered Fibers and Fabrics, 7(4), Mirjalili, S. A., & Ekhtiyari, E. (2010). Wrinkle Assessment of Fabric Using Image Processing. Fibres & Textiles in Eastern Europe, 18(5), Semnani, D., Yekrang, J., & Ghayoor, H. (2009). Analysis and Measuring Surface Roughness of Nonwovens Using Machine Vision Method. International Journal of 165

194 Chemical, Molecular, Nuclear, Materials and Metallurgical Engineering, 3(9), Rodraksa, W., & Tharmmaphornphilas, W. (2013). Appearance Defective Reduction in Nonwoven Process. In International MultiConference of Engineers and Computer Scientists (Vol. II). 64. Bresee R. R. & Danulik T (1996). Characterizing Nonwoven Web Structure Using Image Analysis Techniques, Proceedings of Nonwovens Conference-TAPPI 65. Hariharan, S., Sathyakumar, S. A., & Ganesan, P. (n.d.). Measuring of Fibre Orientation in Nonwovens Using Image Processing. Fibre2Fashion. 66. Dimassi, Koehl, M., & Zeng, L. (2006). Modeling of the Pore network by Image Processing : Application to the Nonwoven Material. In Computational Engineering in Systems Applications, IMACS Multiconference (pp ). 67. Ressom, H., Voos, H., Litz, L., & Schmitt, P. (2000). On-line Estimation of Key Quality Parameters in Nonwoven Production. In Systems, Man, and Cybernetics, 2000 IEEE International Conference on, Nashville, TN (pp ). 68. Gonzalez R C, Woods, R. E. (2002). Digital image processing 2 nd Edition. Prentice- Hall, Inc. Upper Saddle River, New Jersey Nixon, M., & Aguado, A. (2008). Feature Extraction and Image processing. Academic Press, UK. 70. Latif-Amet, A., Ertüzün, A., & Erçil, A. (2000). Efficient method for texture defect detection: Sub-band domain co-occurrence matrices. Image and Vision Computing, 18(6), Raj, V. D., Hariprasad, Y., & Pradesh, A. (2013). Electronics And Communication A Matlab Based Texture Feature Recognition. Journal of Information, Knowledge and Research in Electronics and Communication, 02(02), check 72. Nisha, Kumar, S. (2013). Image Quality Assessment Techniques. International Journal of Advanced Research in Computer Science and Software Engineering, 3(7), Wang, Z., Bovik, A. C., & Simoncelli, E. P. (2005). Structural Approaches to Image Quality Assessment. Handbook of Image and Video Processing. 74. Materka, A., & Strzelecki, M. (1998). Texture Analysis Methods A Review. Technical University of Lodz, Institute of Electronics, COST B11 report, Brussels 1998 Texture (Vol. 11). 75. Gadkari D (2000). Image Quality Analysis Using GLCM. Dsc. Thesis, University of Pune. 166

195 76. Shin S-J, Tsai I-S. & Led P-D (1996). Automatic faults detection and recognition for static plain fabrics : applying the theorem of texture tuned masks. International Journal of Clothing Science and Technology, 8(1), Tolba, a. S., Khan, H. a., Mutawa, a. M., & Alsaleem, S. M. (2010). Decision Fusion for Visual Inspection of Textiles. Textile Research Journal, 80(19), Cho, C. S., Chung, B. M., & Park, M. J. (2005). Development of real-time vision-based fabric inspection system. IEEE Transactions on Industrial Electronics, 52(4), Malek, A. S. (2012). Online Fabric Inspection by Image Processing Technology. Dsc Thesis, Univeristy of Haute Alsace. 80. Kumar A.: Computer vision-based fabric defect detection: a survey, IEEE, Transactions on Industrial Electronics, Vol. 55, Issue 1, 2008, pp J. L. Liu, B.Q. Zuo, X. Y. Zeng, P. Vroman, B. Rabenasolo, & G. M. Zhang, G. (2011). A comparison of robust Bayesian and LVQ neural network for visual uniformity recognition of nonwovens. Textile Research Journal, 81, H. Y. Lai, J. H. Lin, C. K. Lu, S. C. Yao. (2005). An Image Analysis for Inspecting Nonwoven Defect. INJ FALL 2005, TAPPI- Technical Association Of The Pulp And Paper Industry 83. Yousefzadeh, M., Payvandy, P., Seyyedsalehi, S. A., & Latifi, M. Defect Detection And Classification In Nonwoven Web Images Using Neural Network. [Online]. Available: on_in_nonwoven_web_images_using_neural_network 84. Ruuska, H., & Akerberg, I. (1995). Practical inspection systems for nonwoven diaper liner fabrics. Tappi Journal, 78(6), Scharcanski, J. (2006). Stochastic texture analysis for measuring sheet formation variability in the industry. IEEE Transactions on Instrumentation and Measurement, 55(5), Iivarinen, J., & Rauhamaa, J. (1998). Surface Inspection of Web Materials Using the Self-Organizing Map. In Intelligent Robots and Computer Vision XVII: Algorithms, Techniques, and Active Vision, Proc. SPIE 3522 (pp ). 87. Liu, J., Zuo, B., Zeng, X., Vroman, P., & Rabenasolo, B. (2011). Wavelet energy signatures and robust Bayesian neural network for visual quality recognition of nonwovens. Expert Systems with Applications, 38(7),

196 88. Understanding Line Scan Camera Applications. (2014). [Online]. Available : ww.teledynedalsa.com. 89. Comparing Line Scan and Area Scan Technologies.(2014). [Online]. Available : Conor O Neill, R. M. / I. V. (2010). Reduce Waste - Save Time and Cost Application benefits of automated inspection for roll to roll packaging converters. In 2010 Place Conference, New Mexico, USA. 91. Abou-taleb, H. A., & Sallam, A. T. M. (2008). On-Line Fabric Defect Detection And Full Control In A Circular Knitting Machine. Autex Research Journal, 8(1), Aksoy, S. (2012). Texture Analysis. [Online]. Available : C.H. Chen, L.F. Pau, P.S.P. Wang, (1993) Handbook of Pattern Recognition & Computer Vision, 2nd ed. World Scientific Publishing Co., Singapore. 94. J. Chen, A.K. Jain, A structural Approach to Identify Defects in Textured Images, Proc. IEEE Int'l Conf. Systems, Man & Cybernetics (SMC1998), vol. 1, 8 12 Aug 1988, pp M. Bennamoun, A. Bodnarova, Automatic Visual Inspection and Flaw Detection in Textile Materials: Past, Present and Future, Proc. IEEE Int'l Conf. Systems, Man, & Cybernetics (SMC), 1998, pp Bodnarova, M. Bennamoun, K.K. Kubik, Defect Defection in Textile Materials Based on Aspects on The HVS, Proc. IEEE Int'l Conf. Systems, Man, and Cybernetics, San Diego (US), Oct 1998, pp D. Chetverikov, Structural Defects: General Approach and Application to Textile Inspection, Proc. IEEE 15th Int'l Conf. Pattern Recognition (IAPR2000), vol. 1, 3 7 Sep 2000, pp D. Chetverikov, Pattern regularity as a visual key, Image Vision Computing. 18 (2000) Zhang Y. F. and Bresee R. R. (1995). Fabric Defect Detection and Classification Using Image Analysis, Textile Research Journal, Vol. 65, January, pp Mahure, J., & Y.C.Kulkarni. (2013). Fabrics Fault Processing Using Image Processing Technique in MATLAB. International Journal of Computer Science and Technology, 4(2),

197 101. Shanbhag, P. M., Deshmukh, M. P., & Suralkar, S. R. (2012). Overview : Methods Of Automatic Fabric Defect Detection. Global Journal of Engineering, Design & Technology, 1(2), Han, L. W., & Xu, D. (2010). Statistic learning-based defect detection for twill fabrics. International Journal of Automation and Computing, 7(1), Malek A.S., Drean J.-Y., Bigue L. and Osselin J.-F.(2011) Automatic Fabric Inspection: invention or innovation? International Conference on Intelligent Textiles and Mass Customisation ITMC, Casablanca, Morocco, L. Macaire and J. G. Postaire (1993) Flaw detection on galvanized metallic strips in real-time by adaptive thresholding, Proc. SPIE 2183, pp Xie X.(2008) A Review of Recent Advances in Surface Defect Detection using Texture analysis Techniques, Electronic Letters on Computer Vision and Image Analysis, Vol. 7, no. 3, pp S. Tolba (2012). A novel multiscale-multidirectional autocorrelation approach for defect detection in homogeneous flat surfaces. Machine Vision and Applications, 23(4), Hoseini, E., Farhadi, F., & Tajeripour, F. (2013). Fabric Defect Detection Using Auto-Correlation Function. International Journal of Computer Theory and Engineering, 5(1), R.M. Haralick, Statistical and structural approaches to texture (1979), Proceedings of the IEEE 67 (5), Tunák M. and Linka A.(2005) Planar anisotropy of fibre systems by using 2D Fourier transform, Proceedings of the 12th International Conference (STRUTEX), Liberec, Czech Republic, November G. S. Desoli, S. Fioravanti, R. Fioravanti, and D. Corso (1993) A system for automated visual inspection of ceramic tiles, Proc. Intl. Conf. Industrial Electronics, IECON 93, vol. 3, pp Rahaman, G. M. A., & Hossain, M. M. (2009). Automatic Defect Detection and Classification Technique from Image: A Special Case Using Ceramic Tiles. International Journal of Computer Science and Information Security, 1(1), Behera B. K.(2004) Image-processing in Textiles, A critical 113. appreciation of recent developments, Textile Progress, Vol. 35, No. 2/3/4, pp

198 114. B.B. Chaudhuri, N. Sarkar, An Efficient Approach to Compute Fractal Dimension in Texture Image, Proc. IEEE 11th IAPR, Conference A: Computer Vision & Applications, vol. 1, 1992, pp H.-G. Bu, J. Wang, X.-B. Huang, Fabric defect detection based on multiple fractal features and support vector data description, Engineering Applications of Artificial Intelligence 22 (2) (2009) Gedziorowski M. and Garcia J.: Programmable optical digital processor for rank order and morphological filtering, Optics Communications, Vol. 119, Issues 1-2, August, 1995, pp W. J. Jasper and H. Potapalli, Image analysis of mispicks in woven fabrics, Text. Res. J., vol. 65, pp , B. R. Abidi, H. Sari-Sarraf, J. S. Goddard, and Martin A. Haunt, Facet model and mathematical morphology for surface characterization, Scientific Literature Digital Library, Vergados D., Anagnostopoulos C., Anagnostopoulos I., Kayafas E., Loumos V. and Stassinopoulos G.: An Evaluation of Texture Segmentation Techniques for Real- Time Computer Vision Applications, Advances in Automation, Multimedia and Video Systems and Modern Computer Science, WSES Press, 2001, p.p Jasper W. J. and Potlapalli H.: Image Analysis of Mispicks in Woven Fabric, Textile Research Journal, Vol. 65, November, 1995, pp Unser M. and Ade F.: Feature extraction and decision procedure for automated inspection of textured materials, Pattern Recognition Letters, Vol. 2, No. 3, March, 1984, pp Mallik -Goswami B. and Datta A. K.: Detecting Defects in Fabric with Laser- Based Morphological Image Processing, Textile Research Journal, Vol. 70, September,2000, pp Kwak C., Ventura J. A., and Tofang-Sazi K.: Automated defect inspection andvclassification of leather fabric, Intelligent Data Analysis, May, 2001, pp Monadjemi A.: Towards efficient texture classification and abnormality detection, Ph.D. Thesis, Department of Computer Science, University of Bristol, UK, October, Kang, X., Yang, P., & Jing, J. (2015). Defect Detection on Printed Fabrics Via Gabor Filter and Regular Band. Journal of Fiber Bioengineering and Informatics, 8(1),

199 126. Mak, K., & Peng, P. (2006). Detecting defects in textile fabrics with optimal Gabor filters. International Journal of Computer Science, 1(4), Jing, J., Zhang, H., & Li, P. (2011). Improved Gabor filters for textile defect detection. Procedia Engineering, 15, Sari-Sarraf, H., & Goddard, J. S. (1999). Vision system for on-loom fabric inspection. IEEE Transactions on Industry Applications, 35(6), Ralló, M., Millán, M. S., & Escofet, J. (2009). Unsupervised novelty detection using Gabor filters for defect segmentation in textures. Optical Society of America, 26(9), Harwood D., Ojala T., Pietikäinen M., Kelman S., and Davis L.: Texture classification by center-symmetric auto-correlation, using Kullback discrimination of distributions, Pattern Recognition Letters, Vol. 16, Issue 1, January, 1995, pp Unser M.: Local linear transforms for texture measurements, Signal Processing, Vol. 11, Issue 1, July, 1986, pp Monadjemi A., Mirmedhi M. and Thomas B.: Restructured eigenfilter matching for novelty detection in random textures, Proceedings of the 15th British Machine Vision Conference, September, 2004, pp Pritpal Singh, & Sharma, O. C. (2014). Texture Analysis in Fabric Material for Quality Evaluation. International Journal of Applied Engineering and Technology, 4(2), Salem, Y. B. E. N., & Nasri, S. (2011). Woven Fabric Defects Detection based on Texture classification Algorithm. In 8th International Multi-Conference on Systems, Signals & Devices Woven Raheja, J. L., Ajay, B., & Chaudhary, A. (2013). Real time fabric defect detection system on an embedded DSP platform. International Journal for Light and Electron Optics, Elsevier, 124(21), Singh, P., & Singh, P. (2015). Texture Analysis In Fabric Material For Quality Evaluation Using GLCM Matrix. International Journal of Applied Engineering and Technology, 5(1), Habib, M. T., & Rokonuzzaman, M. (2011). Distinguishing feature selection for fabric defect classification using neural network. Journal of Multimedia, 6(5), Sette, S., & Boullart, M. L. (1996). Fault detection and quality assessment in textiles by means of neural nets. International Journal of Clothing Science and Technology, 8(1/2),

200 139. Bahlmann, C., Heidemann, G., & Ritter, H. (1999). Artificial neural networks for automated quality control of textile seams. Pattern Recognition, 32(6), H. Sari-Sarraf and J. S. Goddard, On-line optical measurement and monitoring of yarn density in woven fabrics, Proc. SPIE 2899, pp , Tsai I.-S. and Hu M.-C.: Automatic Inspection of Fabric Defects Using an Artificial Neural Network Technique, Textile Research Journal, Vol. 66, July, 1996, pp Perez R., Silvestre J. and Munoz J.: Defect detection in repetitive fabric patterns, Proceeding of Visualization, Imaging and Image Processing, September 6-8, Marbella, Spain, Weng Y. S. and Perng M. H.: Periodic Pattern Inspection using Convolution Masks, Proceedings of the Conference on Machine Vision Applications (MVA), Tokyo, JAPAN, May 16-18, 2007, pp Sivabalan, K. N., & D. Gnanadurai. (2011). Efficient Defect Detection Algorithm For Gray Level Digital Images Using Gabor Wavelet Filter And Gaussian Filter. International Journal of Engineering Science and Technology, 3(4), Li, Y., Ai, J., & Sun, C. (2013). Online fabric defect inspection using smart visual sensors. Sensors (Switzerland), 13(4), Sun, G., Li, H., Dai, X., & Feng, W. (2013). Method of Mesh Fabric Defect Inspection Based on Machine Vision. Journal of Engineered Fibers and Fabrics, 8(2), Visual Inspection and Grading of Fabrics. [Online]. Available : Rana, N. (2012). Fabric inspection systems for apparel industry. The Indian Textile Journalndian Textile Journal, Smartview Nonwovens, Cognex brochure, Cognex Corporation One Vision Drive, Natick MA USA 172

201 Bibliography 1. M. Ferreira, C. Santos & J. Monteiro (2009). A Texture Segmentation Prototype for Industrial Inspection Applications Based on Fuzzy Grammar. Sensor Review 2, 29(2), J. Scharcanski (2007). A Wavelet Based Approach for Analysing Industrial Stochastic Textures With Applications. IEEE Transactions on Systems, Man, and Cybernetics - Part A: Systems and Humans, Vol. 37, 37(1), pp Vishwakarma, M. D. D. (2012). Analysis of Fabric Properties Using Digital Fabric Simulator. International Journal of Engineering Research and Development, 4(2), Eldessouki, M., Hassan, M., Qashqari, K., & Shady, E. (2014). Application of Principal Component Analysis to Boost the Performance of an Automated Fabric Fault Detector and Classifier. Fibres & Textiles in Eastern Europe, 22(4), Guha P (2011). Automated Visual Inspection of Steel Surface, Texture Segmentation and Development of a Perceptual Similarity Measure. Dsc Thesis, IIT, Kanpur. 6. Dalwadi, M. N., Khandhar, P. D. N., & Wandra, P. K. H. (2013). Automatic Boundary Detection and Generation of Region of Interest for Focal Liver Lesion Ultrasound Image Using Texture Analysis. International Journal of Advanced Research in Computer Engineering & Technology, 2(7), Burcack, K. (2004). Characterization and Role of Porosity in Knitted Fabrics. Dsc thesis, North Carolina State University. 8. S.Anitha, & Dr.V.Radha. (2010). Comparison of Image Preprocessing Techniques for Textile Texture Images. International Journal of Engineering Science and Technology, 2(12), Güler, H., Zor, G., & Gunes, M. (2015). Comparison of Performances of Spectral Based Approaches on Fabric Defect Detection. Journal of Engineering Research and Applications, 5(5), Singha, K., Maity, S., Singha, M., & pal, S. (2012). Computer Simulations of Textile Non-Woven Structures. Frontiers in Science, 2(2),

202 11. Patel, J., Jain, M., & Dutta, P. (2013). Detection and Location of Defects Fabrics Using Feature Extraction Technique. International Journal of Emerging Trends in Engineering and Development, 5(3), Patel, J., Jain, M., & Dutta, P. (2013). Detection of Faults Using Digital Image Processing Technique. Asian Journal of Engineering and Applied Technology, 2(1), Retief, A., & De Klerk, H. M. (2003). Development of a guide for the visual assessment of the quality of clothing textile products. Journal of Family Ecology and Consumer Sciences, 31, Kumar, U. (2010). Development of Automated Non-Contact Inspection Methodology through Experimentation. Department of Industrial Engineering & Management Indian Institute of Technology, Kharagpur, India. Indian Institute of Technology, Kharagpur, India. 15. Bennamoun, M., & Bodnarova, A. (2003). Digital Image Processing Techniques for Automatic Textile Quality Control. Systems Analysis Modelling Simulation, 43(11), Singh, B., & Singh, A. (2008). Edge detection in gray level images based on the Shannon entropy. Journal of Computer Science, 4(3), Albregtsen, F. (2010). INF 4300 Digital Image Analysis. 18. Alavi, F. F. (2010). In-Line Extrusion Monitoring and Product Quality. Dsc thesis, Chemical Engineering and Applied Chemistry. University of Toronto. 19. Sayed, U., Kadam, N., & Avinash, K. Laser Technology in Textile Industry. Textile Asia, Patel, J., Jain, M., & Dutta, P. (2013). Location of Defects Fabrics Using Feature Extraction Technique. International Journal of Research Management. ISSN , 5(3), Krasula, L., Klíma, M., Rogard, E., & Jeanblanc, E. (2011). MATLAB-based applications for image processing and image quality assessment - Part I: Software description. Radioengineering, 20(4), Albrecht, W., fuchs, H., kittelmann W. (2003). Nonwoven fabrics. 23. Singh M D (2008). Parameter Optimization For Segmenting Structures In CT images. Dsc thesis, Thapar University. 24. Stojanovic, R., Mitropulus, P., Koulamas, C., & Karayiannis, Y. (2001). Real-Time Vision-Based System for Textile Fabric Inspection. RealTime Imaging, 7(6), 174

203 25. Aulinas J, & Garcia F. Scene Segmentation and Interpretation.[Online]. Available : Fazekas Z., K. J. R. I. S. L. (1999). Towards objective visual assessment of fabric features. In Image Processing and its Applications (pp ). 27. Sezer, O. G., Ercil, a., & Ertuzun, a. (2007). Using perceptual relation of regularity and anisotropy in the texture with independent component model for defect detection. Pattern Recognition, 40(1), Xin, B., Hu, J., & Baciu, G. (2010). Visualization of textile surface roughness based on silhouette image analysis. Textile Research Journal, 80(2),

204 List of Publications INTERNATIONAL: 1. Industrial Fabrics used in Conveyor & Power Transmission Belts paper published in the Proceedings of 6th International Conference on Advances in Textiles, Machinery, Nonwovens and Technical Textiles held during 7th -9th of December 2009 at Bannari Amman Institute of Tech., Sathyamangalam, Erode District, Tamilnadu, India, organized jointly with Texas Tech University, Nonwovens & Advanced Materials Laboratory, The Institute of Environmental & Human Health, Lubbock, USA 2. Quality Parameters for Medical Textiles and Their Assessment - paper published in the Proceedings of MEDITEX-2014 International Conference on Current Trends in Medical Textile Research organized by Centre of Excellence In Medical Textiles, The South India Textile Research Association, Coimbatore, Tamil Nadu, India and sponsored by Office of the Textile Commissioner, Ministry of Textiles, Government of India on 1st March, Quality Parameters for Baby Diapers and Their Assessment - paper published in the Proceedings of INDO CZECH INTERNATIONAL CONFERENCE on Advancements in Specialty Textiles and their Applications in Material Engineering and Medical Sciences (ICIC 2014) organized jointly by Department of Textile Technology / Department of Fashion Technology, Kumaraguru College of Technology, Coimbatore and Technical University of Liberec, Faculty of Textile Engineering, Czech Republic during 29th-30th April, Development of Eco Friendly and Cost Effective Solutions for Packaging Industries - paper published in the Proceedings of International Conference on Technical Textiles and Nonwovens organized by IIT Delhi during 6-8 November, 2014 at IIT Delhi. 5. Development of Conductive Fabrics and their Applications in Textiles in TEXTILE ASIA, p.29-32, Dec Developments in Medical Textiles for the Need of the Day - paper published as a Poster at the International Conference on Technical Textiles and Nonwovens organized by IIT Delhi during 6-8 November, 2014 at IIT Delhi. 7. Quality Requirements For Woven Fabrics Used As Functional Textiles, paper published in the Proceedings of the Global Textile Congress organized by The Textile 176

205 Association (India) in association with Thailand Convention & Exhibition Centre, Thailand Theme : Global Textile Opportunities & Challenges in an Integrated Word during February, 2015 at Ambassador Hotel (Convention Hall ), Bangkok, Thailand. 8. High Performance Nonwovens for Infrastructural Developments in India paper published in the Proceedings of the Second International Conference on Nonwovens for High Performance Applications organized by the International Newsletters Ltd., UK during 4-5 March, 2015 at Novotel Hotel, Cannes, France. NATIONAL: 9. Influence of Properties of Back-Up Fabrics on Properties of Synthetic Leather in Journal of the Textile Association. May-June, 2014, Vol. No. 75 No. 1 pg A Review of Detection of Structural Variability in Textiles using Image Processing and Computer Vision in Journal for Research Volume 01 Issue 12 February 2016 ISSN: , pg Patents: Filed Provisional patent for Fabric Quality Monitoring Device with Application No , Transaction ID- N

206 Appendix A 178

207 179

208 180

209 Appendix B Script for learning of Images %% %spunbond clear all, close all I1=imread('Still0024ns7.jpg') %convert gray g=rgb2gray(i1) figure(1), imshow(g) cg = adapthisteq(g); figure(2), imshow(cg); gd=im2double(g) cgd=im2double(cg) m1=mean2(g) sd1=std2(g) m2=mean2(cg) sd2=std2(cg) % msgbox(sprintf('meanimg=%g\nsdimage=%g\nmean2cg=%g\nsdimage2=%g', m1, sd1, m2, sd2)) T1=graythresh(g) T2=graythresh(cg) figure(3), imhist(g) figure(4), imhist(cg) u_max1=mean2(max(g)) u_min1=mean2(min(g)) v_max2=mean2(max(cg)) v_min2=mean2(min(cg)) % figure(5), surfc(g) % figure(6), surfc(cg) % figure(7), boxplot(g) %% %geoclear all, close all I1=imread('Still0027g5s.jpg') %convert gray7 g=rgb2gray(i1) figure(1), imshow(g) B = imgaussfilt(g, 1.5) cg=gradientweight(b, 1.5) figure(2), imshow(cg); m1=mean2(g) sd1=std2(g) m2=mean2(cg) sd2=std2(cg) % msgbox(sprintf('meanimg=%g\nsdimage=%g\nmean2cg=%g\nsdimage2=%g', m1, sd1, m2, sd2)) T1=graythresh(g) T2=graythresh(cg) % figure(3), imhist(g) % figure(4), imhist(cg) u_max1=mean2(max(g)) u_min1=mean2(min(g)) v_max2=mean2(max(cg)) v_min2=mean2(min(cg)) % figure(5), surfc(g) % figure(6), surfc(cg) % figure(7), boxplot(g) 181

210 Script for Defect Detection function varargout = Finalnew_QC_all(varargin) % FINALNEW_QC_ALL MATLAB code for Finalnew_QC_all.fig % FINALNEW_QC_ALL, by itself, creates a new FINALNEW_QC_ALL or raises the existing % singleton*. % % H = FINALNEW_QC_ALL returns the handle to a new FINALNEW_QC_ALL or the handle to % the existing singleton*. % % FINALNEW_QC_ALL('CALLBACK',hObject,eventData,handles,...) calls the local % function named CALLBACK in FINALNEW_QC_ALL.M with the given input arguments. % % FINALNEW_QC_ALL('Property','Value',...) creates a new FINALNEW_QC_ALL or raises the % existing singleton*. Starting from the left, property value pairs are % applied to the GUI before Finalnew_QC_all_OpeningFcn gets called. An % unrecognized property name or invalid value makes property application % stop. All inputs are passed to Finalnew_QC_all_OpeningFcn via varargin. % % *See GUI Options on GUIDE's Tools menu. Choose "GUI allows only one % instance to run (singleton)". % % See also: GUIDE, GUIDATA, GUIHANDLES % Edit the above text to modify the response to help Finalnew_QC_all % Last Modified by GUIDE v Jul :06:57 % Begin initialization code - DO NOT EDIT gui_singleton = 1; gui_state = struct('gui_name', mfilename,... 'gui_singleton', gui_singleton,... 'gui_layoutfcn', [],... 'gui_callback', []); if nargin && ischar(varargin{1}) gui_state.gui_callback = str2func(varargin{1}); end if nargout [varargout{1:nargout}] = gui_mainfcn(gui_state, varargin{:}); else gui_mainfcn(gui_state, varargin{:}); end 182

211 % End initialization code - DO NOT EDIT % --- Executes just before Finalnew_QC_all is made visible. function Finalnew_QC_all_OpeningFcn(hObject, eventdata, handles, varargin) % This function has no output args, see OutputFcn. % hobject handle to figure % eventdata reserved - to be defined in a future version of MATLAB % handles structure with handles and user data (see GUIDATA) % varargin command line arguments to Finalnew_QC_all (see VARARGIN) % Choose default command line output for Finalnew_QC_all handles.output = hobject; % Update handles structure guidata(hobject, handles); % UIWAIT makes Finalnew_QC_all wait for user response (see UIRESUME) % uiwait(handles.figure1); % --- Outputs from this function are returned to the command line. function varargout = Finalnew_QC_all_OutputFcn(hObject, eventdata, handles) % varargout cell array for returning output args (see VARARGOUT); % hobject handle to figure % eventdata reserved - to be defined in a future version of MATLAB % handles structure with handles and user data (see GUIDATA) % Get default command line output from handles structure varargout{1} = handles.output; % function Untitled_1_Callback(hObject, eventdata, handles) % hobject handle to Untitled_1 (see GCBO) % eventdata reserved - to be defined in a future version of MATLAB % handles structure with handles and user data (see GUIDATA) % function Untitled_2_Callback(hObject, eventdata, handles) % hobject handle to Untitled_2 (see GCBO) % eventdata reserved - to be defined in a future version of MATLAB % handles structure with handles and user data (see GUIDATA) % function Untitled_5_Callback(hObject, eventdata, handles) % hobject handle to Untitled_5 (see GCBO) for geo % eventdata reserved - to be defined in a future version of MATLAB % handles structure with handles and user data (see GUIDATA) % getting image file global im1 im1a [path,user_cance]=imgetfile(); if user_cance msgbox(sprintf('error'),'error','error'); return 183

212 end im1=imread(path); axes(handles.axes1); imshow(im1); % function Untitled_6_Callback(hObject, eventdata, handles) % hobject handle to Untitled_6 (see GCBO) %%process for geo % eventdata reserved - to be defined in a future version of MATLAB % handles structure with handles and user data (see GUIDATA) %processing Geotexrile Fabric global im1 g=rgb2gray(im1) [rim cim]=size(g) totalimagearea=rim*cim B = imgaussfilt(g, 1.5) cg=gradientweight(b, 1.5) % figure(2), imshow(cg) m1=mean2(g) m2=mean2(cg) [mm nn]=size(cg); for i=1:mm for j=1:nn if cg(i,j)>0.2 new(i,j)=1; else new(i,j)=0; end end end % figure(3), imshow(new) if m2<0.1 bwk=im2bw(new) else bwk=im2bw(~new) end % figure(4), imshow(bwk) %% cc = bwconncomp(bwk); L = labelmatrix(cc); rgb = label2rgb(l); % imshow(rgb) k1=regionprops(l, 'Area', 'BoundingBox') no=cc.numobjects %% %using area, majoraxis... allblobareas=[k1.area] max1=max(allblobareas) bwk1=bwareafilt2(bwk,[49 max1], 10, 'largest') cc1=bwconncomp(bwk1) L1=labelmatrix(cc1) stats=regionprops(l1, 'Area', 'BoundingBox', 'MajorAxisLength','Perimeter') n2=cc1.numobjects MA=[stats.MajorAxisLength] oriarea=[stats.area] sumoriarea=sum(oriarea) peri=[stats.perimeter] % figure(5), imshow(bwk1) bwk1=imfill(bwk1,'holes') 184

213 figure(6), imshow(bwk1) %% %eliminating non defect regions DT1=('nil');defectno1=0 DT2=('nil');defectno2=0 DT3=('nil');defectno3=0 A=0 B=0 C=0 D=0 for j=1:n2 op1(j)=[stats(j).boundingbox(1)] op2(j)=[stats(j).boundingbox(2)] lth(j)=[stats(j).boundingbox(3)] bth(j)=[stats(j).boundingbox(4)] bbarea(j)=lth(j)*bth(j) if peri(j)<500 i(j)=j else i(j)=0 end end %% %detecting defect type & frequency for j=1:n2 if i(j)~=0 & bth(j)<40 DT1='Warp Wise-Missing End/Stain Line' defectno1=defectno1+1 else if i(j)~=0 & lth(j)<40 DT2='Weft Wise-Missing Pick/Weft Crease' defectno2=defectno2+1 else if i(j)~=0 %&peri(j)<500 DT3='Circular-Pinhole/Stain spots/slubs/gout' defectno3=defectno3+1 end end end if i(j)~=0& MA(j)>50 & MA(j)<=280 %& peri(j)<500 A=A+1 elseif i(j)~=0& MA(j)>280 & MA(j)<=570 %& peri(j)<500 B=B+1 elseif i(j)~=0& MA(j)>570 & MA(j)<=850 %& peri(j)<500 C=C+1 elseif i(j)~=0& MA(j)>850% & peri(j)<500 D=D+1 end end defectno=a+b+c+d %% %Grading if A>B & A>C & A>C & defectno<=2 sumoriarea<800 fabgrade='a' else if B>A & B>C & B>D & defectno<=3 fabgrade='b' else if C>A & C>B & C>D defectno<=6 185

214 fabgrade='c' elseif D>A & D>B & D>C defectno>=7 fabgrade='d' end end end %% %show image axes(handles.axes1); imshow(im1), hold on; himage=imshow(bwk1) set(himage, 'AlphaData',.3) %% % all calculations areabb=sum(bbarea) areacms=areabb/1390 areainch=areabb/9000 perdefarea=areabb/totalimagearea*100 if perdefarea>100 perdefarea=100 end MAmax=max(MA) MAmaxinch=MAmax/95 %% f = figure('position',[ ]); % create the data d = [areainch, perdefarea, MAmaxinch, defectno, {fabgrade}, {DT1},{DT2},{DT3} ]; % Create the column and row names in cell arrays rnames = {'Values'}; cnames = {'Total Defective Area','Total Percentage Defective Area', 'Length/Width of Biggest Defect' 'No. of Defects','GRADE OF FABRIC','Defect Names'}%, 'Probable Defect Name'}; cformat = {'numeric','numeric','numeric', 'numeric','char', 'char', 'char', 'char'}; % Create the uitable t = uitable(f,'data',d,... 'ColumnName',cnames,... 'ColumnFormat',cformat,... 'RowName',rnames); t.position(3) = t.extent(3); t.position(4) = t.extent(4); % function Untitled_3_Callback(hObject, eventdata, handles) % hobject handle to Untitled_3 (see GCBO) % eventdata reserved - to be defined in a future version of MATLAB % handles structure with handles and user data (see GUIDATA) %getting image file global im1 im1a [path,user_cance]=imgetfile(); if user_cance msgbox(sprintf('error'),'error','error'); 186

215 return end im1=imread(path); axes(handles.axes1); imshow(im1); % function Untitled_4_Callback(hObject, eventdata, handles) % hobject handle to Untitled_4 (see GCBO) % eventdata reserved - to be defined in a future version of MATLAB % handles structure with handles and user data (see GUIDATA) % processing Spunbonded Fabrics global im1 %convert gray g=rgb2gray(im1) axes(handles.axes1); imshow(g) [rim cim]=size(g) totalimagearea=rim*cim cg = adapthisteq(g); m1=mean2(g) m2=round(mean2(cg)) sd2=std2(cg) k1=cg>100 % dk k2=cg<180 % lt % figure(1), imshow(k1) % figure(2), imshow(k2) %% % thresholding bwk1=~k1 bwk2=~k2 % figure(3), imshow(bwk1)% dk % figure(4), imshow(bwk2)% lt %% % dark places cc1 = bwconncomp(bwk1); L1 = labelmatrix(cc1); rgb1 = label2rgb(l1); k11=regionprops(l1, 'Area', 'BoundingBox') no1=cc1.numobjects %% %light places cc2 = bwconncomp(bwk2); L2 = labelmatrix(cc2); rgb2 = label2rgb(l2); k12=regionprops(l2, 'Area', 'BoundingBox') no2=cc2.numobjects %% % finding regions dk allblobareas1=[k11.area] max11=max(allblobareas1) bwk11=bwareafilt2(bwk1,[49 max11], 5, 'largest') cc11=bwconncomp(bwk11) L11=labelmatrix(cc11) % stats1=regionprops(l11, 'Area', 'BoundingBox', 'MajorAxisLength', 'MinorAxisLength', 'EquivDiameter') % finding regions lt allblobareas2=[k12.area] max12=max(allblobareas2) bwk12=bwareafilt2(bwk2,[49 max12], 5, 'largest') 187

216 cc12=bwconncomp(bwk12) L12=labelmatrix(cc12) % stats2=regionprops(l12, 'Area', 'BoundingBox', 'MajorAxisLength', 'MinorAxisLength', 'EquivDiameter') %% %combined defects bwk22=imadd(bwk11,bwk12) bwk22=imfill(bwk22,'holes') % figure(5), imshow(bwk22) cc22 = bwconncomp(bwk22); L22 = labelmatrix(cc22); rgb22 = label2rgb(l22); k22=regionprops(l22, 'Area') no22=cc22.numobjects allblobareas22=[k22.area] max22=max(allblobareas22) bwk22a=bwareafilt2(bwk22,[49 max22], 10, 'largest') cc22a=bwconncomp(bwk22a) L22a=labelmatrix(cc22a) stats22=regionprops(l22a, 'Area', 'BoundingBox', 'MajorAxisLength') MA22=[stats22.MajorAxisLength] oriarea22=[stats22.area] n222=cc22a.numobjects sumoriarea22=sum(oriarea22) % figure(6), imshow(bwk22a) %% % finding small obj and eliminating them % lt(1)=0 % dk(1)=0 for j=1:n222 lth22(j)=[stats22(j).boundingbox(3)] bth22(j)=[stats22(j).boundingbox(4)] bbarea22(j)=lth22(j)*bth22(j) perdifflth(j)=abs(oriarea22(j)-bbarea22(j))/bbarea22(j) if bbarea22(j)<1000 i(j)=j else i(j)=0 end end defectno=0 bbarea22a=0 for j=1:n222 if i(j)==0 & oriarea22(j)>300 & MA22(j)>100 lth22a(j)=[stats22(j).boundingbox(3)] bth22a(j)=[stats22(j).boundingbox(4)] bbarea22a(j)=lth22a(j)*bth22a(j) defectno=defectno+1 else defectno=defectno+0 end end db=0 for j=1:n

217 if perdifflth(j)<0.6% & oriarea22(j)<3000 db=db+1 end end axes(handles.axes1); imshow(im1), hold on; himage22=imshow(bwk22a) set(himage22, 'AlphaData',.25) %% %all calculations areabb22=sum(bbarea22) % dk areacms22=areabb22/1390 areainch22=areabb22/9000 perdefarea22=areabb22/totalimagearea*100 if perdefarea22>100 perdefarea22=100 end objarea=sum(bbarea22a) objareacms=objarea/1390 objareainch=objarea/9000 objperdefarea22=objarea/totalimagearea*100 if objperdefarea22>100 objperdefarea22=100 end %% %output if (perdefarea22<=10 & objperdefarea22<=10) & defectno<=2 fabgrade='a' elseif (perdefarea22<=30 & objperdefarea22<=30) & defectno<=4 fabgrade='b' elseif (perdefarea22<=60 & objperdefarea22<=60) & defectno>=2 fabgrade='c' elseif (perdefarea22>=60 & objperdefarea22>=60) & defectno>=1 fabgrade='d' end if fabgrade=='a' DT='No Objectionable Defect' elseif fabgrade=='d' DT='Holes' elseif (db==3 db==4) & defectno>=2 DT='Drop Bond/Point Fusion' else DT='PinHole/CC/HF/Holes' end %% f = figure('position',[ ]); % create the data d = [areacms22, perdefarea22, defectno, objareacms, objperdefarea22, {fabgrade}, {DT} ]; % Create the column and row names in cell arrays rnames = {'Values'}; cnames = {'Total Defective Area','Total Percentage Defective Area','Number of Objectionable Defects', 'Objectionable Area','Percentage Objectionable Defective Area', 'GRADE OF FABRIC', 'Probable Defect Name'}; cformat = {'numeric','numeric','numeric', 'numeric','numeric', 'char'}; % Create the uitable t = uitable(f,'data',d,

218 'ColumnName',cnames,... 'ColumnFormat',cformat,... 'RowName',rnames); % Set width and height t.position(3) = t.extent(3); t.position(4) = t.extent(4); 190

219 Appendix C 191