Available online at www.sciencedirect.com Procedia Engineering 15 (2011) 3516 3520 Abstract Advanced in Control Engineering and Information Science Study on Measuring Microfiber Diameter in Melt-blown WebBased on Image Analysis Chen Zeyun, Wang Rongwu a, Zhang Xianmiao, YinBaopu (Textile College Donghua University, Shanghai 201620, China) Melt-blown nonwovens are widely used in filter materials, fiber diameters and distributions play an important role in affecting the filtration efficiency of melt-blown material. The traditional method of microfiber diameter testing mainly relies on manual measurement, which is time-consuming and inaccurate. A large number of experiments show that the images quality of melt-blown material captured by optical microscope is not appropriate for automatic image processing. These images were not only vague, but are very difficult to get the clear edges of fibers by Edge Detection, such as LAPLACE, LOG and CANNY, etc. In such circumstances, based on scanning electron microscope (SEM) images, it proposes a new image processing method for measuring microfibers diameters. The new method consists of many steps: edge detection, image segmentation, thinning, lines linking and rectangle matching, which is successful in making fast, accurate automated measurements of microfiber diameters. The validation experiments show that the microfibers diameters and its distribution of melt-blown material can be measured accurately by the proposed method. 2011 Published by Elsevier Ltd. Open access under CC BY-NC-ND license. Selection and/or peer-review under responsibility of [CEIS 2011] Keyword: melt-blown nonwovens; microfiber diameter; image processing; SEM 1. Introduction Nonwoven fabric is essentially an assemblage of fibers held together by mechanical or chemical means, resulting in a mechanically stable, self-supporting, web-like structure. Melt-blown nonwovens are formed in process: convergent streams of hot air rapidly attenuate the extruded polymer streams to product extremely fine diameter fibers, and further collected to form a fine-fibered, self-bonded web [1]. Generally, the melt-blown web is highly opaque with fibers vary in diameter and random orientation in its surface. Melt-blown nonwovens are widely used in the filter material, as fibers are the main composition of melt-blown web, their diameters and distributions play an important role in determining the filtration performance. The finer of the fiber, the larger of its surface area, and thus the better filter efficiency of the web [2-3]. a Corresponding author. Tel.: 13816202508. E-mail address: wrw@dhu.edu.cn. 1877-7058 2011 Published by Elsevier Ltd. doi:10.1016/j.proeng.2011.08.658 Open access under CC BY-NC-ND license.
Chen Zeyun et al. / Procedia Engineering 15 (2011) 3516 3520 3517 Routine measurement of microfiber diameter and its distribution are carried out by manual methodusing ruler to gain fiber diameters from SEM images [4].This process is very time-consuming and operator consistency and fatigue may reduce the accuracy [5], so it is necessary to measure the fiber diameters quickly and accurately. At present, there has been no successful method for the automated measurement of microfiber diameter in melt-blown web at present [6-15]. Main difficulties are: 1) how to obtain the distinct images which are suitable for automatic analysis; 2) how to design the particular algorithms that are practical for the processing of melt-blown nonwovens images. 2. Experimental To study the measurement of microfiber diameter in the melt-blown web, the following experiments are carried out respectively. 1) Capture fiber images from optical microscope and SEM, these images are further processed respectively to determine fiber diameter using image analysis. 2) Use measuring tool (a software based on WYSIWYG which is applied for measuring distance in images) to obtain fiber diameters with SEM images. 2.1. Experiment in Optical Microscope 2.1.1. Fiber Images Capturing Melt-blown webs with three different area densities: 1# (25.5g/m2), 2# (24.2g/m2), 3# (23g/m2), are individually cut into 5 samples in size of 5cm 3cm and then spread over the glass slides. Using these slides, images are captured in three different types of optical microscope (ⅰ ⅱ ⅲ) with objective of 40x, 100x separately. For each slide, 100 images are captured. The experimental program is shown in Table 1. Table 1 The experimental program No. Microscope Type Objective Magnification BEION N120A 40 1270(0.347 mm/ pixel) M2000 CMOS Camera 100 3300(0.154 mm/ pixel) LABOMED CXLX 40 1300(0.339 mm/ pixel) SONY E435P CCD Camera 100 3400(0.131 mm/ pixel) NIKON ECLIPSE E20 DSU2 CCD Camera 40 680 (0.333 mm/ pixel) 100 1700(0.134 mm/ pixel) 2.1.2. Results and Discussion A series of images are captured in the experiment, however, when microscope objective adopts 40, images captured are quite vague due to the serious laminated problem caused by the deep focal lengths of microscope. The qualities of images are not improved after repeated adjustment of the microscope focus. We choose relatively clearest images to illustrate; Fig. 4 shows images of minimum area density of meltblown webs 3# captured in microscope ⅰ, ⅱ, ⅲ respectively. It s apparent that these images are vague. When microscope objective is switched to 100, numerous images are captured. Fig. 5 shows these images of melt-blown webs 3# captured in microscope ⅰ, ⅱ, ⅲ respectively when microscope objective is 100. It can be seen these images are also blurred. Fig. 4 Images of 3# captured in microscope,, (40 ) Fig. 5 Images of 3# captured in microscope,, (100 )
3518 Chen Zeyun et al. / Procedia Engineering 15 (2011) 3516 3520 Although we try many proposals and capture lots of images in the experiment, these images are not appropriate for automatic analysis. Still, we select the clearest images, trying to use all kinds of image edge detection algorithm and segmentation algorithm for processing[16-22]. However, after applying many edge detection algorithms in processing images in the experiment, obvious edge still can not be obtained. Among the best processing outcomes, as show in Fig 6 (a), (c), are using LOG operator and Canny operator. These processed images are partly visible, which can hardly be used for image analysis. Further processed with Otsu method and Threshold segmentation in order to produce binary images, but targets still can not be successfully separated from the image background, as show in Fig. 6 (b), (d), which leaves no possibility to measure fiber diameter. (a) LOG operator (b) Otsu method (c) Canny operator (d) Threshold method Fig. 6 Image edge detecting and segmenting 2.1.3. Summary In the experiment, many images of different area densities of melt-blown webs are captured. These images are vague whatever types or objectives of microscope applied, the main reasons are: light scattering happens due to the extremely small size and random orientation of fibers in the melt-blown webs, which causes serious multi-focal planes problem when capturing images. For these images, through a variety of processing, expected result of edge detection and segmentation still can not be obtained. The experiment comes to conclusion that the image quality of melt-blown webs captured by optical microscope is not appropriate for automatic image processing. 2.2. Experiment with SEM Images SEM images have high definition which can be used for measuring fiber diameter. Routine measurement of fiber diameter and its distribution are carried out by manual method using SEM images. The following experiment will focus on the images processing and automatic analysis of SEM images to measure fiber diameter. In order to validate the data obtained by this image processing, data that for validation (called validation data in the following text) is required, therefore, we develop a method carried out by measuring tool to measure fiber diameter as validation data. 2.2.1. Measuring Tool In this part of experiment, measuring tool is used to measure fiber diameter. First the scale is set, then, fiber diameters are obtained by the measuring tool. For each sample, 400 fiber diameters are obtained. Fiber diameters of 7 samples are measured respectively by two persons (A B) using measuring tool, meanwhile traditional method are used to obtain fiber diameters, Table 2 is the main fineness of fiber diameter distribution gained by these two manual methods. The data comparison of measuring tool between person A and person B indicates that fiber diameters measured by measuring tool is stable; while comparing data from traditional method and measuring tool it shows that fiber diameters measured by measuring tool is accurate and trusty. Table 2 Main fineness of fiber diameter distribution gained by manual methods Sample Traditional method Measuring tool Person A Person B 1# 2.25 2.25 2.25 2# 2.00 2.00 2.10 3# 1.75 1.75 1.75
Chen Zeyun et al. / Procedia Engineering 15 (2011) 3516 3520 3519 4# 1.9 1.90 1.9 5# 1.25 1.25 1.25 6# 1.25 1.25 1.25 7# 1.25 1.25 1.25 2.2.2. Image Process Through image processing, the high-quality SEM images can be used to measure fiber diameter rapidly. We propose a series of algorithms for the automated measuring of microfiber diameter in meltblown web: first the SEM images are undergone a series of image pre-processing, useful target edge information thus gained; then after the line extracting and rectangle matching, edge matching information of single fiber is available; finally, fiber diameter is obtained by calculating the distance of two matching straight line. The specific image processes are shown in Fig. 8.Fig. 9 shows the process outcome of SEM images, for each matching rectangle, a fiber diameter is obtained by calculating the distance of two matching straight line. Noise Removal Dilation Line extracting Edge Detection Thinning Line connecting Segmentation Pruning Rectangle matching Fig. 8 The specific image processes Fig. 9 process outcome of SEM images Fig. 10 is the fiber diameter distribution of sample-1 obtained by image process. It is apparently seen that fiber diameter distribution measured by this approach is so close to the validation data. For further validation, Fig. 11 is the main fineness of fiber diameter distribution gained by image process, it is obvious that this proposed method is reliable in measuring main fineness of fiber diameters correctly. Fig. 10 Fiber diameter distribution obtained by image process Fig. 11 Main fineness of fiber gained by image process 3. Conclusion In this study, a large number of experiments first are carried out in optical microscope. The qualities of images originally captured and further processed come to conclusion that optical microscope images of melt-blown web are not qualified in automatic analysis. And the method of using image measurement tool to measure fiber diameter from SEM images, although not completely out of manual operation, it reduces subjectivity compared to the traditional method, and the data of fiber diameters gained by this method is stable and accurate. For SEM images, based on edge detection, image segmentation, thinning, line linking and rectangle matching, this paper proposes a series of algorithms for measuring microfiber diameter. Experiment results indicate that the distribution and main fineness of fiber diameter obtained by our method are
3520 Chen Zeyun et al. / Procedia Engineering 15 (2011) 3516 3520 extremely close to the validation data. The method is automated, accurate, and much faster than traditional method and thus has the capability of being used as an advanced approach for measuring microfiber diameter in melt-blown material. Acknowledgment This research was supported by the Natural Science Foundation of Shanghai, China (Grant No. 11ZR1401100) and Doctoral Fund of Ministry of Education of China (Grant No.20100075120003). References [1] Ke Q F, Jin X Y. Studies of Non-woven[M]. Shanghai: Donghua University Press, 2004: 17-22. [2] Liu W S. The Development and Application of Melt-blown Nonwoven Fabric Technology[J]. Chemical Fiber & Textile Technology, 2007, (4): 33-35. [3] Qu Y H, Ke Q F, Xiangyu Jin. Studey on the Meltblow PLA Nonwoven Process and Filtration Property[J]. Technical Textiles, 2005, (5): 19-22. [4] Guo B C. Nonwoven Fabric Performance and Test[M]. Beijing: China Textile Press, 1994. [5] Zeng Y M, Liu L F. Nonwovens Properties Detection and Control Technology Based on Computer Image Processing[J]. Nonwovens, 2001, (3): 37-40. [6] He Z G. The Latest Development of Image Processing Technology in Testing Textile[J]. Inspection and Quarantine Science, 2002, (1): 59-60. [7] R.L.Baker. Determination of Fiber Cross-sectional Circularity from Measurement Made in A Longitudinal View[J].1979, (101): 59-983. [8] J.D.Berlin. Measuring the Cross-sectional Area of Cotton Fiber with an Image Analyzer[J].1981, (51): 109-113. [9] D. D. Thibodeaux. Cotton Fiber Maturity by Image Analysis Textile[J].1986, (56): 130-139. [10] B. Xu, B. Pourdeyhimi. Evaluation Maturity of Cotton Fiber Using Image Analysis: Definition and Algorithm[J].1994, 64(6), 330-335. [11] Wang R W, Wu X Y, Wang S Y. Automatic Identification of Ramie and Cotton Fibers Using Characteristics in Longitudinal View, Part II: Fiber StriPes Analysis[J]. Textile Research Journal,2007,(9). [12] Gong R H, Newton A. Image-analysis techniques Part1: the Measurement of Pore-size Distribution[J]. Journal of the Textile Institute, 1992, 83(2): 253-268. [13] Z. Yan, Rabdll R. Bresee. Characterizing Nonwovens Web Structure Using Image Analysis Techniques.Part5: Analysis of Shot in Meltblown Webs[J]. Text. Inst, 1998, (2): 320. [14] Wang L, Chen T, Chen X. Measurement of the Diameter of Nonwoven Fiber Based on Image Processing[J]. Test and Measurement Technology, 2008, (3): 279-382. [15] Wang X M. Study on Measurement of Fiber Diameters in Tropical Meltblown Webs[J]. Technical Textiles, 2007, (7): 39-42. [16] Lin H, Zhao C S, Shu N. Edge Detection and Evaluation Based on Canny Operator[J]. Heilongjiang Institute of Engineering, 2003, 17(2): 1-6. [17] Xiong Q J, Yang M S. Study on the Comparison of Different Edge Detection Algorithm in Image Processing[J]. Mechanical Engineering and Automation, 2009, (02). [18] Zeng Y Z, Hu Y, Zeng W Y, Zhen X X. Comparison of Algorithm for Image Edge Extracting[J]. Chengdu Aeronautical Vocation and Technology College, 2009, 25(4): 69-72. [19] Ning T F. Application and Development of the Digital Image Processing Technology [J]. Ship Electronic Engineering, 2009, (1): 38-40. [20] Huang C Z, Wang B, Yang Z. Study on Image Segmentation[J]. Computer Technology and Development, 2009, (6). [21] Song Y T, Ji X. Study on Image Segmentation Algorithm Based Mathematical Morphology of Binary Image[J]. Changchun Institute of Technology, 2008, 9(3): 68-70. [22] Shen R Q. Determining the Threshold of Binary Images of Textile Fibers[J]. Wuhan Institute of Textile Technology, 1994, 7(1): 29-32