SAR IMAGE CLASSIFICATION USING FUZZY C-MEANS

Transcription

1 SAR IMAGE CLASSIFICATION USING FUZZY C-MEANS Debabrata Samanta, Goutam Sanyal Deartment of CSE, National Institute of Technology, Durgaur, Mahatma Gandhi Avenue, West Bengal, India ABSTRACT Image Classification is the evolution of searating or grouing an image into different arts. The good act of recognition algorithms based on the quality of classified image. The good feat of recognition algorithms based on the quality of classified image. An imortant roblem in SAR image alication is accurate classification. Image segmentation is the mainly ractical loom among virtually all automated image recognition systems. Fuzzy c-means (FCM) is one of rominent unsuervised clustering methods, which can be used for Synthetic Aerture Radar (SAR) image classification. In this aer, we consider the roblem of SAR image Classification by Fuzzy c-means technique. Here we roosed satial information with the FCM clustering for imroving the SAR image classification result. Hear two different fuzzy clustering techniques on SAR images that minimize two different objective functions for merging different region to get the classified SAR images. KEYWORDS: FCM, SAR image, color ma. I. INTRODUCTION Synthetic Aerture Radar (SAR) grants such a caability. SAR systems take advantage of the longrange roagation characteristics of radar signals and the comlex information rocessing caability of modern digital electronics to rovide high resolution imagery. Synthetic Aerture Radar (SAR) image classification is becoming more and more increasingly imortant in military or scientific research.. Region secification is a crucial ste towards automatic classification of SAR images. Under some severe conditions of imroer illumination and unexected disturbances, the blurring images make it more difficult for target recognition, which results in the necessity of classification. Color based classification of image is a decisive oeration in image analysis and in many comuter vision, image elucidation, and attern recognition system, with alications in scientific and industrial field(s) such as medicine, Remote Sensing, content based image and video reossession, document analysis, industrial automation and quality control. The erformance of color segmentation may significantly affect the quality of an image understanding system. The motivation of this work is to develo a novel classification algorithm, which can be used to classified the SAR images and imrove the overall accuracy [7] [9]. Classification algorithms for Synthetic Aerture Radar images have been suggested in the field of categorization [6]. During the ast years, different techniques were in use for classification of synthetic aerture radar (SAR) images, based on the Maximum Likelihood (ML) [], artificial Neural Networks (NN) [2] [3], fuzzy methods [4] or other aroaches. The NN classifier deends only on the training data and the discrimination ower of the features. Fukuda and Hirosawa [5] [] roosed the wavelet-based texture feature sets for classification of multi frequency olar metric SAR images. So the Classification accuracy deends on quality of features and the emloyed classification algorithm. [8] [2] Fuzzy c-means is a method of clustering, which allows one iece of data belongs to two or more clusters. 508 Vol. 4, Issue 2,

2 In this aer, we roosed the modification of an unsuervised fuzzy clustering technique to guide the classification rocess by adding a-riori geometrical information in order to imrove the final classification result. II. FUZZY C-MEANS CLUSTERING ALGORITHM The Fuzzy c-means clustering algorithm is an iterative clustering method that roduces an otimal c artition by minimizing the weighted within grou sum of squared error objective function T fc Where P {,,.u } 2 n r c q 2 fc = ( rc) ( r, c) () k = l= T U f X Y q = U is the data set in the -dimensional vector sace, is the number of data items, c is the number of clusters with 2 c < n, u rk is the degree of membershi of P r in the rth cluster, q is the weighting exonent on each fuzzy membershi, v c is the rototye of the centre of cluster r, d2( X r, Y c) is a distance measure between object P c and cluster centre V c of SAR image. A solution of the object function T fc can be follows:: Set values for the SAR image are c,q,, Initialize the fuzzy artition matrix of segmented region. Set a loo counter c=0. Calculate c cluster for { Y } with U (b) c Y n ( X irc ) uc l= c n ( X rc ) l= In this case we can write for a SAR image: = (2) n 2 y ( X Y ) = 0 (3) l= rc r c III. PROPOSED ALGORITHM Select initial class rototye of SAR image { V } i Udate the membershis of U ij Obtain the rototye of clusters of SAR image of weighted average. The termination occurred when V new - V old 509 Vol. 4, Issue 2,

3 IV. PROPOSED WORK FLOW DIAGRAM V. EXPERIMENT RESULT In this thesis work, we have considered synthetic aerture radar images. The SAR images are classified by using novel aroach based on Fuzzy c-means technique. The figure (Fig to Fig 2) shows the original SAR images and corresonding histograms and the figure (Fig (a) to Fig 2(a)) shows the histograms of classified SAR images. Fig: SAR image and Histogram If we taking the ick values we get the no. of class are 4. After using our methodology we get the Histogram given below: Figa: Classified Histogram based on Fuzzy c-means technique 50 Vol. 4, Issue 2,

4 If we taking the ick values we get the no. of class are 0. Fig2: SAR image2 and Histogram If we taking the ick values we get the no. of class are 3. After using our methodology we get the Histogram given below: Fig2a: Classified Histogram based on Fuzzy c-means technique If we taking the ick values we get the no. of class are 9. VI. CONCLUSION SAR IMAGE Table:-Class Table of SAR images. CLASS BASED ON NORMAL HISTOGRAM FUZZY C-MEANS TECHNIQUE HISTOGRAM Image 4 0 Image2 3 9 In this aer, a novel algorithm based on the Fuzzy c-means for classification of SAR images is roosed. This technique is based on considering a 3X3 window and calculates successively the corresonding First, mean and variance of the SAR Images. Then store the color feature using Fuzzy c-means of same SAR image for better result. The roosed algorithm gives better result comared with Histogram based classification of SAR images. REFERENCES [] Lee J S, Grunes M R, Kwok R. Classification of multi-look olarimetric SAR imagery based on comlex Wishart distribution, Int. J. Remote Sensing, 994, 5(): [2] Bischof H, Schneider W, Pinz A J. Multisectral classification of landsat-images using neural networks, IEEE Trans. Geosci. Remote Sensing, 992, 30(3): Vol. 4, Issue 2,

5 [3] Chen C T, Chen K S, Lee J S. The use of fully olarimetric information for the fuzzy neural classification of SAR images, IEEE Trans. Geosci. Remote Sensing, 2003, 4(9 Part I): [4] Du L, Lee J S. Fuzzy classification of earth terrain covers using comlex olarimetric SAR data, International Journal of Remote Sensing, 996, 7(4): [5] Fukuda S, Hirosawa H. A wavelet-based texture feature set alied to classification of multi frequency olarimetric SAR images. IEEE Trans. Geosci. Remote Sensing, 999, 37(5): [6] Debabrata Samanta, and Goutam Sanyal, Develoment of Adative Thresholding Technique for Classification of Synthetic Aerture Radar Images, International Journal of Comuter Science and Technology (IJCST), Vol. 2 Issue , OCT - DEC, 20. [7] Chih-Chang Lai,Ching-Chih Tasi,A Modified Strie-RGBW TFT-LCD with Image-Processing Engine for Mobile Phone Dislays, IEEE Transaction on Comuter Electronics,Vol.- 53,No. 4,Nov [8] Debabrata Samanta, Mousumi Paul and Goutam Sanyal, Segmentation Technique of SAR Imagery using Entroy, International Journal of Comuter Technology and Alications, Vol. 2 (5), , 20. [9] J. Huang, S. R. Kumar, M. Mitra, W. J. Zhu, R. Zabih, Image Indexing Using Color Correlograms, Proc. IEEE Int. Conf. on Comuter Vision and Pattern Recognition, , 997. [0] M. J. Swain, S. H. Ballard, Color Indexing, Int. Journal of Comuter Vision, Vol.7, No.,. - 32, 99. [] W. Y. Ma, H. J. Zhang,, Content-based Image Indexing and Retrieval, In Handbook of Multimedia Comuting, Borko Furht. Ed, CRC Press, 998. [2] Theodor Richardson, Imroving the Entroy Algorithm with Image Segmentation, [Online document] 2003, Available at HTTP: htt:// AUTHORS Debabrata Samanta is a member of the IAENG, Board member of the Seventh Sense Research Grou Journals (SSRGJ), member of Editorial Board of IJSCE. He obtained his B.Sc. (Physics Honors) in the year 2007, from the Vivekananda Collage, Takurukur, under Calcutta University; Kolkata, India.He obtained his MCA in the year 200, from the Academy Of Technology, under WBUT. He is working his PhD in Comuter Science and Engg. In the year 200 from National Institute of Technology, Durgaur, India in the area of Image Processing.He is resently working as a Guest Faculty of CSE in University Institute of Technology, Burdwan, West Bengal, India. His areas of interest are Artificial Intelligence, Natural Language Processing and Image Processing. He has guided 6 PG and 5 UG thesis. He has ublished 34 aers in International Journals / Conferences. Gautam Sanyal is a member of the IEEE. He has received his B.E and M.Tech degree from National Institute of Technology (NIT), Durgaur, India. He has received Ph.D. (Engg.) from Jadavur University, Kolkata, India, in the area of Robot Vision. He has ublished nearly 98 aers in International and National Journals / Conferences. 3 Ph.Ds (Engg) have already been awarded under his guidance. At resent he is guiding eight Ph.Ds scholars in the field of steganograhy, Cellular Network, High Performance Comuting and Comuter Vision. He is resently working as a Professor in the deartment of Comuter Science and Engineering and also holding the ost of Dean (Students Welfare) at National Institute of Technology, Durgaur, India. 52 Vol. 4, Issue 2,