Fabric Defect Detection Based on Glcm and Ica

Volume-1, Issue-1, 23 – 29

FABRIC DEFECT DETECTION BASED ON GLCM AND ICA

G.Anusha1 , S.Mythili 2AP[SRG]/CSE, A.Srisakthi 3AP/CSE

1PG scholar, Computer Science and Engineering, United Institute of Technology, Tamil Nadu, India,

2Assistant Professor, Computer Science and Engineering, United Institute of Technology,

Tamil Nadu, India

3Assistant Professor, Computer Science and Engineering, United Institute of Technology,

Tamil Nadu, India

ABSTRACT

Fabric defect detection plays an important role in inspection of fabric products. Many fabric defects are very small and undistinguishable, which can be detected only by monitoring the variation in the intensity. But when the fabric detection is done manually it requires more man power and more over the process involves a huge amount of caution during the process. To deal with this situation, automatic on-line quality control devices are used so that the defects identification rate can be improved In online fabric defect detection system, the reference image and the input image are taken, features are extracted, compared and the defects are identified using image processing techniques like Fourier transform, Gray level Co-occurrence Matrix (GLCM), Gabor filter, wavelet and Independent component analysis (ICA) methods. The main objective of this proposed work is that GLCM and ICA methods are used separately in fabric defect detection system .These are compared and analyzed to identify the best method in detecting the defects and speed in computation. As it can be done online, the man power is reduced and the product quality is improved to meet both customer demands and reduce the costs associated with off-quality

Volume-1, Issue-1, 23 – 29

Volume-1, Issue-1, 23 – 29

Volume-1, Issue-1, 23 – 29

Volume-1, Issue-1, 23 – 29

keywords: Gray level Cooccurence Matrix , Independent Component Analysis , Principal component Analysis

1 INTRODUCTION

Volume-1, Issue-1, 23 – 29

Volume-1, Issue-1, 23 – 29

Fabric defect detection is one of the most important and challenging field in textile industry. It is the process of identifying the location of various defects based on the textural properties of the input image .Generally the defect detection process is carried out by making use of the manual effort, during which some of fabric defects are very small and undistinguishable and cannot be identified easily. The intensive labor cannot always give consistent evaluation of products. Fabric Automatic Visual Inspection (FAVI) system is an attractive alternative to human vision inspection. It can provide reliable, objective and stable performance on fabric defects inspection. A good automated system means lower labor cost and shorter production time. Texture analysis is defined as a picture that is composed of high number of similar elements that are located at different positions of the image. It is one of the important characteristics for identifying defects or faults. Texture defect detection is the process of identifying the location of various defects based on the textural properties of the input image

Fabric faults or defects constitute nearly 85% of the defects found by the garment industry. Manufacturers recover only 45 to 65 % of their profits from seconds or off-quality goods.Broken ends, float, waste, Missing ends, picks, oily weft were the most costly defects that occurs in the fabrics. so when the defects are identified manually is requires more man power and more concern during the process

Image analysisis concerned with making quantitative measurements from an image to produce a description of it. Image analysis techniques require extraction of certain features that aid in the identification of the object. One way to reduce the total manufacturing cost and to provide a more reliable, objective, and consistent quality control process is to use an automated visual inspection system to detect possible defects in textile fabrics. However, automated visual inspection becomes a significant challenge due to some specific features pertaining to textile fabrics, for example: (a) large variety of fabric surfaces has to be examined. (b) Defects may take different forms that are usually difficult to classify. (c) New classes of defects arising from possible changes or aging of machineries in the production process

In order to analyse the texture defects in the fabrics, Gray level co-occurrence matrix and independent component analysis is used, so that the texture defect can be identified easily. The main contributions of this paper is

The two methods are introduced namely GLCM and ICA to identify the defect easily in the minimum time period .The input and the reference images are taken and this methods are used separately and the defects are identified.

After the defects are identified, the analysis is taken to find which method is best for analysing the defect easily with the low computation speed. The best work is suggested online.

2BACKGROUND AND RELATED WORK

The research in texture analysis has gained its attention due to its popularity in the manufacturing industry. The defects in fabrics are the main drawback in the product quality. Gray level cooccurence matrix and Gabor filter[1] to identify the defects in the fabric products. Both the methods are used separately and the defects are identified. Finally the comparison analysis is taken to identify the best method to identify the defects.

Glcm [2] method is used to improve the image quality. The images are divided into sub-windows and the parameters are calculated for the matrix. Image segmentation is done on the images so that the features are extracted and the defects are identified.

Glcm [3] is calculated with the angular second order moment and the entropy is calculated for the images. Correlation is calculated for the images for the image reduction. Homogeneity is calculated by the angular second order moment. Inverse difference moment is also calculated for the matrix.

Spatial domain gray level co-occurrence matrix (SDCM) [4] is used for the feature extraction that leads to effective solutions for the texture analysis problems.

In Gabor filter [5] technique different scales and orientations is generated and fabric images are filtered with convolution mask. Gabor transform localize the defects both in spatial and frequency domain. Morphological technique is used to filter the images

The co-occurrence matrix [6] is calculated for the images with its parameters such as entropy, contrast, homogeneity and correlation. The image is decomposed and the features are extracted in the images.

Independent Component Analysis[7] is used with the wavelet transform .sub-bands are extracted by two level wavelet transform. Pyramid structured wavelet is used. Wavelet packet decomposition is done where the energy is concentrated at the middle frequency and the bands are splitted.

Independent Component Analysis [8]technique is used for revealing the hidden factors in the patterned fabric. It is used for indicating and locating the defects on patterned fabric images. Ica pre-processing is done where the mean of every image is subtracted from itself and the every image is divided by its variance

ICA with Regular band method [9] is used to identify the defects in fabrics. RB method is the regularity approach and if there is any defect it will irregular. It is less sensitive to noise. Gabor filter with Ica [10] is used to identify the defects and the comparison analysis is taken to identify the defects

Principal component Analysis [11] is used for converting correlated into uncorrelated variables and it is analysed for its performance. Principal component analysis [12] [13] algorithm is used where the covariance analysis is calculated with Eigen values and vectors for dimensionality reduction

3 PROPOSED WORK

Fabric defect detection is important aspect in the fabric industry where the defect is the main cause for the low quality products. So if it is done online, the defects can be identified easily. Both GLCM and ICA are used to identify the defects with feature extraction and dimensionality reduction which is done by PCA

fig3.1 Process diagram for defect detection

In the defect detection work the reference image and the input image is taken .After that Gray level co-occurrence matrix and Independent component analysis method are applied on that images and the features are exacted. PCA is used for the dimensionality reduction of the images then the defects are identified by comparing both the images.

Both GLCM and ICA methods are applied separately on the images and the comparative analysis is taken to find which method is best in the computation speed and can be suggested online

Gray Level Cooccurence Matrix

The GLCM matrix is calculated with the four choices of degrees such as 0 0 , 45 0 , 90 0 , 135 0 for every subwindow of the images .The parameters such as entropy, contrast, homogeneity ,energy, inertia and co-relation are calculated for the subwindows. Euclidean distance is calculated between the subwindows of reference and the input image. Based on the parameter values calculated, the suitable threshold value is set. when the obtained value is lesser or equal to the threshold value then it is not defective. when it is greater than the threshold value it is defective.

Glcm Parameters

Energy

Contrast

Entropy

Homogeneity

Correlation

Choice of radius δ

The δ values ranges from 1, 2 to 10.The classification accuracies with δ = 1, 2, 4, 8 gives the best results for δ = 1 and 2.When the displacement value equal to the size of the texture element, it improves classification of the images

Choice of angle θ

Each pixel has eight neighboring pixels allowing eight choices for θ, which are 0°, 45°, 90°, 135°, 180°, 225°, 270° or 315°. The co-occurring pairs obtained by choosing θ equal to 0° would be similar to those obtained by choosing θ equal to 180° and so on

Independent Component Analysis

The independent component analysis is applied for every sub window of the images. The original image matrix is multiplied with the mixing matrix to get the observed matrix. The mixed matrix is inversed to pre-process and the original matrix is vectorized. The each sub window is also vectorized.Euclidean distance is calculated between the original matrix and the each sub window matrix. The threshold value is set and based on that the image is classified whether it is defective or not

Basic Ica Model

The basic linear mixture model can be expressed mathematically as

x=As(3.1)

Where,

x – observed vector

s – source vector

A– mixing matrix

To preprocess the image matrix A is formed and it is inversed ( A-1 ) .Once A is estimated, the sources can be computed as

s=Wx (3.2)

where

W is the (pseudo)inverse of the mixing matrix A and is called the demixing matrix

Dimensionality Reduction Using Pca

Volume-1, Issue-1, 23 – 29

Volume-1, Issue-1, 23 – 29

Principal Component Analysis

Volume-1, Issue-1, 23 – 29

Volume-1, Issue-1, 23 – 29

Principal component Analysis is one of the statistical technique used for dimensionality reduction and data decorrelation. It is the mathematical technique which transforms the original image that is highly correlated into uncorrelated principal components. These components are the linear combinations of original image.

In PCA the covariance matrix is calculated for the zero mean matrixes. From that matrix Eigen values and Eigen vectors are obtained

Here pca is used for dimensionality reduction .For example if there is four images of size 25x25. Each image is read as a single column vector so that the image size will be 625x1 .like this all the four images are read as a column vector in the single matrix so that the matrix size will be 625x4 .so this image matrix should be reduced as 625x2 using the Eigen value matrix. After the image matrix is reduced the ICA method is applied on the image

The four images are read into a column vector in the single matrix

img1 img2 img3 img4

625x4

fig 3.2 The four images are read into a column vector in the single matrix

These four images are reduced into two images by the dimensionality reduction.

625 x 2

fig 3.3 These four images are reduced into two images with the image size 625x2

4 EXPERIMENTS AND RESULTS

This chapter focuses on the demonstration of the proposed model with the parameter calculation of the GLCM matrix. The dimensionality reduction is done by using PCA and the pre-processing is done on that reduced matrix using ICA

Glcm Matrix Calculation

The GLCM matrix is calculated for the smooth, coarse and the regular image. The entropy value will be less for the smooth image compared to the coarse image because for the coarse image the uniformity will be less. The regular image will have the middle range of uniformity.

Table 4.1 Cooccurence matrix of the images with 00 horizontally

Glcm matrix is calculated for the images with the statistical featuresAfter the feature is extracted by the statistical parameters the Euclidean distance is calculated between the reference and the input images and based on that the threshold value is set and the defect is identified

Ica Pre-processing

In independent component analysis the pre-processing is done by the two methods centering and whitening centering is done by subtracting the mean of each columns of the input matrix Xwith zero mean. It is then passed through the whitening matrix V to remove the second order dependency. The whitening matrix Vis twice the inverse square root of the covariance matrix of the input matrix which is given by

V = 2[cov(x)] -1/2

After pre-processing, theIca method is applied in the images and the defects are identified

5 CONCLUSION

Fabric defect detection is the essential task in the fabric industry where the defects are to be identified and cleared .here Gray level cooccurence matrix and Independent component analysis is used and the defects are identified. Gray-scale statistical method is the most common method, which split image by calculating average gray value and standard deviation, and then compared with the abnormal region, which excess the defect. Gray Level Co-occurrence Matrix is the most classical second-order statistical method for texture analysis. In the proposed work Gray level Co-occurrence matrix is constructed for smooth, coarse and regular images. The parameters are then calculated for the three images. The defects are identified using the two methods Gray level Co-occurrence matrix and Independent Component Analysis and the analysis is taken to find which the best method to identify the defects based and speed in computation. The performances of the schemes developed are to be extensively evaluated by using an offline test database, which contains a variety of fabrics images, with and without defects chosen from various sources. Further the output result shows that, most of the defects in fabrics can be successfully detected and can be successfully used online

REFERENCES

[1] Ankit Chaudharyc, Jagdish Lal Rahejaa, Sunil Kumarb , “Fabric Defect Detection Based On Glcm And Gabor Filter A comparison ”, 14 May 2013.

[2] Dhanashree Gadkari , “Image quality analysis using glcm”, University of Pune, 2000 College of Arts and Sciences at the University of Central Florida ,Orlando, Florida

[3]L.GuruKumar ,P. Mohanaiah, P. Sathyanarayan , “Image Texture Feature Exaction Using Glcm Approach”, International Journal of Scientific and Research Publications, Volume 3, Issue 5, May 2013 1 ISSN 2250-3153

[4] Alaa Eleyan, Hasan Dem˙Irel , “Co-Occurrence Matrix And Its Statistical Features As a new Approach For Face Recognition”, Turk J Elec Eng & Comp Sci, Vol.19, No.1, 2011

[5]Baqir Ali ,TanveerSajid1,“Fabric Defect Detection In Textile Images Using Gabor Filter” (IOSR Journal of Electrical and Electronics Engineering (IOSR-JEEE)ISSN: 2278-1676Volume 3, Issue 2 (Nov. – Dec,2012), PP 33-38

[6]Ahmet Latif Amet, Aysın Ertüzün, Aytül Erçil“An efficient method for texture defect detection: sub band domain co-occurrence matrices”, Bogazici University ,Istanbul TURKEY

[7] A. Ertuzun,Serdaroglu, “Defect detection in textile fabric images using wavelet transforms and independent component analysis”Bogazici University, Turkey Vol. 16 No. 1 2006

[8] Aysın Ertüzün,, Aytül Erçil, O. Gökhan Sezer,“Independent Component Analysis For Texture Defect Detection” ,Bogazici University, Electrical and Electronics Engineering Department, Istanbul-Turkey

[9] R K Rao Ananthavaram, O.Srinivasa Rao,MHM Krishna Prasad, “Automatic Defect Detection of Patterned Fabric by using RB Method and Independent Component Analysis”International Journal of Computer Applications (0975 – 8887) Volume 39– No.18, February 2012

[10]Rashmi S Deshmukh , Dr P RDeshmukh “Comparison Analysis For Efficient Defect Detection Algorithm For Gray Level Digital Images Using Median Filters Gabor Filter And Ica” ,Volume 2, Issue 1, January 2012 ISSN: 2277 128X International Journal of Advanced Research in Computer Science and Software Engineering International Journal of Computer Applications (0975 – 8887) Volume 39– No.18, February 2012

[11] Pramod Kumar Pandey, Sweta Tripathi, Yaduvir Singh, “Image Processing using Principle Component Analysis” , International Journal of Computer Applications (0975 – 8887) Volume 15– No.4, February 2011

[12]Lindsay I Smith, “A tutorial on Principal Components Analysis “ ,February 26, 2002

[13]Du-Ming Tsai, Ping-Chieh Lin and Chi-Jie Lu , “An independent component analysis based filter design for defect detection in low-contrast surface images “ , Yuan-Ze University, Taiwan

Volume-1, Issue-1, 23 – 29