This paper deals with a procedure that recognizes common defects occurring in woven fabric. Images of woven fabric are considered as having a directional texture due to their periodical nature. We used a statistical approach based on the analysis of periodicity of texture images in horizontal and vertical directions. These periodicities correspond to the periodicity of second-order grey level statistical features obtained from a grey level co-occurrence matrix. A set of five significant features is extracted from the matrix: energy, correlation, homogeneity, cluster shade and cluster prominence. The presence of a defect over texture causes regular structure changes and consequently, statistical changes. Detection algorithm is based on the sliding window technique; the window is moved over the whole image area. We counted the test statistic for every window and the multivariate control charts are used as a tool for judging the existence of defects. The results show that the statistical approach is suitable for detection of directional defects or changes in regular structure in analysed simulated and real fabrics.
CitationDownload as .RIS
Emerald Group Publishing Limited
Copyright © 2008 Emerald Group Publishing Limited