32 Similarity Based Interactive Defect Detection for Periodic Patterned Fabrics

The purpose of this paper is to detect the defects and the flaws in patterned textile fabrics at textile industries’ production line. In our proposed method, the approach uses the distribution of pixel intensities to detect the defects of fabrics. The main idea behind this study is to recognize defected pieces of the fabric from the impeccable ones based on the similarity of each piece with a standard piece as measure of correctness. Two different similarity measures are introduced in this study. We call this method “Interactive Defect Detection” since the user chooses the standard piece used for comparison. The idea is described in detail and the results of the performance of this method in an experiment are presented. The results show that the method has potential for future research and progress.