Recognition of Plain Fabric Defects by Image Processing and Hierarchical Classification System

• Main Authors: LI, CHIH-HSIN, 李智信
• Other Authors: Chang-Chiun Huang
• Format: Others
• Language: zh-TW
• Published: 2010
• Online Access: http://ndltd.ncl.edu.tw/handle/27913970772471644193

碩士 === 國立臺灣科技大學 === 高分子系 === 98 === In textile industry, the detection of the surface defects is important because it can improve quality of the textile products. However, plain fabric defects are still examined by human now. The common surface defects of plain fabric include oiled defect, west bar, misdrawing and bore. In this thesis, we will present an automatic plain fabric defect detection method and develop the surface defect inspection and classification system, which will assist hand-actuated examination. The surface defect inspection and classification system use statistical threshold value decision method to choose two optimal threshold values for separating defect areas. We use hierarchical classification in our system. In the first phase, we use the defect point number to determine defect samples or not. In the second phase, finding the gray level mean of oiled defect by the black region is the main work in order to judge the existence of defects. In the last phase, the entropy and gradient direction are used as defect features to identify the three kinds of fabric defects, which are west bar, misdrawing and bore. In the first phase, there are 25 studying samples and 78 testing samples. In the second phase, there are 20 studying samples and 53 testing samples. In the third phase, there are 15 studying samples and 38 testing samples. The experiment results show that the recognition rates are 100% in all the phases. For the results above, hierarchical classification system can be used to inspect the surface defects of plain fabrics effectively.