FABRIC DEFECT DETECTION METHODS BASED ON GRAY-VALUE STATISTICS
In this paper, three methods are applied to fabric defect detection based on the gray value statistics of defect images and their corresponding defect-free images. In the first method, the fabric sample image is divided into square blocks, by which the background texture of fabric is attenuated and the defect part is accentuated, then the defect is detected by thresholding. In the second method, the threshold value is determined by calculating the maximum of the revised variance expression which is obtained by introducing a weight coefficient to the between-class variance expression of the OTSU method, the defect part is segmented by binarization. In the third method, the gray value features of defect areas and histogram of defect image are used to obtain the threshold value for defect segmentation. The three methods make the calculations simple and fast, and the experimental results indicate that they are effective. Especially the last two methods reduce computational cost significantly, which make them be suitable for on-line real-time detection.
defect detection block between-class variance histogram statistics
Lingmin Zhang Runping Han Surong Sun
Department of Industrial Design and Information Engineering, Beijing Institute of Fashion Technology , Beijing 100029, China
国际会议
北京
英文
1-5
2008-09-26（万方平台首次上网日期，不代表论文的发表时间）