Abstract:

Focusing on the problems of low detection efficiency, poor stability and slow processing speed of traditional artificial fabric detection, a patterned fabric defect detection method based on alternating direction method with Gaussian back substitution (ADMG) image decomposition was presented. Firstly, histogram equalization as preprocessing was first conducted for the sampled images to eliminate the influence of background texture of fabric defects. Secondly, ADMG image decomposition method based on combination of the total variation norm and semi-norm in negative Sobolev space was employed, the patterned fabric images could be decomposed into defect structure u and texture structure v. Finally the defect structure u was segmented by using a two-dimensional Otsu thresholding, the fabric defects could be identified. The experimental results demonstrate that method based on ADMG image decomposition is feasible and effective in patterned fabric defect detection contained star-, box- and dot- patterned fabric images and satisfactory identification results could be achieved.