Defect Detection on Patterned Fabrics Using Distance Matching Function and Regular Band

In this paper, regular band is presented to detect defects on patterned fabrics. Patterned fabrics are firstly disposed by fabric average to form object images mixed with positive and negative pixels in this proposed method. Distance matching function is computed to determine the periodic distance of patterned fabrics. The obtained periodic distance would be the length and width of regular band on row and column. Two features are calculated with regular band. The threshold of defect segmentation is extracted from the training step. Two features of regular band negotiating the threshold are considered as a fabric defect. Regular band method in this paper can avoid patterned interference and obtain perfect results in defect detection of patterned fabrics. Various defects on patterned fabrics can be inspected, and speedy defect detection could prove the performance ability in practice.