Design of Optimal Gabor Filters for Textile Defect Detection

A design approach based on the characteristics of textile texture is developed for the optimal Gabor filters.Two types of optimization design models of Gabor filters which can match normal texture and defect texture respectively are formulated,and the niche genetic algorithm is employed to solve the models.By comparing and analyzing the detection results,it is found that the normal texture matched Gabor filters are more suitable than the defect texture matched Gabor filters,and the center frequency of the normal texture matched Gabor filters is consistent with the highest energy harmonic element in the power spectrum of the texture image,which greatly reduces time spent on solving the optimization model.