Fabric defect detection algorithm using RDPSO-based optimal Gabor filter

Abstract

A new fabric defect detection algorithm is presented in this paper. Instead of a Gabor filter bank, an optimal Elliptical Gabor filter (EGF) is used in the proposed algorithm. For a given non-defective template image, the parameters of the optimal EGF is determined by using the random drift particle swarm optimization (RDPSO) algorithm during the training procedure. During the detecting procedure, a sample image containing the same texture background with the template image is convolved with the optimal EGF. Then, defects discrimination can be carried out to determine whether the sample image contains defects or not. Finally, if the sample image is distinguished to have defects, an adaptive thresholding technique can be applied to locate the defects. Extensive experiments have been carried out to demonstrate the effective detecting performance of the proposed algorithm.