Automatic Defect Detection of Patterned Fabric via Combining the Optimal Gabor Filter and Golden Image Subtraction

Abstract

A new algorithm based on optimal Gabor filter and the basic Golden Image Subtraction (GIS) is\r presented for patterned fabric defect detection. Firstly, the defect-free patterned fabric images are\r processed to search optimal real Gabor filter parameters using traditional Genetic Algorithm (GA). Then\r test patterned fabric images are filtered according to the obtained optimal real Gabor filter. Furthermore,\r the basic GIS are adopted to perform subtractions between golden images from referenced fabric images\r and test images to get resultant images. Finally, thresholding is obtained by training a large amount\r of defect-free patterned fabric samples to segment defects from fabric background. Experiment results\r indicate that the average detection success rate is up to 96.31% with ninety defective patterned images\r and ninety defect-free patterned images. It demonstrates that the proposed method is more efficient.