Abstract:

In order to overcome the poor adaptability of existing fabric defect detection algorithms on numerous kinds of defects, especially the ones appearing as minor texture changes, a fabric defect detection algorithm based on monogenic wavelet analysis was proposed. The monogenic wavelet analysis on fabric images works with the Riesz?Laplace wavelet, which is generated by performing Riesz transform to an isotropic Laplace wavelet constructed by combining a fractional Laplacian and a polyharmonic spline. For the multiresolusional orientation and amplitude subbands outputted by monogenic wavelet analysis, respective criteria for the best responses and segmentation method on the best response subbands were designed. Experiment results showed that the proposed detection algorithm could effectively segment various kinds of defects in different fabric textures, consequently demonstrating the position and shape of defects, and achieved a detection rate of 97.37% on 342 experimental sample images, bearing a sound self-adaptability and robustness.

Key words: fabric defect detection, monogenic wavelet analysis, Riesz rtansform, Lablace wavelet, polyharmonic spline