Abstract: In order to realize the automatic woven fabric defect inspection, the paper firstly constructs the fabric adaptive orthogonal wavelets libraries, and then takes the wavelets libraries as the group population of Genetic Programming Algorithm. After selecting the better one from four different kinds of Fitness Functions, the wavelet devices matching fabric texture are successfully founded from the group population. In comparison with the four fitness functions, taking the fabric texture fluctuations as the fitness functions can get the best wavelet device. The experiments verified the effectiveness of detection on some related defects, and the position of defects is located by window segmentation method. All of these show that the genetic programming algorithm combined with the fitness function optimization can find the optimal wavelet devices matching fabric texture, and realize the automatic fabric defect inspection.

Key words: wavelet basis, defect detection, genetic programming, fitness function