Research on Defect Recognition of Textile Products

摘要

This paper presents an approach of visual inspection for textile surface defects. The approach is the combination of feature extraction method and pattern recognition algorithm. Five important texture feature extraction methods and two pattern recognition algorithms are discussed in order to accomplish the real time recognition to the defects. The experiments are carried out in principle and results of them are given. Experimental results show that the combination of the wavelets analysis method and the K-means nearest neighbor classifier is better than others and it is effective.