Surface defect inspection based on wavelet statistical analysis
小波统计法的表面缺陷检测方法

According to the characteristics of defect image on copper strips surface, we design a surface defect detection system on the basis of wavelet-based multivariate statistical approach. First, the surface image is divided into sub-images; each sub-image is further segmented into multiple wavelet processing units. Then, each wavelet processing unit is decomposed by 1-D db4 wavelet function. The multivariate statistics of Hotelling T2 is then applied to detect the defects, and Support-Vector-Machines(SVM) is used as the defect classifier. The defect detection performances of the proposed approach are compared with those of the grayscale- difference method. Experimental results show that the proposed method has higher performances on identification; the recognition rate for the ripple defects achieves 96.7% which is unattainable by common algorithms.