Title: Inspection of surface defects in copper strip using multivariate statistical approach and SVM
Authors: Xue-Wu Zhang; Fang Gong; Li-Zhong Xu
Addresses: College of Computer and Information, Hohai University, Nanjing 210098, China. ' College of Computer and Information, Hohai University, Nanjing 210098, China. ' College of Computer and Information, Hohai University, Nanjing 210098, China
Abstract: The surface quality would directly influence the capability and quality of the final product, but there is little domestic research focused on surface defects inspection for copper strip based on automated visual inspection. According to the gradual change of intensity levels of copper strips surface defect, a defect detection algorithm is proposed using wavelet-based multivariate statistical analysis. First, the image is divided into several sub-images, namely statistical units, and then each unit is further decomposed into multiple wavelet processing units. Then each wavelet processing unit is decomposed by 1D db4 wavelet function. Then, multivariate statistics of Hotelling T² are applied to distinguish the existence of defects and classify the defects using support vector machine (SVM). During SVM design, the authors used cross-validation method to get the best parameters and then used the parameters to train and test the samples. Finally, the defect detection performance of the proposed approach is compared with the traditional method based on greyscale. Experimental results demonstrate that the proposed method has better performance on identification, especially its application in the ripple defects can achieve a 96.7% probability of detecting the existence of micro defects, which was poor in common algorithms.
Keywords: copper strip; surface defects; machine vision; defect inspection; wavelet decomposition; Hotelling T2 multivariate statistics; support vector machines; SVM; surface quality; automated inspection; visual inspection; ripple defects; micro defects.
International Journal of Computer Applications in Technology, 2012 Vol.43 No.1, pp.44 - 50
Published online: 13 Mar 2012 *Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article