Automated visual inspection of ripple defects using wavelet characteristic based multivariate statistical approach

This paper presents a wavelet characteristic based approach for the automated visual inspection of ripple defects in the surface barrier layer (SBL) chips of ceramic capacitors. Difficulties exist in automatically inspecting ripple defects because of their semi-opaque and unstructured appearances, the gradual changes of their intensity levels, and the low intensity contrast between their surfaces and the rough exterior of a SBL chip. To overcome these difficulties, we first utilize wavelet transform to decompose an image and use wavelet characteristics as texture features to describe surface texture properties. Then, we apply multivariate statistics of Hotelling T², Mahalanobis distance D², and Chi-square X², respectively, to integrate the multiple texture features and judge the existence of defects. Finally, we compare the defect detection performance of the three wavelet-based multivariate statistical models. Experimental results show that the proposed approach (Hotelling T²) achieves a 93.75% probability of accurately detecting the existence of ripple defects and an approximate 90% probability of correctly segmenting their regions.