Classification of specular object based on statistical learning theory

Abstract

This paper has presented an efficient solder joint inspection technique through the use of wavelet transform and Support Vector Machines. The proposed scheme consists of two stages: A feature extraction stage for extracting features with wavelet transform, and a classification stage for classifying solder joints with a support vector machines. Experimental results show that the proposed method produces a high classification rate in the nonlinearly separable problem of classifying solder joints. © Springer-Verlag Berlin Heidelberg 2001.