An Efficient Weld Image Classification System Using Wavelet And Support Vector Machine

Abstract

A weld defect is a flaw occurs during the weldment. These defects are unavoidable during welding process. In this paper, an efficient weld image classification system for the classification of weld images into defect or no defect is presented. It uses GD X-ray weld image database for the evaluation. Discrete Wavelet Transform (DWT) is applied to GD X-ray weld images to obtain the wavelet coefficients of low and high frequencies. Then, energy and entropy features are computed. Support Vector Machine (SVM) classifier with different kernels is used for classification of flaw images into defect or no defect. Result show that DWT and SVM classifier provides 95% accuracy for weld image classification.