Detection of line weld defects based on multiple thresholds and support vector machine

Because X-ray images of weld contain uncertain noise and also the defects inside them have low contrast to their background, it is difficult to detect weld defects of X-ray images. The goal of this paper is to locate and segment the line defects in X-ray images. Firstly, we present an approach to extract features of X-ray images with multiple thresholds. Then, use the support vector machine (SVM) technique to classify the defect and non-defect features to obtain a coarse defect region. Furthermore, perform the Hough transform to remove the noisy pixels in the coarse defect region. Then the defect is located and segmented. The experimental results show that the proposed approach is effective and feasible to segment and locate defects in noisy and low contrasted X-ray images of weld.