A real-time weld defect inspection system in X-ray images

Detection of weld defects is still done by the human interpreter. The problems of this process are subjective, inconsistent, labor intensive and fatigue of the interpreter. It is desirable to develop a computer-aided system to assist the human interpreter in evaluating the quality of welded part. In this paper, a real-time weld inspection system is presented. The system consists of three steps. First, extracting the weld part within the whole image by a fast algorithm based on the gray level, then the defect detection is applied on this part. Second, reducing the affection of the noise and making the defect more evident. A method to average several consecutive frames is used in the preprocessing part. Last, detecting the flaw. A more effective fuzzy theory that is based on the local image characteristics such as spatial contrast and spatial variance is used to determine the existence of a weld defect. The result shows success of the system in detecting weld defects.