Automatic Weld Defect Detection in Real-time X-ray Images Based on D-S Evidence Theory

摘要

Automatic weld defect inspection in real-time X-ray image has the advantages of objective, normative and efficient. But real-time X-ray image usual has low signal-to-noise ratio, and most of the existing automatic inspection systems only use single defect segment method, which lead to the outstanding conflict between false alarm and missed detection. In this paper, the redundancy and complementarity are analyzed in two defect segmentation methods of weld X-ray image, i.e., background subtraction and gray-level wave analysis. Then, the information fusion method of them is proposed based on D-S evidence theory. Results show that the false alarm and missed detection are decreased effectively by the information fusion method.