Abstract: It is important to detect defects of welding lines on an X-ray film for improving welding techniques and the product quality. This paper proposes a detection method based on support vector machine based on analysis and comparison with several other methods. In the proposed method, the film is first converted into digital format, and characteristics of the defect are extracted. A “one against one” SVM clustering structure is constructed according to the characteristics, and recognition is made. Experiments show high accuracy and processing speed, and easy implementation of the method. The method is suitable for detection and recognition of defects of welding lines from X-ray films.

Key words: X-ray film, defect of welding line, support vector machine, recognition