基于数字化图像处理的超声TOFD法D扫描图像中缺陷自动识别

In ultrasonic time of flight diffraction (TOFD) D-scan image, only a small fraction represents defects, whereas the majority of the data is considered redundant. Because of the low contrast between defect and background image, TOFD suffers from the difficulty of large amounts of time and effort consuming in manually interpretation of D-scan image. In addition to this, due to the nature of the weak diffracted signals, the human factor introduces inconsistency into the interpretation. In order to distinguish weld defects from the D-scan data automatically and reliably, a defect detection method based on image processing technique is proposed in the paper. Firstly, image pre-processing including clutter and noise suppression is conducted. Secondly, information entropy based image segmentation technique is employed to extract defect in the pre-processed image. At last, mathematical morphological based post-processing is carried out. The experiment shows that with the proposed method, TOFD can be used as an automatic weld defect detection technique.