Research About Weld Defect Detection Algorithm

With the advancement of image processing arid computer technology, which rapidly promotes the digitization of X-ray weld images. But now manual assessment is the main way of X-ray detection, which is vulnerable to personal subjective factors and easily lead to miscarriage of justice and missed sentence in the process of detection. Therefore, it is very necessary to realize the automatic detection of X-ray weld defection.In this paper, X-ray weld images are regarded as the research object and a study on weld defect detection algorithm is carried out. Aiming at the problem of complex texture, contrast difference and background ups and downs in weld image, the image preprocessing algorithm is firstly studied. Then the corresponding defect detection method is studied for the different characteristics of the defect. Finally, the complementarity of different algorithms is used to avoid false detection and miss detection.Firstly, aiming at the problem of low signal-to- noise ratio and poor contrast in weld image,image enhancement and noise reduction are used to improve the image quality and reduce the noise in the image. Aiming at the problem of dynamic video detection, the scheme of real-time display of acquisition and detection is solved by using gray scale normalization and continuous detection of N frames every M frames. At the same time, in order to reduce the interference and improve the detection efficiency, a kind of automatic extraction method of weld boundary for DR images is proposed, which has good adaptability and practicability.According to the characteristics of different weld defects, this paper designs the corresponding defect detection algorithm based on Canny, Lapalace, frame difference method and Butter Worth filter, but a method only works for the specific defect type. In order to prevent the defects from being missed, the complementarity of different detection method is used to fuse the test results. The corresponding detection scheme is designed for dynamic video and still pictures, which has good adaptability.Based on the analysis of the characteristics about the defect area and non-defect area in the weld image, the model vector is constructed by using the spatial characteristics, and the samples are trained by SVM, and then the defect detection and corresponding result analysis are carried out.Finally, this paper designs and implements a DR imaging and defect detection system,which achieves the image acquisition, picture capture, defect detection and other functions through the computer multi-thread technology.