Research On Extraction Technology Of Weld Defects In X-ray Image

In industrial nondestructive testing, X-ray inspection of welding defects is a very important tool, which can directly detect the potential weld porosity, slag, cracks and other defects, so as to facilitate the management and control of welding quality. But the traditional X-ray inspection of welds mainly depends on manual assessment for films, which made test results vulnerable to be affected by the technical quality of the personnel, and subjective experience. Also, the manual assessment of X-ray films is characterized by a low detection efficiency, high operational complexity, difficulty to preserving films, uneasy automation, which is difficult to meet the requirements of industrial modernization. Therefore, automatic recognition of weld defects will be the mainstream direction of X-ray inspection technology in the future development.The paper focused on the process of X-ray auto-detection of weld defects, and discussed about automatic extraction of weld defects as the key technology of automatic recognition. The main contents of this paper are as follows:(1)Digitization of X-ray film : Discussed the requirements of X-ray film digitization, described the principle and system components, and analyzed the factors which affect digital images. (2) Research of image preprocessing techniques: By comparison of image denoising and enhancement algorithms commonly used in weld image enhancement application, select Gaussian-filter and fuzzy enhancement method to improve the weld images, which can get better results. Then, a method for determining the boundary points based on the analysis of gray curve was proposed, by which the ideal weld can be segmented from the original image. On this basis, in order to avoid loss of weld defects near the border, the method combined with gray projection, can accurately locate the weld area, and achieve fully automatic extraction of the weld .(2) Extraction of weld defects based on edge detection methods: This paper compares several classical edge detection algorithms used in the detection of weld defects, and analyze theirs performance. On this basis, an edge detection algorithm based on multi-scale morphological gradient operator was proposed. By experiments, it is proved that the method has better noise immunity, and meanwhile performs well especially for the extraction of line defects, which laid a good foundation for the future research of automatic recognition of weld defects.