Hull Weld Defects In X-ray Images Of Automatic Identification Method

With the development of shipbuilding industry and the improvement of welding technology, X-ray detection of welding seams has encountered more opportunity and challenge. The quality judgement of X-ray film is an important part in X-ray detection. In a long time the judgement work was carried out by judging workers who analyse the defect of welding seams on X-ray film and decide the quality grade of the film. The traditional method has many disadvantages, for example, no good effect resulted from man's operation, low efficiency, damaging film, and so on. It cannot meet the requirement of productivity enhancement. So many new judgement technologies on X-ray film occur. Intelligent judging technology by computer is a cut-edged research field now. The digitization of X-ray film and the digital image processing have become an important part in the automatic film judging. Much research about shipbuilding X-ray digitized film has been done in this paper, which includes preprocess of image, analysis of image, extracting the characteristic information of defects and the identification of defect.Preprocess of image is the base of image processing. After analyzing the noise model of image, effective treatment method is put forward. The method of gray equilibrization aims at X-ray film's character of low gray contrast and faint edge, which intensifis object's edge in image, to a certain extent, it widens gray distributing. Edge detector is an effective method to line out defect. This paper analyzes the theory of some edge detecting algorithms, and presents a new edge detecting method that is based on morphology. Selection and the extraction of characteristic parameter of defects are precondition for defect identification. It directly affects the result. By studing the character of defects, an eigenvector is presented that factually reflected defect's essential properties and their respective computation. For the identification of defects, the proposed method of based on statistic data is verified very simple and effective.