Research On Defect Detection Technology For Welding Seam Based On Machine Vision Technology

X-ray defect detection is one of the primary methods in non-destructive testing of industrial products and is also widely applied. At present, primary methods of detection defects in the weld is still film imaging and image intensifier real time imageing and done by human interpreter both foreign and domestic, The problems of this process are subjective, inconsistent, labor intensive and fatigue of interpreter. So computer-aided detection of weld defects has been concerned. However, as a result of the imageing condition, the object as well as the algorithm complexity and the limitations, in the practical application, there are still many issues to be studied further.The machine vision technology, which can extraordinarily reduce the inspection cost, effectively ensure the quality of products, strongly raise the speed and efficiency of production, has become one hot topic in the industrial online inspection in recent years and is widely used in the filed of industrial inspection. In this paper, defect detection technology for welding seam image based on machine vision technology is researched. First of all, based on analyzing and studying the x-ray select system, a x-ray image select system is designed. Then, In photography and transmission process, Welding seam image usually is corrupted by guass noise and salt and pepper noise simultaneously, so the effect is not ideal if we only use mean filtering and med filtering. In order to denoise both of noise simultaneously, In the paper, combined with morphology filtering and mean filtering, an multiscale morphology filtering which can denoise the image have gause noise and salt and papper noise method was presented; Then, edge detection algorithm is studied. According to the characteristics of weld radiographic image, combining to edge detection theory based on mathematic morphology, improving this algorithm which are eristed, this thesis presents a method of Multi-scale auto-adapted entropy edge detection that adapts to the X-ray welding seam; Finally, combining fuzzy set theory and random Hough transform, using a improved random Hough transform, circular defect which may exist is selected form binary welding seam image which is selected, The experimental results also show that the effectiveness of the algorithm.