Research Of Welds Crack Defect Recognition Technology Of Magnetic Particle Testing Based On The Morphology

Evaluation of testing result made by Magnetic particle detection methods rely on the magnetic mark formed around defects. At present, Magnetic particle testing relies on the detection of macroscopic observation mainly. But in the actual application, undetected, false dismissal and some other situations often appear, because of visual fatigue and other reasons. This paper puts forward a method that can identify magnetic particle detection crack defects automaticly. Under laboratory conditions, we established a weld crack defects image automatic data collection system, and carried out the weld crack defects image research with the application of modern image processing technology, and realize recognition of weld crack defects.In the actual detection, a jitter between the acquisition system and defects caused blurred images because of poor environmental conditions and Vibration and other reasons,which contributes to a fuzzy image and makes certain errors exist in the image analysis results compared with the measured. A weld crack image restoration method is presented in this paper, and through the establishment of defect image degradation model, and then restoring the fuzzy image of defects according to the inverse process of that model, we improve the degree of approximation of the image. But after the restoration of crack defect image signal is weak and the noise is excessive, which influence the crack identification heavily. Therefore,it is important to the defect image preprocessing to be gray level transformation, histogram equalization, the mean filter smoothing and image enhancement, to make it more suitable for computer analysis and identification of the crack. Then using the analysis method of similar morphological gradient image to processing the defects, and remove the background that has nothing to do with the defects in the original image, maximum limit retains the morphological characteristics of crack. Aiming at the problem of weld crack defects extraction, this paper adopts the Ostu adaptive threshold algorithm, it is a kind of special used in segmentation image grey value of defects and background when large difference image algorithm, this algorithm can retain more crack the original details. Because defects image still have large numbers of false defects after segmentation, this paper puts forward a set of concise and effective characteristic parameters cracks is based on the morphological characteristics of crack Through this set of parameters to compare connected domain filtering of defect image and ultimately get meaningful real crack defect figure. In the process of identification of the defect, the image often morphology transformation and transformation of space, defect edge may occur distortion. This paper use morphology restoration and region growing method to repair the possible distortion problem, prevent discontinuity of crack and deformation..Finally we adopted discriminant method of 8 connected to each crack distribution corresponding label in the images, and used for marking, record location and length of meaningful information of each crack. Realize visualization of weld defects.