Striation Segmentation Of Metal Fatigue Fracture Based On MRF

Research of metal fracture surface is an important part of fracture discipline. Apair of mutual matching fracture surfaces and their appearance morphologies whichobtained from metal torn are called fracture surface. Fracture always happened inmetal tissue weakest place that recorded the irreversible distortion of materials underthe environment action and crack from initiation and propagation to the whole processof fracture.The every phase of breaking process will leave the corresponding trace,morphology and characteristics on the fracture. And fatigue strip is an importantfracture characteristics of fatigue fracture image, but also the important basis forfurther research on the fatigue fracture,so the texture segmentation images are veryimportant.Now fracture surface analysis has become one of the important methods offailure analysis for metal components.Therefor the analysis of fatigue fracture isimportant to determine the failure models,causes,mechanisms.Image segmentation is one of the most basic and important research problems ofimage processing and computer vision areas.All the time it has been hot spot whichpeople research. The traditional texture segmentations have many methods,such asStatistical method, structural method, spectrum method and so on. Statistical methordis based on the basis of gray histogram, but the calculation process is verycomplicated; Structural method only applies to artificially texture, and it’s scope isvery narrow; And the spectrum method is based on image processing frequencydomain features, but the defect is the feature extraction variety.Because of thedisadvantages of the traditional methods, This paper adopts Markov Random Field(MRF) model method. Markov random field model is a probability model describingthe structure of image. It makes full use of the image of space-related information andcan be achieved with low signal to noise ratio of metal fatigue fracture striationsegmentation.This research content as follows:(1)Introduced emphatically basic theory of the MRF and several common models.(2)Parameter estimation of image segmentation algorithm were studied, Brieflyanalyzes some common parameters estimation methods. Comparing their advantagesand disadvantages,the segmentation algorithm is proposed in this paper. (3)Gaussian distribution are introduced combining with the MRF,and designed a kindof specific segmentation algorithm for texture image segmentation.(4)Finally, compare the effects of MRF method segmentation results to wavelettransform and EMD method. MRF method is significantly better than the latter two.According to the fatigue fracture images, a Markov Random Field(MRF)-basedstriation segmentation methord for metal fatigue fracture image is proposed. MarkovRandom Field model of image is built and image segmentation algorithm is proposedin this article. The experiment results indicate that the algorithm has fasterconvergence and better stability.