Research Of Steel Fracture Image Analysis Method Based On Iamge Segementation And Pattern Recognition

Drop Weight Tear Test(DWTT)is a test of the falling hammer of the prefabricated gap material,which is used to evaluate the toughness of the material by calculating the shear percentage of the fracture surface.At present,it mainly relies on artificial visual judgment to determine the percentage of the ductile area,the subjective factors influence the accuracy,the detection efficiency is low,and the automatic detection instrument is urgently needed.But drop weight tear specimen fracture mode of the image is very complicated,toughness,brittleness,high fluctuation can be up to 30 mm,lighting in imaging,especially image automatic identifying a great technical challenge.This article through to drop weight tear fracture characteristics and in-depth study of machine vision technology,puts forward the evaluation method based on image segmentation and pattern classification,set up a platform of image acquisition,test software is developed and conducted experiments on a complete set of testing system is verified.Firstly,this paper solves the lighting system design of the fracture image by combining the optical reflection and 3d features of the fracture surface.Then the image is pretreated by means of threshold segmentation,mean filtering and image fusion.Image segmentation is performed for different fracture type images.The characteristics of digital image segmentation after area is extracted,Gaussian mixture model(GMM)and support vector machine(SVM)classifier was trained,get the right image classification model,the realization of ultimate fracture zone of the images in fracture identification and classification.Algorithm of the evaluation result comparing with human experts to assess the results and the experimental results show that this design of automatic assessment algorithm with human experts to assess the results of absolute error within 4%,can realize automatic evaluation of drop weight tear fracture.