Research On Defects Detection Of Break Pad Based On Machine Vision

Brake pad is the friction material which contacts the brake disk when the vehicle brakes,the braking performance of the vehicle is closely related to the quality of the brake pad,which directly affects the safety of vehicles and personnel.Machine vision technology is an interdisciplinary subject involving many fields,the surface defect detection of product is a kind of application of machine vision detection technology,it greatly improves the efficiency of production operations and the accurately of detection.Based on the basic principle of machine vision,combined with image processing and pattern recognition technology,the detection methods of the brake pad defects were researched by this paper.The main work is as follows:(1)For brake pads three kinds of defects,combined with the characteristics of different image acquisition system,the image acquisition device and light sources were selected and lighting systems were designed.(2)For the detection of the contour defects of the brake pad,the variable comparison model was established.The variable comparison model was trained by several qualified brake pad images aligned by temple matching technique,the brake pad image which to be detected was compared with average image acquired by the model.(3)For the detection of the character defects of the brake pad,an improved relational matching was proposed.Every character was regarded as a component,the contours and coordinates of all characters,the distance and angle relationship between the adjacent characters were stored in the model.At the time of detection,the position of the root component was searched in the image,and the neighboring element was searched according to the distance and angle relation of two neighboring components in the model,According to the search results to determine whether the defect of characterwas exist or not.(4)For the detection of the surface defects of the brake pad,DAGSVM structure was constructed.At the first,22 features of non-defect and three types of defects were extracted.Then reduce the dimensions of data by the principal component analysis of 22 kinds of defect features.Based on the results of principal component analysis,cluster analysis of genealogy was carried out and the decision tree was constructed according to the distance between various defects.Finally,the training samples were used to optimize the SVM,then the decision tree was constructed to detect the brake pad image.This paper researched the detection methods about contour defects,character defects and surface defects of brake pads based on the machine vision technology.Experiments had shown that the methods were stable and effective,Which provided a new idea for the application of machine vision technology in the defects detection of brake pads.