The Reserch On Image Detection And Its Recognition Of Railway Freight Bearing Surface Defects

With the development of the railway, defect inspection of roller bearing becomes moreand more important to ensure the safety. But in our country, the roller bearing surface defect isstill defected using hands leading low efficiency. In developed countries such as US and Japan,advanced research has been done and many defect equipments has been developed, thus highefficiency and economics could be reached. With the development of the industrial controltechnology and information technology, railway transport meets great challenges.According to the surface defect image's traits and using digital image processing, thepaper abstracts the defects from the roller bearing surface defect image, makes classificationof them according to their feature. The paper is composed of the following parts.1. At the aspect of the image pretreatment, the thesis analyzes the characteristics and theway surface defect formed and difficulties of its recognition. Besides, it designs the basicframework of the processing and recognition system. The multiwavelet transform and K-Ltransform methods are used to improve an image, which help to enhance the contrast andremove the noise.2. At the aspect of edge detect, at first, Fuzzy c-means (FCM) clustering algorithm isused to separate edge from noise, then multiwavelet transform algorithm is used to makeprecision segmentation.3. The neural network is applied to the recognition of surface detect. Defect's area, itseccentricity, its solidity, its roundness, and its edge smoothness are extracted as 5 characters,also a BP neural network sorter is trained.At last the paper provides the results of image processing and pattern recognition todifferent types of defects, which prove that the system could inspect roller bearing surfacedefects exactly and is applicable.