Analysis And Recognition Of Rail Surface Defects In Rail Detection Image

Rail surface defects not only affect the stability and comfort of driving,but also endanger the safety of driving.With the increase of existing line speed,and the development of heavy haul railway and high-speed railway,more and more new defects of rail surface generate.Therefore,it is very important to detect the defect of rail surface and take corresponding measures for maintaining a safe,comfortable and continuous railway operation.The method of rail surface defect detection based on machine vision has become the main method because of its advantages such as high speed,automatic and non-contact.Therefore,by analysing and processing the rail detection images,the defects in the rail surface is detected and identified.First of all,according to the structure and principle of machine vision system,combining with the camera imaging principle and the railway environment,and taking the detection requirement of rail detection images as standard,the lighting source,lighting way,camera,lens,computer and other devices in rail detection system are selected.Next,the bilateral filter is used to filter rail detection images by comparing several common image filtering algorithms.As for the filtered rail detection images,the rail surface area is extracted by using the gray difference between the rail surface area and non-rail surface area at first.And then,the Hough transformation is used to extract the boundary line of the rail surface area,which is based on the regular shape and the obvious edge of rail features.At last,according to the results of the extraction,combining the least square method with affine transformation,the level and the integrity of the rail surface area is extracted.Moreover,combining the principle of the spatial structure of the image can be measured by using the excess entropy,with the fuzzy theory can be used to describe and process the fuzzy set,the total fuzzy excess entropy function of image is obtained by using the fuzzy excess algorithm after defect segmentation.Then,the optimal parameter combination of the maximum entropy is calculated by using genetic algorithm.And according to the membership function in fuzzy theory and the optimal parameter combination,the optimal threshold of rail surface defect segmentation is calculated.Then,establishing the positive and negative defect sample space,according to the Haar-Like features of the sample space set,the AdaBoost algorithm is used to design the rail surface defect recognizer,to exclude the interference images of defect images,and according to the low level features of the positive sample space set,the C4.5 algorithm is used to classify the defect images.Finally,the overall performance of the system is verified on the experimental platform of the rail detection system.In addition,the three main functions of the system are as follows: rail surface area extraction,rail surface defect segmentation and rail surface defect recognition,which are verified by experiment and analyzed statistically.The experimental results show that the rail surface defect analysis and recognition technology can be applied to the rail detection in complex environment.