Automatic Segmentation Of Rail Defects Based On Improved Markov Random Field

Railway defects damage detection-of rail infrastructure security maintenance is of great significance as the high speed railway train operation density increase running speed increase and heavy freight line load increase,increases the degree of impact load is extruded and rail,the rail failure and increase incidence of injuries,damage types and failure modes of rail is changing,which brings the new challenges to traditional rail detection and testing technology machine vision technologies are fast non-contact precision is high It has been widely used in rail surface defect detection due to its strong anti-interference.In this paper,the visual technology of rail defects is studied,and the rail defect segmentation method based on improved markov random field is proposed.According to the target of rail defect segmentation,the overall scheme of rail defect segmentation system is designed,and the hardware of image acquisition module,such as imaging system,lighting system and motion platform,is selected to complete high-quality image acquisition.According to the requirement of rail surface defect detection system based on machine vision,the basic flow of rail defect detection algorithm is given.Secondly,the improved BM3D algorithm is used to remove the original noise in the process of rail image acquisition.Based on the analysis of the gray level characteristics of each region of rail image,an algorithm of rail area extraction based on vertical projection is proposed,which makes use of the difference between the gray level of rail area and non-rail area,and achieves the extraction of rail area.Then,by studying the gray distribution characteristics of the image and combining the background difference method,the gray level of the image is preprocessed to keep the gray level of the image basically consistent and enhance the defect features.Combining with the real rail image,the filtering algorithm,the rail region extraction algorithm and the gray processing algorithm are experimented,and the image pre-processing before image segmentation is completed.Based on the analysis of the theory of fuzzy set and Markov random field,an improved Markov rail defect segmentation method is proposed.By using Markov random field to utilize the spatial constraints in the image in the premise part of the fuzzy IF-THEN rule,while the result part specifies the distance map of the pixels,and using Markov random field to segment the edges accurately can improve the accuracy of the segmentation.Robustness to noise suppression and reduction of misclassification.Finally,the segmenting contrast experiments of FCM,GMM and improved Markov algorithm are established,and the segmenting performance indexes of three methods for different rail defect types are compared,and the robustness experiments of three methods to noise after adding different·levels of noise are carried out.The experimental results show that the improved Markov random field segmenting method for rail defect is more accurate and effective than the other two methods.And the robustness to noise is the best.