Research on Image-Based Detection and Recognition Technologies for Cracks on Rail Surface

Abstract

At present, rail flaw inspection relies on ultrasonic technology, which requires manual analysis of mass data. This detection method faces defects of low efficiency, long cycle, and this method requires a professional with high level of experiences on data analysis. This paper proposes a visual detection and recognition technology for the surface crack of steel rail. Firstly, the article mainly discuss according to the composition and principle of the machine vision inspection system, combined with the camera imaging and the railway viewing environment, the acquisition device is selected based on the inspection requirements of the track inspection image. Secondly, by comparing several common algorithms of image filtering, it improves the weighted median filtering algorithm and filter the track detection image. The filtered image is enhanced by a histogram equalization algorithm. By using the coefficient of variation, the crack area and the shadow area are divided from the image, and the similarity is calculated to distinguish the crack as well as the shadow area. Then the original image is distinguished from the background image. At the end the combination of the custom and iterative methods is used to calculate the threshold and extract crack area. Finally, the Graphical User Interface (GUI) in MATLAB will be called to generate the required software framework, and the related member variables of each step are called to display, storage and with other operations. The detection and recognition system includes: the image display interface, the module control Interface and the parameter storage. The experimental results show that the analysis and identification technology of rail cracks in the track inspection image can quickly locate the rail area and accurately extract the crack image, which can meet the accuracy and speed requirements of the track inspection image detection, Also it has high theoretical value and practical application prospects.