Research On The Detection System Of Rail Surface Defects Based On Computer Vision

With the rapid development of the modern railway, it becomes increasingly important toaccurately and efficiently detect the defect of in-service rails. Existing rail detection methods mainlyconcern visual inspection, ultrasonic testing method and the eddy current method. These methods areall manual inspection requiring intensive labor, so there is an urgent need for an automated testingsolution. This paper presents a computer-vision method consists of image processing and patternrecognition technology to achieve an automatic detection of defects on track surface, with advantagesof high efficiency and low cost.The paper designs a testing system according to the traditional theoretical model of computervision, including an illumination device, an image acquisition system and a control system. Theillumination device uses linear light source and chooses low-angle illumination in order to highlightthe defect information. The image acquisition system utilizes a high-speed linear CCD to meet theneeds of fast acquisition. The control system adjusts the image acquisition speed to the train speed.According to the specification of the system and the characteristics of rail surface defects, the paperfocuses on the design of image processing algorithms, which is divided into three modules: imagepre-processing, defect extraction and defect characterization. Each module will improve and optimizethe processing algorithm, greatly improving the efficiency and precision of the system.At last, an experimental platform is designed in laboratory instead to verify the systemperformance. Results show that the system can achieve automatic detection of surface defects at ahigh speed and can correctly classify defections, so it has certain applicability.