Method of Real-time Recognition of Effective Rail Profiles from Complex Track Structures

Rail profile detection equipment is widely used in measuring the wear of rail, which provides important reference for railway maintenance. Whether the effective rail profile can be recognized quickly and accurately from the complex rail conditions such as ordinary tracks, switches and joints, restricts the practical application performance of the rail profile detection equipment. In this paper, a recognition method for rail profiles was proposed to identify and classify the effective profile and abnormal profile from the sampling data of statistical analysis of 2D laser sensors. First, the rail profiles in different sections of the line were compared to select two important characters, the discontinuity and the morphological correlation, between rail head and waist as the basis for the identification of effective contour. Then, the abnormal profile data in the original sampling data of rail profiles that are very difficult to classify for evaluating the wear, can be identified and classified by analyzing their distribution of 2D point cloud for varying rail types. The field experiment results show that the effective profile distribution recognized by this method is basically consistent with the actual railway conditions. The proposed method can effectively identify the rail profile with identification accuracy of more than 97% and average recognition time per profile of 9.78 ms under the condition of running velocity of 8.3 km/h. © 2023 Science Press. All rights reserved.