Rail Profile Matching Research Based On Linear Structure Laser

With the increase of railway transportation speed and the opening of a large number of lines,more stringent requirements are put forward for rail damage detection,which includes many aspects,such as surface defects,rail roof wear,etc.At this stage,there are more mature methods for the detection of surface defects,and the detection of wear still needs to be further explored to meet the increasing detection needs,so this paper studies the rail contour image matching method based on line structure laser and builds a set of wear detection devices that can be used for actual measurement,the main research content and results are as follows:1.A line laser scanning system is built to obtain the contours of the left and right sides of the rail and calculate the wear.The system includes an image acquisition device and a computer control terminal.There are three line lasers in the acquisition device to illuminate the entire rail section,after the camera calibration,the calibration plate is replaced with the rail,and then the left and right parts of the rail are photographed by the industrial camera equipped with a filter,and the image is transmitted to the computer through the data line and processed by the algorithm to display the matching results,while calculating the wear of the rail top and judging the degree of wear.2.By analyzing the limitations of traditional calibration methods during image calibration,a new calibration plate is designed,and a semi-automatic calibration method is used to complete the calibration of the camera.By combining Zhang’s calibration method and Harris corner point detection,the automatic calculation of camera parameters can be realized,but when using the traditional checkerboard calibration plate,the calibration plane is difficult to coincide with the laser plane,so a stepped calibration plate is designed,and then the calibration parameters are calculated by polynomial transformation,and the image is reconstructed by bilinear interpolation based on backward mapping.After experiments,when 13 groups of feature points were collected,the best calibration accuracy was obtained,so as to complete the rail contour image correction.3.The rail profile matching method was studied.The shortcomings of the traditional point cloud iterative algorithm and the geometric feature matching method are confirmed by experiments.And then a contour matching method based on the improved SURF algorithm is proposed.Compared with the traditional contour matching method based on image features,the improved algorithm matching running time is fast,the average matching time is 933 ms when matching the left image,the correct matching logarithm accounts for 90.9%,and the average matching time is 962 ms when matching the right image,and the correct matching logarithm accounts for 91.3%.In addition,due to the randomness of the random sample consensus(RANSAC)algorithm within the number of iterations,the stability of the correct fusion effect of the algorithm is tested by setting different iteration times and running the program 100 times respectively,and the results show that the stability is more than 90% when matching with the algorithm.