Rail Longitudinal Displacement Monitoring Method Based On Image Recognition

With the rapid development of high-speed railway in China,continuous welded rail has become the inevitable choice of high-speed railway track structure.Due to the elimination of steel joints,the huge temperature stress generated inside the long rail strip may cause rail expansion runway and rail breaking accidents,resulting in potential safety hazards for railway operation.The longitudinal displacement of rail is a typical manifestation of the state evolution of continuous welded rail.Monitoring the longitudinal displacement of rail has become the key work of continuous welded rail maintenance.At present,the commonly used continuous welded rail condition monitoring methods are based on fixed equipment,which is difficult to realize rapid monitoring,and most of them have the disadvantages of high cost and low efficiency.By investigating the cause mechanism and actual operation environment of continuous welded rail diseases,combined with the implementation basis of machine vision technology in track scene image acquisition,a new method of rail longitudinal displacement monitoring based on image recognition is proposed in this paper.Set up on track markers at the rail bottom and roadside markers on the fixed foundation piles on both sides of the line.Take the manually implemented acquisition platform as the carrier and complete the information acquisition of the whole line through the marker image acquisition equipment mounted on the platform.Based on the two-dimensional gray value matrix and image filtering technology,the gray features of each measuring mark are extracted.By restoring the pixel areas before and after the roadside base mark,the relative position of the on track measuring mark in the restored scene and the reference scene can be tracked,so as to obtain the longitudinal displacement of the rail according to the calibrated object-image geometric relationship.In this paper,the imaging error of the calibration acquisition equipment is analyzed theoretically and experimentally,and the correction method of the calibration error caused by lens distortion is proposed.At the same time,the influence of acquisition equipment swing on the imaging position of roadside landmark is studied,and the pixel correction formula of error image under different ranging is put forward.The test shows that the horizontal camera group should be started to shoot the roadside base mark as far as possible to reduce the image error;When the acquisition equipment swings and deflects horizontally,under the same deflection angle,the farther the horizontal shooting distance,the smaller the displacement error of the pixel area of the roadside landmark image,and the swing direction of the equipment has no effect on the displacement measurement error.The application test on the test section shows that the proposed monitoring method has strong overall applicability,the collected image is clear,the marker image algorithm is reliable,and the measurement accuracy can be controlled within ± 1mm.At the same time,in order to realize the effective application and management of monitoring data,this paper establishes the rail longitudinal displacement monitoring information management software based on B/S structure,which has a certain reference value for railway track information management.