Research On Rail Light Band Geometric Features Detection Based On Computer Vision System

With the rapid development of China’s high-speed railway and the continuous improvement of the speed of high-speed trains,railway departments have more stringent requirements on wheel-rail contact state.In order to judge the wheel-rail contact state more intuitively,before the advanced measurement and control technology and infrastructure have not been fully popularized,railway technicians have adopted the traditional naked eye detection method to observe the rail light band in the work of track maintenance and repair,through the phenomenon to see the essence,to solve the problem of wheel-rail contact on the railway.Although this method is reliable,it only qualitatively judges the wheel-rail contact state by the outline characteristics of the light belt along the train’s traveling direction,which is much worse in accuracy and accuracy.Therefore,there is an urgent need for a rail strip detection method with high accuracy and small error to replace the previous naked eye recognition method in order to guide the line maintenance and ensure good wheel-rail contact.The main research work in this paper is as follows:1.Based on Roberts-Marr’s computer binocular vision recognition method and combined with the actual working conditions of a metro line,a detection car for measuring rail ribbon by binocular camera is designed.The light band detection car adopts the combination of linear array camera and linear light source,with a self-made light band detection program,which can clearly photograph the geometric shape of the light band on the top of the rail.2.Analyzing the relationship between the change of eigenvalues and the image in the imaging process of the optical band image,expounding the difference of the field of view between the front projection model and the back projection model,establishing the geometric linear model of binocular camera imaging,and studying the transformation relationship between the world coordinate system,the camera coordinate system and the image coordinate system.3.Several methods of optical band image preprocessing are compared,and the optimal solution is selected to process the optical band image,so as to selectively highlight the geometric features of optical band image without destroying the details and clarity of the image.4.Threshold segmentation of optical band image can effectively divide rail area and optical band area,so that the information of optical band edge and bandwidth can be extracted completely.Finally,the abnormal light bands of a metro line are briefly classified,and the abnormal light bands of some metro lines are summarized.The theoretical analysis shows that the main reason for the abnormal condition of rail ribbon is the deviation of the action point and force between rail and wheel.The fundamental reason is that the existence of different types of track irregularities leads to irregular serpentine movement of wheels on the rail support along the traveling direction,which aggravates rail wear and makes abnormal light bands on the top and side of the rail.Among them,the serpentine ribbon formed between wheel and rail is particularly serious.The continuous waveform wear and the distortion of rail are present in the field,which may be caused by the plastic deformation of rail surface.