Research Of The Detection Algorithm About The Distance Between Railway Maintenance Machine And Fixed Application

With the increase of the density of railway transportation in China, most of the repairmen equipment in the busy line is Railway Maintenance Machine (hereinafter referred to as ’RMM’), such as reshaping machine, cleaning machine. The RMM usually operates under "appointed time limit" mode. Most of the "appointed time limit" mode is arranged in the night with harsh environment, low visibility and easy fatigue, which leads to the mechanism of RMM(like cleaning machine’s rotative dirty belt) can’t retract timely in the process of moving. Mechanism of RMM and fixed application(mainly the catenary pillar) collided frequently, which threaten people’s and property safety. Therefore, it’s necessary and urgent to research a warning device of real-time distance detection, which makes the mechanism retract timely. The central matter of the device is the detection algorithm about the distance between the front end of operation mechanism and catenary pillar. The detection algorithm of parallax image distance of binocular is researched in this paper. Calculating the distance between RMM and fixed application (the research object is catenary pillar in this paper) is based on the binocular stereo vision system, the operation characteristic of RMM and binocular parallax ranging principle. Feedback the real-time distance information to remind the driver avoid the catenary pillar timely to ensure the operation safety of the RMM.Firstly, the theory of binocular vision, camera image model and the mathematical conversion between three kinds of coordinate have been subjected to intensively. In addition, the principle of the method for common binocular calibration was analyzed, a kind of adaptive algorithm was improved, binocular vision system location was realized by program, and the internal and external parameters of the camera were obtained:the focus a and/?, base point uo and v0, distortion coefficient k1, k2,p1,p2, rotation matrix R and translation vector T.Secondly, because in common binocular vision algorithms, the distance of only one feature point is calculated. the distance between two feature points is calculated in this paper, Binocular vision algorithms is improved, distance detection algorithm based on binocular vision between two feature points is programmed.Thirdly, according to shape edge features of the catenary pillar on both sides of the railway line, the common edge detection algorithm based on several image content is analysed and compared, edge detection algorithm of Hough transform is adapted, Gray-scale, binarization, denoising of image is preprocessed, the image edge detection algorithm is programmed, feature points of the catenary pillars and job agencies in the image is determined.Finally, under different illumination and complex background, a lot of test of the algorithm have been taken. The result shows, this algorithm can meet the actual demand, and it has a certain value of reference.