Research On The Detection System Of Rail Wear Based On Machine Vision

There is complicated interaction between the wheel and the rail, which directly resulting in thevertical and side wear of the rail. It will be a great threat to the safety of the vehicle when the wearvalue exceeds a certain limit. Therefore, it’s very important for the railway maintenance departmentsto measure the rail wear accurately and periodically. Traditional detection methods are based onmechanical devices in contacted with the track. This kind of methods requires much experience ofworkers and high labor intensity but with low efficiency and accuracy, which could not meet thecurrent high-speed rail development. So, there is an urgent need for an automated testing solution.The hardware and software platform for rail wear detection device is developed based on thebasic theory of machine vision technology, optoelectronic measuring technology and digital imageprocessing technology. The hardware includes a CCD camera, light source, image acquisition card,speed and position measurement unit. The parameters of each hardware are calculated in detail. Theexternal trigger module is introduced to keep the frequency of the camera synchronized with trainspeed. Phase detecting method based on photoelectric encode and single-chip is designed to realizethe speed and position function. A zhang camera calibration algorithm is analyzed to solve the keyproblem of image distortion of the rail contour. The profile image processing, including thresholdsegmentation, denoising, contour searching, restoration, contour matching, is discussed. With the helpof OpenCV algorithms library and Access database, the system software is realized on the VS2008platform. The function of data storage, inquiry, report generation and data analysis is also achieved.Finally, by building a simulation platform in the laboratory, static and dynamic testing isconducted to verify the performance of the system. Results show that the system can measure thewear value accurately and automatically. So, it has a better applicability.