Upgrade And Optimization Of Non-contact Rail Measuring System

The development of railways requires more high-quality rails.The profile size of the rail is an important parameter for the quality of the rail.Whether the profile size parameter of the rail is qualified will directly affect the quality of the rail.my country has developed from the contact-type manual measurement of rail profile dimensions to non-contact three-dimensional measurement.Many scientific research institutions have designed and developed rail profile measurement systems based on machine vision.Although these rail measurement systems can replace manual measurement,there is still room for improvement.In the manufacturing action plan announced by the state,it is clearly stated that:building an intelligent platform for rail transit,building a full life cycle management of rails and other development directions.Therefore,this research topic optimizes and upgrades the original rail profile detection system from two aspects,improving the image processing algorithm respectively,and developing and upgrading the rail cloud data platform added by the industrial Internet.After two sets of experiments of static measurement and dynamic measurement,the repeated absolute error is lower than0.02 mm,the measurement absolute error is lower than 0.09 mm,and the average measurement absolute error is reduced from the original 0.06 mm to 0.04 mm,which further improves the measurement accuracy.The development and deployment of the rail cloud data platform simplifies the management of rail production data and improves the efficiency of rail detection.The research contents of this paper are as follows:1.According to the principle of non-contact 3D measurement,comprehensively analyze and introduce the rail profile measurement system before the upgrade and optimization,and give an upgrade and optimization plan based on the problems in the actual deployment and application.2.According to the laser line coincidence problem of the rail measurement system and the measurement of different rail types,a calibration method of multiple laser profile sensor cameras using a rectangular calibration block is designed,and the contour stitching redundancy is quickly removed according to the gray level of the rail profile image.2.Introduce and analyze the preprocessing method of rail profile image,design experiments to analyze different processing methods,compare the experimental results,and finally determine the image preprocessing method of median filter combined with gray correlation degree.4.Optimize the least squares method used for fitting the straight line part and the arc part of the rail,and use the Gauss-Newton method to optimize the least squares method for the arc part,experiment with different algorithms,and select the best operator according to the results.5.Designed and developed a rail measurement cloud data platform,including onsite,server,client,cloud server and cloud database,web page and mobile application APP,to realize the daily management and maintenance of rail data,and timely through the mobile APP View rail data.