Research On The Measurement Of Rail Wear Based On Machine Vision

With rapid development of railway transport and increasing of operating frequency in China,rail wear becomes serious increasingly.The over wear rails is risk to the safety of running train.Today,the measurement method of rail profile has contact and non-contact,the domestic measurement method is to detect with artificial caliper mainly,its efficiency is low,labor intensity is high,the measurement result is vulnerable to be affected by human factors,it cannot meet the requirement of rapid railway development.So,it is very important to develop high efficient?accurate and suitable for small grinding machines measurement system of rail wear.Firstly,this paper introduces origin and background of the research subject and current situation of steel wear measurement at home and abroad.According to processing principle of machine vision and image,it designs the overall measurement system.Though studying several common used camera calibration methods,it uses distortion mode in NI Vision to calibrate current perspective errors and lens distortion errors in the system,designs process of contour prefabrication of steel profile,including threshold segmentation,image denoising,expansion and corrosion,and image refinement.Secondly,it calls calibrated data to restore profile of steel rail,and uses SSDA algorithm for image matching and wear calculation,and at last,it stores measurement results of data and prints reports.Based on Lab VIEW software platform,this paper completes design and development of rail wear measurement software system by means of modular programming.This paper builds simple experimental platform for rail wear measurement system,and measures system statically and dynamically.The experimental result shows that the error is within ± 0.5mm basically,verifies the accuracy,stability and repeatability of the system,it can meet the measurement requirements and has certain practical significance.At last,this paper analyzes the source of error,and put forward further suggestions to eliminate the error.