Study Of Inspection Technology For Rail Surface Defects Based On Machine Vision

The rails support the train body and lead the direction of the train,and their states are directly related to the safety of the train.With the development of Chinese modern railway enterprise,the operating mileage,speed and density increase greatly,which require higher frequency and speed rail inspection.How to efficiently and accurately inspect the in-service rails becomes increasingly important.This thesis has a study on defect inspection methods of rail surface based on machine vision,and presents an automatic inspection system for rail surface defects.The main works are as follows.1.High speed and high resolution hardware system for rail surface defects inspection is designed.Sub-system for high speed imaging,capturing,storaging is conducted around line scan camera Dalsa Piranha4 4K whose speed is currently the fastest.LED illumination sub-system is designed to deal with the lighting problem of short exporture time,and the influence of LED color is discussed.Synchronous trigger sub-system is designed to stabilizing the aspect ratio of image at various speeds.By estimating the experimental data collected on the high-speed wheel simulation platform,the system can achieve the image acquisition of 0.22mm/pixel at the speed of 160km/h.According to the theoretical caculation,the speed of 360km/h can be reached at 0.5mm/pixel resolution.An experiment trolley is designed and built with the validated hardware,which acquired clear sample images of actual rails with cooperation of Beijing Railway Bureau.2.By analyzing the rail samples acquired from experiment trolley,this thesis proposed the rail surface defect inspection algorithm.Rail egde locating algorithm based on LSD is designed to precisely locate the edge of rail surface in images,which filters out various distractions from rail side.Rail surface outline refine algorithm based on OTSU is designed to refine the previous rail edge location to the pixel-scaled outline,and rail surface outline inspection methods is designed to measure the geometric information like rail width,track gauge and gauge-alignment,detect rail defect like severe wear and crackle.Rail surface abrasion inspection based on MSER is investigated to detect the abrasion on the surface of complex textured rails.This thesis lays the foundation of realizing vehicular online detection system.