Study Of Imaging Method And Apparatus On Rail Surface Defects Detect With Machine Vision

Rail surface defects has become the main cause of railway traffic accident.Timely maintenance can effectively reduce the surface defect of track railway trafficaccidents. With the rapid development of machine vision technology, the utilizationof machine vision technology to track surface detection is becoming easier, and theimaging method and the apparatus existing on rail surface defects detect with machinevision has large limitation. Based on the above background, combined with practicalengineering, the imaging method and the apparatus on rail surface defects detect withmachine vision are studied, and test the imaging effect of two different imagingapparatus. The main content of this paper are as follows:(1) Rail surface defects detection methods existing at home and abroad arereviewed, then against the limitations of traditional methods, track surface defectdetection method based on line scanning imaging is proposed and illustrates theadvantages of line scanning imaging and working principle in detail.(2) In order to analyze the line scan imaging device module selection andworking principle, lab indoor simulation imaging system is designed, and then usinglaboratory simulation imaging system for image acquisition of the rail surface defectsin different imaging conditions, which test reasonable of system module selection andconfiguration parameters.(3) Because of the great gap between surface defects man-made of indoor railwith the surface defects of actual rail, so to further test the effect imaging of surfacedefects of actual rail with machine vision inspection technology, outdoor trolleyimaging system was developed, and designed software for imaging acquisition detail,which test real tracks with this system.(4) Changes of external environment as well as difference of the rail surface lightreflection effect will cause luminance uniformly of imaging, which increases thedifficulty of defect recognition. Algorithm that the information entropy as imagequality evaluation criteria, using fuzzy PI controller for controlling integration timeof camera exposure adaptive has been provided in this paper, which is enable thebrightness of all acquisition images in a constant range, which is conducive to thesubsequent identification of defects.The experimental results shows imaging method and apparatus for rail surface defects detected with machine vision can be imaging well for real online tracks; thewhole system can run normally, parameter calculation accuracy and reasonableselection of each component module, laid the foundation for product development inthe future.