Study On Foreign-Body Detection Methods At Railway Crossing Based On Machine Vision

With the rapid development of machine vision theory,machine vision has been applied to various fields.Railway crossing is the intersection of the road and the railway,casualties and economic losses often occur in traffic accident because of foreign body intrusion.In this paper,a method based on machine vision is proposed to detect the intrusion in railway crossing and conducive to the safety of railway crossing.This paper analyzed the calibration,the camera parameters based on binocular vision detection system of image correction,stereo matching,3D reconstruction,object localization and tracking problems,and gives some improvement measures which achieved good results.The main work of this paper is as follows:(1)Based on the analysis of railway crossing of the existing intrusion detection methods,a general scheme of the intrusion detection system is designed.And the description of the structural function of the detection system is given.(2)Aiming at the problem of low accuracy of traditional intruder location of foreign body,presents a method based on semi-global matching algorithm of binocular reconstruction,by the way of experimental test,for the distance between the camera center and intrusion body within the range of 7 meters,the relative error between the depth of information extraction and the actual distance less than 3%,and the problem of accuracy of traditional 3D reconstruction based on sum of absolute differences is improved.(3)Aiming at the problem of high false positive rate of traditional intrusion detection method,this paper presents a new method of intrusion detection based on binocular vision system and Mean-Shift tracking technology,and new kind of binocular intrusion system is realized,it can improve the positioning error of the moving target by the tracking of the dual cameras.The experimental results show that the system can detect the movement state of intrusion bodies in a certain time,and the problem which the traditional method is easy to midjudge is solved.