Design And Implementation Of Train Positioning System Based On Computer Vision

Rail transit is an important urban transportation infrastructure.In order to manage and control trains dynamically,the precise location of trains in the operating line needs to be determined in real time.Traditional positioning systems have problems such as cost.Computer vision technology can be used as a supplement to the traditional solution due to its low cost and high flexibility.This thesis firstly introduces the selection of pattern and the image processing technology.According to the actual factors,Aruco pattern is selected as the image identification beacon.The recognition process is affected by the environment and imaging equipment,which requires the use of computer vision techniques such as morphological processing,image filtering,and color space conversion.This thesis also proposes a binarization algorithm for the uneven lighting environment.The core of the positioning system is distance measurement.Binocular stereo vision obtains parallax information through the camera,calculates the depth map to makes reliable distance measurement.In this thesis,we design and implement a positioning system based on computer vision,which requires image pattern to be placed next to the track to determine the rough position of the train through real-time recognition process and further determine the precise position through binocular distance measurement.The overall system has been verified by simulated environment and onboard actual environment,and is capable of accurate,real-time and reliable positioning of trains.