Research On Clearance Intrusion Detection System For High-speerd Railway Based On Binocular Stereo Vision

With the rapid development of high-speed railway, the operating and being constructed high-speed railway are increasing in number, so the operational security of rail transport has been given higher standard. The intrusion situation might happen occasionally either in the existing railway lines where geological disasters occur frequently or the railway section near construction. If the train is in high-speed operation, the intrusion will badly threaten the safety of train's running. The clearance intrusion detection system based on binocular stereo vision researched in this thesis can judge the existence of intrusion accurately, and the research has important meanings to ensure the operational security of lines.Firstly, the overall struction of clearance intrusion detection system which was integrated with laser scanning and binocular stereo vision was described in this thesis. On these bases, this thesis mainly studied on the clearance intrusion detection subsystem based on binocular stereo vision. The principles and techniques of the binocular stereo vision, the hardware architecture and software design of system and the identification algorithm of intrusion as well as the processes and results of experiments were discussed in this thesis.In the aspect of binocular stereo vision, the important content included camera calibration, image preprocessing, stereo matching and 3D reconstruction. The corresponding experiment results were also given in this thesis. The system's hardware included two cameras, image collection card, industrial PC and so on, and these hardware modules built the environment of image collecting, real-time processing, analyzing and storing. This system's software realized the functions of data processing and remote alarm displaying by using VC++ and Matlab mixed-programming. To identify the intrusion, we adopted the method that combined 2D image recognition and 3D stereo vision recognition. Firstly, we drew the image background and judged the existence of questionable objects, then established the bounds of the detection area, and then calculated 3D coordinates of questionable objects, finally, judged the existence of intrusion.In the lab and field experiments, the system could detect the intrusion accurately and timely, and we achieved all that we expected.