Analysis On Metro Gauge Intrusion Based On Stereo Vision

With the advancement of urbanization process, the traffic problem has become increasingly prominent. In order to improve the urban traffic carrying capacity, many cities have increased the intensity of the metro construction with the problem derived from metro security drawing people’s attention. The metro gauge is a spatial extent in which the subway can run in safety and if intrusions exist, vehicles can have a big threat. So quick and accurate detection of intrusions using dynamic non-connect measurement methods has become more important to the subway safety operation.The detection methods combing the laser scanning and stereo vision were determined after analyzing both the advantages and disadvantages of the two detection techniques. Since clearance detection system based on laser scanning has been more developed, the stereo vision detection was studied mainly and it could be used as a supplementary method for gauge detection.The following research has been accomplished:(1) Video camera was calibrated by using the Halcon vision software and the model of camera imaging was built after getting the internal and external parameters.(2) Feature point matching was formulated by using the method of combing automatic and manual feature point matching. Among automatic feature point matching includes the following three parts:the extraction of feature points, feature points description and the matching. Considering the number of feature points extracted and the computation time, a combination of three algorithms met the experimental requirements. Among SURF algorithm was utilized for quickly searching for matching points, while SURF algorithm was chosen for describing feature points and Flann algorithm was selected for the feature point match.(3) World coordinates for feature points were evaluated based on the camera model built and the following feature points matched, thereby the3D reconstruction model of feature points was built.(4) The clearance space was determined on the basis of metro clearance coordinate system and the metro standard of gauge. Identification of the intrusion was studied and determined based on the feature points converted to the clearance space.The results show that the clearance detection system built can not only accurately extract the feature points of intrusion but identify the intrusion quickly. With the advantages of avoiding the mismatch and leakage match, the system can reach centimeter-level accuracy which achieve the intended purpose.