Research On Real-time Monitoring System Of Metro Network Passenger Flow Based On Computer Vision

The urban rail transit is developing rapidly in recent years in China, which facilitates people to travel and alleviate road traffic jam at the same time. But the high dense crowds in underground confined space bring a great challenge to the subway operation safety. Accessing the real-time data and information of network passenger flow accurately, including the number of boarding and descending, the number of station waiting and the number of passengers on the on-line operational trains, will provide powerful decision support data for the operation department to improve the operation and organization efficiency, enhance the capability of responding to the operation in peak hours and passenger flow outburst significantly and ensure the operation safety and security finally.The computer vision technology which is advancing rapidly will be applied to metro passenger volume monitoring in this paper. The passenger volume data within the specific scene is obtained after the real-time acquisition, processing and analysis of passenger flow video, and passenger volume statistical computing model will be built to realize the volume data acquisition and monitoring finally. On the basis of stating the framework of computer vision combined with the practical application in subway station, the main work in this paper as follows:(1) Taking the monocular camera as the tool of passenger flow video capture and both the methods of background subtraction algorithm based on mixture Gaussian background modeling and improved temporal difference method respectively are applied to detect moving passengers. And the experimental results of these two methods in similar scene were contrasted. Multi-targets tracking and counting are realized by consecutive match of object features based on improved cost function.(2) In order to solve the problem that the passenger volume data acquisition will be inaccuracy because of the mutual occlusion in high-density populations, binocular stereo vision is introduced. Take the top of pedestrian's head as the feature region, combined with depth information of objects to achieve passenger flow detection, recognition and counting, and compared the experimental result with which conducted by monocular camera and corresponding algorithms;(3) On the basis of the real-time passenger flow data in specific scene acquired by the computer vision technology, the real-time statistical computing model of the passenger volume in stations and on the online operational trains is built. The framework of the real-time monitoring system of the network passenger volume is presented, and take Beijing Subway Line 1 and 2 as a case to achieve a simple simulation of this system based on VC++ development platform and real-time database.