Research On The Extraction Of Pedestrian Characteristic Data Based On Subway Station Video Analysis

The subway station works as the center of subway lines and the whole line network, service facility planning, designing and optimization of which will influence the whole operation effectiveness of the line network. Sometimes the subway station jams and imbalance between supply and demand will happen whether the real pedestrian flow is bigger than the design flow or not, which reflects the importance of real pedestrian related data. So the planning, designing and optimization of subway station service facilities should not only depend on related specifications and pedestrian simulation softwares, but also take the real pedestrian arrival distribution rule and the real facility service capability into consideration. Nowadays, surveillance videos which contain much pedestrian characteristic data are mostly used for security monitoring, then how to pick up those data efficaciously from those videos is important.Therefore, this paper extracts pedestrian characteristic data via the video of the service facility in the subway station. Specifically, research contents are as followed:(1) Considering the need of pedestrian characteristic data, this paper takes the arrival time interval, cumulative total and service time of pedestrian in the passageway, the automatic ticket vending machine and the automatic machine of ticket examination as the pedestrian characteristic data types and defines them.(2) For the pedestrian occlusion problem, this paper analyzes the monitoring environmental characteristic and compares three shooting modes, which are the vertical, the horizontal and the inclined, and finds that the vertical won’t cause occlusion. So this paper chooses vertical shooting mode and determines the video capture program.(3) This paper deals with the video using three kinds of image pre-processing technology, including smooth processing, morphological processing and connected domain processing, and makes sure of the value of the parameter for the preparation of pedestrian detection.(4) For the non rigid body problem, this paper chooses the head to describe the pedestrian. Common target detection methods are applied to the pedestrian detection of this paper, including background subtraction, watershed algorithm, Hough circle detection, and so on. After comparing advantages and disadvantages of those methods, this paper combines background subtraction, threshold segmentation, edge detection and Hough circle detection and puts forward an improved method for detecting pedestrians using subway station videos. The test accuracy is bigger than 90%.(5) The pedestrian tracking method is put forward using point coordinates based on the pedestrian detection and its effect is 100%. Then this paper designs data mining methods of three types of pedestrian characteristic data. Finally, this paper realizes a system of extracting pedestrian characteristic data in the subway station based on monitoring videos with OpenCV, MFC and Visual Studio 2010.