Research On The Video Moving Objects Detecting And Tracking In Automatic Passenger Counting

It is important to develop the public transportation for improving the urban traffic. The Advanced Public Transport System (APTS), except for the installation of foundational establishment, plays a significant role in improving the service. As a basal technology of APTS, Automatic Passenger Counting (APC), which can be used to record the passengers boarding and alighting in each period and bus stops, can effectively help to make decisions for arranging the buses'schedule, optimizing the bus routes, controlling and forecasting the traffic. Consequently, the buses can be dispatched dynamically, and then the efficiency and service can make a rapid progress.At present, APC using infrared and pressure equipment have been widely used in the world. However, those methods perform not well in congestive buses. As a result, a new solution of APC to the situation in our country needs to be researched. Due to the development of image processing and pattern recognition, a new means is applied.This paper introduces the video technology to bus automatic passenger counting, and it mainly includes the study on the technology of detection and tracking for moving video object in image sequence, and by this technology, implements the count for passengers'boarding and alighting.This paper firstly analyzes the process objects of moving objects detecting and trace algorithms, the characteristic of video image sequence, and studies their statistical character. After introducing the current methods of detecting and tracking moving objects, and analyzing the surveillance scene, an advanced approach on moving objects is put forward based on the joint of background subtraction and delayed frame subtraction. A statistical algorithm is used to obtain the background model, which is updated real time in order to adapt to changes of illumination and objects in the scene. The results of experiments show that this approach restrain the affection such as the variant shinning, noise and .realizes the detection and fetch of moving object.Secondly, an effective approach is adopted such as the image filtering, image segmentation, morphology processing and so on. After analyzing the area measurement of connected region, the invariable features of the moving object are chosen. According to these statistic features, false objects are cleared and the true objects of the heads are extracted,Thirdly, the Kalman filter is adopted to implement the predication of the moving object's location, so the searching areas of the objects[0] are reduced greatly. It builds match module of fetched motive objects, searches objects and matches in the predicting area. At the same time, as the result of the moving state and feature value of the objects in current frame are recorded to ensure the continuity of the dynamic tracking, so a kind of cost function is put forward which is applied in the tracking match of the image.Lastly, this paper finishes the design of bus APC's system structure and implements the simulation experiment system. Based upon the field video sequence collected on the buses, it implements experiment according to the detecting and tracking algorithms of moving objects proposed in this paper and the results of experiments showed a satisfied performance.