Research On Recognition And Tracking Of Moving Vehicle In Traffic Video Surveillance

As a research cutting edge involving multidisciplinary in the field of image processing and computer vision, identification and tracking technology of moving targets in traffic video surveillance has great scientific significance and broad application prospect. Identification and tracking of moving targets are two kinds of core technologies to achieve intelligent traffic video surveillance. The result of identification and tracking is the basis of follow-up high-end applications, such as identification of moving vehicle behaviors, analysis of moving vehicle behavior, detection of traffic incidents, unusual alarm for traffic, as well as the key to the realization of intelligent and automated system of traffic video surveillance. Currently, it remains an unsolved subject for system of traffic video surveillance to explore algorithms that can automatically adapt to environmental change, and identify and track moving vehicles in a real-time, accurate and stable way.Based on the background of achieving a system of traffic video surveillance, this essay focuses and researches on two key technologies- identification and tracking of moving targets in traffic video surveillance, to obtain more reliable and practical algorithms that can satisfy an intelligent video surveillance, and completed the design and achievement of the system of traffic video surveillance.In the detection of moving target, the background substrate difference method of background updating algorithm is mainly researched , combined with the traffic video background of stationary and monotonic, an improved background updating algorithm is proposed based on Surendra's background updating algorithm; In the shadow detection and removal algorithms, according to the Movement of vehicles in traffic video shadow characteristic, an elimination method is proposed based on the HSV color space and dynamic space binary image shadow; In moving target classification and recognition, the comprehensive analysis of the geometry and velocity of moving objects, according to these features can be effectively divided into a large moving target vehicles, cars, motorcycles and pedestrians on the road in our country. In the identification process, according to the phenomenon of vehicles merge caused by traffic congestion, zoning method base on virtual lane is used to separate vehicles in different lanes separate. Eight connected method is used to tag the moving targets, obtained the minimum enclosing rectangle; In tracking moving targets, focusing on the problem of tracking failure cause by vehicles blocking each other, This article using a tracking method based on matching matrix. According to the traffic video easily affected by environment, adopt a general exception tracking approach;According to the practical running tests in highway monitoring and control center, the system can achieve quite high accuracy under various conditions. Results of the tests verify the correctness of the algorithms used and the feasibility of the system. The algorithms can satisfy a real-time, intelligent and stable system of traffic video surveillance. The prospected research goals has been realized.