The Research On Robust Vehicle Detection And Fast Tracking Method Based On Traffic Video

With the traffic management system becoming more and more automatic and intelligent,intelligent traffic system has been the hot spot of the research. Vehicle detection and trackingare the core modules of the intelligent traffic management system, involved in many aspectsof technology such as the video monitoring, computer vision technology, computer imageprocessing, and machine learning. Reliable and real-time vehicle detection and trackingalgorithms are able to deal with many problems such as the change of background and light,shadow, objects’ shelter and so on. About vehicle detection and tracking algorithms,predecessors have put forward many ideas, but to design robust and real-time vehicledetection and tracking algorithm are still a very challenging task. Based on this, this paper hascarried on the exploration and improvement about vehicle detection and tracking technology.In this paper, in the vehicle detection section, we use the method of Gaussian mixturemodel to extract the moving object. The detection result is not ideal due to the complexchange of the scene. The existence of the shadow seriously affects the detection effect. Basedon this, this paper puts forward a shadow elimination algorithm based on dark channel byusing the reverse figure with similar properties of translucent mist and meanwhile combiningwith classic mist elimination algorithm, which can effectively eliminate the shadow ofvehicles and is also robust to the shadow of the dark vehicle. Thus, we can extract the outlineof the vehicle more accurately. In the vehicle tracking section, this paper proposes a fastvehicle tracking algorithm based on the maximum overlap area and the cam-shift. Thisalgorithm can effectively solve the vehicle adhesion and division problems, and meanwhileensure the accuracy and the real-time of vehicle tracking. Finally, this article has also carriedon detection about some relevant traffic events based on the result of vehicle tracking andstudied the main practical value of vehicle detection and tracking.