Traffic Vehicle Detection And Tracking In Video Streaming Technology Research

Intelligent Transportation System (ITS) is a traffic management system usinginformation technology, communication technology, automatic control technology etc.ITS can alleviate traffic congestion, reduce traffic accidents and improve road resourceutilization rate. To get traffic information by computer vision technology has theadvantage of easy installation and it can provide a large amount of information. So it isa subject that is worth studying. Automatic vehicle detection and tracking play animportant part in ITS. This paper aims at studying video vehicle detection and tracking.For vehicle detection, background subtraction method is used in this paper to detect vehicle invideo,which can get the whole contour of vehicle. To build real-time background by frame averagemethod, combining with inter-frame difference method to adjust the background update rate. Thismethod is fast and it can accelerate the background updating and cut down the smearing of vehicle.Building background by this method can meet the need for vehicle extraction. In order to improvethe accuracy of vehicle extraction, this paper processes the binary image of vehicle by mathematicalmorphology operation.For vehicle tracking, region based object tracking algorithm has the advantage ofsimpleness and easy to realize. Combining the Kalman filter to predict the location ofthe vehicle, the region based tracking algorithm has a lower matching time of vehicleand a higher tracking efficiency. A methed combing Mean-shift algorithm and Kalmanfilter to detect partially occluded vehicles is put forward. Using the Kalman filter toforecast the location of vehicles firstly, then serch the vehicle from the predictedlocations and get the accurate location with Mean-shift algorithm. Getting the nuclearbandwidth through the changing ratio of vehicle’s size in adjacent frames,which canimprove the search ability of Mean-shift algorithm. The result of the experiments withtraffic videos show that this tracking algorithm can segment occluded vehiclesefficiently, and enhance the stability of the vehicle tracking.