Research And Realization Of Vehicle Detection And Tacking Algorithm Based On Video

The number of private car has made a sharp increase with the development of social economy.The importance of intelligent transportation system which is on many countries'agenda is moreapparent. The algorithm of vehicle detection and tracking based on video is an important part of ITS.We collect traffic videos by camera, and then extract traffic flow and traffic congestion informationthrough a series of video processing.This article proposed an algorithm for round-the-clock vehicle counting and congestion leveljudgment. The main works have been accomplished are as follows:1. Vehicle tracking and counting methods of day. The train of thought of target tracking forexisting targets is matching approximately first and selecting the optimal one second and thenupdating the information of targets with the features of the optimum. For those targets whichdisappear temporarily, we update their information using the predicted module of kalman and thenupdate their characteristics with themselves when then appear again.2. Vehicle tracking and counting methods of night. This paper introduces two methods forvehicle counting of night, which are vehicle counting method based on virtual test line and vehiclecounting method based on lamp tracking. For the former, we set a virtual line in the middle of thevideo first, and then scan this line. If there are lights on the line, light number increased by one. Forthe latter, in order to count vehicle, we first track the lights and then match the lights.3. The method of congestion level judgement. The algorithm of congestion level judgement isdivided into two parts in order to matching the algorithm of vehicle g. And we Set up three levelsfor the algorithm of congestion level judgement.4. Switching algorithm for day and night. This paper puts forward a scheme of switchingalgorithm for day and night. According to the system time, we confirm a time slot for algorithmswitching and then extract the the brightness of current video frame in order to determine the exacttime for algorithm switching between day and night.