The Research On Video-based Detection And Tracking Method Of Vehicle In Intelligent Transportation Systems

Intelligent transport system is currently being studied and worldwide focused on in the area of world traffic transport. These years, the application of ITS brings great economical benefit to traffic transport, and it plays a more and more important role in road design, traffic surveillance and highway charge. In this paper, as a important part of ITS, video vehicle detection and tracking technology is researched, Compared with traditional vehicle detection methods, video vehicle detection has a lot of advantages such as easy installation and maintenance, wide monitoring areas, and also through image analysis and processing it Can obtain much useful traffic information, including the amount of vehicle in certain period, the position of vehicle at any moment, and so on.After the study and research on the classic algorithms of vehicle detection and vehicle tracking based on video sequence, some algorithms and techniques for vehicle detection and tracking have been analyzed and improved to deal with the existing problems. As a result, a new method for vehicle detection and tracking in municipal transportation was presented. The main contents of this paper include such aspects as follows:(1)A new method of background model building was presented. According to the method, background models are built by using background subtraction and Inter-frame subtraction. Firstly, only the frames according with the demands are chosen for background initialization frame so that the accuracy of background model could be enhanced, and the number of the chosen frames is variable. Secondly only the pixels which are regarded as background pixel through background subtraction can be chosen for background model updating, and the coefficient used for updating the exist background model is dynamic. The variation of coefficient depends on the variation of actual background. So the background model call be updated in time even when the actual background changes suddenly. In addition, the image segmentations ale done by dynamic thresholds to avoid serious mistake caused by sudden change in actual background. The experimental results show that this algorithm can detect the vehicles on video sequence more accurately.(2)A new method of Vehicle tracking was presented, which is multi. Feature matching based on region algorithm. According to the past moving features, the center position of tracking objects Call be located approximately in the next frame. Accordingly, the field of object tracking is narrowed down. In the process of vehicle tracking, both gray images and binary images ale used for the tracking of the moving Vehicle. In the matching stage, the features of center position, size and gray-value Can be used in the object matching. Experimental results demonstrate that this algorithm can track the moving vehicle efficiently.