Research On Vehicle Detection Method Of Intelligent Transportation System

City transportation system is one of important basic- facilities in city. It's a necessary project in economy growth and people's life. City traffic is not only meets the need of residential trip, in some way, it plays a significant role in normally bringing city function into play. In recent years, Traffic Demand is growing day by day with the growing of economy. Crowded city traffic, frequent traffic accidents and worsening traffic environment become a common problem that all the countries in the world are facing. The traffic problem is a complicated and integrative problem. Considering it only from road or vehicle will fail in solving the traffic problem. In such context, the theory integrating vehicle with background to considering the method solving traffic problems comes forth. And vehicle detection technologies are just the important aspect of this theory.Real-time detection and positioning is an important part of Intelligent Transportation System. Up to now, numerous scholars are doing research work on correlative domain, at the same time, facing many problems, like the shadow of object, shooting noise in the real scene, the variety of light and weather, and so on. All of above will affect the precision of detection.In recent years, the scholars home and abroad have done extensive research on moving object detection based on video. Traditional methods of vehicle detection include: background subtracting, time difference, optical flow method and so on. Among these methods, the background subtracting method has been widely used in moving object detection area because of its little calculation and being able to use background updating technology to implement self-adaptive background updating. This method can precisely segment the moving object from background. However it still has many disadvantages, many improvements have been done according to the need. The time difference method adopts the difference based on pixel among two frames or three frames in sequential image sequence. However it can not obtain all the characteristic pixels, and bring holes in the moving entity. Due to the great calculation, bad de-noise ability and depending on hardware, The application of optical flow method is restricted.In this paper a new method for vehicle detection is proposed, that is the self-adaptive background updating technique based on the moving region. This method apply the self-adaptive background updating technique to moving object detection and segmenting based on video image. It can different the moving object fast and precisely, and then find the shadow of vehicle and the wheel part. Experiment results show It can improve largely the precision and quality, achieving our expectation.