Research On Vehicle Tracking Technology Based On Video

With the development of computer technology and image processing technology, video technology which can detect and track vehicle has become the core and the key to intelligent transportation systems. It has many advantages such as covering large area and obtaining rich information of traffic.The real-time tracking moving vehicle is an important way for obtaining reliable traffic information. After analyzing and comparing various methods of moving vehicle detection, moving shadow removal and moving vehicle tracking, an method for tracking moving vehicle with a static camera is proposed, it not only can deal with complex traffic scenes such as global illumination changes and occlusion, but also can robust to solve noise problem.The thesis'emphasizes are to detect moving vehicle, to remove moving shadow and to track moving vehicle which under occlusion. Firstly, for detecting moving vehicle, under analyzing the choose of the number of Gaussian and the adjust of the learning rate in the mixture Gaussian background model, an adaptive moving vehicle detection method based on mixture Gaussian background model is proposed, this method can obtain complete information of vehicle; Secondly, for removing shadow, under analyzing the pixel's photometric properties and gradient properties in its neighbouring pixels, a method based on photometric properties and multi-gradient analysis is proposed, this method has preferable adaptive performance, a high detection rate and a high recognition rate; Lastly, for tracking moving vehicle which under occlusion, the thesis has improved the conventional mean shift method, and combined gray prediction model to track vehicle, this method is robust to solve occlusion problem.In order to track moving vehicle, the thesis puts forward three improved methods which are moving vehicle detection method, moving shadow removal method and moving vehicle tracking method. These three methods not only have some reference value for the application of tracking rigid targets, but for tracking non-rigid targets.