Researches On Video-based Automatic Detection And Tracking Of Highway Vehicles

Digital image processing and the analysis of computer vision&intelligent video isa new science research field which has tremendous development potential. Videosurveillance technology provides the traffic system with a direct and convenientanalysis method. Therefore, the video surveillance technology based on video imageprocessing, analyzing as well as understanding has successfully attracted the public’sclose attention, which has also been widely used in the Intelligent TransportationSystem in recent years. Among the system the real-time traffic detection and trackingis a very important project.With the development of computer technology and the prosperity of artificialintelligence, pattern recognition has rapidly developed into a new subject. This paperdoes some research on the issues mentioned above and makes some analysis andpartial improvement in vehicle detection, shadow removal and vehicle tracking. As aresult, we have designed an intelligent monitoring system which is based on thedetection and tracking of traffic image orders, which has shows its effectiveness andreal time in the experiments. The main researches are as follows.(1) Moving Vehicle detection. This paper first analyzes the existing detectionalgorithm, and then makes a deep research on Gaussian Mixture Model (GMM). Afterthis work, some improvement is made on GMM to overcome its defects andshortcomings. In order to rise the model’s convergence rate,we use an adaptivelearning rate threshold method to update the model, thus making it possible to extractclear background real-timely image and updating the background image effectively.(2) Moving Vehicle’s Shadow’s Removal. Moving shadow detection is the key stepto segmental and identifies the target vehicle. The paper analyzes the shadow’s opticalproperties and then proposes a rapid shadow detection algorithm based on YUV colorspace, which uses the shadow region brightness and color variation to remove theshadow. Experiment results show that this method is effective.(3) Moving Vehicle Tracking. In the complex environment of the expressway, theseproblems exist that the moving vehicles will disappear or be partially occludedtemporarily. In order to track moving vehicles effectively, this paper comes up with aregional matching vehicle tracking methods according to the Kalman PredictionModel, hoping to track the vehicles by using this method. Experiments show that the algorithm has good robustness and high accuracy which can track the vehiclesreal-timely.This thesis puts forward different ways to solve the problems existing inexpressway video vehicle tracking and traffic information extraction system. What’smore, experiments conducted show that these methods are adaptable.