Research On Automatic Tracking Algorithm Of Dynamic Traffic

Digital image processing and the analysis of computer vision & intelligent video is a new science research field which has great development potential. Video monitor technology provides the traffic system with a direct and convenient analysis method. Therefore, the video monitor technology based on video image processing, analyzing has successfully attracted much more attention of public, which has also been widely used in real-time monitoring, tracking and vehicle assorting in the Intelligent Transportation System in recent years. With the development of computer technology and artificial intelligence, pattern recognition has rapidly developed as a new subject. This paper takes some research on the issues mentioned above and makes some analysis and partial improvement in vehicle detection, shadow removal and vehicle tracking. As a result, we have designed an intelligent monitoring system based on tracking of traffic image orders, which has verified its effectiveness and real-time in the experiments. The main researches are as follows:(1) Moving vehicle detection. This paper first analyzes the existing detection algorithm, then makes a deeply research on Gaussian Mixture Model(GMM). After this work, some improvement is taken on GMM to overcome its defects and deficiency. In order to rise the model’s convergence rate, this paper use threshold method of an adaptive learning rate to update the model, then make it possible to extract clear background real-time image and update the background image effectively.(2) Shadow Removal of Moving Vehicle. Moving shadow detection is the key step to segment and identifie the target vehicle. This paper analyzes the shadow’s optical properties and proposes a rapid shadow detection algorithm based on YUV color space, which uses the shadow region brightness and color variation to remove the shadow. Experiment results show that this method is effective.(3) Moving Vehicle Tracking. In the complex situation of the high-way,the moving vehicles will disappear temporarily or be partially sheltered. In order to track moving vehicle effectively, this paper comes up with a regional matching vehicle tracking methods according to the Kalman Prediction Model. Experiments show that algorithm has good robustness and high accuracy which can track the vehicles realtime.This paper proposes different ways to solve the problems of vehicle tracking and identifying existing in high-way traffic information system. In the end, experiments show that these methods arefeasible.