Research On Moving Object Tracking For Intelligent Traffic Monitoring

In the modern society, the traffic problems, generated by the rapid development of ground traffic, are gradually increasing. In order to know about the traffic, the vehicles and pedestrian flow around the monitoring road timely and accurately, a lot of video monitoring system have been used extensively. As an important part of the Intelligent Transportation System(ITS), its importance is increasingly highlighted. However, with the development of new transport infrastructure construction and the increase of the vehicles and pedestrian flow, the video monitoring system is becoming huger, and there are more and more information need to monitor. Just relying on the limited human resources to implement the real-time and comprehensive monitoring has become impossible. Therefore, the research of the traffic video analysis technology which based on the computer vision technology is of great significance. This article uses the expressway monitoring video as a data source, and researches the related key technologies of the video monitoring system. The concrete research contents are as follows:In the video image preprocessing, in view of the characteristic that the expressway monitoring videos are greatly influenced by the fog, this paper studies the theory of dark channel prior and clears about the strengths and weakness of the algorithm, then on this basis, uses a kind of algorithm based on the dark channel prior and wavelet transform to defog the monitoring video. Through the wavelet transform, we can get the low frequency and high frequency component of the image, and then use the dark channel prior to defog the low frequency component of the image. The experimental results show that the method can effectively reduce the time consumption, and improve the execution efficiency of the algorithm.In terms of moving object detection algorithm, this paper uses background difference method for the video image processing. Firstly, we use the adaptive Gaussian model to build the background, and then use the background to make difference with the video frames, and then we use the adaptive binarization processing and mathematical morphology method successively for the difference image to deal with the noise; At the last, we use the area of 5*5 search algorithm to extract the moving objects. The experimental results show that the method we adopted can accurately detect the moving objects.In terms of moving object tracking algorithm, due to the algorithm already has some time consumption in the previous image preprocessing, and in order to ensure the real-time of the whole algorithm, this paper uses a feature matching tracking algorithm based on the kalman filter. First of all, we extract the feature of the moving objects detected before, such as the size, shape, position of the center and the motion estimation, etc. Then, search the objects in the area of the kalman filtering prediction and match the objects searched before according to the object matching rules. Finally, we choose the optimal matching object and achieve the tracking of the object. The experimental results show that the method is relatively better in the real-time performance, and can achieve the tracking of the object accurately.