The Abstraction Study Of Traffic Parameters Based On Video Technology

This paper firstly introduces some domestic and overseas' general development situations of intelligent traffic system(ITS), and studies on the traffic surveillance based on video technology. In order to obtain exact traffic parameters, author has done some researches of division of foreground and background from video frames, vehicle shadow segmentation, vehicle abstraction and recognition, abstraction of traffic parameters, recognition of traffic congestion degree and so on, and gotten good results from experiments. In the study of division of foreground and background, this paper mainly researches the background forming algorithm in background differencing method, improves the background forming algorithm based on common information in frame differencing, solves vehicle wrong recognition problems arisen from close intensity but differences in color between vehicle and background. In order to get automatic thresholds and solve image's trembling problems, author presents the background forming algorithm based on HIS color space's single Gauss model and the background forming algorithm based on sliding window's color space and single Gauss model. Experiments have proved that the improved algorithms can solve problems well. Otherwise, author presents the background forming algorithm based on texture space and discusses updating methods of background. In the study of vehicle shadow segmentation, this paper compares among several common vehicle shadow segmentation algorithms, uses the vehicle shadow segmentation method based on spatial characteristics that has better segmentation results in the traffic detection system. In the study of vehicle abstraction and recognition, author contrasts the vehicle recognition method based on background differencing to the vehicle recognition method based on texture space's background differencing. Experimental results present the vehicle recognition method based on background differencing can obtain more vehicle characteristics parameters. In the study of vehicle recognition based on background differencing, author constructs the rules of region combination and the model of vehicle figure's recognition, improves the vehicle recognition method based on edge projection and gets better accuracy in vehicle recognition. In the study of traffic parameters abstraction, the traffic detection system can exactly obtain traffic parameters of video frame, according to vehicle characteristics parameters obtained from vehicle abstraction and recognition and designed traffic parameters' computation models. On the basis of former study, this paper designs and realizes traffic detection system based on video technology and gets no bad experimental results. In the study of recognition of traffic congestion, this paper adopts a novel approach to synergetic pattern recognition which is very different in contrast to the traditional pattern method. Because the standard synergetic pattern recognition algorithm spends more time on computation and is not fit for huge data and relearning of model, author introduces the reconstruction of order parameter, exert the restrict synergetic pattern recognition algorithm based on reconstruction of order parameter for the recognition of traffic congestion based on video technology. This approach gets good effect and has certain innovation.