This paper analyzes the development status of the techniques to acquire traffic information. According to the application requirements on current traffic systems, a motion estimation method adopting three-frame-differencing is used to construct initial background, and the background is updated in real time with the strategy of statistical scoring. A simple and effective method of shadow removal is put forward to increase the accuracy of traffic flow parameters detection. In addition, aiming at Flow Rate Calculation error caused by existing traffic flow detection system having no vehicle tracking system, a matching pursuit algorithm of the vehicle based on the location information, the color information and the fractal dimension information of the vehicle is proposed. A great deal of experiments demonstrate the system can detect various traffic flow parameters quickly and effectively.