Video-based Traffic Information Detection System In ITS And Its Camera Calibration Strategy

The research on intelligent transport system is always very hot all over the world. Among them, video-based traffic information acquisition and detection is being attached more and more importance because of its simplicity,convenience and real-time characteristic.In this paper the camera set above the road is used to collect the road images. Then traffic information is obtained by image processing by the video acquisition board. Finally these parameters and video are uploaded to the surveillance center. Thus by using video-based method the traffic parameters detection is realized, which is very important to traffic intelligent management.On the basis of the research of the DSP-based detection system, the deep research on the video acquisition board– core of the system—is presented. Beginning from the analog video signal input, the following aspects are introduced in detail: video A/D, video cache, timing control, image processing by DSP and so on.In the system, the pixel distance is necessarily converted to the actual road distance. Therefore camera calibration is applied to the conversion from image coordinates to world coordinates. After reviewing all kinds of calibration methods, an algorithm based on Direct Linear Transform (DLT) is presented, whose calculation is accomplished by the least square iteration and the conversion is realized by 3-D reconstruction. Based on the anlysis of MATLAB simulation results,it is concluded that the algorithm has the advantage of high-accuracy and its parameters can be easily estimated. It is proved that the algorithm could meet the requirement of calibration accuracy in the application of the video-based traffic information detection. The algorithm is finally applied to vehicle velocity detection program.The detection system adopting video acquisition board and camera calibration algorithm is tested on real-world scenes and experimental results prove that the system can effectively aquisite the traffic parameters such as the velocity,occupation ratio and car flow quantity. Real-time detection and high accuracy is achieved.