Research On Vehicle Speed Measurement Method Based On Single Camera Video Sequence Matching

Since the concept of intelligent transportation system(ITS) had been proposed, collecting traffic information, especially dynamic real-time information, has been the focus of intelligent transportation system research. Real-time speed of vehicles, as a crucial parameter, has become focal point of dynamic real-time information in traffic. This paper discussed a new vehicle speed measurement techniques based on single camera video sequence. The main work as follows.This paper firstly analyzes the basic mechanisms, advantages and disadvantages, and applicable situations of existing moving target detection algorithm. Among the existing methods, the background subtraction method is selected as the basic method of moving target detection, and Gaussian Mixture Model is used as the main tool for video background modeling.The second part of this paper presents the research on the algorithms of video moving target vehicle matching in depth. On regarding to the paper proposed a new algorithm of dynamic sliding window object matching through double windows on the basis of color histogram sequence of video image. The feature sub sequences of the target vehicle’s color histogram in the first window will be firstly detected and paired with the feature sequences obtained from the image target in the second window via correlation matching. With the best matching point, it can obtain the matching sub feature sequences when target appears in the second window.. By detecting the best matching point, it can obtain the matching sub feature sequences when target appears in the second window. This paper has proved the validity of this algorithm with examples and experiments.Furthermore, with the method above, this paper puts forward a complete video vehicle speed measurement scheme, including the establishment of data collection environment, camera calibration, setting window, vehicle target detection and matching, and. A new camera calibration method is designed for this scheme. By establishing a real road plane coordinate system and an image coordinate system, the transformation matrix and the mapping relationship of those two coordinate systems can be calculated with 4 or more pairs of known matching points. Considering the principle of camera imaging will cause the video scene to zoom in and out along with the camera’s view we designed an adaptive window setting method to ensure the two windows are in the same size in the real road plane coordinate system.In the last part, a video vehicle speed measurement system is designed and implemented on the basis of the matching algorithm and speed measurement scheme mentioned above. The system consists of five modules: collection module, display module, detection module, matching module and vehicle speed calculation module.