Research On Video-based Traffic Flow Parameters Extraction Method And System Realization

In recent decades, with the rapid increase of the number of motor vehicles, the urban traffic has faced more and more pressure, which results in heavier traffic congestion and many other problems. And it’s not a workable solution to simply increase road construction and manual traffic management. Currently, the intelligent transportation system is an effective solution to solve the problem of traffic congestion. To satisfy easier camera calibration and full-time employment of the traffic flow parameters extraction based on video in ITS, the thesis has studied from the following aspects:Firstly, in order to obtain real vehicles’ speed, the camera should be calibrated, while many current calibration methods still rely on manual spot operation, which results in low efficiency, so this thesis proposed a calibration method using reference images and road information to avoid manual spot operation. The method uses some known roadway information which includes road markings and so on to get two vanishing points which are on the directions of parallel and perpendicular of the road, and also uses the known road width to realize camera calibration. As for the ill condition which a vanishing point approaches infinity, the thesis calibrates the camera through a group of reference images including two vanishing points at the same time by rotating the camera.Secondly, the paper utilizes different methods of extracting the traffic flow parameters for the two different daytime and nighttime traffic scenes. 1) For the daytime scene, as it will results in high error to use single vehicle detection and tracking on the condition of high density of traffic flow, we obtain the flow parameters by using optical flow. After getting the velocity and road density, the vehicle flow is obtained by using traffic flow theory for the condition of high density of traffic flow, while it is calculated by the theory of impulse counting for the other conditions. 2) For the nighttime scene, due to the poor lighting conditions, the thesis has proposed an effective method based on the histograms of vehicles lamps’ luminance to realize vehicle detection and tracking. We calculate the histogram according to the lane which the centroid of the vehicle lamp belongs to. After that, we can obtain vehicles location and accomplish detection by obtaining the beginning and ending intervals of the histogram’s peaks, and realize vehicles tracking by using histogram matching of the intervals between the image frames. After vehicle detection and tracking, we can get the information of car numbers and velocities, which can help extract the traffic flow parameters on the nighttime scene.Finally, we have programmed to implement the functions of video capture and traffic flow parameters extraction, and tested on some traffic videos in different scenes. The results of the experiments indicate that the system has achieved high accuracy in traffic flow parameters extraction, and obtained the expected goal.