Video-based Vehicle Speed Monitoring System Design And Research Of Algorithm

Transportation is very important in national economy and people’s lives. Vehicle speed detection is an important part of Intelligent Transportation System. Video vehicle speed detection is widely used now, because of the Simplicity and convenience of image processing technology. In this paper, the vehicle speed detection system based on video is designed as a part of intelligent traffic.The experimental platform of this system includes hardware part and software part The hardware are composed by the Industrial computers and high-definition camera fixed on the road. Then the software system with the function of image acquisition, background reconstruction, vehicle detection, object match and prediction, speed detection, realizes the automatic monitoring of the vehicle speed. In the software system, firstly, the images captured by the camera are stored in a specified location, then according to the camera coordinate model, uses the Radon transform algorithm to find the calibration points in order to correct the images, and divide the images by lanes, then we get the background using an arithmetic that combined the edge projection statistics and Surrendra Arithmetic. Secondly, in order to obtain accurate and complete moving vehicle silhouette, the paper proposes a method of edge projection gradient statistics, which could correct the object silhouette getting by the background subtraction and edge detection. Thirdly, we use regional characteristics matching algorithm to track the movement of vehicles, then we use the Kalman filter to predict the location of vehicles in the next frame. Finally, we convert the image distance to actual distance and calculate the speed of vehicles.The innovative points of the paper are: (1) arithmetic that used edge projection statistics can roughly identify the background image and non-background, which make the subsequent calculations more convenient and easie.(2) using the Radon transform algorithm to correct the images. (3) We put forward a method of edge projection gradient statistics to obtain the target profile with high accuracy.