Research Of Video-based Vehicle Velocity Measurement System

In order to modernize our transportation and make our traffic system more efficient, we have put significant effort into improving China's intelligent transportation system, so it is developing rapidly during these years. This paper is written under this background.Through the research of the development of ITS and image processing technology, a video-based vehicle velocity measurement system is designed. The core idea of the system is to reduce the processing time on condition that the accuracy of vehicle detection and velocity measurement can also be kept.In order to guarantee the accuracy of video-based velocity measurement, the position of the vehicle has to be accurately measured by monocular camera. The traditional method requires many parameters such as focal length, mounting height, angle of the camera and so on. It is complicated and inaccuracy. Therefore, a new road coordinate system is designed to get the position, length and width of the vehicles. Firstly, in order to get each of the connections weight of trained artificial neural network, a set of points whose distances are already known are selected. Secondly, these weights can be used as coefficients to create a polynomial which is the function used to calculate the road coordinate values from the pixels'position in frame images. Experiments show that this method is relatively simple, and the precision can meet the requirements. This method is simpler and more accuracy to meet requirements. The road coordinate system laid the foundation for real-time velocity measurement and created conditions for the vehicle detection and tracking.In the step of vehicle detection, frame images are firstly processed by mean filter. And then images are segmented by two thresholds. In this approach, shadow elimination is done at the same time. After that, the small noises are eliminated rapidly by taking area as a condition. In the step of vehicle tracking, a feature-based method is employed to realize the tracking efficiently. The features are divided into primary and assistant ones, and they're used in the order of their importance to match the vehicles which are already detected. In the step of velocity measurement, the exactly position of the vehicle can be got from the road coordinate system established, and the time interval can be got from the frame rate of the video; the ratio of the distance and the time interval is the speed of the vehicle.To sum up, some active explorations concerning with monocular video-based distance measurement, vehicle detection and vehicle tracking are made in this paper; the way of the road coordinate system establishment and the method of elimination of noises and shadows have high practical values in real-time systems.