The traffic tracking is very important in intelligent transportation systems (ITS), and more and more vision systems have been widely installed in monitoring traffic. The video image processing and computer vision technologies can not only conveniently detect the common parameters of traffic flows (e.g. velocity and flow of the traffic) like traditional methods, but also retrieve more information from traffic video images (e.g. traveling trajectory). Most commercial region-based traffic tracking systems lack the capability in object segmentation, especially in heavy traffic situations, where partial image of one vehicle is covered by another. The system described in this paper is a real-time traffic tracking system based on a feature detection method. During the tracking process, a spatial filtering velocimeter is devised to detect the velocity of the entire traffic flow, which is of very great importance in tracking. It alleviates the computation, which is one of the main obstacles of many tracking algorithms. Based on this, other parameters like velocity, traffic flow, trajectory and vehicle type can be easily detected. In this paper, the outline of this real-time traffic tracking system is first introduced, then, three main parts of this system, especially spatial filtering used in the traffic tracking, are described. Finally the testing results of a simulation experiment are given.
Abstract