The Research on Vehicle Flow Detection in Complex Scenes

Vehicle flow detection plays an important role in ITS. In the process of vehicle flow detection, the vehicle is contiguous with another and the same vehicle counted repeatedly are common problems, especially the problem of changing lanes, which is very difficult to solve. This paper uses the method that combines background difference and virtual-loop sensor to detect vehicle flow, which based on the fast adaptive background updating method, and it introduces a new method of multi-lane sharing one detecting region. Experiment results show that this method can update the background exactly along with the variance of illumination; it also can solve vehicle orientation and vehicle changing lane problems. In real traffic scenes, it can detect vehicle flow accurately and meet the real-time processing demand.