A real-time vehicle flow-measuring algorithm for complex urban intersection in the daytime

For the problem of tracking vehicles on freeways using computer vision has been existed and worked well in free-flowing traffic. However, the traffic experts and the department of transportation management are more interested in monitoring the vehicle flow in the complex urban intersections, where the current systems break down for uneven vehicle velocity, congestion, vehicles partially occluded and pedestrians. This paper will develop a real-time vehicle flow-measuring algorithm for complex urban intersection based on multi-detection windows, instead of tracking entire vehicles or extracting the three-dimensional features of the vehicles. For the validity of the multi-detection windows background, we adopt a new background refresh strategy corresponding to the background changing. Traffic light will also be detected before vehicle flow measuring. This real-time measuring algorithm can meet the demand of the practical application.