Abstract:A method using virtual trapline is proposed for traffic flow detection. For improving the detection accuracy, the system sets up several models, such as road model, illumination model, shadow model and relationship model among them to assist detection process. Through introducing the hue saturation degree information, adaptive background updating and motion detection, vehicles can be distinguished from shadow. Binary decision tree is used to analyze pressing line of vehicles to improve the reliability of the system. The problem of the vehicle velocity detection is resolved by one dimension video tracing. The results indicate that the method is better than existing methods in accuracy rate (can be improved to 98%) and real-time performance.

New video detection method for traffic flow