Multi-type vehicles´ traffic data collection using video processing

A multi-type vehicles´ traffic data collection algorithm based on video processing is presented. After traffic scene calibration and coordinate transformation, adaptive background subtraction is used to detect vehicles, and several geometric parameters of vehicles are obtained for classification and vehicles are tracked for speed. This algorithm records flow and speed of multi-type vehicles. The result of experiments shows that the accuracy of vehicle counting and vehicle classification is 90.7% and 90% respectively, which indicates that the multi-type vehicles´ traffic data collected by this algorithm are reliable.