In this paper, an automatic vehicle counting and classification system based on image processing is presented. In this method, by defining virtual loops and detecting vehicles from the image background, vehicles are counted according to the fullness or emptiness of the virtual loops.
It is worthy of mention that the proposed algorithm shows a good performance when a vehicle passes simultaneously on the two virtual loops or a vehicle that occupies only a small part of the virtual loop. Vehicles are classified into five distinct classes without any need for calibration. The proposed method has high speed because of no need to track objects in the counting stage and can be implemented in real time.
Link of paper: