Analysis of road image sequences for vehicle counting
Jung Soh, Byung Tae Chun, Min Wang · 2002
This paper describes a computer vision system to count moving vehicles on roads. The system operates on a road image sequence captured during a time period and computes the total number of vehicles which passed the road during the period. Vehicle counting is accomplished by tracking vehicle movements within a tracking zone. The system consists of four image processing modules: (i) road structure extraction, (ii) vehicle detection, (iii) vehicle speed measurement, and (iv) vehicle movement tracking. The system limits the search area for vehicle detection using the road structure information and adjusts the vehicle movement tracking zone based on vehicle speeds. The primary feature of the system is efficient use of acquired knowledge about the road structure and vehicle speeds. The system was tested on a sequence of images of a road that has a medium-level traffic volume.