Positioning sensors, video tool for counting pedestrian

This paper describes a method to detect, count and calculate the speed and the direction of pedestrians in urban areas or buildings using a video system. This system is called OVICOP^(c) (Outil Video pour le Comptage de Personnes). It processes a picture from a CCD camera using image processing. This camera is set vertically to the spot where the counting has to be achieved. The layout of the camera represents a critical step, the best results have been achieved in a constantly lit environment, with low luminescence variations and the camera axis set as near to vertical as possible. However, our system is robust. It is only slightly sensitive to external light variations. An algorithm has been developed which achieves a low computational time. Heavy calculations such as speed detection, direction and counting are only performed at regular intervals or when something moves in the shot (e.g. a pedestrian). The system allows one to ascertain the direction and the speed of pedestrians with only one matrix camera. Furthermore, it is much cheaper than the other counting methods based on video