Motion estimation and tracking for urban traffic monitoring

Abstract

The use of traffic monitoring techniques based on image processing algorithms for supervising urban vehicle flows could be very useful. Classical inductive loops can only compute traffic density on a single lane but are unable to estimate, for example, the behaviour of the vehicle flow at a crossroad. In this paper a system that estimates the turning rates at an urban crossroad by processing the sequence of images taken by a videocamera is presented. Block matching motion estimation, segmentation, and moving object tracking techniques are used. The good results obtained are presented and their relation to the camera position is discussed.