Monitoring pedestrians in a uncontrolled urban environment by matching low-level features

In this paper, the problem of non-rigid objects tracking in a uncontrolled urban environment is discussed. We first present a specific motion detection algorithm used to detect objects in the field of view of a fixed camera, such as pedestrians, vehicles, buses, etc. This real-time motion detection algorithm is based on the construction of a reference edge image of the background, composed of all stationary edges in the scene. The paper then deals with a specific matching procedure developed to extract the 2D motion parameters of the objects. This approach is used to estimate the mean crossing-time of pedestrians at urban road crossings. Finally, results on a real traffic image sequence are given in order to evaluate the performance of the system