Tracking multiple objects using particle filters and digital elevation maps

Abstract

Tracking multiple objects has always been a challenge, and is a crucial problem in the field of driving assistance systems. The particle filter-based trackers have the theoretical possibility of tracking multiple hypotheses, but in practice the particles will cluster around the stronger one. This paper proposes a two-level approach to the multiple object tracking problem. One particle filter-based tracker will search the whole state space for new hypotheses, and when a hypothesis becomes strong enough, it will be passed to an individual object tracker, which will track it until the object is lost. The initialization tracker and the individual object trackers use the same state models and the same measurement technique, based on stereovision-generated elevation maps, and differ only in their use of the estimation results. The proposed solution is a simple and robust one, adaptable to different types of object models and to different types of sensors.