A probabilistic exclusion principle for tracking multiple objects

Tracking multiple targets is a challenging problem, especially when the tar gets are "identical", in the sense that the same model is used to describe each target. In this case, simply instantiating several independent 1-body trackers is not an adequate solution, because the independent trackers tend to coalesce onto the best-fitting target. This paper presents an observati on density for tracking which solves this problem by exhibiting a probabili stic exclusion principle. Exclusion arises naturally from a systematic deri vation of the observation density, without relying on heuristics. Another i mportant contribution of the paper is the presentation of partitioned sampl ing, a new sampling method for multiple object tracking. Partitioned sampli ng avoids the high computational load associated with fully coupled tracker s, while retaining the desirable properties of coupling.