A track-oriented multiple hypothesis multitarget tracking algorithm

An algorithm for tracking multiple targets simultaneously in the presence o f false signals is described. The method is an improvement of multiple hypo thesis tracking (MHT) which has recently been reported. MHT tracks multiple targets by maintaining data association hypotheses regarding whether an ob servation vector represents a new target, clutter, or an already-tracked ta rget. The idea of clusters is introduced in MHT, because a large problem in a wide region is thereby divided into independent small problems. However, no theoretical basis for the structure of MHT in which a cluster can be de composed has been published. In this paper, a track is defined by a time se ries of observed vectors; the hypothesis is defined by a combination of tra cks; and a cluster is defined by whether tracks share an observation vector . This is a proposal of a new method of constructing MHT. With this method, the necessary and sufficient conditions for separating a track from a clus ter are described so that a method of imitating tracking of multiple target s is established. The usefulness of the proposed method is confirmed by exa mples. (C) 1999 Scripta Technica.