Three formalisms of multiple hypothesis tracking

This paper describes three mathematical formalisms for defining multiple target tracking (MTT) problems and deriving multiple hypothesis tracking (MHT) algorithms. The three formalisms are based on (1) random finite sequences, (2) finite point processes, and (3) random finite sets, respectively. The main objective of this paper is to show the theoretical foundations of MHT, emphasizing equivalence among these three possible approaches, and discussing some subtle differences among them. We will also show that the MHT algorithms can be derived on solid theoretical foundations, and are not necessarily heuristically derived.