Multi-hypotheses tracking algorithm based on the belief theory

The aim of this paper is to enhance the research works developed since 1998 by Gruyer and Royere. They established a tool for detections of vehicles and tracking based on the belief theory. Their approaches focus on the use of generalized Dempster's combination rules in the extended open world and they provide two partial views of the situation with two belief matrices. The first one represents the belief that targets are associated with tracks and the second one represents the belief that tracks are associated with targets. To take the final decision they used the Hungarian algorithm which gives the better associations between targets and tracks according to the criteria of maximal global belief. But this global association can lead to some suspicious local associations. This paper provides a solution in order to avoid theses disturbances with the development of a new MHT algorithm based on belief theory.