A possibilistic exclusion principle for tracking multiple objects

Tracking multiple targets in a dense environment where there is a high density of false alarms has received recently a wide interest in the literature due to its important repercussions in many radar and military type applications. The challenge arises from how to determine the evidence with which a given measurement is associated to a given track. Typically, the class of algorithms based on joint probabilistic data association filter (JPDAF) investigated by Y. Bar-Shalom and his team is widely explored. In an attempt to use some possibilistic tools for determining the evidence of measurement-track association, we have recently proposed a hybrid approach that keeps track of the structure of JPDAF while one makes use of fuzzy clustering methodology. In this paper. we extend the proposal to the case where the track itself may not be observable, or does not exist. This refers in the literature to track initiation problem. In probability theory, such purpose is handled, for instance, using Integrated probabilistic data association filter (IPDAF) proposed by D. Musicki (1994). Two proposals are put forward, the first one makes use of multiple model approach while the second one is based on adding a second class of noise prototype which captures the lack of perceivability of the target. Some simulation results are performed to prove the feasibility of the proposals