Extended and multiple target tracking: Evaluation of an hybridization solution

In the usual multiple target tracking systems, detections associated to the targets are considered as issued from a single point source. This hypothesis is true if the size of the sensor resolution cells is bigger than the size of the target and if there is only one target in the resolution cell. Due to the increasing resolution capabilities of modern sensors this hypothesis is considered valid for the small targets (like ground vehicles). However, in real situations observed with modern GMTI (Ground Moving Target Indicator) sensors we cannot neglect the sensor resolution phenomenon: for littoral surveillance applications, the large targets (or extended targets) can generate more than one detection at a time; in addition for ground surveillance applications the distance between the individual targets can often be less than the size of the resolution cell which produces only one detection for a group of targets. On those considerations, we must adapt our individual targets' tracking algorithm with the extended and group tracking algorithms. In this paper, we test a very simple hybridization between a multiple target tracking algorithm and the recent bayesian approach for extended object tracking and group tracking represented by a random symmetrical positive definite matrix.