default Multi-sensor extended target tracking


Multi-sensor extended target tracking.Vivone, Gemine; Braca, Paolo. CMRE-FR-2017-002. June 2017.

In this report, the multiple sensor extended target tracking problem is studied into the random matrix framework. We deal with data from a network of multiple sensors and focus on the spatial density modelling. The multiple sensor extension for extended target tracking is not straightforward. Indeed, the extent of the target differs when it is observed from different perspectives. The scope of this report is limited by the assumption that there are no clutter measurements and there is exactly one target present in the surveillance area. Four different multi-sensor measurement updates are presented; three updates based on a parametric density representation of the extended target state distribution, and one update based on a Rao-Blackwellized (RB) particle representation of the extended target state distribution. The updates are evaluated and compared in an extensive simulation study using two kinds of simulators evaluating the performance varying the numbers of particles, the numbers of detections per target, the numbers of sensors, and the noise levels. Results show that the RB particle filter gives the best performance, at the price of higher computational cost.