Multi-Sensor Joint Detection and Tracking with the Bernoulli Filter

Abstract

This paper proposes a filter for joint detection and tracking of a single target using measurements from multiple sensors under the presence of detection uncertainty and clutter. To capture the target presence/absence in the surveillance region as well as its kinematic state, we represent the target state as a set that can take on either the empty set or a singleton. The uncertainty in such a set is modeled by a Bernoulli random finite set (RFS), and Bayes optimal estimators for joint detection and tracking are presented. A closed-form solution for the linear-Gaussian model is derived and an analytic implementation is proposed for nonlinear models based on the unscented transform. We apply the technique to tracking targets constrained to move on roads with time difference of arrival/frequency difference of arrival (TDOA/FDOA) measurements.