Distributed Multitarget Tracking via Heterogeneous Fusion with 3D Radars and Infrared Sensors

This paper addresses distributed multitarget tracking over a network of heterogeneous sensors. We consider a 3D heterogeneous track-to-track fusion (T2TF) problem using radars and infrared (IR) sensors. The radar trackers use the Cartesian state vector with 3D position and velocity. The IR trackers use modified spherical coordinate (MSC) for the state vector, since use of MSC leads to improved state estimation accuracy compared with Cartesian coordinate. In order to fuse state estimates, the state estimates and associated covariances in MSC are first converted to corresponding quantities in Cartesian coordinates via one-to-one mapping function and the unscented transformation (UT). Different from the single target tracking scenario, the T2TF for multitarget tracking scenario is faced with the problem of associating tracks from different sensors. To this end, we propose a track-to-track association method, at the core of which is a 2D assignment problem taking the generalized covariance intersection (GCI) divergence as the cost function. Results of Monte Carlo simulations highlight the accuracy of the proposed distributed heterogeneous T2TF.