Fusion of Asynchronous Radar and Infrared Sensors Data on a Moving Platform Using the PHD Filter

We consider heterogeneous asynchronous sensor fusion of radar and infrared (IR) data on a moving platform. Existing work seldom considers heterogeneous and asynchronous problems simultaneously. First, we build models for radar and IR sensors. Since they have different coordinates, dimensions and data rates, some valid solutions are proposed in this paper. The unscented transformation (UT) is used to unify their coordinates. We use a batch processing method to handle the asynchronous arrival time. For efficient multitarget tracking (MTT), we extend the heterogeneous trackers into the Probability Hypothesis Density (PHD) filter even though their models are distinct. In view of the unknown dependence between sensors on the same platform, the local posteriors are fused using the Generalized Covariance Intersection (GCI) rule. Furthermore, combined with the distinct models, a concrete form of heterogeneous asynchronous GCI rule is derived. Finally, numerical results confirm the excellent performance of the proposed fusion scheme.