Heterogeneous Multi-Sensor Fusion With Random Finite Set Multi-Object Densities

This paper addresses the density based multi-sensor cooperative fusion using random finite set (RFS) type multi-object densities (MODs). Existing methods use scalar weights to characterize relative information confidence among local MODs, and in this way portion of contribution each MOD fused global can be tuned via adjusting these weights. Our analysis shows that mechanism a coefficient oversimplified for practical scenarios, as an is complex usually space-varying due imperfection sensor ability various impacts from surveillance environment. Consequently, severe performance degradation observed when fail reflect actual situation. We make two contributions towards addressing problem. Firstly, we propose novel heterogeneous method perform averaging RFS MODs. By factorizing MODs into number smaller size sub-MODs, it transform original complicated problem much easier parallelizable multi-cluster Secondly, proposed strategy general procedure without any particular model assumptions, further derive detailed equations, with centralized network architecture, both probability hypothesis (PHD) filter multi-Bernoulli (MB) filter. The Gaussian mixture implementations algorithms are also presented. Various numerical experiments designed demonstrate efficacy methods.