Random Finite Sets and Sequential Monte Carlo Methods in Multi-Target Tracking

摘要

Random finite sets provide a rigorous foundation for optimal Bayes multi-target filtering. The major hurdle faced in Bayes multi-target filtering is the inherent computational intractability of the method. Even the Probability Hypothesis Density (PHD) filter, which propagates only the first moment (or PHD) instead of the full multi-target posterior, still involves multiple integrals with no closed forms. In this paper, the authors highlight the relationship between the Radon-Nikodym derivative and set derivative of random finite sets that enable a Sequential Monte Carlo (SMC) implementation of the optimal multi-target filter. In addition, a generalized SMC method to implement the PHD filter also is presented. the SMC PHD filter has an attractive feature -- its computational complexity is independent of the (time- varying) number of targets.