Gibbs Sampling of Measurement Partitions and Associations for Extended Multi-Target Tracking

Abstract

In this paper we propose a novel approach to handle the extended target association problem in multi-target tracking for conjugate priors like the $\delta$-GLMB, LMB, PMBM and MBM. By introducing dependencies between partition cells we are able to employ a Gibbs sampler to simultaneously sample from partitions of the measurement set and association mappings. This formulation allows for a reduction of the approximation error as well as a more efficient implementation.