The histogram Poisson, labeled multi-Bernoulli multi-target tracking filter

Abstract

A Random Finite Set (RFS) based multi-target filter is proposed, which utilizes a labeled Multi-Bernoulli distribution to model the multi-target state, together with a Poisson RFS distribution to model target birth. Referred to as the Poisson Labeled Multi-Bernoulli (PLMB) filter, results show that, in simulated environments, it outperforms the Labeled Multi-Bernoulli (LMB), δ-Generalized Labeled Multi-Bernoulli (δ-GLMB) and Labeled Multi-Bernoulli Mixtures (LMBM) filters under general target birth scenarios. An algorithm based on a histogram of Gibbs samples is also proposed which efficiently generates a posterior labeled Multi-Bernoulli distribution in a simple manner using a histogram of the state-measurement associations obtained by a Gibbs sampler. The histogram approach is readily applicable to all Multi-Bernoulli based filters and is demonstrated in the form of the Histogram-PLMB (HPLMB) filter.