Multi-object Tracking for Generic Observation Model Using Labeled Random Finite Sets

摘要

This paper presents an exact Bayesian filtering solution for the multi-objecttracking problem with the generic observation model. The proposed solution isdesigned in the labeled random finite set framework, using the product styledrepresentation of labeled multi-object densities, with the standardmulti-object transition kernel and no particular simplifying assumptions on themulti-object likelihood. Computationally tractable solutions are also devisedby applying a principled approximation involving the replacement of the fullmulti-object density with a labeled multi-Bernoulli density that minimizes theKullback-Leibler divergence and preserves the first-order moment. To achievethe fast performance, a dynamic grouping procedure based implementation ispresented with a step-by-step algorithm. The performance of the proposed filterand its tractable implementations are verified and compared with thestate-of-the-art in numerical experiments.