MEAP: Approximate optimal estimate extraction for the SMC-PHD filter

Abstract

Multi-estimate extraction (MEE), also referred to as displaying tracks, lies at the core of any multi-target tracking systems, but remains a challenge for the sequential Monte Carlo implementation of the probability hypothesis density (SMC-PHD) filter. In this paper, we recall decision and association techniques to distinguish real measurements of targets from clutter and to associate particles to measurements. The MEE problem is then formulated as a family of parallel single-estimate extraction problems, where the expected a posteriori (EAP) estimator can be employed, namely the multi-EAP (MEAP) estimator. The MEAP estimator is free of iterative clustering computation, computes fast and yields accurate and reliable estimates. Classical simulation scenarios are employed to demonstrate the superiority of the MEAP estimator over existing methods in terms of fast processing speed and best estimation accuracy.