A Decomposed Probability Hypothesis Density (DPHD) Tracker in a Multi-target Tracking System Verified with a Linear Gaussian Model and a Simulated Laser Tracking System
L-L. Zhao, X-H. Su, P-J. Ma, C-M. Shi and X-L. Dong
The probability hypothesis density (PHD) filter is a practical method to jointly estimate the number and states of multiple targets from the observations from laser and other types of sensors including clutter. Still, the PHD filter cannot provide identities of individual target trajectories. In this paper we propose a decomposed probability hypothesis density (DPHD) multitarget tracker with the track management. This scheme uses PHD components to represent each target distribution respectively instead of PHD itself, and track their motions by associating the PHD components between frames, meanwhile alleviating the false tracks by detecting and pruning possible clutter located at the confirmed targets. Simulation results demonstrate that the proposed PHD tracker has provided more accurate and efficient target trajectory estimating.
Keywords: Probability hypothesis density (PHD) filter, multi-target tracking, decomposed probability hypothesis density (DPHD) filter, state extraction, track continuity, track management, Gaussian, laser tracking