Particle filtering for multitarget detection and tracking

Abstract

This paper presents a particle filter approach to recursively estimating the joint multitarget probability density (JMPD) for the purposes of simultaneous multitarget detection and tracking. The JMPD is a conditional probability density that characterizes uncertainty in both target state and target number given the measurements. Estimation of the JMPD presents a formidable computational challenge due to the high dimensionality of the state space needed to explicitly model the correlations between target states and between target states and target number. We address this challenge with an importance density that is measurement directed and which adaptively factorizes the problem into a set of smaller sub-problems when possible. We demonstrate the algorithm on a set of real targets whose motion is taken from a set of military battle exercises