An improved probability hypothesis density filter for multi-target tracking

Abstract

For the problem that the standard probability hypothesis density is unable to correctly estimate the states of targets and their number when tracking multiple targets in possible missed detection environments, an improved probability hypothesis density filter based multi-target tracking algorithm is proposed under the linear Gaussian conditions. Two assisted parameters, namely label and probability of existence, are introduced to expand the standard target state in the proposed algorithm which includes three robust schemes compared with the Gaussian mixture probability hypothesis density filter. Firstly, the extended component set of target states representing the target intensity can be correctly updated in the proposed target intensity update scheme. Secondly, by optimizing the component set that approximates the target posterior intensity, the excess and invalid components are effectively reduced in the improved component fusion scheme. Lastly, a new target state extraction scheme is able to accurately estimate the states of targets by comprehensively utilizing both the weight and existent probability of the target. Simulation results show that the proposed algorithm not only provides relatively accurate multi-target estimates, but also has a relatively low computational burden.