Measurement-driven multi-target tracking filter under the framework of labeled random finite set

Abstract

Multiple clusters of particles characterized in terms of fixed-position and fixed-distribution are utilized for target capture, target state estimation, and track maintenance in traditional particle filters. It is often unsustainable in target acquisition and track maintenance since the prior information of target is hard to acquire. In addition, the obtained measurement information may be fuzzy (interval measurement) in practical engineering applications. This is also a challenge in traditional particle filters. In view of these two problems, a novel measurement-driven AParticle-δ-GLMB-IM filter is proposed. The AParticle-δ-GLMB-IM filter uses RLF to determine the affiliation between the survival targets and the interval measurements. A newborn label or a survival label will be given to the newborn particles driven by the same single interval measurement. Furthermore, this paper presents a new box particle application named ABox-δ-GLMB filter. Numerical experiments verify the effectiveness and superiority of the proposed methods.