Multi-sensor multi-target tracking using domain knowledge and clustering

This paper proposes a novel joint multi-target tracking and track maintenance algorithm over a sensor network. Each sensor runs a local joint probabilistic data association (JPDA) filter using only its own measurements. Unlike the original JPDA approach, the proposed local filter utilises the detection amplitude as domain knowledge to improve the estimation accuracy. In the fusion stage, the DBSCAN clustering in conjunction with statistical test is proposed to group all local tracks into several clusters. Each generated cluster represents the local tracks that are from the same target source and the global estimation of each cluster is obtained by the generalized covariance intersection (GCI) algorithm. Extensive simulation results clearly confirms the effectiveness of the proposed multisensor multi-target tracking algorithm.