Multiple-detection multi-target tracking with labelled random finite sets and efficient implementations

Traditional multi-target tracking algorithms assume that each target can generate at most one detection per scan. However, a target may produce multiple detections (MDs) in many practical applications, e.g. over-the-horizon radar (OTHR), tracking for extended target and tracking with multiple sensors. In this study, the authors propose a new algorithm for tracking multiple targets with MD observation. The proposed technique is based on the labelled random finite set (RFS), which estimates the number of targets and the trajectories of their states. Furthermore, they propose two methods, pre-partition and joint partition, to implement the labelled RFS density recursion. The joint partition method derives a joint prediction, partition and update formulation of the MD filtering and extends the Gibbs sampler to provide global optimal results with low computational cost. The authors' algorithm is demonstrated on an OTHR simulation compared with the previous approach, such as the MD probability hypothesis density filter and the multipath cardinality-balance multi-target multi-Bernoulli filter.