Generalised distance partitioning for multiple-detection tracking filter based on random finite set

When a target generates multiple detections, standard tracking algorithms are not applicable. Unlike extended target tracking, which has received considerable attention, studies on generalised multiple-detection tracking (MDT) are still scarce. MDT algorithms proposed in recent years suffer from high computational complexity. In this work, based on distance partitioning, the authors propose a generalised distance partitioning (GD partitioning) algorithm to reduce the number of partitions, and thus decrease the computational complexity of MDT filters. In the algorithm, after defining a multi-detection distance and calculating the corresponding distance threshold, the authors calculate the original partition of measurement set by using distance partitioning. Subsequently, on the basis of original partition, they can obtain informative subsets to replace all subsets in a multiple-detection Bernoulli filter, as well as a collection of partitions to replace all partitions in multiple-detection probability hypothesis density filter. Simulations show that the algorithm can notably reduce the number of subsets and partitions. Thus, by means of a partitioning algorithm, computational complexity of random finite set-based MDT filters is decreased successfully, which implies that a GD partitioning algorithm may play an important part in real-time MDT systems.