Improved GM PHD tracking algorithm for neighbor targets

DOI:	10.11809/bqzbgcxb2024.04.014
Keywords:	multi target tracking; probability hypothesis density; weight reallocation; adjacent target tracking
Abstract:	Aiming at the problem that the target tracking system is difficult to make accurate estimation in the neighboring target scenario, an improved neighboring target GM PHD tracking algorithm is proposed. The algorithm effectively avoids the huge iterative burden of clutter on the update step of the algorithm by constructing an adaptive threshold based on the predicted weights and velocity parameters. At the same time, we fully consider the possible distribution of the measurements when the target is in the vicinity, and propose a new correction method of weight allocation for the phenomena of “one to zero” and “one to many” between the target and the measurements. The results of simulation experiments show that the improved algorithm outperforms the traditional algorithm in terms of target number and target state estimation when the target is neighboring, and can significantly improve the tracking accuracy.
Published:	2024-04-30
PDF