Tracking Multiple Maneuvering Targets Hidden in the DBZ Based on the MM-GLMB Filter
Weihua Wu, Hemin Sun, Yichao Cai, Surong Jiang, Jiajun Xiong
Abstract
Tracking multiple maneuvering targets hidden in the Doppler blind zone (DBZ) is a challenging problem. To overcome the complicated problem, we proposed a tracker based on the multiple model probability hypothesis density (MM-PHD) filter. However, the PHD filter is only the first-order moment approximation of the multi-target Bayesian filter, and it cannot output track labels in principle. In order to improve the tracking performance, another novel tracker is proposed in this paper. To track multiple maneuvering targets, the proposed tracker is built on the latest multiple model generalized labeled multi-Bernoulli (MM-GLMB) filter. Moreover, it incorporates the minimum detectable velocity (MDV) to suppress the DBZ masking. Finally, a measurement-driven adaptive track initiation is introduced to address the fixed track initiation problem of the standard MM-GLMB filter. It is demonstrated through numerical examples that the proposed tracker outperforms the existing work significantly, especially in terms of both accuracy and robustness of cardinality estimation.