基于标签多伯努利滤波的衍生目标跟踪算法

针对现有随机有限集(RFS)滤波器在低信噪比环境下对衍生目标跟踪性能严重下降的问题,提出了一种基于Delta扩展标签多伯努利(δ-GLMB)滤波器的改进算法。基于随机集理论和伯努利衍生模型,推导了新的预测方程,并采用了假设裁剪及分组手段和多伯努利近似技术以降低算法的计算量。针对假设增多引起的虚警问题,将多帧平滑思想和算法相结合,利用标签信息对新目标进行回溯处理。仿真结果表明,所提算法能对目标数目进行无偏估计,在低探测概率和强杂波环境下性能明显优于概率假设密度(PHD)算法,计算开销在衍生初始阶段增长快于PHD,目标较分散时低于PHD。

Spawned target tracking algorithm based on labeled multi-Bernoulli filtering

For the problem that the performance of existing random finite set (RFS) based filters serious degrades when tracking spawned targets in low signal-to-noise ratio condition, an improved algorithm based on δ-generalized labeled multi-Bernoulli (δ-GLMB) filter is proposed. According to RFS theory and Bernoulli spawning model, a new predicted formulation is derived. Hypotheses pruning and grouping method along with Bernoulli approximation are adopted to cut down the computational load. To reduce the false alarms introduced by extra hypotheses, the multi-frame smooth method is employed through a new target confirmation step with label information. Simulations show that the proposed method can provide unbiased estimation of cardinality, and significant outperforms the probability hypothesis density (PHD) filter in the low detection probability and dense clutter environment. The computational cost of proposed algorithm grows faster than PHD in the early stage of spawning, while is lower than PHD when targets separate well.