论文部分内容阅读
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。
Aiming at the problem that the existing random limited set (RFS) filter seriously degrades the tracking performance of the derivative target in the low signal-to-noise ratio environment, an improved algorithm based on Delta-Extended Label Multi-Bernoulli (δ-GLMB) . Based on the random set theory and the Bernoulli derivative model, a new prediction equation is deduced, and the hypothesis clipping and grouping method and the Dobre-Bernoulli approximation technique are adopted to reduce the computational complexity. Aiming at the problem of false alarm caused by the increase of hypothesis, the multi-frame smoothing idea and algorithm are combined, and the new target is retrospectively processed by label information. The simulation results show that the proposed algorithm can make unbiased estimation of the number of targets, performance is superior to probability hypothesis density (PHD) algorithm in low detection probability and strong clutter environment, and the computational cost grows faster than PHD in the initial stage of derivation. More dispersed than PHD.