基于多伯努利概率假设密度的扩展目标跟踪方法

摘要

Extended targets usually generate multiple measurements in high resolution radar systems. Existing algorithms of the Random Finite Set (RFS) assume that the measurement number of extended targets follows Poisson distribution in a general way. However, this assumption is inconsistent with actual situations. Considering this issue, a Multi-Bernoulli Extended Target Probability Hypothesis Density (MB-ET-PHD) tracking method is proposed. First, this method assumes that the measurement number of extended targets is Multi-Bernoulli (MB) distributed. Then, its update equation is derived by using the FInite Set STatistics (FISST) multi-target calculus. Finally, simulated results of Gaussian Mixture (GM) framework are given. The simulation results show that the proposed method can obtain better tracking performance compared with the Poisson ET-PHD method.%高分辨率雷达系统中，扩展目标一般会产生多个量测。现有随机有限集(RFS)类算法一般假定扩展目标的量测数目服从泊松分布，然而这个假设与实际情况不符。针对这一问题，该文提出一种多伯努利扩展目标概率假设密度(MB-ET-PHD)跟踪算法。该算法首先假设扩展目标的量测数目服从多伯努利分布，然后通过有限集统计(FISST)理论的多目标微积分推导得到校正等式，最后给出了高斯混合(GM)框架的仿真结果。仿真结果表明该算法能够获得比泊松ET-PHD算法更好的跟踪性能。