Pairwise马尔科夫模型下的势均衡多目标多伯努利滤波器

由于在实际应用中目标模型不一定满足隐马尔科夫模型（Hidden Markov model，HMM）隐含的马尔科夫假设和独立性假设条件，一种更为一般化的Pairwise马尔科夫模型（Pairwise Markov model，PMM）被提出.它放宽了HMM的结构性限制，可以有效地处理更为复杂的目标跟踪场景.本文针对杂波环境下的多目标跟踪问题，提出一种在PMM框架下的势均衡多目标多伯努利（Cardinality balanced multi-target multi-Bernoulli，CBMeMBer）滤波器，并给出它在线性高斯PMM条件下的高斯混合（Gaussian mixture，GM）实现.最后，采用一种满足HMM局部物理特性的线性高斯PMM，将本文所提算法与概率假设密度（Probability hypothesis density，PHD）滤波器进行比较.实验结果表明本文所提算法的跟踪性能优于PHD滤波器.

Cardinality Balanced Multi-target Multi-Bernoulli Filter for Pairwise Markov Model

Because the Markovian and independence assumptions, which are implicitly implied in hidden Markov model (HMM), may not be satisfied by the target model in some practical applications, a more general pairwise Markov model (PMM) has been proposed. PMM relaxes the structural limitations of HMM and can effectively deal with more complex target tracking scenarios. In this paper, a cardinality balanced multi-target multi-Bernoulli (CBMeMBer) filter in the framework of PMM is proposed for multi-target tracking in clutter environment, and a closed-form solution to the CBMeMBer filter under linear Gaussian PMM is presented. Finally, the proposed algorithm is compared with the probability hypothesis density (PHD) filter via simulations using a particular linear Gaussian PMM, which keeps the local physical properties of HMM. Simulation results show that the tracking performance of the proposed algorithm is better than that of the PHD filter.