未知杂波条件下样本集校正的势估计概率假设密度滤波算法

在贝叶斯框架下的多目标跟踪算法中，总是假设杂波的先验信息是已知的。然而，实际应用中，杂波分布一般是未知的，假设的杂波分布往往与实际情况匹配度差，难以保证滤波精度。针对该问题，该文研究了未知杂波势估计概率假设密度(CPHD)滤波算法。首先，提出一种基于狄利克雷过程混合模型(DPMM)类的未知杂波CPHD算法，该算法能够自动选取合适的类数对杂波进行描述，有效降低了杂波空间分布估计的误差。此外，提出样本集校正的思想，并将其引入所提算法，通过去除样本集中由真实目标产生的量测，较好地解决了杂波数过估和目标数低估的问题。与传统算法相比，所提算法的滤波精度更接近于杂波信息匹配情况下的性能，仿真结果验证了其优越性与鲁棒性。

A Cardinalized Probability Hypothesis Density Filter with Unknown Clutter Estimation Using Corrected Sample Set

In multi-target tracking algorithms under the Bayesian filtering framework, it is usually assumed that the priori knowledge of clutter is known. However, in practice, the knowledge of clutter is usually unknown, and the assumption of clutter may not agree with the truth, resulting in the filtering precision declining. For this problem, this paper addresses the problem of Cardinalized Probability Hypothesis Density (CPHD) filter with clutter estimation. Firstly, this paper presents a new CPHD filter with clutter estimation based on Dirichlet Process Mixture Model (DPMM). Thus, this DPMM--CPHD algorithm can reduce the estimation error of the clutter spatial distribution effectively by selecting an appropriate class number. Secondly, to solve the clutter overestimation and cardinality underestimation problems, a correction idea of the sample set via CPHD filter recursion is proposed. By introducing this idea to the DPMM--CPHD algorithm, an improved DPMM--CPHD algorithm is proposed to solve this intractability of errors on clutter number and target number. Simulation results show that the proposed algorithm can effectively estimate the unknown parameters of clutter and has a good performance of multi-target tracking.