量化量测下的多目标变分贝叶斯PHD滤波.PDFVIP

中国青岛 第七届中国信息融合大会 Qingdao China 2015 年10 月 The Seventh Chinese Information Fusion Conference 46 量化量测下的多目标变分贝叶斯PHD 滤波 1 1 1 王明杰，姬红兵，张永权 1）西安电子科技大学，西安 710071 摘 要：为了满足数字传输和有限带宽的要求，原始量测在发送到信息处理中心之前需要进行量化。在此基础上，研究了量 化量测下的多目标跟踪问题，提出了变分贝叶斯量化概率假设密度(VBQPHD)滤波。该方法可联合估计量化量测噪声方差和目 标状态，并采用一个方差上界确定量化噪声方差估计的有效性。实验结果表明，对于量化量测下的多目标跟踪问题，本文所 会 提方法的跟踪性能要优于传统的滤波方法。 关键词：多目标跟踪；量化量测；变分贝叶斯；概率假设密度 中图分类号：TN953 文献标识码：A 大 VBPHD for Multi-target tracking with quantized measurement 合 1 1 1 Mingjie Wang ，Hongbing Ji ，Yongquan Zhang 1）Xidian University, Xi’an 710071 融 Abstract ：To meet the requirements of digital transmission with limited bandwidth, the raw measurements need to be quantized before transmitted to the information center. However, the quantized noise will degrade the target tracking performance. To t ackle this problem, multi-target tracking with quantized measurements is investigated in this paper, and the variational Bayesian quantized 息 probability hypothesis density (VBQPHD) filter is proposed. Both the variance of the quantized measurement noise and system state are jointly estimated, and the effectiveness of the variance estimate is verified by a known upper bound. Simulation results show that the proposed algorithm outperforms conventional quantized PHD method for multi-target tracking wi