一种检测概率未知时的MeMBer跟踪算法  

A MeMBer tracking algorithm with unknown detection probability

作　　者：刘泽为 郑岱堃[1] 袁俊泉[1] 黄亮 LIU Zewei;ZHENG Daikun;YUAN Junquan;HUANG Liang(Air Force EarlyWarning Academy,Wuhan 430019,China)

机构地区：[1]空军预警学院,武汉430019

出　　处：《空天预警研究学报》2022年第4期235-241,共7页JOURNAL OF AIR & SPACE EARLY WARNING RESEARCH

基　　金：国家自然科学基金青年科学基金项目(62001512);军队“双重”建设项目。

摘　　要：针对检测概率未知的多目标跟踪场景中跟踪性能会受目标检测概率估计结果影响的问题,提出了一种对未知检测概率实时估计的MeMBer跟踪算法.采用Beta分布对未知检测概率进行建模,然后通过测量分类来区分目标性质并实现对检测概率的实时估计,利用多伯努利随机有限集跟踪算法实现多目标跟踪.仿真结果表明,在检测概率未知情况下该算法有良好的多目标跟踪效果,能对未知检测概率进行实时估计;与传统MeMBer算法相比,该算法在低检测概率下的检测概率估计精度更高且更加稳定.Aiming at the problem that the tracking performance in the multi-target tracking scene with unknown detection probability is affected by the estimation result of target detection probability, this paper proposes a MeMBer tracking algorithm for real-time estimation of unknown detection probability. The paper first uses Beta distribution to model the unknown detection probability, and then employs the measurement classification to distinguish the target properties and realize the real-time estimation of the detection probability. Finally, the paper uses multi-Bernoulli random finite set tracking algorithm to realize multi-target tracking. The simulation results show that when the detection probability is unknown, the proposed algorithm has good multi-target tracking effect, and can estimate the unknown detection probability in real time, and that compared with the traditional MeMBer algorithm, the detection probability estimation accuracy of this algorithm is higher and more stable under low detection probability.

关 键 词：检测概率模型 多目标跟踪 随机有限集 多目标多伯努利滤波器 BETA分布 

分 类 号：TN957[电子电信—信号与信息处理]