未知量测噪声分布下的CBMeMBer算法  

CBMeMBer Algorithm Under Unknown Measurement Noise Distribution

作　　者：胡建旺 董青 张铁 吉兵 HU Jian-wang;DONG Qing;ZHANG Tie;JI Bing(Shijiazhuang Campus of Army Engineering University,Shijiazhuang 050003,China;North Lianchuang Communiction Limited Company,Nanchang 330000,China)

机构地区：[1]陆军工程大学石家庄校区,石家庄050003 [2]北方联创通信有限公司,南昌330000

出　　处：《火力与指挥控制》2019年第6期111-115,共5页Fire Control & Command Control

摘　　要：针对目标跟踪过程中量测噪声概率分布等先验知识无法准确获取的问题,提出一种基于风险评估的势均衡多目标多伯努利(RE-CBMeMBer)滤波算法。采用CBMeMBer算法的序贯蒙特卡洛实现,在粒子预测后利用风险函数和评估函数计算粒子风险值,并用评估结果更新粒子权值。避免了计算似然函数且不依赖量测噪声的概率分布。仿真表明:与SMC-CBMeMBer算法相比,RE-CBMeMBer算法具有更好的实时性,特别是当量测噪声分布未知时,具有更高的跟踪精度和稳定性。Aiming at the problem of probabilistic distribution of measurement noise and so on are unable to be acquired accurately,a Risk Evaluation-based Cardinality Balanced Multi-Target Multi-Bernoulli(RE-CBMeMBer)filter algorithm is proposed.The algorithm adopts the Sequential Monte Carlo implementation of CBMeMBer algorithm,calculates the particle risk value using the risk function and evaluation function after particle prediction,and updates the particle weight value with the evaluation result.The algorithm avoids the likelihood function calculation and does not depend on the probabilistic distribution of measurement noise.The simulation results show that RE-CBMeMBer filter algorithm has better real-time performance in comparison with SMC-CBMeMBer filter algorithm,especially the algorithm has higher robustness and stability in unknown measurement noise distribution environment.

关 键 词：目标跟踪 随机集 概率分布 势均衡多目标多伯努利 风险评估 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]