摘要
概率假设密度(PHD)滤波是实现多目标跟踪(MTT)的一种有效算法,然而由于其推导过程使用了泊松近似,因而带来了较大的势估计误差。提出了最大熵模糊多目标粒子‑PHD(MEFMOP‑PHD)滤波算法,利用最大熵模糊聚类算法构造联合概率数据关系(JPDA)算法中的关联概率矩阵,同时结合标签技术与聚矩阵对关联概率矩阵进行修正,从而使算法在节省运算量的同时达到较好的跟踪效果。该滤波算法对PHD密度的近似采用了高斯混合(GM)形式,并通过仿真试验与传统的高斯混合滤波算法进行了比较,最终验证了算法在低检测概率下的优越性。
Probabilistic hypothesis density(PHD)filter is an efficient algorithm for multi-target tracking(MTT).However,due to the Poisson approximation used in its derivation,the potential estimation error is larger.Maximum entropy fuzzy multi-objective particle‑probabilistic hypothesis density(MEFMOP‑PHD)filtering algorithm is proposed,using the maximum entropy fuzzy clustering algorithm to construct the associated probability matrix of joint probability data association(JPDA)algorithm,and the associated probability matrix is modified by combining the tag technology with polymer matrix,so that the algorithm can achieve a better tracking effect while saving computation.The filtering algorithm approximates PHD density in the form of Gaussian mixture(GM).Compared with the traditional Gaussian mixture filtering algorithm through simulation experiments,the algorithm has advantage in low detection probability environment.
作者
吴娜
WU Na(College of Internet of Things,Nanjing University of Posts and Telecommunications,Nanjing 210003,China)
出处
《指挥信息系统与技术》
2020年第2期27-32,共6页
Command Information System and Technology
基金
南京邮电大学科研启动基金(XK0160919133)资助项目。
关键词
多目标跟踪
概率假设密度
最大熵模糊聚类
关联概率矩阵
multi-target tracking
probabilistic hypothesis density
maximum entropy fuzzy clustering
association probability matrix
