摘要
针对传统粒子滤波的数据融合和粒子贫乏问题,提出一种结合非参数信念传播和粒子滤波(NBP-RPF)的分布式WSN目标跟踪方法。首先检测目标的节点,然后对检测数据进行核密度估计(KDE)得到目标估计信息,最后,通过非参数信念将信息传播到簇首节点,簇首节点对信息乘积进行Gibbs采样和正则化粒子滤波,实现了对目标的精确跟踪。仿真结果表明,NBP-RPF法在增加粒子多样性和有效融合数据等方面具有优势,同时也提高了目标的跟踪精度。
Aiming at the problem of data fusion and particle degeneracy in traditional particle filtering,the target tracking method of distributed WSN based on the combination of nonparametric belief propagation and regularized particle filtering（NBP-RPF） is proposed.First,the node of target is detected,then kernel density estimation（KDE） is conducted for detected data and the target estimation information is obtained.Finally,the information is transmitted to the cluster head node;the product of information is Gibbs sampled and regulation particle filtered for realizing precise tracking.The result of simulation indicates that the method proposed possesses superiority in increasing particle diversity and effective fusion data;and enhances the accuracy of target tracking.
出处
《自动化仪表》
CAS
北大核心
2011年第1期19-22,共4页
Process Automation Instrumentation
基金
广东省自然科学基金资助项目(编号:9151052101000013)
茂名市重点科技计划资助项目(编号:20091010)
关键词
无线传感器网络
非线性模型
滤波
均方根误差
多样性
Wireless sensor network（WSN） Nonlinear model Filtering Root-mean square error Diversity
