摘要： 标准粒子滤波算法用于无线传感器网络运动目标跟踪时，非高斯噪声环境会降低其跟踪精度和计算效率。针对该问题，结合多传感器测量模型和Kullback-Leibler距离（KLD）采样方法，提出一种自适应粒子滤波算法。在满足预设阈值条件时，引入补偿函数对重要性概率密度函数（IPDF）进行迭代更新，同时利用具有自适应退火参数的模拟退火算法使粒子快速接近高似然区域。在此基础上，结合KLD采样动态调整粒子规模，在保证跟踪精度的同时减少运算量。仿真结果表明，与KLD-PF算法相比，该算法的IPDF分布接近真实后验概率密度分布，跟踪精度较高，能够在不同参数的非高斯噪声下进行有效跟踪。

关键词: 自适应粒子滤波, Kullback-Leibler距离采样, 目标跟踪, 无线传感器网络, 模拟退火算法

Abstract: When standard Particle Filtering(PF) algorithm is applied to the Wireless Sensor Network(WSN) based moving object tracking,the non-Gaussian noise environment can reduce its tracking accuracy and calculation efficiency.So,this paper proposes an adaptive PF algorithm based on multi-sensor measurement model and Kullback-Leibler Distance(KLD) sampling.After the preset threshold condition is met,this paper introduces the compensation function to iteratively update the Importance Probability Density Function(IPDF).Then,this paper uses the simulated annealing algorithm with adaptive annealing parameters to make particles move fast to the high likelihood region.On this basis,the KLD sampling is used to adjust the particle size,which reduces the amount of computation while maintaining the tacking accuracy.Simulation results show that compared with the KLD-PF algorithm,the IPDF distribution of the proposed method is closer to the true posterior probability density distribution.Besides,its tracking accuracy is higher,which enables an effective tracking under different parameters of non-Gaussian noise.

Key words: adaptive Particle Filtering(PF), Kullback-Leibler Distance(KLD) sampling, target tracking, Wireless Sensor Network(WSN), simulated annealing algorithm

中图分类号: 

• TN971