Abstract: Aiming at solving the problem that PF tracking performance reduces in the situation of strong-nonlinear，non-Gaussian and high precision measurements，an improved algorithm is proposed.The algorithm of PF has small calculation but low precision，however，the algorithm of UPF is just opposite.As a result，an improved algorithm is proposed which combines the advantages of the algorithm of PF and UPF.Simulations about PF，UPF and the improved PF are performed.Results show that the tracking precision of the improved algorithm is almost the same as that of UPF，but with as 35% long as the time of UPF.

Key words: particle filtering, unscented particle filtering, posterior probability density function, non-linear filtering

摘要： 在强非线性、非高斯系统、高精度测量的环境下，针对粒子滤波（PF）算法的跟踪性能降低问题，提出一种PF的改进算法。由于PF算法的计算量虽然小但精度不高，而无迹粒子滤波（UPF）算法精度虽然很高但计算量过大，结合PF算法计算量小和UPF算法精度高的优势，提出一种PF改进算法。对PF、UPF和PF改进算法三种跟踪算法进行了仿真，结果表明，改进PF算法的跟踪精度和UPF的跟踪精度相当，但所需运算时间仅为UPF算法的35%左右。

关键词: 粒子滤波, 无迹粒子滤波, 后验概率密度函数, 非线性滤波