中文摘要：

核函数粒子滤波（KPF）是小噪声动态系统目标跟踪的一种有效方法，核窗宽选择是该方法中核密度估计的核心问题。本文提出了一种基于协方差的变窗宽核粒子滤波算法。该方法首先通过粒子集的协方差矩阵估计粒子的粗略核窗宽和其粗略的后验概率密度，然后调节全局核窗宽获得适用于每一个粒子的精确核窗宽，提高核密度的估计精度；然后，通过迭代寻找后验概率模型，使得粒子集能够在核密度估计后向后验概率密度的真实分布移动，从而提高跟踪精度。通过这种方法生成的新粒子是对后验概率密度的一个更加近似的表达。实验结果表明，在小噪声动态系统中，本文提出的变窗宽核函数粒子滤波在光电目标跟踪的性能和效率（PF的20％粒子数目）上都优于传统的粒子滤波（PF）、UPF（Unscented Particle Filter）以及KPF方法。

英文摘要：

Kernel Particle Filter （KPF） is an effective method for target tracking of a dynamic system with small noises, in which the selection of the kernel bandwidth is a critical step of Kernel Density Estimation（KDE） in KPF. In this paper, a Variable Bandwidth Kernel Particle Filter （VBKPF） based on covariance matrix is proposed. Firstly, the covariance matrix of particle sets is used to compute the coarse bandwidth and the coarse posterior Probability Density Functions （PDFs）. Then, each particle can acquire its own accurate bandwidth by adjusting the global kernel bandwidth to improve the precision of the KDE. Finally, to get a more effective particle allocation, the variable bandwidth KDE in the VBKPF is used to approximate the PDFs by moving particles toward the posterior, which gives a closed-form expression of the true distribution. Experimental results show that the proposed VBKPF performs better than the standard particle filter（PF）, Unscented Particle Filter（UPF） and the Kernel Particle Filter（KPF） both in efficiency（20% particle number of PF）and estimation precision for optoelectronic target tracking systems.