Abstract: In order to solve nonlinear global optimization problems of particle filter, on the basic of small weighted particles are removed and the number high weighted particles are increased by re-sampling, Particle Filter of Neighborhood Iteration Re-sampling（NIRPF） is proposed for tracking. Firstly, particles are predicted and Sequential Importance Sampling（SIS） is used for particle empowering value. Then, in the process of searching a high probability of posterior probability density, the position of single particle is updated. Gaussian weighted neighborhood search is adopted for weighting all the particles iteratively. Finally, the current status is estimated. Bearings target tracking problem involves two static observations and two types of targets non-motorized and motorized. The effectiveness of proposed method is verified by Monte Carlo simulation results. The proposed method has a faster initial convergence speed in comparison with square-root cubature Kalman filter, cubature particle filter and random search-particle filter, and the evaluation of Root Mean Square Error（RMSE） and Root Time Averaged Mean Square（RTAMS） on non-maneuvering target and maneuvering is much better.

Key words: particle filter, target tracking, re-sampling, Gaussian weighted neighborhood search, Sequential Importance Sampling（SIS）

摘要： 为了解决粒子滤波的非线性全局优化问题，基于重采样的思想是移除权重小的粒子，增加权重大的粒子数量，提出利用邻域搜索重采样的粒子滤波（NIRPF）进行目标跟踪。首先，预测粒子，并利用重要序列采样（SIS）给粒子赋权值；然后，在搜索后验概率密度的高概率区过程，更新单个粒子位置，利用高斯-邻域搜索迭代地加权所有粒子；最后，进行当前状态的估计。纯方位目标跟踪问题涉及两个静态观察器和非机动和机动两类目标。蒙特卡罗仿真结果验证了提出方法的有效性，与均方根容积卡尔曼滤波、容积粒子滤波和随机搜索的粒子滤波相比，提出的方法拥有更快的初始收敛速度，非机动目标和机动目标的根均方误差（RMSE）和时间根均方差（RTAMS）的评估更优。

关键词: 粒子滤波, 目标跟踪, 重采样, 高斯-邻域搜索, 重要序列采样