中文摘要：

在目标跟踪中为达到目标的运行模型与实际轨迹相符，本文提出基于时间序列自适应建模的粒子滤波算法（TS_PF）。采用时间序列方法动态构建预测模型，并将粒子滤波算法中一系列加权粒子以该模型进行状态转移，运用粒子滤波重采样技术，使预测误差进一步减小，预测精度逼近最优估计。仿真实验表明在粒子滤波算法中采用时间序列自适应建模，能够实时反映目标的运行轨迹，克服了单一模型无法准确跟踪动态目标，以及交互多模型需要先验知识的缺陷，提高了动态目标跟踪的准确性。

英文摘要：

In order to make good match for a target moving model and its actual track, time series-basect adaptive modeling in particle filter （TS PF） is presented in this paper. The prediction model is dynami cally made by the time series analysis. The states of weighted particles in particle filter are transferred according to the prediction model. By the resample technique of particle filter, the prediction error is further reduced and the prediction accuracy approximates to the optimal estimation. The simulations show that the time series adaptive modeling in particle filter can make good match with its actual track and overcome the defects of a single model＇ s inaccuracy and IMM＇s apriorism. The accuracy of the dynamic target tracking is improved by TS_PF.