| 摘要: |
| 针对现有交互多模型箱粒子滤波(Interacting Multiple Model Box Particle Filter,IMMBPF)算法在区间量测目标跟踪过程中模型切换和跟踪精度方面的不足，结合自适应交互多模型算法，提出了一种自适应交互多模型箱粒子滤波(Adaptive IMMBPF,AIMMBPF)算法。该算法利用模型似然后验信息构建修正因子，并结合阈值对马尔可夫转移概率矩阵进行自适应修正，使得匹配模型的概率快速增大，并且可以减小非匹配模型的影响，即使在目标运动模型先验信息不足或者不准确情况下，也能对模型转移概率进行自适应更新。对于量测常受到未知分布和偏差的区间误差所影响而呈现区间形式的问题，将箱粒子代替普通粒子，拟合后验概率密度从而进行滤波。仿真结果表明，相比于原有算法，该算法在区间量测机动目标跟踪的应用中，拥有更优的模型匹配度和目标跟踪精度。 |
| 关键词: 机动目标跟踪 箱粒子滤波 自适应交互多模型 区间量测 转移概率矩阵 |
| DOI：10.20079/j.issn.1001-893x.221206008 |
|
| 基金项目:宁夏自然科学基金资助项目（2021AAC03226） |
|
| An Adaptive IMMBPF Algorithm for Maneuvering Target Tracking with Interval Measurement |
| ZHANG Jungen |
| (School of Electrical and Information Engineering,North Minzu University,Yinchuan 750021,China) |
| Abstract: |
| In order to solve the defect of model switching and tracking accuracy of the existing Interacting Multiple Model Box Particle Filter(IMMBPF) algorithm in interval measurement target tracking,combined with Adaptive Interacting Multiple Model(AIMM) algorithm,an adaptive IMMBPF(AIMMBPF) algorithm is proposed.The algorithm uses posteriori information of model likelihood to construct a coefficient and combines the threshold value to adaptively modify the Markov probability transition matrix,which makes the probability of the matching model increase rapidly,and can reduce the effects of the mismatched model.Even if the prior information of the target motion model is insufficient or inaccurate,the model transition probability can be updated adaptively.To solve the problem that the measurement is often affected by the interval error of unknown distribution and deviation and can be expressed in interval form,box particles are used instead of ordinary particles to fit a posterior probability density for filtering.The simulation results show that the adaptive IMMBPF algorithm has better model matching performance and target tracking accuracy than the original algorithms in the application of maneuvering target tracking with interval measurement. |
| Key words: maneuvering target tracking box particle filter adaptive interacting multiple model interval measurement probability transition matrix |
