Particle Filter via Chopthin Resampling for Target Tracking with Glint Noise

2018 2nd International Conference on Artificial Intelligence Applications and Technologies

计算机科学

Liu, C.^1 ; Yang, S.C.^1 ; Wang, L.D.^2

Army Engineering University, No.97 Heping West Road, Shijiazhuang, Hebei

050003, China^1

Natl. Key Lab. of the Complex Electromagnetic Environment Effect of Electronic Information System, Luoyang, Henan

471003, China^2

关键词: A-particles;    Filtering accuracies;    Glint Noise;    Observation noise;    Parameter setting;    Particle filter;    Resampling method;    Sample sizes;   

Others  :  https://iopscience.iop.org/article/10.1088/1757-899X/435/1/012038/pdf
DOI  :  10.1088/1757-899X/435/1/012038

学科分类：计算机科学（综合）

来源: IOP

【 摘 要 】

Traditional resampling methods in particle filter suffer from the disadvantage of the low filtering accuracy, the severe fluctuation of effective sample size (ESS) and the great influence of parameter setting. A particle filter algorithm based on chopthin resampling is proposed in this paper. Traditional resampling methods produce a set of equally weighted particles and only occur when ESS below a threshold. In contrast to this, the proposed algorithm enforces an upper bound on the ratio between the weights and can be implemented in every step. Simulation shows that the new method outperforms traditional resampling methods in filtering accuracy, ESS stability and computational efficiency, especially when sample size or observation noise is small.