两类典型多目标跟踪算法的性能分析与比较

王芝; 徐晓滨; 刘伟峰; 文成林

doi:10.3724/SP.J.1146.2009.00917

两类典型多目标跟踪算法的性能分析与比较

doi: 10.3724/SP.J.1146.2009.00917

计量
- 文章访问数: 3900
- HTML全文浏览量: 116
- PDF下载量: 1027
- 被引次数: 0
出版历程
- 收稿日期: 2009-06-23
- 修回日期: 2009-12-02
- 刊出日期: 2010-07-19

Performance Analysis and Comparison of Two Classic Algorithms in Multi-target Tracking

摘要

摘要: 在处理目标跟踪的两类主要方法中，一类是通过数据关联来解决，如PDA和JPDA等；另一类则是绕过关联直接处理，如随机集、GM-PHD等。该文从两类典型方法中各选取一种有代表性的方法，如JPDA与GM-PHD，首先通过分析两种算法主要步骤的计算量，得到相应算法总计算量的解析表达式；然后根据观测与目标状态之间关联复杂程度，分3种情况对两类算法的计算量进行比较；最后以仿真说明算法的跟踪效果，并以算法运行时间来验证计算量公式的正确性。
- 多目标跟踪; 联合概率数据关联(JPDA); 随机集; 概率假设密度(PHD)
Abstract: There are two primary ways to process multi-target tracking problem. One is data association method, whose deputies are PDA and JPDA. The other is direct method without the data association, whose deputies are random sets theory and GM-PHD. Two representational algorithms are chosen from aforementioned two kinds of methods respectively, that is, JPDA and GM-PHD. Firstly, general analytical forms to evaluate calculation complexity of each algorithm are formulated by analyzing and totaling their major operation steps. Secondly, the calculation complexity of two algorithms is compared through three cases respectively, which are divided on the basis of associated complexity between states and the measurements. Finally, one example, including tracking effect and the running time, is utilized to illustrate the analytical forms of evaluating calculation complexity proposed in this paper.

HTML全文

参考文献(1)

李良群. 信息融合系统中的目标跟踪及数据关联技术研究[D].[博士论文], 西安电子科技大学, 2007.Li Liang-qun. Research on the target tracking and dataassociation techniques with the information fusion system[D].[Ph. D. dissertation], Xidian University, 2007.[2]韩崇昭, 朱洪艳, 段战胜. 多源信息融合[M]. 北京: 清华大学出版社, 2006, 297-303.Han Chong-zhao, Zhu Hong-yan, and Duan Zhan-sheng.Multi-source Information Fusion [M]. Beijing, TsinghuaUniversity Press, 2006: 297-303.[3]李正周, 金钢, 董能力. 基于改进概率数据关联滤波的红外小运动目标跟踪[J].电子与信息学报.2008, 30(4):953-956浏览[4]樊国创, 戴亚平, 许向阳. 基于神经网络的混合双滤波器自适应目标跟踪算法. 火力与指挥控制, 2009, 34(2): 120-123.Fan Guo-chuang, Dai Ya-ping, and Xu Xiang-yang. Neuralnetwork-based hybrid double filters for adaptive targettracking. Fire Control Command Control, 2009, 34(2):120-123.[5]Goodman I, Mahler R, and Nguyen H. Mathematics of DataFusion[M]. Dordrecht: Kluwer Academic Publishers, 1997,Chapter 2, 4-8.[6]徐晓滨, 文成林, 刘荣利. 基于随机集理论的多源信息统一表示与建模方法. 电子学报, 2008, 36(6): 1174-1181.Xu Xiao-bin, Wen Cheng-lin, and Liu Rong-li. The unifiedmethod of describing and modeling multi-source informationbased on random set theory. Acta Electronica Sinica, 2008,36(6): 1174-1181.[7]Mahler R. Multi-target Bayes filtering via first-ordermulti-target moments[J].IEEE Transactions on Aerospace andElectronic System.2003, 39(4):1152-1178[8]Mahler R. An introduction to multi-source multi-targetstatistics and its applications. Technical Monograph.Lockheed Martin, Eagan MN, 2000.[9]Ulmke M, Franken D, and Schmidt M. Missed detectionproblems in the cardinalized probability hypothesis densityfilter. 11th International Conference on Information Fusion.Cologne, 2008: 1-7.[10]Vo Ba-ngu and Ma Wing-kin. A closed-form solution for theprobability hypothesis density filter. 8th InternationalConference on Information Fusion. Philadelphia, 2005:856-863.[11]Zhang Hong-jian, Jing Zhong-liang, and Hu Shi-qiang. Tracksextraction of the probability hypothesis density filter forsurvival targets. Proceedings of the 27th Chinese ControlConference, Kunming, 2008: 343-347.[12]Liu Wei-feng, Han Chong-zhao, Lian Feng, Xu Xiao-bin, andWen Cheng-lin. Multitarget state and track estimation forthe probability hypothesis density filter[J].Journal ofElectronics (China.2009, 26(1):2-12[13]Wen Cheng-lin and Xu Xiao-bin. Random sets in data fusion:a new framework for multi-target tracking. Systems andControl in Aerospace and Astronautics, Harbin, 2006:999-1004.[14]K Chang and Y Bar-shalom. Joint probabilistic dataassociation for multi-target tracking with possibly unresolvedmeasurements and maneuvers[J].IEEE Transactions onAutomatic Control.1984, 29(7):585-594[15]Clark D, Vo Ba-tuong, Vo Ba-Ngu, and Godsill S. Gaussianmixture implementations of probability hypothesis densityfilters for non-linear dynamical models. 2008 IET Seminar onTarget Tracking and Data Fusion: Algorithms andApplications, Birmingham, 2008: 21-28.[16]文成林, 吕冰, 葛泉波. 一种基于分步式滤波的数据融合算法[J]. 电子学报, 2004, 32(8): 1264-1267.Wen Cheng-lin, L Bing, and Ge Quan-bo. A data fusionalgorithm based on filtering step by step. Acta ElectronicaSinica, 2004, 32(8): 1264-1267.

施引文献

资源附件(0)

访问统计