·706· 智能系统学报 第14卷 4结束语 25(4):1779-1792 [5]WANG Dong,LU Huchuan.On-line learning parts-based 针对目标跟踪的运动模型，本文提出了一种 representation via incremental orthogonal projective non- 智能群体优化滤波(SIF)算法。在贝叶斯滤波的 negative matrix factorization[J].Signal processing,2013. 基础上，本文提出的算法融入了智能群体优化中 93(6):1608-1623. 的3种智能群体优化思想，即内聚、分离和排列 [6]吴吴，孙晓燕，郭玉堂，等.保持粒子多样性的非退化粒 运动。在当前时刻能够准确地估计后验状态的情 子滤波方法研究[J】.电子学报，2016,44(7)：1734-1741. 况下，内聚运动是将权值较低的粒子聚合在高权 WU Hao,SUN Xiaoyan,GUO Yutang,et al.Non-degener- 值粒子周围，以增加其权值并保留了粒子多样 acy particle filtering method research for particle diversity 性，再结合排列运动准确地预测下一时刻的先验 preserving[J].Acta electronica sinica,2016,44(7): 状态，能够有效地增加算法对遮挡和形变的适应 1734-1741. 性。分离运动是在当前时刻无法准确估计后验状 [7]常天庆，李勇，刘忠仁，等.一种改进重采样的粒子滤波 态的情况下，通过扩大搜索范围来增加粒子多样 算法J.计算机应用研究，2013.30(3)少：748-750 性，能够有效处理快速移动和运动模糊导致的粒 CHANG Tianging,LI Yong,LIU Zhongren,et al.Particle 子权值退化问题，提高了下一时刻的先验滤波概 filter algorithm based on improved resampling[J].Applica- 率密度。 tion research of computers,2013,30(3):748-750. 实验结果表明，相比于广泛使用的粒子滤波 [8]CAO Bei,MA Caiwen,LIU Zhentao.Particle filter with 算法，智能群体优化滤波算法更能准确地估计后 fine resampling for bearings-only tracking[J].Procedia en- 验状态，当实际运用在目标跟踪中，更加有效地 gineering,2012,29:3685-3690. 应对复杂多变的跟踪环境。同时本文提出的算 [9]DU Kelin,SWAMY M N S.Swarm intelligence[M]//DU 法思想还可以使用在任何基于采样的跟踪算法 Kelin,SWAMY M N S.Search and Optimization by Meta- 中，因此该算法具有很好的适用性。本文的实验 heuristics.Cham:Birkhauser,2016. [10]彭喜元，彭宇，戴毓丰.群智能理论及应用刀.电子学 只将算法应用到了IPONMF算法和IVT算法中， 报.2003,31(S1):1982-1988 为了进一步提高跟踪效果，下一步的工作将考虑 PENG Xiyuan,PENG Yu,DAI Yufeng.Swarm intelli- 将智能群体优化滤波算法应用到其他的跟踪算 法中。 gence theory and applications[J].Acta electronica sinica, 2003,31(S1)1982-1988. 参考文献： [11]CHENG Shi,ZHANG Qingyu,QIN Quande.Big data analytics with swarm intelligence[J].Industrial manage- [1]LGUENSAT R,TANDEO P,FABLET R,et al.Non-para- ment and data systems,2016,116(4):646-666. metric Ensemble Kalman methods for the inpainting of [12]XIA Junbo.Coverage optimization strategy of wireless noisy dynamic textures[Cl//Proceedings of 2015 IEEE In- sensor network based on swarm intelligence algorithm[Cl// ternational Conference on Image Processing.Quebec City, Proceedings of 2016 International Conference on Smart Canada,2016:4288-4292 [2]王法胜，鲁明羽，赵清杰，等.粒子滤波算法刀.计算机学 City and Systems Engineering.Hunan,China,2016: 报，2014,37(8)：1679-1694 179-182. WANG Fasheng,LU Mingyu,ZHAO Qingjie,et al. [13]DEVI K U,SARMA D,LAISHRAM R.Swarm intelli- Partilce filtering algorithm[J].Chinese journal of com- gence based computing techniques in speech enhance- puters,.2014,37(8:1679-1694. ment[C]//Proceedings of 2015 International Conference [3]BAO Chenglong,WU Yi,LING Haibin,et al.Real time on Green Computing and Internet of Things.Noida,India, robust LI tracker using accelerated proximal gradient ap- 2015:1199-1203 proach[C]//Proceedings of 2012 IEEE Conference on [14]KRONANDER J,SCHON T B.Robust auxiliary particle Computer Vision and Pattern Recognition.Providence, filters using multiple importance sampling[C]//Proceed- USA.2012:1830-1837. ings of 2014 IEEE Workshop on Statistical Signal Pro- [4]ZHANG Kaihua,LIU Qingshan,WU Yi,et al.Robust cessing.Gold Coast,Australia,2014:268-271 visual tracking via convolutional networks without train- [15]WANG Naiyan,SHI Jianping,YEUNG D Y,et al.Un- ing[J].IEEE transactions on image processing,2016, derstanding and diagnosing visual tracking systems[C]//4 结束语 针对目标跟踪的运动模型，本文提出了一种 智能群体优化滤波 (SIF) 算法。在贝叶斯滤波的 基础上，本文提出的算法融入了智能群体优化中 的 3 种智能群体优化思想，即内聚、分离和排列 运动。在当前时刻能够准确地估计后验状态的情 况下，内聚运动是将权值较低的粒子聚合在高权 值粒子周围，以增加其权值并保留了粒子多样 性，再结合排列运动准确地预测下一时刻的先验 状态，能够有效地增加算法对遮挡和形变的适应 性。分离运动是在当前时刻无法准确估计后验状 态的情况下，通过扩大搜索范围来增加粒子多样 性，能够有效处理快速移动和运动模糊导致的粒 子权值退化问题，提高了下一时刻的先验滤波概 率密度。 实验结果表明，相比于广泛使用的粒子滤波 算法，智能群体优化滤波算法更能准确地估计后 验状态，当实际运用在目标跟踪中，更加有效地 应对复杂多变的跟踪环境。同时本文提出的算 法思想还可以使用在任何基于采样的跟踪算法 中，因此该算法具有很好的适用性。本文的实验 只将算法应用到了 IPONMF 算法和 IVT 算法中， 为了进一步提高跟踪效果，下一步的工作将考虑 将智能群体优化滤波算法应用到其他的跟踪算 法中。 参考文献： LGUENSAT R, TANDEO P, FABLET R, et al. Non-para￾metric Ensemble Kalman methods for the inpainting of noisy dynamic textures[C]//Proceedings of 2015 IEEE In￾ternational Conference on Image Processing. Quebec City, Canada, 2016: 4288–4292. [1] 王法胜, 鲁明羽, 赵清杰, 等. 粒子滤波算法 [J]. 计算机学 报, 2014, 37(8): 1679–1694. WANG Fasheng, LU Mingyu, ZHAO Qingjie, et al. Partilce filtering algorithm[J]. Chinese journal of com￾puters, 2014, 37(8): 1679–1694. [2] BAO Chenglong, WU Yi, LING Haibin, et al. Real time robust L1 tracker using accelerated proximal gradient ap￾proach[C]//Proceedings of 2012 IEEE Conference on Computer Vision and Pattern Recognition. Providence, USA, 2012: 1830–1837. [3] ZHANG Kaihua, LIU Qingshan, WU Yi, et al. Robust visual tracking via convolutional networks without train￾ing[J]. IEEE transactions on image processing, 2016, [4] 25(4): 1779–1792. WANG Dong, LU Huchuan. On-line learning parts-based representation via incremental orthogonal projective non￾negative matrix factorization[J]. Signal processing, 2013, 93(6): 1608–1623. [5] 吴昊, 孙晓燕, 郭玉堂, 等. 保持粒子多样性的非退化粒 子滤波方法研究 [J]. 电子学报, 2016, 44(7): 1734–1741. WU Hao, SUN Xiaoyan, GUO Yutang, et al. Non-degener￾acy particle filtering method research for particle diversity preserving[J]. Acta electronica sinica, 2016, 44(7): 1734–1741. [6] 常天庆, 李勇, 刘忠仁, 等. 一种改进重采样的粒子滤波 算法 [J]. 计算机应用研究, 2013, 30(3): 748–750. CHANG Tianqing, LI Yong, LIU Zhongren, et al. Particle filter algorithm based on improved resampling[J]. Applica￾tion research of computers, 2013, 30(3): 748–750. [7] CAO Bei, MA Caiwen, LIU Zhentao. Particle filter with fine resampling for bearings-only tracking[J]. Procedia en￾gineering, 2012, 29: 3685–3690. [8] DU Kelin, SWAMY M N S. Swarm intelligence[M]//DU Kelin, SWAMY M N S. Search and Optimization by Meta￾heuristics. Cham: Birkhäuser, 2016. [9] 彭喜元, 彭宇, 戴毓丰. 群智能理论及应用 [J]. 电子学 报, 2003, 31(S1): 1982–1988. PENG Xiyuan, PENG Yu, DAI Yufeng. Swarm intelli￾gence theory and applications[J]. Acta electronica sinica, 2003, 31(S1): 1982–1988. [10] CHENG Shi, ZHANG Qingyu, QIN Quande. Big data analytics with swarm intelligence[J]. Industrial manage￾ment and data systems, 2016, 116(4): 646–666. [11] XIA Junbo. Coverage optimization strategy of wireless sensor network based on swarm intelligence algorithm[C]// Proceedings of 2016 International Conference on Smart City and Systems Engineering. Hunan, China, 2016: 179–182. [12] DEVI K U, SARMA D, LAISHRAM R. Swarm intelli￾gence based computing techniques in speech enhance￾ment[C]//Proceedings of 2015 International Conference on Green Computing and Internet of Things. Noida, India, 2015: 1199–1203. [13] KRONANDER J, SCHÖN T B. Robust auxiliary particle filters using multiple importance sampling[C]//Proceed￾ings of 2014 IEEE Workshop on Statistical Signal Pro￾cessing. Gold Coast, Australia, 2014: 268–271. [14] WANG Naiyan, SHI Jianping, YEUNG D Y, et al. Un￾derstanding and diagnosing visual tracking systems[C]// [15] ·706· 智 能 系 统 学 报 第 14 卷