正在加载图片...
444 智能系统学报 第6卷 降低,在第3步中,能提高吸烟这种特定动作的识别 ceedings of CVPR:IEEE Conference on Computer Vision 正确率,这样系统对于吸烟活动的识别率也就得到 and Patterm Recognition.San Francisco,USA,2010::1959- 1966. 了大幅度的提升.· 10].CAO Liangliang,LU Zicheng,HUANG T.Cros-dataset 5结束语 action detection[C]//Proceedings of CVPR:IEEE Confer- ence on Computer Vision and Pattern Recognition.San 本文主要对真实电影中的人的抽烟行为进行识 Francisco,USA,2010::1998-2005. 别,与之前在特定场景中分析人的活动相比,这里是 [11]NATARAJAN P.NEVATIA R.View and scale invariant action recognition using multiview shape-flow models 在包括人物外表改变、场景变换、镜头视角变换和动 [C]//Proceedings of CVPR:IE Conference on Com- 作时间改变的真实场景中进行活动分析与识别.在 puter Vision and Pattern Recognition.Anchorage,USA. 真实场景的识别活动中,由于各种因素的影响,导致 2008::1-8. 现在很多在特定视频中识别效果比较好的方法在真 [12]VITALADEVUNI S N.KELOKUMPU V.DAVIS L s. Action recognition using ballistic dynamics[Cl//Proceed- 实电影中的识别效果很低.考虑到若只使用单独运 ings of CVPR:IEEE Conference on Computer Vision and 动信息或形状信息在真实场景中识别效果不高,因 Patten Recognition.Anchorage,USA,2008:1-8. 此采用了一种纯贝叶斯互信息最大化组合分类器作 [13]YILMAZ A,SHAH M.Actions sketch::a novel action rep- 为统一的计算框架,实验结果证明此方法相比于传 resentation[C]//Proceedings of CVPR::IEEE Conference on Computer Vision and Patten Recognition.San Diego, 统方法提高了识别率· USA,2005::984-989. 但是,使用视频中帧的信息的方法,对于包含物 [4]]YUAN Junsong,LIU Zicheng,WU Ying.Discriminative 品的运动此较有效,如吸烟、喝水,而对于诸如走路、 subvolume search for efficient action detection[C]//Pro- 慢跑、跑步这样动作相似的行为识别效果一般.如何 ceedings of CVPR:IEEE Conference on Computer Vision and Pattern Recognition.Miami,USA,2009:2442-2449. 将这种方法运用到其他所有动作以及如何减少运算 15]]LAPTEVI,PEREZ P.Retrieving actions in movies[C]// 时间都将是今后研究的重点方向. Proceedings ofICCV:IEEE International Conference on Computer Vision.Rio de Janeiro,Brazil,2007::18 参考文献: [16]YWU Pin,HSIEH JH,CHENG J C,et al.Human smok- [1]LAPTEV I,MARSZALEK M,SCHMID C,et al.Learning ing event detection using visual interaction clues[C]y/Pro- realistic human actions from movies C]//Proceedings of ceedings of ICPR:.IEEE International Conference on Pat- CVPR:IEEE Conference on Computer Vision and Pattern tern Recognition.Istanbul,Turkey,2010:4334-4347. Recognition.Anchorage,USA,2008::H& [2]GAIDON A.MARSZALEK M.SCHMID C.Mining visual 作者简介 actions from movies[C]]//Proceedings of BMVC:British 叶果,男,1990年生,本科生,主要 Machine Vision Conference.London,UK,2009::1-11. 研究方向为人的活动识别、计算机视觉 3WANG JZ,GEMAN D.LUO Jiebo,et al.Real-world im- 与模式识别. age annotation and retrieval:an introduction to the special section[J]].IEEE Transactions on Patten Analysis and Ma- chine Inelligence,2008,30(11)1:1873-1876. 4 LAPTEV I.On space-time interest points[J].Interational Journal of Computer Vision,2005,64(2/3)::107-123. [5]LOWE D G.Distinctive image features from scale-invariant 程洪,男,1973年生,教授,博士生 keypoints[J].International Journal of Computer Vision, 导师,博士,EEE和ACM会员,2006- 2004,60(2):91-110. 2009年在美国卡内基-梅隆大学计算 [6]ALI S.BASHARAT A.SHAH M.Chaotic invariants for 机学院进行博士后研究.主要研究方向 human action recognition[C]//Proceedings of ICCV:IEEE 为机器人、计算机视觉与模式识别、 International Conference on Computer Vision.Rio de Janei- ro,Brazil,2007:14-21. 机器学习.先后主持和参与包括国家“ [7]JNGUYEN N T.PHUNG D Q,VENKATESH S.et al. 973“计划项目、国家"863"计划项目,以 Leaming and detecting activities from movement trajectories 及重要企业横向项目等10余项科研项目.发表学术论文 using the hierarchical hidden Markov models [C]//Pro- 40余篇,出版教材和专著各1部. ceedings of CVPR:IEEE Conference on Computer Vision and Pattern Recognition.San Diego,USA,2005::955- 960. 赵洋,男,1988年生,硕士研究生,主要 [8]MOESLUND T B.HILTON A.KRUGER V.A survey of 研究方向为计算机视觉与模式识别 advances in vision-based human motion capture and analysis [I.Computer Vision and Image Understanding,2006, 1042②):90-126. 9]DUAN Liin,XU Dong,TSANG IW,et al.Visual event ecognition in videos by learring from web data[C]//Pro-降低,在第3步中,能提高吸烟这种特定动作的识别 正确率,这样系统对于吸烟活动的识别率也就得到 了大幅度的提升. [2] GAIDON A,MARSZALEK M,SCHMID C. Mining visual actions from movies[C] //Proceedings of BMVC: British Machine Vision Conference. London,UK,2009: 1-11. 444· 本文主要对真实电影中的人的抽烟行为进行识 别,与之前在特定场景中分析人的活动相比,这里是 在包括人物外表改变、场景变换、镜头视角变换和动 作时间改变的真实场景中进行活动分析与识别.在 真实场景的识别活动中,由于各种因素的影响,导致 现在很多在特定视频中识别效果比较好的方法在真 实电影中的识别效果很低.考虑到若只使用单独运 动信息或形状信息在真实场景中识别效果不高,因 此采用了一种纯贝叶斯互信息最大化组合分类器作 为统一的计算框架,实验结果证明此方法相比于传 统方法提高了识别率 [8] MOESLUND T B,HILTON A,KRUGER V. A survey of advances in vision-based human motion capture and analysis [J] . Computer Vision and Image Understanding,2006, 104(2) :90-126. 4 ] LAPTEV I. On space-time interest points[J]. Interational Journal of Computer Vision,2005,64(2/3) : 107-123. [14] YUAN Junsong,LIU Zicheng,WU Ying. Discriminative subvolume search for efficient action detection[C] //Pro￾ceedings of CVPR: IEEE Conference on Computer Vision and Pattern Recognition. Miami,USA,2009: 2442-2449. [11] NATARAJAN P,NEVATIA R. View and scale invariant action recognition using multiview shape-flow models [C]//Proceedings of CVPR: IE Conference on Com￾puter Vision and Pattern Recognition. Anchorage,USA, 2008: 1-8 [1] LAPTEV I,MARSZALEK M,SCHMID C,et al. Learning realistic human actions from movies [ C] //Proceedings of CVPR: IEEE Conference on Computer Vision and Pattern Recognition.Anchorage,USA,2008: 1-8. [5] LOWE D G. Distinctive image features from scale-invariant keypoints[J]. International Journal of Computer Vision, 2004,60(2) : 91-110. 赵洋,男,1988年生,硕士研究生,主要 研究方向为计算机视觉与模式识别 5结束语 参考文献: [13] YILMAZ A,SHAH M. Actions sketch: a novel action rep￾resentation[ C] //Proceedings of CVPR: IEEE Conference on Computer Vision and Patten Recognition. San Diego, USA,2005: 984-989. [7] NGUYEN N T,PHUNG D Q,VENKATESH S,et al. Learning and detecting activities from movement trajectories using the hierarchical hidden Markov models [C] //Pro￾ceedings of CVPR: IEEE Conference on Computer Vision and Pattern Recognition. San Diego,USA,2005: 955- 960. [12] VITALADEVUNI S N,KELOKUMPU V,DAVIS L s. Action recognition using ballistic dynamics[ C]//Proceed￾ings of CVPR: IEEE Conference on Computer Vision and Patten Recognition.Anchorage,USA,2008: 1-8. [6] ALI S,BASHARAT A, SHAH M. Chaotic invariants for human action recognition[ C] //Proceedings of ICCV: IEEE International Conference on Computer Vision. Rio de Janei￾ro,Brazil,2007: 14-21. 但是,使用视频中帧的信息的方法,对于包含物 品的运动比较有效,如吸烟、喝水,而对于诸如走路、 慢跑、跑步这样动作相似的行为识别效果一般.如何 将这种方法运用到其他所有动作以及如何减少运算 时间都将是今后研究的重点方向. 15] LAPTEVI,PEREZ P. Retrieving actions in movies[ C] / / Proceedings of ICCV: IEEE International Conference on Computer Vision. Rio de Janeiro,Brazil,2007: 1-8. 智 能 系 统 学 程洪,男,1973年生,教授,博士生 导师,博士,EEE和 ACM会员,2006- 2009年在美国卡内基-梅隆大学计算 机学院进行博士后研究.主要研究方向 为机器人、计算机视觉与模式识别、 机器学习.先后主持和参与包括国家" 973"计划项目、国家"863"计划项目,以 及重要企业横向项目等10余项科研项目.发表学术论文 40余篇,出版教材和专著各1部. 报 第6卷 叶果,男,1990年生,本科生,主要 研究方向为人的活动识别、计算机视觉 与模式识别. ceedings of CVPR: IEEE Conference on Computer Vision and Patterm Recognition. San Francisco,USA,2010: 1959- 1966. 10] CAO Liangliang,LU Zicheng,HUANG T. Cros-dataset action detection[ C] //Proceedings of CVPR: IEEE Confer￾ence on Computer Vision and Pattern Recognition. San Francisco,USA,2010: 1998-2005. [3] WANG JZ,GEMAN D,LUO Jiebo,et al. Real-world im￾age annotation and retrieval: an introduction to the special section[J] .IEEE Transactions on Patten Analysis and Ma￾chine Inelligence,2008,30(11) : 1873-1876. [9] DUAN Liin,XU Dong,TSANG IW,et al. Visual event ecognition in videos by learring from web data[C] //Pro- 作者简介: 16] WU Pin,HSIEH JH,CHENG J C,et al. Human smok￾ing event detection using visual interaction clues[C] //Pro￾ceedings of ICPR: IEEE International Conference on Pat￾tern Recognition. Istanbul,Turkey,2010: 4334-4347
<<向上翻页
©2008-现在 cucdc.com 高等教育资讯网 版权所有