›› 2011,Vol. 26 ›› Issue (1): 187-194.doi: 10.1007/s11390-011-1121-z

• • 上一篇    下一篇

一种有效的动作影片中打斗镜头的识别方法

  

  • 收稿日期:2009-09-30 修回日期:2010-11-23 出版日期:2011-01-01 发布日期:2011-01-01

Effectively Discriminating Fighting Shots in Action Movies

Shu-Gao Ma1,2 (马述高) and Wei-Qiang Wang1,3,* (王伟强), Member, ACM, IEEE   

  1. 1. School of Information Science and Engineering, Graduate University of Chinese Academy of Sciences Beijing 100049, China;
    2. Computer Science Department, Boston University, Boston MA 02215, U.S.A.;
    3. Key Laboratory of Intelligent Information Processing, Institute of Computing Technology, Chinese Academy of Sciences Beijing 100190, China
  • Received:2009-09-30 Revised:2010-11-23 Online:2011-01-01 Published:2011-01-01
  • About author:Shu-Gao Ma received his M.S. degree in computer science from Graduate University of Chinese Academy of Sciences in 2009. He is currently a Ph.D. candidate in Boston University, US. His current research interests include multimedia content analysis, computer vision.
    Wei-Qiang Wang is a professor in School of Information Science and Engineering, Graduate University of Chinese Academy of Sciences, Beijing, China. He is a member of IEEE, ACM. He received his Ph.D. degree from Institute of Computing Technology, Chinese Academy of Sciences, in 2001. His current research interests include multimedia content analysis, computer vision and machine learning.
  • Supported by:

    This work was supported in part by the National High Technology Research and Development 863 Program of China under Grant No. 2006BAH02A24-2 and by the National Natural Science Foundation of China under Grant No. 60873087.

1.本文的创新点
在动作电影中打斗场面是吸引观众眼球的重要组成部分,有效地识别相关镜头是潜在的许多实际应用需要解决的一个技术问题。 本文针对该问题利用计算机视觉的相关技术给出了解决该问题的一种基于视觉信息的方法。不同于现存的其他基于视觉的方法,我们的方法将视频帧中的前景运动与背景运动鲁棒地加以区分,并对提取出的前景关键点构造出更丰富的特征(角速度、加速度等)对其运动进行刻画,最后通过机器学习来建立识别打斗镜头的分类模型。
2.实现方法
我们的方法首先在视频帧上提取局部关键特征点,并通过帧间的不变特征匹配对特征点进行跟踪。在局部特征点匹配计算中,我们不仅考虑两个关键点之间的外表相似性还考虑它们之间运动轨迹的平滑性。为了提高计算效率,我们的算法自适应地调节搜索范围根据前面一个时刻对应点的运动速度,并动态调节视频帧的采样间隔。
然后通过一个高级的投票过程将前景特征点与背景关键特征点加以区分。在投票计算中, 每个关键点的运动分成静止、旋转、缩放、以及8个不同方向的平移10个类别,每个关键点参与投票权重并不相同,会根据前面时刻对应点的类别进行自适应的调整,背景点会得到不断的加强,前景点相反,并利用选取代表运动向量让投票的关键点具有位置分布的均衡性,最后获得最多票数的运动类别作为摄像机的运动类别,并给出该判别的信度值。那些与摄像机类型相一致的关键点被标记为背景点,用于估计摄像机运动模型的参数。
获得背景运动模型后,我们进一步选取出可靠的前景关键点,以获得的背景模型作为参照物计算出它们的实际运动向量。基于提取出的前景关键点这些实际运动向量,我们可以计算描述两个采样帧间前景对象运动的特征向量,该向量包含的特征有:运动速度大小与方向归一化直方图,角速度归一化直方图,加速度归一化直方图,由静到动比,由动到静比,连续性比。
最后,我们通过求取一个镜头内的前景运动特征向量集的均值向量来描述该镜头内前景对象的运动特点,并利用支持向量机学习出用于识别打斗镜头的分类器模型。
3.结论及未来待解决的问题
我们的方法能够有效地识别动作片打斗镜头,给出的摄像机运动分析算法非常鲁棒,非常适合动作片中大幅度摄像机运动。希望结合电影中的音频信息,进一步提高识别打斗镜头的鲁棒性,并检测其他与运动相关的事件。
4.实用价值或应用前景
所提出的方法可用于自动生成动作片的精彩片断用于广告宣传以及用户对影片内容的快速预览,特定电影内容过滤用于防止暴力场面对儿童以及青少年成长的不良影响,动作片内容的检索与索引等。

Abstract:

Fighting shots are the highlights of action movies and an effective approach to discriminating fighting shots is very useful for many applications, such as movie trailer construction, movie content filtering, and movie content retrieval. In this paper, we present a novel method for this task. Our approach first extracts the reliable motion information of local invariant features through a robust keypoint tracking computation; then foreground keypoints are distinguished from background keypoints by a sophisticated voting process; further, the parameters of the camera motion model is computed based on the motion information of background keypoints, and this model is then used as a reference to compute the actual motion of foreground keypoints; finally, the corresponding feature vectors are extracted to characterizing the motions of foreground keypoints, and a support vector machine (SVM) classifier is trained based on the extracted feature vectors to discriminate fighting shots. Experimental results on representative action movies show our approach is very effective.

[1] Adams B, Dorai C, Venkatesh S. Novel approach to determiing tempo and dramatic story sections in motion pictures. In Proc. the 7th IEEE International Conference on Image Proessing, Vancouver, Canada, Sept. 10-13, 2000, pp.283-286.
[2] Liu A, Li J, Zhang Y, Tang S, Song Y, Yang Z. An innovative odel of tempo and its application in action scene detection or movie analysis. In Proc. the 9th IEEE Workshop on pplications of Computer Vision, Copper Mountain, USA, an. 7-9, 2008, pp.1-6.
[3] Chen H W, Kuo J H, Chu W T, Wu J L. Action movies egmentation and summarization based on tempo analysis. In Proc. the 6th ACM SIGMM International Workshop on ultimedia Information Retrieval, USA, Oct. 15-16, 2004, p.251-258.
[4] Chen L, Ozsu M T. Rule-based scene extraction from video. In Proc. the 9th IEEE International Conference on Image Processing, Rochester, USA, Sept. 22-25, 2002, pp.737-740.
[5] Cheng W, Liu C, Xu D. An approach to action scene detecion in martial arts movies. Acta Electronica Sinica, 2006, 34(5): 915-920.
[6] Gong Y, Wang W Q, Jiang S Q, Huang Q M, Gao W. Deecting violent scenes in movies by auditory and visual cues. In Proc. the Pacific-Rim Conference on Multimedia, Tainan, hina, Dec. 9-13, 2008, pp.317-326.
[7] Datta A, Shah M, Lobo N D V. Person-on-person violence etection in video data. In Proc. the 16th International Conerence on Pattern Recognition, Quebec, Canada, Aug. 11-15, 002, pp.433-438.
[8] Mecocci A, Micheli F. Real-time automatic detection of iolent-acts by low-level colour visual cues. In Proc. the 4th IEEE International Conference on Image Processing, an Antonio, USA, Sept. 16-19, 2007, pp.345-348.
[9] Wang S, Jiang S, Huang Q, Gao W. Shot classification for ction movies based on motion characteristics. In Proc. the 5th IEEE International Conference on Image Processing, an Diego, USA, Oct. 12-15, 2008, pp.2508-2511.
[10] Lowe D G. Distinctive image features from scale-invariant eypoints. International Journal of Computer Vision, 2004, 60(2): 91-110.
[11] Battiato S, Gallo G, Puglisi G, Scellato S. SIFT features racking for video stabilization. In Proc. the 14th Internaional Conference on Image Analysis and Processing, Modna, Italy, Sept. 10-14, 2007, pp.825-830.
[12] Kraaij W, Ianeva T. TRECVID-2005 low-level (camera otion) feature task. http://www.nlpir.nist.gov/projects/vpubs/tv5.papers/tv5.llf.slides.final.pdf, 2005.
[13] Fan R, Chen P, Lin C. Working set selection using the secnd order information for training SVM. Journal of Machine earning Research, 2005, 6(12): 1889-1918.

No related articles found!
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
[1] 张钹; 张铃;. Statistical Heuristic Search[J]. , 1987, 2(1): 1 -11 .
[2] 孟力明; 徐晓飞; 常会友; 陈光熙; 胡铭曾; 李生;. A Tree-Structured Database Machine for Large Relational Database Systems[J]. , 1987, 2(4): 265 -275 .
[3] 林琦; 夏培肃;. The Design and Implementation of a Very Fast Experimental Pipelining Computer[J]. , 1988, 3(1): 1 -6 .
[4] 孙成政; 慈云桂;. A New Method for Describing the AND-OR-Parallel Execution of Logic Programs[J]. , 1988, 3(2): 102 -112 .
[5] 张焰; 何继潮;. Data Dependencies in Database with Incomplete Information[J]. , 1988, 3(2): 131 -138 .
[6] 张钹; 张恬; 张建伟; 张铃;. Motion Planning for Robots with Topological Dimension Reduction Method[J]. , 1990, 5(1): 1 -16 .
[7] 许志明;. Discrete Interpolation Surface[J]. , 1990, 5(4): 329 -332 .
[8] 王鼎兴; 郑纬民; 杜晓黎; 郭毅可;. On the Execution Mechanisms of Parallel Graph Reduction[J]. , 1990, 5(4): 333 -346 .
[9] 蔡士杰; 张福炎;. A Fast Algorithm for Polygon Operations[J]. , 1991, 6(1): 91 -96 .
[10] 周权; 魏道政;. A Complete Critical Path Algorithm for Test Generation of Combinational Circuits[J]. , 1991, 6(1): 74 -82 .
版权所有 © 《计算机科学技术学报》编辑部
本系统由北京玛格泰克科技发展有限公司设计开发 技术支持:support@magtech.com.cn
总访问量: