计算机应用 ›› 2011, Vol. 31 ›› Issue (02): 410-412.

• 模式识别 • 上一篇    下一篇

基于动态特征的突发暴力事件辨识算法

蒲静1,傅饶2,胡栋3,4,卢璇5   

  1. 1. 西华师范大学
    2.
    3. 浙江网新恒天软件有限公司
    4. 电子科技大学计算机科学与工程学院
    5. 成都电子科技大学
  • 收稿日期:2010-06-04 修回日期:2010-07-19 发布日期:2011-02-01 出版日期:2011-02-01
  • 通讯作者: 蒲静

Sudden violence identification algorithm based on motion feature

  • Received:2010-06-04 Revised:2010-07-19 Online:2011-02-01 Published:2011-02-01

摘要: 在智能监控系统中,公共场所发生的暴力事件应及时预警,否则可能造成严重的后果。针对该情况,对暴力事件发生时图像中反映出来的信息进行了全面的提取,提出一种基于动态特征的能量算法。该算法从运动的剧烈程度、运动的不规则程度和目标间的位置关系3个角度较为全面地描述了该事件发生时图像的数字特征,并通过引入动能和势能总结出能量函数,用于辨识是否发生突发暴力事件。最后通过对大量包含各类事件的视频进行实验,并得出结论,该方法在处理突发暴力事件时有较高的识别率。

关键词: 暴力事件, 光流特征, 动态特征, 能量函数, 质量权重

Abstract: In the intelligent monitoring system, violence behavior in public place should be warned in time; otherwise, it can lead to serious consequences. In this paper, the information of image was abstracted comprehensively when violence behavior happened, and an energy algorithm based on motion feature was proposed. The digital features of image from moving intensity, moving irregularity and location between targets when violence behavior happened were described, and the energy function using kinetic energy and potential energy was summarized to identify violence behavior. Finally, lots of videos that include all kinds of behaviors were used to carry out experiments and draw the conclusion that this method has a higher identification rate when handles sudden violence incident.

Key words: sudden violence, optical-flow feature, motion feature, energy function, quality weight