计算机工程与应用 ›› 2013, Vol. 49 ›› Issue (18): 164-166.

• 图形图像处理 • 上一篇    下一篇

基于离散曲率熵的徘徊行为检测

刘  强1,2,罗  斌1,2,翟素兰1,3,涂铮铮1,2,4   

  1. 1.安徽省工业图像处理与分析重点实验室,合肥 230039
    2.安徽大学 计算机科学与技术学院,合肥 230039
    3.安徽大学 数学科学学院,合肥 230039
    4.安徽大学 计算智能与信号处理教育部重点实验室,合肥?230039
  • 出版日期:2013-09-15 发布日期:2013-09-13

Loitering detection based on discrete curvature entropy

LIU Qiang1,2, LUO Bin1,2, ZHAI Sulan1,3, TU Zhengzheng1,2,4   

  1. 1.Key Lab of Industrial Image Processing & Analysis of Anhui Province, Hefei 230039, China
    2.School of Computer Science & Technology, Anhui University, Hefei 230039, China
    3.School of Mathematical Sciences, Anhui University, Hefei 230039, China
    4.Key?Lab?of?Intelligent?Computing?and?Signal?Processing?of?Ministry?of?Education, Anhui?University, Hefei?230039, China
  • Online:2013-09-15 Published:2013-09-13

摘要: 针对公共重点区域的智能监视问题,提出了一种新的徘徊行为异常检测方法。该方法利用视频目标跟踪算法得到可疑行人的运动轨迹,通过曲线拟合对运动目标的离散点轨迹进行平滑,计算离散点的离散曲率,计算感兴趣区域内运动目标轨迹点的离散曲率的熵及方差,通过离散熵阈值比较进行徘徊行为判断,该方法只需计算运动目标的轨迹,无需建立样本库,实验证明了该方法的有效性、实时性。

关键词: 徘徊行为, 异常检测, 曲线拟合, 离散熵

Abstract: In order to solve the intelligent monitoring problem, a novel method is introduced for loitering detection. Moving object is detected and tracked using tracking algorithm. This mothod gets the trajectory of moving object, smoothes the trajectory using curve fitting, and gets the discrete curvature and the entropy. Loitering detection is accomplished by judging threshold value. The method has no database for samples. Experiments demonstrate the efficiency of the proposed method.

Key words: loitering detection, abnormal detection, curve fitting, discrete curvature entropy