中文摘要：

随着近年来危害公共安全的群体性事件频繁发生,对人群场景下的人群状态分析与异常行为检测成为计算机视觉领域研究的热点问题。目前提出的诸如纯光流法,社会力模型,时空运动模型等算法,在检测准确率方面已能满足需求,但是算法复杂度普遍较高,运算量较大,在实际应用中难以保证实时性。鉴于此,首先,引入了人群移动区域面积的概念并定义了人群状态指数,来描述人群状态的变化,通过光流法获得人群运动矢量场,基于人群运动矢量场定义了人群运动强度指数来描述人群运动强度,基于人群运动矢量场与信息熵定义了人群混乱指数来描述人群运动方向的混乱程度。其次,基于降低算法运算量的考虑,根据上面提到的三个描述人群运动状态的特征变量设计了一种分层处理的人群异常行为检测方案,实验结果证明方案具有很好的效果。

英文摘要：

With the frequent occurrence of mass incidents in recent years,the analysis of crowd status and abnormal behavior detection in the crowd scene become the hot topic in the field of computer vision. At present,most of the algorithms can achieve good detection accuracy,but these algorithms have very high complexity and large computation,so it is difficult to guarantee real-time character of the monitor system,such as pure optical flow,social force model and spatial-temporal model. In view of this,the concept of crowd moving area was introduced and the index of crowd status was defined base on crowd moving area to indicate the changes of crowd status. The crowd motion vector field is obtained by optical flow method. The index of crowd motion intensity was defined base on crowd motion vector field to describe the intensity of crowd motion. The index of crowd chaos was defined base on crowd motion vector field and entropy to express the degree of crowd chaos. Furthermore,in order to reduce the computation of algorithm,a hierarchical processing scheme was designed to detect the abnormal crowd behavior. The experimental results show that the method has good effect.