摘要
群体异常行为的发生将对社会公共安全构成危害,而当发生异常行为的场景中人群密度越高时,对应的危害程度也就越高。因此,通过对群体的密度特征进行学习分类,提出一种适合于中、高密度人群的群体异常行为分级算法。算法首先根据ViBe算法对场景中人群密度比例进行阈值判断,当该比例超过阈值时,采用基于纹理的分析方法对该场景进行密度估计,并以此为依据,对场景中的异常行为进行分级。实验结果表明,该算法能对中、高密度场景中的人群异常行为进行有效合理分级,具有较强的鲁棒性。
Crowd abnormal behaviors would risk the social public security,when the crowd density grows in the scenario where abnormal behaviors take place,the corresponding risk degree would also go up.By studying and sorting the density characteristic of the crowds,the article tries to propose a classification algorithm suitable for the abnormal behaviors of medium and high density crowds.The algorithm first carries a threshold judgment on the proportion of the medium and high density crowds in the scenario according to the ViBe algorithm,when the proportion exceeds the threshold.It evaluates the density of the scenario with a texture-based analytical method,and takes this as the basis to classify the abnormal behaviors in the scenario.The experimental result shows that the algorithm has performed an effective and rational classification to the crowd abnormal behaviors of the medium and high density scenario and with a strong robustness.
作者
汪晓飞
杨龙成
胡俊
杨晓玲
WANG Xiaofei;YANG Longcheng;HU Jun;YANG Xiaoling(College of Computer Science,Chengdu Normal University,Chengdu 611130,China;Editorial Department of Journal,Chengdu Normal University,Chengdu 611130,China;High School Key Laboratory on Pattern Identification and Intelligent Information Processing,Chengdu University,Chengdu 610106,China)
出处
《成都师范学院学报》
2019年第3期118-124,共7页
Journal of Chengdu Normal University
基金
四川省科技厅应用基础项目"基于粒子流的大型公共场合人群流疏散模型应用研究"(2016JY0199)
成都大学模式识别与智能信息处理高校重点实验室开放基金项目"基于流体力学的突发事件下人群行为分析研究"(MSSB-2016-7)
成都师范学院科研培育项目"基于视频的人群骚乱行为检测"(CS15ZD02)
成都师范学院2018年科研培育项目"基于群体行为特征的人群骚乱事件检测"(CS18ZD03)
