摘要
针对城市轨道交通监控等方面的应用需求,设计并实现了一个人群异常行为检测系统。目前,大部分的监控系统只能进行运动检测和跟踪,而本文设计的系统主要是对待检视频进行人群密度分析和人群运动分析,根据密度分析和运动分析得到的结果推断人群行为是否异常。其中密度分析部分主要是利用纹理特征分析结合支持向量机分类,实现了密度分级。运动分析部分主要是改进传统的块匹配算法,再结合支持向量机分类,实现人群运动分析。
Focusing on the application requirement of urban rail transit surveillance,this paper designed and implemented Crowd Abnormal Behavior Detection System.Most of Surveillance Systems could only detect or track movement.This System could make an analysis of crowd density and crowd motion for videos.According to the results,the System infered a conclusion whether the crowd behavior was abnormal.About crowd density analysis,it combined texture characteristic analysis with support vector machine classification to classify the crowd density.About crowd motion analysis,it improved traditional block matching algorithm and combined block matching with support vector machine classification to analyze crowd motion.
出处
《铁路计算机应用》
2010年第7期37-41,共5页
Railway Computer Application
基金
国家自然科学基金项目(60973061)
教育部创新团队计划(IRT0707)
关键词
人群行为
异常
块匹配
纹理
支持向量机
crowd behavior
abnormal
block matching
texture
support vector machine