欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
为有效识别视频监控中的人体行为,提出了新的人体行为识别模型和前景提取方法.对前景提取,采用背景边缘模型与背景模型相结合的前景检测方法,有效避免了光照、阴影等外部因素的影响.为了快速发现人体运动过程中产生的新行为,采用分层Dirichlet过程聚类人体特征数据来判断是否有未知人体行为产生,用无限HMM对含有未知行为模式的特征向量进行有监督学习,由管理者将其添加到知识库中.当知识库的行为模式达到一定规模时,系统可以无监督地对人体行为进行分析.通过仿真实验证实了提出的方法在人体行为识别方面较其他方法具有独特的优势.
英文摘要：
For effectively recognizing human behaviors in video surveillance,a novel behavior recognition model and a foreground extraction method are presented.For foreground detection,combining background edge model and background model,a foreground detection method is proposed,which can effectively avoid the light,shadows and other external factors.To quickly find new behaviors produced in the process of human motion,a hierarchical Dirichlet process is adopted to aggregate monitored feature data for human body to determine whether unknown behaviors are produced or not.The infinite hidden Markov model（HMM） is adopted to learn unknown behavior patterns with supervised method,and then update the knowledge base.When knowledge base reaches a certain scale,system can analyze human behaviors with unsupervised method.Simulation experiments show that the proposed method has unique advantage over others for human behavior detection in real-time video surveillance.
同期刊论文项目
同项目期刊论文
位置：