基于视听信息融合的智能监控系统  

An intelligent Surveillance System Based on A/V Information Fusion

作　　者：李超[1] 熊璋[1] 孟岩[1] 韩军[1] 

机构地区：[1]北京航空航天大学计算机科学与工程系,北京100083

出　　处：《计算机工程与应用》2004年第31期218-221,共4页Computer Engineering and Applications

摘　　要：传统的监控系统往往基于分离的视觉和听觉信息,没有充分利用场景中视听媒体间存在的天然时空相关性。文章在对传统监控系统进行实际研究的基础上,提出了视听信息融合的智能化监控系统。利用运动对象检测与跟踪技术提取视觉特征,通过音频内容分析抽取音频场景特征,并结合视听信息融合技术进行联合场景事件判断、决策和产生警报信息,实现多种媒体信息结合的监控与报警。实验结果证明这种方法可有效提高监控系统对事件和警报的处理效率与整体的自动化水平,具有实用价值,最后给出了原型设计方案。Conventional surveillance systems are always constructed with separated audio or video information.No existing spatial/temporal relationship between the two medias is adopted in such systems.An intelligent surveillance system based on audio /video information fusion technique is proposed,in which motion object detecting /tracking and audio content analysis techniques are used to abstract the individual perceptual features of a given scene.Then an appropriate layered alarm is activated from a decision generator.Experiments have proved the practical ability of such a system in improving the automatic response level and high efficiency on handling alarms.At last the design and implementation of such a system is given.

关 键 词：监控系统 特征提取 信息融合 运动对象检测 

分 类 号：TP319[自动化与计算机技术—计算机软件与理论]