CAJ | 학술논문

本文系统地研究了视频监控系统中多目标多特征实时检测及其实现技术。通过在视频带中完整记录可疑目标的行为，为此类事件定义一套标准。本系统能够自动的从视频序列中提取可疑人物或事件的若干初级特征，而这些特征的有机融合包含足够的信息，可以决定是否发生了非常事件。
본문계통지연구료시빈감공계통중다목표다특정실시검측급기실현기술。통과재시빈대중완정기록가의목표적행위，위차류사건정의일투표준。본계통능구자동적종시빈서렬중제취가의인물혹사건적약간초급특정，이저사특정적유궤융합포함족구적신식，가이결정시부발생료비상사건。
The main task of the dissertation is the whole studying of the multi-features detection methods aiming at multi-surveillance objects.By analyzing objects behaviors in Video sequences it should define a set of criteria for events,which deviate from normal.The system automatically extracted several simple features from the objects in their entire presence in the scene.A combination of the low level features did contain sufficient information to make a decision if the abnormal event has occurred.