基于DOG的异常行为监测模型的设计

Detection model of abnormal behavior based on DOG

为了在监控系统中识别行人的动作,提出了一种用于异常行为预测和检测的实时视频监控系统模型,模型基于动态有向图(DOG)进行实时无人监控。通过单向连接点结构方式来描述观测行为,其中每个点定义了被观测的运动物体在标准属性多维空间的区域,并对产生异常行为的人确定可能性。实验结果表明,该方法能够成功跟踪运动的目标并区分其行为类型,跟踪效果可靠、精确。

In order to identify pedestrian movement in monitoring system, an abnormal behavior analysis module of the system is presented, using real-time unsupervised. It called dynamic oriented graph （DOG） is used to predict and detect abnormal behaviors. The DOG method characterizes observed actions by a structure of unidirectional connected nodes, each one node defining a region in the hyperspace of attributes measured from the observed moving objects and having assigned a probability to generate an abnormal behavior. Experimental result show that the DOG method can track successfully moving objects, and it is robust and accurate.