结合多属性的视频中全局异常事件检测方法

The Global Abnormal Event Detection Method with Multiple Attributes in Video

针对视频序列中的全局异常事件,提出了一种结合多属性的异常事件检测方法.首先对视频序列按位置提取目标时空块,以时空块为节点,采用三维尺度不变特征变换描述子(3D-SIFT)作为时空特征,计算节点的光流属性与熵属性,并结合各个节点之间的KL距离构建时空检测模型.实验结果表明,该方法具有良好的鲁棒性,对全局异常事件具有较好的检测性能.

This paper proposed a novel approach that the global abnormal event detection with multiple attributes in video sequences. At first the video sequence is cropped into spatio-temporal volumes, based on locations, each of which is considered to be a node. Extract 3-dimensional SIFT of targets in nodes as spatio- temporal features, and a detection model for abnormal behavior is constructed using the entropy attribute, optical flow velocity of each node and the KL divergence between nodes. Experimental results show that the proposed method have better detection performance in detecting the global abnormal events than state-of-the- art approaches.