基于视频区域特征的行人异常行为检测

从区域特征分析的角度检测行人的异常行为，先进行背景建模，然后通过运动目标检测，得到运动目标区域信息，采用最小外接矩形框作为区域特征，最后计算该矩形长宽比并进行曲线拟合、预测，进而完成异常行为检测。实验结果显示，该方法简单、快速。

Pedestrian abnormal behavior detection based on region feature

Abnormality of pedestrian was detected based on region feature in this paper. The principle of the method is simple but effective. Firstly, moving object was detected with mixture of Gaussians background model and the minimal rectangle of the object whose ratio of height to width was taken as the feature to measure abnormality of the pedestrian. Experiments demonstrate the approach proposed is fast and effective.