基于运动轨迹的徘徊行为检测

Wandering behavior detection based on trajectory of moving object

针对金融场所(比如银行)的智能视频监控系统中徘徊行为的检测进行了研究,采用了基于运动轨迹的方法检测徘徊行为。利用混合高斯模型和背景差分法,画出运动物体的外接轮廓,通过计算外接轮廓质心移动的路程和方向来判断是否进行徘徊行为。实验结果表明,这种方法能够简单快速的检测出徘徊行为,并且不用进行任何的样本进行训练,大大地节约了时间和空间。

The detection of the event of wandering in intelligent video surveillance systems in banks is studied by the paper. The trajectory of moving object is used to detect wandering behavior. By using the multiple Gaussian model and background subtraction,an external contour of the object is drawn,and whether wandering behavior or not is determined by calculating the distance and direction of centroid trajectory. The experimental results show that the algorithm has a faster detection speed without training samples to save the computing time and storage space.