利用归一化前景和二维联合熵的人群聚集检测方法

Pedestrians Gathering Detection Based on Normalized Foreground and Two-dimension Joint Entropy

摘要: 基于摄像机透视效应提出了一种场景归一化前景面积的计算方法;然后结合联合概率密度的概念设计出前景二值图的二维联合概率密度计算公式,并据此进一步计算出二维联合熵;最后,根据归一化前景面积和二维联合熵提出了一种人群聚集检测模型。实验表明,该模型可以实现对监控场景下人群聚集现象的快速有效检测。

Abstract: A normalized foreground computing approach based on the camera perspective effect is presented.A two-dimensional probability density for the binary foreground is calculated through classic joint probability distribution theory,and then the two-dimension joint entropy is calculated.Finally,a novel model based on normalized foreground and two-dimension joint entropy for detecting pedestrian gathering is proposed.Experimental results show that this model can quickly and effectively detect a pedestrian gathering event in a surveillance scene.