人群的密度估计对于建筑物的设计和人群安全至关重要。论文提出了一种新的人群密度自动估计方法。对于低密度人群图像采用基于像素统计的方法,对于较高密度人群图像采用基于多尺度分析和分形的纹理分析方法.并应用支撑向量机进行人群密度等级分类。对人群图像集的实验表明,该方法较以前的方法更为准确有效。
Crowd density estimation is very important to building designs and crowd safety.In this paper,a new crowd density estimation technique is proposed.A pixel-counting based method is applied to low density images whereas a texture analysis method based on multi-scale analysis and fractal is implemented to higher density images which are classified by SVM.The results from experiments on the crowd image set indicate that the new technique presented by this paper is more accurate and efficient than previous ones.