基于可变矩形框的人群密度估计算法

Crowd Density Estimation Algorithm based on Changeable Rectangle

随着经济的发展,在一些公共场合人流量越来越大。传统的人工监控因为其局限性,已不能满足实际需要,于是出现了基于图像处理技术的智能视频场景监控系统。针对高人流密度场景的现状,在基于像素数的人群密度估计算法的基础上,提出了基于可变矩形框的人群密度估计算法。该算法主要由最小二乘法线性拟合和检测算法2部分组成。实验结果表明该算法不仅能减小视频中行人大小变化所造成的像素数非线性变化误差,而且能较有效地消除噪声点的影响,具有较高的人群密度估计准确度和鲁棒性。

With economic development,in some public places appeares more and dense stream of people.Because of its limitations the traditional manual control could not meet the actual needs,so based on image processing technology the intelligent video surveillance system takes its place.In the light of crowd density estimation algorithm based on pixel number,a crowd density estimation algorithm based on a changeable rectangle is proposed for the high pedestrian density scene.The algorithms mainly composed of linear fitting algorithm based on least square method and detection algorithm.Experimental results show that the proposed algorithm could reduce the nonlinearity error of pixel number caused by the variable size of pedestrian in video sequence,and effectively overcome the effect of noisy points.The algorithm is of high crowd density estimation accuracy and robustness.