复杂监控场景下的人群密度估计与实现

作者：苏丹 陈永生

来源：《电脑知识与技术》2016年第35期

        摘要：人群密度作为智能监控中的重要特性，逐渐成为人们关注的焦点。针对复杂的监控场景，提出一种结合人群像素特征以及纹理特征的混合人群密度监控算法。首先引入Vibe对提取人群的前景图像，并通过形态学处理消除人物远近造成的误差。然后统计前景像素占整幅图像的比例，最后，对比例上下的两种情况分别用基于像素特征的算法以及基于纹理特征的算法。试验结果表明，该算法可以应对各种密度场景下的人群密度检测，准确率高，计算简单。

        关键词：人群密度估计；阈值分割；Vibe小波变换 ；灰度共生矩阵

        中图分类号：TP391 文献标识码：A 文章编号：1009-3044（2016）35-0202-04

        Crowd Density Estimation Algorithm under Complicated Monitoring Scenarios

        SU Dan，CHEN Yong-sheng

        （Department of Computer Science，Tongji University， Shanghai 201804， China）

        Abstract：As an important characteristic of intelligent monitoring， crowd density has gradually become the focus of attention. Aiming at the complex monitoring scene， this paper proposes a hybrid population density monitoring algorithm which combines the pixel features of the population and the texture features. First， Vibe was used to extract the foreground image of the crowd， and the error caused by the distance between characters was eliminated by morphological processing. Then the proportion of the foreground pixel is calculated. Finally， the two cases are compared with the algorithm based on the pixel feature and the texture feature. The experimental results show that the algorithm can deal with the population density detection under various density scenes， with high accuracy and simple calculation.