基于鱼眼视频图像的人群密度估计算法的研究  

Research on Estimation Algorithm of Crowd Density Based on Fisheye Video Image

作　　者：杨前华[1] 李尤 赵力[2] YANG Qianhua;LI You;ZHAO Li(Communication Institute,Nanjing College of Information Technology,Nanjing 210023,China;School of Information Science and Engineering,Southeast University,Nanjing 210018,China)

机构地区：[1]南京信息职业技术学院通信学院,南京210023 [2]东南大学信息科学工程学院,南京210018

出　　处：《电子器件》2020年第3期646-651,共6页Chinese Journal of Electron Devices

基　　金：南京信息职业技术学院基金项目(YK20180504)。

摘　　要：随着公共活动在规模和数量上的不断增长,人群的密度估计已成为人群监控安全领域的重要问题。采用基于纹理特征的估计方法,并针对常规相机在人群监控领域的局限性,采用鱼眼镜头对人群进行监控,通过对鱼眼图像进行了预处理、划分图像ROI区域等操作,进而提取基于图像纹理的9维特征向量,再利用SVM分类器,解析人群图像的密度等级信息。通过对比实验验证,该算法人群密度分类准确率明显高于仅使用灰度共生矩阵参数的分类结果,可见,基于鱼眼视频图像的人群密度估计方法是有效的,且适用于人群安全监控领域。With the continuous growth of public activities in scale and quantity,crowed density estimation has become an important issue in the field of crowd monitoring and security. To deal with the limitation of common camera involved in crowd monitoring,the fisheye lens have been applied to capture videos. In the process of crowd density estimation based on fisheye video image,after image preprocessing and division of ROI,nine-feature vectors involved in image texture are extracted,and then SVM classifier is applied to obtain density level information. The comparative experiment and analysis demonstrate the effectiveness of this algorithm.

关 键 词：鱼眼视频图像 人群密度估计 纹理特征分析 支持向量机分类器 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]