基于个体特征和纹理特征的视频人数统计算法  

People counting based on individual feature and texture feature in video surveillance

作　　者：刘红[1] 宋茹[1] 王保兴[1] 

机构地区：[1]安徽大学计算智能与信号处理教育部重点实验室,安徽合肥230039

出　　处：《安徽大学学报（自然科学版）》2015年第3期47-50,共4页Journal of Anhui University(Natural Science Edition)

基　　金：安徽大学博士科研启动基金资助项目(33190049)

摘　　要：视频人数统计利用视频图像特征,通过监测公共场所中的人群密度,可防止公共场所人群拥堵,确保行人安全.提出一种改进的视频人数统计算法,对于中低密度人群,利用个体特征法实现人数统计,对于高密度人群,利用纹理特征法实现人数统计.使用提出的算法,设计了视频人数统计系统,分别对多组视频进行了测试,测试结果表明该算法误差较低.Image feature is used in video people counting to monitor the crowd density in public place.Getting the information of the crowd density in real-time is an effective way to prevent personnel congestion in public places.An improved algorithm of people counting was proposed in this paper.For middle and low crowd density,the algorithm was based on using individual features to achieve people counting.For high crowd density,the algorithm was based on using texture features to achieve people counting.Using the proposed algorithm,a people counting system was designed.Several videos were tested respectively.Experimental results showed that the algorithm had an insignificant error.

关 键 词：视频监控 人数统计 HOG行人检测 灰度共生矩阵 最小二乘法 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]