基于HOG的快速行人计数算法  

Fast Pedestrians Counting Algorithm Based on HOG

作　　者：李峰松[1] 张运楚[1,2] 杨红娟[1,2] 王业篷 

机构地区：[1]山东建筑大学信息与电气工程学院,济南250101 [2]山东省智能建筑技术重点实验室,济南250101

出　　处：《计算机系统应用》2014年第5期172-176,共5页Computer Systems & Applications

基　　金：国家自然科学基金(61303087);住房和城乡建设部科学技术项目(2011-K9-20)

摘　　要：为解决智能监控中空间受限情景下的行人计数问题,设计了基于HOG的行人识别与计数算法.通过限制HOG目标检测算法中图像的缩放范围,并设置感兴趣的检测区域,相比直接应用OpenCV函数实现的目标检测,其时间效率大幅提升;采用基于单计数线算法完成双向流量统计.三个不同走廊环境的实验表明,本文算法具有不依赖帧间运动信息的特点,可快速稳定的对空间受限情景下的行人进行计数.A rapid pedestrian detection and counting algorithm based on HOG is proposed for getting the pedestrian flow passing the confined space. The time efficiency is increased remarkable by restricting window shrink range and setting ROI. Counting method based on single line is presented for getting tow directions people number. Experiments in three different passageway show that the proposed method don＇t depend on moving information, besides it has the ability of fast and stable in confined space pedestrian counting.

关 键 词：智能监控 目标检测 行人计数 HOG 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]