基于头部检测和跟踪的人数统计方法  

A HEAD COUNT METHOD BASED ON HEAD DETECTION AND TRACKING

作　　者：郑翔翔[1] 张怡[1] 

机构地区：[1]上海交通大学电子信息与电气工程学院,上海200030

出　　处：《计算机应用与软件》2011年第2期44-46,102,共4页Computer Applications and Software

基　　金：国家高技术研究发展计划项目(2007AA01Z157)

摘　　要：提出一种基于头部检测和跟踪的人数统计方法。方法利用Adaboost算法进行头部检测,将头部检测分成了头顶检测、侧面检测、正面检测和后脑勺检测等多个子检测过程。此外,利用图像特征直方图的信息进行头部跟踪,排除了对静态误检目标的统计,提高了检测的精度。实验结果验证了该方法的有效性。This paper presents a novel method of head count based on head detection and tracking.This method detects head with Adaboost algorithm,and divides the head detection into several sub detecting procedures,i.e.overhead,front-head,side-face and back-head detections.Besides,the information of image feature histogram is employed to track the target,and this excludes the false calculation with regard to static objects and improves the accuracy of counting.Experiments confirm the validity of the proposed method.

关 键 词：ADABOOST算法 类HAAR特征 运动轨迹跟踪 人数统计 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]