基于位置特征的运动行人检测与跟踪方法  

A method of detecting and tracking moving pedestrians based on position features

作　　者：姚亚夫[1] 刘佰昂[1] 

机构地区：[1]中南大学机电工程学院,湖南长沙410083

出　　处：《广西大学学报（自然科学版）》2009年第6期769-773,共5页Journal of Guangxi University（Natural Science Edition）

基　　金：国家863项目(2009AA11Z206)

摘　　要：针对在静态背景视频中行人检测和跟踪时受行人运动状态影响较大,容易产生误识别,提出了一种基于位置特征的运动行人检测与跟踪方法。检测阶段得到运动行人的二值化差分图后,引入水平融合值、垂直融合值将满足融合要求的非连通区域外接矩形融合得到一个新矩形,以新矩形及其中心作为行人的位置特征,对行人进行检测与跟踪。用该方法跟踪行人视频中的两个行人,跟踪准确率分别为98%与95%。In this paper, a method of detecting and tracking moving pedestrians is studied, based on position features in static scenes, to solve problem of fluctuated image caused by moving that can easily resulted in false identification. Firstly getting the binary image segmentation of moving pedestrian during the process of tracking, then introducing horizontal merging value, vertical merging value to merge the contour-rectangle of some separated areas, lastly taking the new rectangle and its center as the position character of moving pedestrian to detect and track moving pedestrian. Experimental results of two moving pedestrians＇ scenes show that 98% of one pedestrian can be detected and tracked and the other is 95%.

关 键 词：行人检测 行人跟踪 水平融合值 垂直融合值 位置特征 

分 类 号：U491.226[交通运输工程—交通运输规划与管理]