一种基于多特征融合的头部检测算法  

A Head Detection Algorithm Based on Multi-feature Fusion

作　　者：刘恋[1] 贺赛先[1] 熊杰[1] 

机构地区：[1]武汉大学电子信息学院,武汉430072

出　　处：《电光与控制》2016年第9期106-110,共5页Electronics Optics & Control

基　　金：国家"九七三"计划资助项目(2012CB725301);空间位置智能传感技术集成与应用研究(210600001)

摘　　要：在基于图像识别技术的教学录播系统中,通常需要对发言学生进行自动识别和定位,以便摄像机给出该学生的特写。针对该问题提出了一种基于多特征融合的室内学生头部检测算法。该算法首先对头发的颜色特征设定阈值二值化图像来得到头部的候选区域,然后融合人体头发区域的形状特征以及头发下方脸部的肤色特征确定全体学生头部区域。当发言学生起立后,可通过头部轮廓中心点在视频图像竖直方向的坐标变化判断该学生。应用该算法对课堂多个角度的视频图像进行处理,实验结果表明该算法能有效检测出学生头部区域。The teaching recording and broadcasting system based on image recognition technology usually requires the close-up of the indoor students from the camera by automatic identification and position.A head detection algorithm based on multi-feature fusion for the indoor students is proposed to solve this problem.Firstly,head candidate regions are extracted through setting a threshold value based on the color characteristics of human hair.Then the head region of all the students can be detected by shape features of the hair region and skin color features of the face region.Finally,the speaking student can be captured by detecting the position change of the center of the contour in the vertical direction of the image when he stands up.The method was used for image processing of several angles of classroom video,and the experiment results show the effectiveness of it in detecting the head region of students.

关 键 词：多特征融合 头部检测 轮廓提取 教学录播系统 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]