摘要
为了构造一个精度高、处理快速的人脸检测系统,本文结合颜色信息和人脸构造特征,提出了一种从序列图像中实现人脸分割与识别的新方法。该方法基于混合高斯模型与贝叶斯判别技术,分两步实现,即运动目标分割和人脸特征识别。首先,通过建立自适应的混合高斯模型,对背景图像建模,分割出前景运动人体区域;然后采用贝叶斯判别法对人脸区域进行识别。实验结果显示,该方法具有很高的检测精度和较低的漏警率,具有良好的实用价值和工程应用前景。
In order to build up a effective and rapid system for detection by combining the skin color information and basic human face features, this paper presents a new method for human face segmentation and recognition in color image series. There are mainly two steps in our algorithm： moving object segmentation and facial feature recognition. First, Gaussian models are built for background. Then, Bayesian discriminating classifier is used for making judgment of the possible face area and gets the face detection result. Experimental illustrate the high detection precision and low false alarm of the proposed algorithm. It shows this new algorithm has practical worthiness and good outlook engineering application.
出处
《微计算机信息》
北大核心
2008年第27期278-279,273,共3页
Control & Automation
基金
上海市高等学校科技发展基金
项目名称:基于计算机视觉的人体运动检测
跟踪
识别技术的研究颁发部门:上海市教育委员会(06AZ017)项目申请人:王唯一
关键词
运动检测
人脸识别
混合高斯模型
贝叶斯判别
motion detection
face recognition
Gaussian modeling
Bayesian discriminating
