基于多点模型和改进BFGS的人脸姿态估计  

FACE ATTITUDE ESTIMATION BASED ON MULTI-POINTS MODEL AND MODIFIED BFGS

作　　者：邱丽梅[1] 李青虹[1] 

机构地区：[1]三明学院物理与机电工程系,福建三明365000

出　　处：《计算机应用与软件》2011年第9期126-129,共4页Computer Applications and Software

基　　金：福建省教育厅科技项目(JK2009037)

摘　　要：针对人脸姿态估计往往存在姿态解不稳定和不能唯一确定人脸三维空间姿态的缺陷,准确提取人脸特征点及进行相应特征点深度值估计后,以人脸的多个特征点建立人脸模型,并利用人脸特征点近似估计人脸姿态,通过改进的BFGS(Broyden-Fletch-er-Goldfarb-Shanno)算法精确估计三维人脸空间姿态。实验结果表明,该方法不仅可以获得稳定和唯一的3D人脸空间姿态,而且与同类方法比较具有良好的姿态估计精度。To overcome the defects in face attitude estimation that there sometimes has the unstable attitude solution and that the 3D space face attitude can not be uniquely determined,by using the modified algorithm of BFGS the 3D space face attitude was able to be estimated accurately after the face feature points were extracted accurately and were estimated the depth value of their corresponding feature points,establishing the face model with multiple face feature points as well as estimating the face attitude approximately with face feature points.Experimental results show that the method can acquire steady and unique 3D face attitude,and can also attain better attitude estimation accuracy compared with methods of the same kind.

关 键 词：人脸姿态估计 多点模型 特征点 BFGS 

分 类 号：TN911.73[电子电信—通信与信息系统]