摘要
为消除表情变化对人脸识别结果的影响,构造相对恒定的人脸识别特征,提出了一种恒定特征构造方法:在人脸曲面上分割出基本不受表情变化影响的区域;对该区域原始数据进行统一规格的二次采样,保证数据精度的一致性;分析特征区域的数据,得到原始特征向量;变换、组合特征向量,构造此区域的恒定特征向量;将此特征用于人脸的分类识别,以克服表情变化的影响。通过在实际三维人脸数据库上的实验,验证了该方法的有效性。
In order to eliminate the variations of the facial expression and construct the constant characteristic for 3D face re-cognition,the paper presented a method for constructing the characteristic.First,analyzed the 3D face data and found the nose region and other similar regions.Second, resampled the regions data for uniform the data’s precision.Third, processed the data in these regions to get the original characteristic.Finally,transformed and combined the original characteristic for ultimately constant one. Using this characteristic in face recognition for improving the recognise precision . The experiment results based on the actual 3D face data demonstrate the validity of the method.
出处
《计算机应用研究》
CSCD
北大核心
2010年第5期1938-1940,1943,共4页
Application Research of Computers
基金
国家"863"计划资助项目(2009AA01Z315)
高等学校科技创新工程重大项目培育基金资助项目(708085)
关键词
人脸识别
坐标校正
重采样
主分量
恒定特征
face recognition
coordinate normalize
resampling
primary components
constant characteristic
