摘要
针对多姿态人脸识别中的姿态人脸生成问题,提出了一种局部加权平均的多姿态人脸生成算法.应用多项式拟合的方法,得到正面人脸与姿态人脸图像局部特征点之间的映射函数集,采用相邻特征点映射函数加权平均的方法获得每个像素的形变函数,据此形成多姿态的人脸图像;再利用主成分分析提取人脸特征矢量;最后应用支持向量机实现多姿态的人脸识别,从而解决了多姿态人脸识别中难以获得多姿态人脸图像和姿态变化导致的识别率迅速下降的问题.实验结果表明,使用局部加权平均算法生成的多姿态人脸能够保持人脸的局部特性,与ORL人脸库中的人脸的相似度高,能有效地提高多姿态人脸的识别率.
To solve the problem of generating multi-pose faces in multi-pose face recognition,we propose a multi-pose face generation algorithm based on local weighted mean method.We find the mapping function set of local feature points between front face and multi-pose face by applying polynomial fitting,and then design the transform function of each pixel for generating the multi-pose face by using weighted mean of mapping functions of neighboring feature points.At last,we chose principal component analysis to extract face feature vectors and utilize support vector machine to recognize the multi-pose faces.Our approach overcomes the difficulty of obtaining multi-pose face images in multi-pose face recognition and solves the problem of the rapid fall of recognition rate due to the face pose change.Experimental results show that the multi-pose faces produced by local weighted mean algorithm preserve the local features of faces and have a high similarity with the original faces in library ORL,thus effectively improving multi-pose face recognition rate.
出处
《江苏科技大学学报（自然科学版）》
CAS
2013年第1期74-79,共6页
Journal of Jiangsu University of Science and Technology:Natural Science Edition
关键词
多姿态
人脸识别
形变函数
局部加权平均
特征点
multi-pose
face recognition
transform function
local weighted mean
feature point
