零空间保局判别本征脸  

Null Space Locality Preserving Discriminant Intrinsicface

作　　者：楼宋江 张国印 

机构地区：[1]哈尔滨工程大学计算机科学与计算学院,哈尔滨150001

出　　处：《电子与信息学报》2011年第4期962-966,共5页Journal of Electronics & Information Technology

摘　　要：本征脸从人脸自身的差别出发,将每一人脸分为脸部共同差别、个体类间差别和个体类内差别,取得了较好的识别效果。但是它未考虑人脸的流形结构,并且会遇到矩阵的奇异性,即小样本问题。针对这些问题,该文提出了零空间保局判别本征脸,该算法充分考虑了个体类内差别和个体类间差别,结合流形学习思想并借助于判别准则使得投影后个体类内之间保持一定的相似性而个体类间之间的区分度有所增加。通过在个体类内保局差异散度矩阵的零空间中求最优特征向量,避免了矩阵的奇异性问题,解决了小样本问题。在人脸识别上的实验验证了算法的正确性和有效性。Based on the image differences,Intrinsicface is proposed,which divides the face image into three parts,common facial differences,intrapersonal differences and individual differences,and shows desirable performance.But it does not consider the manifold structure and suffers from the singular problem,which is also called Small Sample Size（SSS） problem.To solve these problems,Null Space Locality Preserving Discriminant Intrinsicface（NSLPDI） is proposed,which makes full use of intrapersonal differences and individual differences and employs the idea of manifold learning so that the similarity in the intra-class is preserved while the separability of samples from different classes is enlarged by discriminant criterion.The optimal feature vectors are extracted from the null space of intrapersonal locality preserving difference scatter matrix,which avoids the singularity and the SSS problem is solved.Experiments on face recognition demonstrate the correctness and effectiveness of the proposed algorithm.

关 键 词：人脸识别 本征脸 保局算法 小样本问题 零空间 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]