摘要
为了提高局部保持映射算法(locality preserving projections,LPP)对外部因素如光照、表情等变化的鲁棒性,提出了一种子模式局部保持映射算法(subpattern locality preserving projections,SPLPP)用于人脸识别。该算法首先将样本划分为多个子模式集,然后对每个子模式运用LPP算法,最后采用最近邻分类器分类。在Yale,ORL,FERET人脸图像测试库上的实验表明:SPLPP算法的性能优于PCA,LDA,LPP算法。
To improve the robustness of the Locality Preserving Projections（LPP）to variation in illumination,expression and so on,a novel approach called Subpattern Locality Preserving Projections（SPLPP） was proposed for face recognition.In the SPLPP framework,samples were first divided into many non-overlapping subpattern sets and then LPP method was used on each of set,nearest neighbor classifier was used for classification in the end.The experimental results on Yale,ORL,FERER face databases illustrate that the proposed SPLPP method is superior to PCA,LDA and LPP in terms of recognition rate.
出处
《重庆理工大学学报（自然科学）》
CAS
2011年第6期84-89,共6页
Journal of Chongqing University of Technology：Natural Science
基金
国家"863"高技术研究发展项目(2007AA01Z423)
重庆市科技攻关项目(CSTC
2008AB5038)
关键词
人脸识别
鲁棒性
局部保持投影
子模式
face recognition
robustness
locality preserving projections
subpattern
