基于支持向量机的人脸识别应用研究  

Research on Face Recognition Application Based on SVM

作　　者：叶晓波[1] 秦海菲 吕永林 YE Xiaobo;QIN Haifei;LV Yonglin(Institute of Network & Information Systems, Chuxiong Normal University, Chuxiong, Yunnan Province 675000;School of Information Sciences & Technology, Chuxiong Normal University, Chuxiong, Yunnan Province 675000;School of Economics & Management, Chuxiong Normal University, Yunnan Province 675000)

机构地区：[1]楚雄师范学院网络与信息系统研究所,云南楚雄675000 [2]楚雄师范学院信息科学与技术学院,云南楚雄675000 [3]楚雄师范学院经济与管理学院,云南楚雄675000

出　　处：《楚雄师范学院学报》2019年第3期121-127,共7页Journal of Chuxiong Normal University

基　　金：云南省教育厅科学研究基金项目“基于网络教学平台的大学生行为习惯研究”(NO:2012Y131)

摘　　要：使用C-SVC和nu-SVC支持向量机模型,对人脸进行识别。以支持向量机分类原理为基础,以LIBSVM集成软件为工具,以剑桥大学计算机实验室的ORLFaces数据库和本文第一作者的人脸图片作为实验数据,对数据标准化实现方法,最优惩罚系数与核函数参数的查找方法,线性核函数、多项式核函数、径向基核函数、Sigmoid核函数进行对比研究,选择三组随机数进行人脸识别实验。实验结果显示,nu-SVC模型的人脸识别正确率高,且波动性小,更适合用于人脸识别;C-SVC模型用于人脸识别时指定惩罚参数C和核函数参数γ可获得更高的分类正确率。结论,SVM分类器用于人脸识别效果较好。In the research C-SVC and nu-SVC SVM models were used for face recognition. Three sets of random numbers were selected based on the classification principles of SVMs for experiment on face recognition to conduct a comparative study on realization methods of data standardization, identification methods of optimal penalty coefficient and kernel function parameters, linear kernel functions, polynomial kernel functions, radical basis functions and Sigmoid kernel functions with LSBSVM integration software, ORL Faces database of the Computer Laboratory of Oxford and the face pictures of the first author of the present paper. The experiment results show that nu-SVC model is the most suitable for face recognition for its high accuracy and low fluctuation while C-SVC model can achieve a higher classification accuracy due to its specified penalty parameter C and kernel function parameter γ when applied to face recognition. Thus the conclusion that SVM classifiers can obtain better face recognition results.

关 键 词：人脸识别 支持向量机 LIBSVM 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]