[1]严云洋,等.融合多尺度多特征的人脸识别方法[J].南京理工大学学报(自然科学版),2009,(01):47-52.
 YAN Yun-yang,GUO Zhi-bo,CHEN Fu-bing,et al.Face Recognition Based on Fusion of Multi Features with Multi Scales[J].Journal of Nanjing University of Science and Technology,2009,(01):47-52.
点击复制

融合多尺度多特征的人脸识别方法
分享到:

《南京理工大学学报》(自然科学版)[ISSN:1005-9830/CN:32-1397/N]

卷:
期数:
2009年01期
页码:
47-52
栏目:
出版日期:
2009-02-28

文章信息/Info

Title:
Face Recognition Based on Fusion of Multi Features with Multi Scales
作者:
严云洋1 2 郭志波2 3 陈伏兵2 杨静宇2
1. 淮阴工学院计算机工程系, 江苏淮安223001;2. 南京理工大学计算机科学与技术学院, 江苏南京210094; 3. 扬州大学信息学院, 江苏扬州225009
Author(s):
YAN Yun-yang12GUO Zhi-bo23CHEN Fu-bing2YANG Jing-yu2
1.Department of Computer Engineering,Huaiyin Institute of Technology,Huaian 223001,China;2.School of Computer Science and Technology,NUST,Nanjing 210094,China;3.School of Information Engineering,Yangzhou University,Yangzhou 225009,China
关键词:
多尺度 低频 奇异值 特征融合 人脸识别
Keywords:
multi scales low frequency singular values feature fusion face recognition
分类号:
TP391.41
摘要:
为降低光照、噪音、姿态等变化的影响,减少有效局部信息的损失,提出了使用图像的变换特征,及多尺度分块线性鉴别分析的算法。将图像进行多尺度划分,对划分后的每个子图像分别抽取其低频部分或奇异值,组合起来作为该图像的特征向量,进行线性鉴别分析。针对单一特征表示图像时的局限性,又提出了融合多尺度低频特征和多尺度奇异值特征进行人脸识别的方法。在ORL和Yale人脸库上的实验结果显示,所提出的算法识别精度明显提高,泛化能力较强。
Abstract:
To avoid some influences caused by lighting condition,noise and gesture,modular linear discriminant analysis(MLDA) is developed here for face recognition based on the transform feature of each sub-image after multi-scale division.Using the new LDA,local effective discriminant character is lost little.An image is divided with different scales.The coefficients with low frequency and the singular values of each sub-image are derived respectively.All coefficients or singular values are combined together and used as the feature vector of the image.LDA is implemented based on these feature vectors.In view of the limitations to express a face using only one feature,fusion of low frequency characters with singular values of an image for face recognition is presented.The experimental results performed on ORL or Yale face data set show that the proposed method is obviously superior to corresponding algorithms in terms of recognition rate and generalization ability.

参考文献/References:

[ 1] 杨静宇, 金忠, 郭跃飞. 人脸图像有效鉴别特征抽取与识别[ J] . 南京理工大学学报( 自然科学版),2000, 24( 3): 193- 198.

[ 2] Belhum eur P N, H espanha J P, Kr iegm an D. E igenfacesvs. fisherfaces: Recogn ition using c lass spec ific linearpro jection[ J]. IEEE Transactions on Pattern AnalysisandMach ine Intelligence, 1997, 19( 7): 711- 720.
[ 3] H ong Z-i quan. A lgebra ic feature extraction of im age forrecogn ition[ J]. Pa ttern Recogn ition, 1991, 24 ( 3):211- 219.
[ 4] 洪子泉, 杨静宇. 用于图像识别的图像代数特征抽取[ J] . 自动化学报, 1992, 18( 2): 233- 237.
[ 5] 洪子泉, 杨静宇. 基于奇异值特征和统计模型的人脸识别算法[ J]. 计算机研究与发展, 1994, 31 ( 3):60- 65.
[ 6] Tian Yuan, Tan T ie-n iu, W ang Yun-hong, e t a .l Dosingu lar values conta in adequate inform ation for facerecognition[ J] . Pattern Recogn ition, 2003, 36 ( 6 ):649- 655.
[ 7] 王蕴红, 谭铁牛, 朱勇. 基于奇异值分解和数据融合的脸像鉴别[ J]. 计算机学报, 2000, 23( 6): 649- 653.
[ 8] 甘俊英, 张有为. 一种基于奇异值特征的神经网络人脸识别新途径[ J]. 电子学报, 2004, 32( 1): 170- 173.
[ 9] 刘小军, 王东峰, 张丽飞, 等. 一种基于奇异值分解和隐马尔可夫模型的人脸识别方法[ J] . 计算机学报, 2003, 26( 3) : 340- 344.
[ 10] 杜干, 朱雯君. 基于局部奇异值分解和模糊决策的人脸识别方法[ J]. 中国图象图形学报, 2006, 11( 10):1456- 1459.
[ 11] Fo lye D H, Sammon JW. An optim al set o f discr im -inant vec to rs [ J ]. IEEE T ransaction on Com puter,1975, 24( 3): 281- 289.
[ 12] Chen L F, L iao H Y M, Ko J C, et a.l A new LDAbasedface recogn ition system wh ich can so lve thesm a ll samp le size problem [ J]. Pa ttern Recognition,2000, 33: 1713- 1726.
[ 13] Ra jk iran G, V ijayan K A. An improv ed face recogn-ition techn ique based on m odular PCA appro ach [ J].Pa ttern Recognition Letters, 2004, 25: 429- 436.
[ 14] 胡钟山, 娄震, 杨静宇, 等. 基于多分类器组合的手写体数字识别[ J]. 计算机学报, 1999, 22( 4): 369- 374.
[ 15] 杨健, 杨静宇, 王正群, 等. 一种组合特征抽取的新方法[ J]. 计算机学报, 2002, 25( 6): 570- 575.
[ 16] Y ang Jian, Y ang Jing-yu, Zhang Dav id, et a.l Featurefusion: Pa ra lle l strategy vs. serial strategy [ J]. Pa-ttern Recogn ition, 2003, 36: 1369- 1381.
[ 17] Ludm ila I K. A theoretica l study on six classifier fusionstra teg ies[ J]. IEEE Trans on PAM I, 2002, 24 ( 2):281- 286.
[ 18] Jin Zhong, Yang Jing-yu, H u Zhong-shan, e t a .l Facerecognition based on uncorre lated d iscrim inan t transformation[ J]. Pa ttern Recogn ition, 2001, 34 ( 7 ):1405- 1416

相似文献/References:

[1]陆春义,周彦煌,余永刚,等.底排装置低频振荡燃烧和永久熄灭[J].南京理工大学学报(自然科学版),2009,(01):112.
 LU Chun-yi,ZHOU Yan-huang,YU Yong-gang,et al.Low Frequency Oscillation Combustion and Extinguishment of Base Bleed Unit[J].Journal of Nanjing University of Science and Technology,2009,(01):112.
[2]王健,吴军基,陶钢,等.射流低频高声强发生器设计与实验研究[J].南京理工大学学报(自然科学版),2009,(03):344.
 WANG Jian,WU Jun-ji,TAO Gang.Design and Experiment on Fluidic Low-frequency and High-intensity Sound Generator[J].Journal of Nanjing University of Science and Technology,2009,(01):344.
[3]丁红星,戴丽莉,沈中华,等.准一维声子晶体中兰姆波低频禁带特性研究[J].南京理工大学学报(自然科学版),2013,37(06):869.
 Ding Hongxing,Dai Lili,Shen Zhonghua,et al.Low-frequency band gap performance for Lamb waves in quasi-one-dimensional phononic crystal[J].Journal of Nanjing University of Science and Technology,2013,37(01):869.

备注/Memo

备注/Memo:
基金项目: 国家自然科学基金( 60632050 ); 江苏省高校自然科学基金( 08KJB520001, 06KJD520024,07KJD510021); 淮安市科技发展基金(HAG07063); 江苏省/ 青蓝工程0项目
作者简介: 严云洋( 1967- ), 男, 教授, 博士, 主要研究方向: 数字图像处理, 模式识别, E-m a il:areyyyke@ 163.com; 通讯作者: 杨静宇( 1941- ), 男, 教授, 博士生导师, 主要研究方向: 计算机视觉、模式识别、智能机器人等, E-m ail:yang jy@ m a il. njust. edu. cn。
更新日期/Last Update: 2012-11-19