计算机科学与探索 ›› 2022, Vol. 16 ›› Issue (12): 2851-2859.DOI: 10.3778/j.issn.1673-9418.2104088
收稿日期:
2021-04-26
修回日期:
2021-06-18
出版日期:
2022-12-01
发布日期:
2021-06-08
通讯作者:
+E-mail: wu_xiaojun@jiangnan.edu.cn作者简介:
蔡雨虹(1996—),女,湖南常德人,硕士研究生,CCF会员,主要研究方向为模式识别、人工智能。基金资助:
Received:
2021-04-26
Revised:
2021-06-18
Online:
2022-12-01
Published:
2021-06-08
About author:
CAI Yuhong, born in 1996, M.S. candidate, member of CCF. Her research interests include pattern recognition and artificial intelligence.Supported by:
摘要:
线性回归作为简单有效的工具,在模式识别中已得到广泛使用。但是直接从高维数据到二元标签可能无法得到灵活的投影和适合分类问题的数据表示。针对这一问题,标签松弛技术被提出,虽然已经证明其有效性,但仍然存在增大同类差异的问题。因此,提出类内低秩的子空间学习(ICLRSL),不同于原始线性回归和基于标签松弛的方法,在使用原始二元标签的同时采用两个投影矩阵分别完成类内低秩子空间投影和标签空间投影。ICLRSL将类内低秩子空间作为高维数据空间到标签空间之间的桥梁,得到对数据的初步编码,通过类内低秩约束使其与最终的回归目标拥有类似的类内相关性。同时,行稀疏约束保证子空间投影关注与类内低秩最相关的少数特征,在一定程度上降低冗余信息带来的负面影响。通过中间子空间的连接,一方面比直接学习单个投影矩阵具备更多灵活性,另一方面也能得到判别的数据表示。在四个公开人脸数据集上的实验验证了ICLRSL算法的有效性。
中图分类号:
蔡雨虹, 吴小俊. 用于人脸识别的类内低秩子空间学习[J]. 计算机科学与探索, 2022, 16(12): 2851-2859.
CAI Yuhong, WU Xiaojun. Intra-class Low-Rank Subspace Learning for Face Recognition[J]. Journal of Frontiers of Computer Science and Technology, 2022, 16(12): 2851-2859.
算法 | 4个样本 | 6个样本 | 8个样本 | 10个样本 |
---|---|---|---|---|
DLSR[ | 81.85 | 89.66 | 92.43 | 94.60 |
ReLSR[ | 82.96 | 89.75 | 93.19 | 94.61 |
GReLSR[ | 82.54 | 89.96 | 92.98 | 94.58 |
RLR[ | 81.30 | 91.05 | 94.62 | 96.13 |
DRR[ | 81.35 | 88.70 | 94.47 | 96.14 |
RLSL[ | 82.19 | 87.50 | 92.79 | 95.61 |
CDPL[ | 80.96 | 86.64 | 90.44 | 92.35 |
LC-KSVD[ | 68.97 | 81.19 | 85.91 | 88.68 |
LCPDL[ | 61.69 | 64.17 | 79.69 | 87.71 |
ICLRSL | 84.40 | 91.60 | 94.74 | 96.39 |
表1 各算法在AR数据集上的识别率 单位:%
Table 1 Classification accuracy of different algorithms on AR dataset
算法 | 4个样本 | 6个样本 | 8个样本 | 10个样本 |
---|---|---|---|---|
DLSR[ | 81.85 | 89.66 | 92.43 | 94.60 |
ReLSR[ | 82.96 | 89.75 | 93.19 | 94.61 |
GReLSR[ | 82.54 | 89.96 | 92.98 | 94.58 |
RLR[ | 81.30 | 91.05 | 94.62 | 96.13 |
DRR[ | 81.35 | 88.70 | 94.47 | 96.14 |
RLSL[ | 82.19 | 87.50 | 92.79 | 95.61 |
CDPL[ | 80.96 | 86.64 | 90.44 | 92.35 |
LC-KSVD[ | 68.97 | 81.19 | 85.91 | 88.68 |
LCPDL[ | 61.69 | 64.17 | 79.69 | 87.71 |
ICLRSL | 84.40 | 91.60 | 94.74 | 96.39 |
算法 | 15个样本 | 20个样本 | 25个样本 | 30个样本 |
---|---|---|---|---|
DLSR[ | 94.41 | 96.45 | 97.67 | 98.05 |
ReLSR[ | 93.98 | 96.14 | 97.75 | 98.28 |
GReLSR[ | 93.13 | 95.25 | 97.06 | 98.06 |
RLR[ | 94.79 | 96.58 | 97.92 | 98.32 |
DRR[ | 93.75 | 95.67 | 97.10 | 98.10 |
RLSL[ | 93.29 | 95.18 | 96.69 | 98.16 |
CDPL[ | 89.71 | 92.74 | 94.71 | 95.92 |
LC-KSVD[ | 93.23 | 95.11 | 96.65 | 97.23 |
LCPDL[ | 90.41 | 92.11 | 93.21 | 94.08 |
ICLRSL | 94.93 | 96.77 | 98.05 | 98.42 |
表2 各算法在Extended Yale B数据集上的识别率 单位:%
Table 2 Classification accuracy of different algorithms on Extended Yale B dataset
算法 | 15个样本 | 20个样本 | 25个样本 | 30个样本 |
---|---|---|---|---|
DLSR[ | 94.41 | 96.45 | 97.67 | 98.05 |
ReLSR[ | 93.98 | 96.14 | 97.75 | 98.28 |
GReLSR[ | 93.13 | 95.25 | 97.06 | 98.06 |
RLR[ | 94.79 | 96.58 | 97.92 | 98.32 |
DRR[ | 93.75 | 95.67 | 97.10 | 98.10 |
RLSL[ | 93.29 | 95.18 | 96.69 | 98.16 |
CDPL[ | 89.71 | 92.74 | 94.71 | 95.92 |
LC-KSVD[ | 93.23 | 95.11 | 96.65 | 97.23 |
LCPDL[ | 90.41 | 92.11 | 93.21 | 94.08 |
ICLRSL | 94.93 | 96.77 | 98.05 | 98.42 |
算法 | 10个样本 | 15个样本 | 20个样本 | 25个样本 |
---|---|---|---|---|
DLSR[ | 88.03 | 92.23 | 93.61 | 94.53 |
ReLSR[ | 87.87 | 91.62 | 93.40 | 94.65 |
GReLSR[ | 87.13 | 91.30 | 93.07 | 94.20 |
RLR[ | 89.90 | 92.28 | 93.69 | 94.66 |
DRR[ | 88.34 | 91.74 | 93.54 | 94.79 |
RLSL[ | 88.00 | 91.09 | 93.08 | 94.07 |
CDPL[ | 81.91 | 89.00 | 92.07 | 93.87 |
LC-KSVD[ | 77.83 | 81.97 | 84.15 | 85.22 |
LCPDL[ | 77.56 | 82.08 | 86.57 | 89.36 |
ICLRSL | 90.24 | 93.17 | 95.10 | 95.64 |
表3 各算法在CMU PIE数据集上的识别率 单位:%
Table 3 Classification accuracy of different algorithms on CMU PIE dataset
算法 | 10个样本 | 15个样本 | 20个样本 | 25个样本 |
---|---|---|---|---|
DLSR[ | 88.03 | 92.23 | 93.61 | 94.53 |
ReLSR[ | 87.87 | 91.62 | 93.40 | 94.65 |
GReLSR[ | 87.13 | 91.30 | 93.07 | 94.20 |
RLR[ | 89.90 | 92.28 | 93.69 | 94.66 |
DRR[ | 88.34 | 91.74 | 93.54 | 94.79 |
RLSL[ | 88.00 | 91.09 | 93.08 | 94.07 |
CDPL[ | 81.91 | 89.00 | 92.07 | 93.87 |
LC-KSVD[ | 77.83 | 81.97 | 84.15 | 85.22 |
LCPDL[ | 77.56 | 82.08 | 86.57 | 89.36 |
ICLRSL | 90.24 | 93.17 | 95.10 | 95.64 |
算法 | 识别率 |
---|---|
DLSR[ | 92.73 |
ReLSR[ | 91.82 |
GReLSR[ | 91.91 |
RLR[ | 93.32 |
DRR[ | 94.09 |
RLSL[ | 94.18 |
CDPL[ | 86.77 |
LC-KSVD[ | 82.41 |
LCPDL[ | 92.23 |
ICLRSL | 95.27 |
表4 各算法在FRGC数据集上的识别率 单位:%
Table 4 Classification accuracy of different algorithms on FRGC dataset
算法 | 识别率 |
---|---|
DLSR[ | 92.73 |
ReLSR[ | 91.82 |
GReLSR[ | 91.91 |
RLR[ | 93.32 |
DRR[ | 94.09 |
RLSL[ | 94.18 |
CDPL[ | 86.77 |
LC-KSVD[ | 82.41 |
LCPDL[ | 92.23 |
ICLRSL | 95.27 |
Algorithm | AR | Extended Yale B | CMU PIE | FRGC |
---|---|---|---|---|
LSR | 93.10 | 97.97 | 94.51 | 90.55 |
Eq. (5) | 95.90 | 97.87 | 95.32 | 94.45 |
Eq. (21) | 95.77 | 98.10 | 95.39 | 94.55 |
Eq. (22) | 96.03 | 97.96 | 95.26 | 95.18 |
ICLRSL | 96.39 | 98.42 | 95.64 | 95.27 |
表5 消融实验结果 单位:%
Table 5 Experimental results of ablation study
Algorithm | AR | Extended Yale B | CMU PIE | FRGC |
---|---|---|---|---|
LSR | 93.10 | 97.97 | 94.51 | 90.55 |
Eq. (5) | 95.90 | 97.87 | 95.32 | 94.45 |
Eq. (21) | 95.77 | 98.10 | 95.39 | 94.55 |
Eq. (22) | 96.03 | 97.96 | 95.26 | 95.18 |
ICLRSL | 96.39 | 98.42 | 95.64 | 95.27 |
[1] |
WRIGHT J, YANG A Y, GANESHA Y, et al. Robust face re-cognition via sparse representation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2009, 31(2): 210-227.
DOI URL |
[2] | ZHANG L, YANG M, FENG X C. Sparse representation or collaborative representation: which helps face recognition?[C]// Proceedings of the 2011 International Conference on Computer Vision, Barcelona, Nov 6-13, 2011. Washington: IEEE Computer Society, 2011: 471-478. |
[3] | HE X, NIYOGI P. Locality preserving projections[C]// Adv-ances in Neural Information Processing Systems 16, Van-couver and Whistler, Dec 8-13, 2003. Cambridge: MIT Press, 2004: 153-160. |
[4] | HE X F, CAI D, YAN S C, et al. Neighborhood preserving embedding[C]// Proceedings of the 10th IEEE International Conference on Computer Vision, Beijing, Oct 17-20, 2005. Washington: IEEE Computer Society, 2005: 1208-1213. |
[5] |
XIANG S M, NIE F P, Meng G F, et al. Discriminative least squares regression for multiclass classification and feature selection[J]. IEEE Transactions on Neural Networks and Learning Systems, 2012, 23(11): 1738-1754.
DOI PMID |
[6] | WANG L, PAN C. Groupwise retargeted least-squares reg-ression[J]. IEEE Transactions on Neural Networks and Lear-ning Systems, 2018, 29(4): 1352-1358. |
[7] |
FANG X, XU Y, LI X, et al. Regularized label relaxation linear regression[J]. IEEE Transactions on Neural Networks and Learning Systems, 2018, 29(4): 1006-1018.
DOI PMID |
[8] |
HAN N, WU J, FANG X Z, et al. Double relaxed regression for image classification[J]. IEEE Transactions on Circuits and Systems for Video Technology, 2019, 30(2): 307-319.
DOI URL |
[9] | CHEN Z, WU X J, KITTLER J. Low-rank discriminative least squares regression for image classification[J]. Signal Processing, 2020, 170: 107485. |
[10] |
WANG J, GUO Y, GUO J, et al. Synthesis linear classifier based analysis dictionary learning for pattern classification[J]. Neurocomputing, 2017, 238: 103-113.
DOI URL |
[11] | FANG X Z, TENG S H, LAI Z H, et al. Robust latent subspace learning for image classification[J]. IEEE Transac-tions on Neural Networks and Learning Systems, 2017, 29(6): 2502-2515. |
[12] | ZHANG Z, JIANG W M, ZHANG Z, et al. Scalable block-diagonal locality-constrained projective dictionary learning[C]// Proceedings of the 28th International Joint Conference on Artificial Intelligence, Macao, China, Aug 10-16, 2019: 4376-4382. |
[13] |
MENG M, LAN M C, YU J, et al. Constrained discrimi-native projection learning for image classification[J]. IEEE Transactions on Image Processing, 2019, 29: 186-198.
DOI URL |
[14] |
RONG Y, XIONG S W, GAO Y S. Low-rank double dic-tionary learning from corrupted data for robust image classi-fication[J]. Pattern Recognition, 2017, 72: 419-432.
DOI URL |
[15] | YANG Y, SHEN H T, MA Z G, et al. l2,1-norm regularized discriminative feature selection for unsupervised learning[C]// Proceedings of the 22nd International Joint Conference on Artificial Intelligence, Barcelona, Jul 16-22, 2011. Menlo Park: AAAI, 2011: 1589-1594. |
[16] | BYOD S P, PARIKN N, CHU E, et al. Distributed optimi-zation and statistical learning via the alternating direction method of multipliers[J]. Foundations & Trends in Machine Learning, 2010, 3(1). |
[17] |
WEN J, FANG X Z, CUI J R, et al. Robust sparse linear discriminant analysis[J]. IEEE Transactions on Circuits and Systems for Video Technology, 2019, 29(2): 390-403.
DOI URL |
[18] |
CAI J F, CANDÈS E J, SHEN Z W. A singular value thre-sholding algorithm for matrix completion[J]. SIAM Journal on Optimization, 2010, 20(4): 1956-1982.
DOI URL |
[19] | MARTINEZ A, BENAVENTE R. The AR face database: CVC Technical Report No.24[R]. Barcelona: Universitat Au-tonoma de Barcelona, 1998. |
[20] | GEORGHIADES A S, BELHUMEUR P N, KRIEGMAN D J. From few to many: illumination cone models for face re-cognition under variable lighting and pose[J]. IEEE Transac-tions on Pattern Analysis and Machine Intelligence, 2001, 23(6): 643-660. |
[21] |
LEE K C, HO J, KRIEGMAN D J. Acquiring linear sub-spaces for face recognition under variable lighting[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2005, 27(5): 684-698.
DOI URL |
[22] |
SIM T, BAKER S, BSAT M. The CMU pose, illumination, and expression database[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2003, 25(12): 1615-1618.
DOI URL |
[23] | PHILLIPS P J, FLYNN P J, SCRUGGS T, et al. Overview of the face recognition grand challenge[C]// Proceedings of the 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, San Diego, Jun 20-26, 2005. Washington: IEEE Computer Society, 2005: 947-954. |
[24] |
ZHANG X Y, WANG L, XIANG S, et al. Retargeted least squares regression algorithm[J]. IEEE Transactions on Neural Networks and Learning Systems, 2015, 26(9): 2206-2213.
DOI URL |
[25] |
JIANG Z L, LIN Z, DAVIS L S. Label consistent K-SVD: learning a discriminative dictionary for recognition[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2013, 35(11): 2651-2664.
DOI PMID |
[1] | 杨军, 雷喜文. 弱监督学习下的三维点云模型簇协同分割[J]. 计算机科学与探索, 2022, 16(9): 2121-2131. |
[2] | 刘雅芬, 郑艺峰, 江铃燚, 李国和, 张文杰. 深度半监督学习中伪标签方法综述[J]. 计算机科学与探索, 2022, 16(6): 1279-1290. |
[3] | 李勇, 高灿, 刘子荣, 罗金涛. 动态一致自信的深度半监督学习[J]. 计算机科学与探索, 2022, 16(11): 2557-2564. |
[4] | 刘宇, 孟敏, 武继刚. 一致性约束的半监督多视图分类[J]. 计算机科学与探索, 2022, 16(1): 242-252. |
[5] | 马玉琨, 徐姚文, 赵欣, 徐涛, 王泽瑞. 人脸识别系统的活体检测综述[J]. 计算机科学与探索, 2021, 15(7): 1195-1206. |
[6] | 李会荣, 张林, 赵鹏军, 李超. 带有局部坐标约束的半监督概念分解算法[J]. 计算机科学与探索, 2021, 15(2): 379-388. |
[7] | 姚晓红, 黄恒君. 非负半监督函数型聚类方法[J]. 计算机科学与探索, 2021, 15(12): 2438-2448. |
[8] | 马永杰, 徐小冬, 张茹, 谢艺蓉, 陈宏. 生成式对抗网络及其在图像生成中的研究进展[J]. 计算机科学与探索, 2021, 15(10): 1795-1811. |
[9] | 曹佳伟,钱鹏江. 流形学习与成对约束联合正则化非负矩阵分解[J]. 计算机科学与探索, 2020, 14(7): 1211-1220. |
[10] | 刘莹莹,王士同. 弹力理论传播的半监督学习新方法[J]. 计算机科学与探索, 2020, 14(4): 606-618. |
[11] | 刘颖,朱丽,林庆帆,李莹华,王富平,卢津. 图像超分辨率技术的回顾与展望[J]. 计算机科学与探索, 2020, 14(2): 181-199. |
[12] | 张典,汪海涛,姜瑛,陈星. 基于轻量级网络的实时人脸识别算法研究[J]. 计算机科学与探索, 2020, 14(2): 317-324. |
[13] | 梁俊杰,韦舰晶,蒋正锋. 生成对抗网络GAN综述[J]. 计算机科学与探索, 2020, 14(1): 1-17. |
[14] | 王小玉,韩昌林,胡鑫豪. 加权特征融合的密集连接网络人脸识别算法[J]. 计算机科学与探索, 2019, 13(7): 1195-1205. |
[15] | 郭伟,白文硕,曲海成. PCANet下的遮挡定位人脸识别算法[J]. 计算机科学与探索, 2019, 13(12): 2149-2160. |
阅读次数 | ||||||
全文 |
|
|||||
摘要 |
|
|||||