《计算机应用》唯一官方网站 ›› 2023, Vol. 43 ›› Issue (4): 1013-1020.DOI: 10.11772/j.issn.1001-9081.2022030329
所属专题: 人工智能
收稿日期:
2022-03-21
修回日期:
2022-06-18
接受日期:
2022-06-27
发布日期:
2023-04-11
出版日期:
2023-04-10
通讯作者:
万鸣华
作者简介:
葛孟婷(1998—),女,江苏连云港人,硕士研究生,CCF会员,主要研究方向:模式识别、特征提取、人脸识别、低秩学习;
基金资助:
Mengting GE1, Minghua WAN1,2()
Received:
2022-03-21
Revised:
2022-06-18
Accepted:
2022-06-27
Online:
2023-04-11
Published:
2023-04-10
Contact:
Minghua WAN
About author:
GE Mengting, born in 1998, M. S. candidate. Her research interests include pattern recognition, feature extraction, face recognition, low-rank learning.
Supported by:
摘要:
针对无监督的局部不变鲁棒主成分分析(LIRPCA)算法未考虑样本间的类别关系的问题,提出了一种基于近邻监督局部不变鲁棒主成分分析(NSLIRPCA)的特征提取模型。所提模型考虑了样本间的类别信息,并以此构建关系矩阵。对所提模型进行公式求解和公式的收敛性证明,并将所提模型应用于各种遮挡数据集。实验结果表明,在ORL、Yale、COIL-Processed和PolyU数据集上,与主成分分析(PCA)算法、基于L1范数的主成分分析(PCA-L1)算法、非负矩阵分解(NMF)算法、局部保持投影(LPP)算法和LIRPCA算法相比,所提模型在原始图像数据集上的识别率分别最高提升了8.80%、7.76%、20.37%、4.72%和4.61%,在遮挡图像数据集上的识别率分别最高提升了30.79%、30.73%、36.02%、19.65%和17.31%。可见,所提模型提高了算法的识别性能,降低了模型复杂度,明显优于对比算法。
中图分类号:
葛孟婷, 万鸣华. 基于近邻监督局部不变鲁棒主成分分析的特征提取模型[J]. 计算机应用, 2023, 43(4): 1013-1020.
Mengting GE, Minghua WAN. Feature extraction model based on neighbor supervised locally invariant robust principal component analysis[J]. Journal of Computer Applications, 2023, 43(4): 1013-1020.
图4 各个算法在ORL、Yale和COIL-20-Processed数据集上遮挡下的分类识别率
Fig. 4 Classification recognition rate of each algorithm on ORL, Yale, and COIL-20-Processed datasets under occlusion
算法 | ORL | Yale | COIL-20-Processed | |||||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
原始图像 | 5×5遮挡块 | 10×10遮挡块 | 原始图像 | 5×5遮挡块 | 10×10遮挡块 | 原始图像 | 5×5遮挡块 | 10×10遮挡块 | ||||||||||
识别率/% | 维度 | 识别率/% | 维度 | 识别率/% | 维度 | 识别率/% | 维度 | 识别率/% | 维度 | 识别率/% | 维度 | 识别率/% | 维度 | 识别率/% | 维度 | 识别率/% | 维度 | |
PCA | 89.00 | 29 | 69.30 | 40 | 50.37 | 38 | 87.01 | 46 | 72.33 | 40 | 68.30 | 37 | 61.10 | 32 | 61.03 | 36 | 60.18 | 30 |
PCA-L1 | 89.12 | 40 | 69.34 | 36 | 50.45 | 31 | 87.85 | 41 | 72.66 | 40 | 68.33 | 35 | 61.51 | 32 | 61.33 | 32 | 60.83 | 40 |
LPP | 89.35 | 42 | 69.36 | 40 | 50.52 | 34 | 90.40 | 21 | 76.00 | 33 | 74.66 | 27 | 62.69 | 20 | 62.16 | 24 | 61.90 | 24 |
NMF | 79.62 | 31 | 67.50 | 30 | 50.25 | 34 | 86.66 | 36 | 66.66 | 39 | 66.33 | 34 | 62.93 | 28 | 62.71 | 30 | 62.02 | 29 |
LIRPCA | 89.36 | 22 | 69.37 | 14 | 50.62 | 28 | 90.50 | 12 | 89.67 | 12 | 88.33 | 11 | 63.18 | 13 | 63.02 | 12 | 62.80 | 12 |
NSLIRPCA(p=1.5) | 89.38 | 29 | 69.38 | 27 | 50.63 | 31 | 91.50 | 22 | 89.78 | 26 | 88.80 | 29 | 63.19 | 16 | 63.10 | 20 | 63.03 | 18 |
NSLIRPCA(p=1.0) | 90.63 | 29 | 69.40 | 29 | 51.25 | 28 | 92.00 | 22 | 89.93 | 26 | 89.24 | 30 | 63.33 | 18 | 63.26 | 19 | 63.11 | 21 |
NSLIRPCA(p=0.5) | 91.25 | 31 | 81.25 | 57 | 59.38 | 51 | 94.67 | 22 | 90.67 | 27 | 89.33 | 26 | 63.41 | 19 | 63.33 | 20 | 63.18 | 22 |
表1 块遮挡下ORL、Yale和COIL-20-Processed数据集上各算法的最高识别率
Tab. 1 The highest recognition rate of each algorithm on ORL, Yale and COIL-20-Processed datasets under block occlusion
算法 | ORL | Yale | COIL-20-Processed | |||||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
原始图像 | 5×5遮挡块 | 10×10遮挡块 | 原始图像 | 5×5遮挡块 | 10×10遮挡块 | 原始图像 | 5×5遮挡块 | 10×10遮挡块 | ||||||||||
识别率/% | 维度 | 识别率/% | 维度 | 识别率/% | 维度 | 识别率/% | 维度 | 识别率/% | 维度 | 识别率/% | 维度 | 识别率/% | 维度 | 识别率/% | 维度 | 识别率/% | 维度 | |
PCA | 89.00 | 29 | 69.30 | 40 | 50.37 | 38 | 87.01 | 46 | 72.33 | 40 | 68.30 | 37 | 61.10 | 32 | 61.03 | 36 | 60.18 | 30 |
PCA-L1 | 89.12 | 40 | 69.34 | 36 | 50.45 | 31 | 87.85 | 41 | 72.66 | 40 | 68.33 | 35 | 61.51 | 32 | 61.33 | 32 | 60.83 | 40 |
LPP | 89.35 | 42 | 69.36 | 40 | 50.52 | 34 | 90.40 | 21 | 76.00 | 33 | 74.66 | 27 | 62.69 | 20 | 62.16 | 24 | 61.90 | 24 |
NMF | 79.62 | 31 | 67.50 | 30 | 50.25 | 34 | 86.66 | 36 | 66.66 | 39 | 66.33 | 34 | 62.93 | 28 | 62.71 | 30 | 62.02 | 29 |
LIRPCA | 89.36 | 22 | 69.37 | 14 | 50.62 | 28 | 90.50 | 12 | 89.67 | 12 | 88.33 | 11 | 63.18 | 13 | 63.02 | 12 | 62.80 | 12 |
NSLIRPCA(p=1.5) | 89.38 | 29 | 69.38 | 27 | 50.63 | 31 | 91.50 | 22 | 89.78 | 26 | 88.80 | 29 | 63.19 | 16 | 63.10 | 20 | 63.03 | 18 |
NSLIRPCA(p=1.0) | 90.63 | 29 | 69.40 | 29 | 51.25 | 28 | 92.00 | 22 | 89.93 | 26 | 89.24 | 30 | 63.33 | 18 | 63.26 | 19 | 63.11 | 21 |
NSLIRPCA(p=0.5) | 91.25 | 31 | 81.25 | 57 | 59.38 | 51 | 94.67 | 22 | 90.67 | 27 | 89.33 | 26 | 63.41 | 19 | 63.33 | 20 | 63.18 | 22 |
算法 | 原始图像 | 5×5遮挡块 | 10×10遮挡块 | |||
---|---|---|---|---|---|---|
识别率/% | 维度 | 识别率/% | 维度 | 识别率/% | 维度 | |
PCA | 73.00 | 40 | 68.34 | 42 | 62.83 | 46 |
PCA-L1 | 76.75 | 38 | 69.25 | 40 | 64.75 | 42 |
LPP | 77.26 | 36 | 73.20 | 38 | 65.30 | 40 |
NMF | 65.63 | 30 | 63.12 | 32 | 61.87 | 36 |
LIRPCA | 77.70 | 17 | 74.80 | 29 | 65.81 | 45 |
NSLIRPCA(p=1.5) | 78.00 | 20 | 75.00 | 37 | 66.00 | 45 |
NSLIRPCA(p=1) | 78.50 | 24 | 76.00 | 39 | 67.03 | 46 |
NSLIRPCA(p=0.5) | 79.00 | 29 | 76.50 | 38 | 68.00 | 47 |
表2 块遮挡下PolyU掌纹数据集上各算法的最高识别率
Tab. 2 The highest recognition rate of each algorithm on PolyU palmprint dataset under block occlusion
算法 | 原始图像 | 5×5遮挡块 | 10×10遮挡块 | |||
---|---|---|---|---|---|---|
识别率/% | 维度 | 识别率/% | 维度 | 识别率/% | 维度 | |
PCA | 73.00 | 40 | 68.34 | 42 | 62.83 | 46 |
PCA-L1 | 76.75 | 38 | 69.25 | 40 | 64.75 | 42 |
LPP | 77.26 | 36 | 73.20 | 38 | 65.30 | 40 |
NMF | 65.63 | 30 | 63.12 | 32 | 61.87 | 36 |
LIRPCA | 77.70 | 17 | 74.80 | 29 | 65.81 | 45 |
NSLIRPCA(p=1.5) | 78.00 | 20 | 75.00 | 37 | 66.00 | 45 |
NSLIRPCA(p=1) | 78.50 | 24 | 76.00 | 39 | 67.03 | 46 |
NSLIRPCA(p=0.5) | 79.00 | 29 | 76.50 | 38 | 68.00 | 47 |
算法 | 训练时间 | 算法 | 训练时间 |
---|---|---|---|
PCA | 0.24 | LIRPCA | 10.40 |
PCA-L1 | 39.68 | NSLIRPCA(p=1.5) | 1.67 |
LPP | 10.57 | NSLIRPCA(p=1) | 1.65 |
NMF | 0.42 | NSLIRPCA(p=0.5) | 1.57 |
表3 各个算法在ORL数据集上的训练时间 (s)
Tab. 3 Training time of different algorithm on ORL dataset
算法 | 训练时间 | 算法 | 训练时间 |
---|---|---|---|
PCA | 0.24 | LIRPCA | 10.40 |
PCA-L1 | 39.68 | NSLIRPCA(p=1.5) | 1.67 |
LPP | 10.57 | NSLIRPCA(p=1) | 1.65 |
NMF | 0.42 | NSLIRPCA(p=0.5) | 1.57 |
1 | KIM W, LEE R, PARK M, et al. Low-light image enhancement using volume-based subspace analysis[J]. IEEE Access, 2020, 8: 118370-118379. 10.1109/ACCESS.2020.3005249 |
2 | NAPOLÉON T, LFALOU A A. Pose invariant face recognition: 3D model from single photo[J]. Optics and Lasers in Engineering, 2017, 89: 150-161. 10.1016/j.optlaseng.2016.06.019 |
3 | TURK M A, PENTLAND A P. Face recognition using eigenfaces[C]// Proceedings of the 1991 IEEE Computer Society on Computer Vision and Pattern Recognition. Piscataway: IEEE, 1991: 586-591. |
4 | BELHUMEUR P N, HESPANHA J P, KRIEGMAN D J. Eigenfaces vs. Fisherfaces: recognition using class specific linear projection[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1997, 19(7): 711-720. 10.1109/34.598228 |
5 | BELKIN M, NIYOGI P. Laplacian eigenmaps for dimensionality reduction and data representation[J]. Neural Computation, 2003, 15(6): 1373-1396. 10.1162/089976603321780317 |
6 | LEE D D, SEUNG H S. Learning the parts of objects by non-negative matrix factorization[J]. Nature, 1999, 401(6755): 788-791. 10.1038/44565 |
7 | 王平,王宜怀,余文森,等. 基于非负矩阵分解和极限学习机分类的合成孔径雷达图像变化检测[J]. 科学技术与工程, 2020, 20(5): 1959-1966. 10.3969/j.issn.1671-1815.2020.05.038 |
WANG P, WANG Y H, YU W S, et al. Change detection in SAR images based on NMF and ELM classification[J]. Science Technology and Engineering, 2020, 20(5): 1959-1966. 10.3969/j.issn.1671-1815.2020.05.038 | |
8 | JIAN Y, ZHANG D, FRANGI A F, et al. Two-dimensional PCA: a new approach to appearance-based face representation and recognition[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2004, 26(1): 131-137. 10.1109/tpami.2004.1261097 |
9 | LUO D J, DING C, HUANG H. Symmetric two dimensional Linear Discriminant Analysis (2DLDA)[C]// Proceedings of the 2009 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2009: 2820-2827. 10.1109/cvpr.2009.5206635 |
10 | YANG J, ZHANG D, YONG X, et al. Two-dimensional discriminant transform for face recognition[J]. Pattern Recognition, 2005, 38(7): 1125-1129. 10.1016/j.patcog.2004.11.019 |
11 | 胡姿岚,王海贤. L1-范数子空间技术的鲁棒建模综述[J]. 安徽大学学报(自然科学版), 2017, 41(5): 9-16. 10.3969/j.issn.1000-2162.2017.05.003 |
HU Z L, WANG H X. L1-norm based subspace techniques for robust modelling: a brief review[J]. Journal of Anhui University (Natural Science Edition), 2017, 41(5): 9-16. 10.3969/j.issn.1000-2162.2017.05.003 | |
12 | KWAK N. Principal component analysis based on L1-norm maximization[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2008, 30(9): 1672-1680. 10.1109/tpami.2008.114 |
13 | NIE F P, HUANG H, DING C, et al. Robust principal component analysis with non-greedy l1-norm maximization[C]// Proceedings of the 22nd International Joint Conference on Artificial Intelligence. Palo Alto, CA: AAAI Press, 2011: 1433-1438. |
14 | NIE F P, TIAN L, HUANG H, et al. Non-greedy L21-norm maximization for principal component analysis[J]. IEEE Transactions on Image Processing, 2021, 30: 5277-5286. 10.1109/tip.2021.3073282 |
15 | MARKOPOULOS P P, KARYSTINOS G N, PADOS D A. Optimal algorithms for L1-subspace signal processing[J]. IEEE Transactions on Signal Processing, 2014, 62(19): 5046-5058. 10.1109/tsp.2014.2338077 |
16 | KWAK N. Principal component analysis by Lp-norm maximization[J]. IEEE Transactions on Cybernetics, 2014, 44(5): 594-609. 10.1109/tcyb.2013.2262936 |
17 | YE Q L, FU L Y, ZHANG Z, et al. Lp- and Ls-norm distance based robust linear discriminant analysis[J]. Neural Networks, 2018, 105: 393-404. 10.1016/j.neunet.2018.05.020 |
18 | HASSAN R J, ABDULAZEEZ A M. Deep learning convolutional neural network for face recognition: a review[J]. International Journal of Science and Business, 2021, 5(2): 114-127. 10.38007/nn.2021.020207 |
19 | 葛轶洲,刘恒,王言,等. 小样本困境下的深度学习图像识别综述[J]. 软件学报, 2022, 33(1): 193-210. |
GE Y Z, LIU H, WANG Y, et al. Survey on deep learning image recognition in dilemma of small samples[J]. Journal of Software, 2022, 33(1): 193-210. | |
20 | FU J L, WU Y, MEI T, et al. Relaxing from vocabulary: robust weakly-supervised deep learning for vocabulary-free image tagging[C]// Proceedings of the 2015 IEEE International Conference on Computer Vision. Piscataway: IEEE, 2015: 1985-1993. 10.1109/iccv.2015.230 |
21 | BAI S, LI Y W, ZHOU Y Y, et al. Adversarial metric attack and defense for person re-identification[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2021, 43(6): 2119-2126. 10.1109/tpami.2020.3031625 |
22 | 刘会,赵波,郭嘉宝,等. 针对深度学习的对抗攻击综述[J]. 密码学报, 2021, 8(2): 202-214. |
LIU H, ZHAO B, GUO J B, et al. Survey on adversarial attacks towards deep learning[J]. Journal of Cryptologic Research, 2021, 8(2): 202-214. | |
23 | BI P F, DU X. Application of locally invariant robust PCA for underwater image recognition[J]. IEEE Access, 2021, 9: 29470-29481. 10.1109/access.2021.3058761 |
24 | 郑宝玉,李昂. 基于快速稀疏低秩和鲁棒主成分分析的图像处理算法的研究[J]. 信号处理, 2020, 36(2): 290-296. |
ZHENG B Y, LI A. Image processing algorithm based on fast sparse low rank and robust PCA[J]. Journal of Signal Processing, 2020, 36(2): 290-296. | |
25 | 杨紫晴,姚加林,伍国华,等. 集成协方差矩阵自适应进化策略与差分进化的优化算法[J]. 控制理论与应用, 2021, 38(10): 1493-1502. 10.7641/CTA.2021.10002 |
YANG Z Q, YAO J L, WU G H, et al. Ensemble optimization algorithm from covariance matrix adaptive evolution strategy and differential evolution[J]. Control Theory and Applications, 2021, 38(10): 1493-1502. 10.7641/CTA.2021.10002 | |
26 | KNOR M, ŠKREKOVSKI R. On the minimum distance in a k-vertex set in a graph[J]. Applied Mathematics and Computation, 2019, 356: 99-104. 10.1016/j.amc.2019.03.050 |
27 | NEUMAYER S, NIMMER M, SETZER S, et al. On the robust PCA and Weiszfeld’s algorithm[J]. Applied Mathematics and Optimization, 2020, 82(3): 1017-1048. 10.1007/s00245-019-09566-1 |
28 | 张悦,刘德山,王姗姗,等. 应用于人脸识别的改进局部保持投影算法[J]. 计算机测量与控制, 2019, 27(10): 176-180. |
ZHANG Y, LIU D S, WANG S S, et al. Improved locality preserving projections for face recognition[J]. Computer Measurement and Control, 2016, 27(10): 176-180. | |
29 | 唐吉深,覃少华. 稀疏表示系数下局部最优重构的SAR图像目标识别算法[J]. 探测与控制学报, 2021, 43(2): 69-75, 80. |
TANG J S, QIN S H. SAR images recognition based on sparse coefficients optimal local reconstruction[J]. Journal of Detection and Control, 2021, 43(2): 69-75, 80. | |
30 | 赵鹏,汪纯燕,张思颖,等. 一种基于融合重构的子空间学习的零样本图像分类方法[J]. 计算机学报, 2021, 44(2): 409-421. 10.11897/SP.J.1016.2021.00409 |
ZHAO P, WANG C Y, ZHANG S Y, et al. A zero-shot image classification method based on subspace learning with the fusion of reconstruction[J]. Chinese Journal of Computers, 2021, 44(2): 409-421. 10.11897/SP.J.1016.2021.00409 | |
31 | ROWEIS S T, SAUL L K. Nonlinear dimensionality reduction by locally linear embedding[J]. Science, 2000, 290(5500): 2323-2326. 10.1126/science.290.5500.2323 |
[1] | 贾洁茹, 杨建超, 张硕蕊, 闫涛, 陈斌. 基于自蒸馏视觉Transformer的无监督行人重识别[J]. 《计算机应用》唯一官方网站, 2024, 44(9): 2893-2902. |
[2] | 杨鑫, 陈雪妮, 吴春江, 周世杰. 结合变种残差模型和Transformer的城市公路短时交通流预测[J]. 《计算机应用》唯一官方网站, 2024, 44(9): 2947-2951. |
[3] | 付帅, 郭小英, 白茹意, 闫涛, 陈斌. 改进的CloFormer模型与有序回归相结合的年龄评估方法[J]. 《计算机应用》唯一官方网站, 2024, 44(8): 2372-2380. |
[4] | 陈彤, 杨丰玉, 熊宇, 严荭, 邱福星. 基于多尺度频率通道注意力融合的声纹库构建方法[J]. 《计算机应用》唯一官方网站, 2024, 44(8): 2407-2413. |
[5] | 龙伍丹, 彭博, 胡节, 申颖, 丁丹妮. 基于加强特征提取的道路病害检测算法[J]. 《计算机应用》唯一官方网站, 2024, 44(7): 2264-2270. |
[6] | 刘瑞华, 郝子赫, 邹洋杨. 基于多层级精细特征融合的步态识别算法[J]. 《计算机应用》唯一官方网站, 2024, 44(7): 2250-2257. |
[7] | 陈学斌, 任志强, 张宏扬. 联邦学习中的安全威胁与防御措施综述[J]. 《计算机应用》唯一官方网站, 2024, 44(6): 1663-1672. |
[8] | 吴郅昊, 迟子秋, 肖婷, 王喆. 基于元学习自适应的小样本语音合成[J]. 《计算机应用》唯一官方网站, 2024, 44(5): 1629-1635. |
[9] | 夏吾吉, 黄鹤鸣, 更藏措毛, 范玉涛. 基于无监督学习和监督学习的抽取式文本摘要综述[J]. 《计算机应用》唯一官方网站, 2024, 44(4): 1035-1048. |
[10] | 董炜娜, 刘佳, 潘晓中, 陈立峰, 孙文权. 基于编码-解码网络的大容量鲁棒图像隐写方案[J]. 《计算机应用》唯一官方网站, 2024, 44(3): 772-779. |
[11] | 江锐, 刘威, 陈成, 卢涛. 非对称端到端的无监督图像去雨网络[J]. 《计算机应用》唯一官方网站, 2024, 44(3): 922-930. |
[12] | 崔晨辉, 蔺素珍, 李大威, 禄晓飞, 武杰. 基于孪生网络和Transformer的红外弱小目标跟踪方法[J]. 《计算机应用》唯一官方网站, 2024, 44(2): 563-571. |
[13] | 范艺扬, 张洋, 曾尚, 曾渝, 付茂栗. 基于分解和频域特征提取的多变量长时间序列预测模型[J]. 《计算机应用》唯一官方网站, 2024, 44(11): 3442-3448. |
[14] | 赵培, 乔焰, 胡荣耀, 袁新宇, 李敏悦, 张本初. 基于多域特征提取的多变量时间序列异常检测[J]. 《计算机应用》唯一官方网站, 2024, 44(11): 3419-3426. |
[15] | 黄杰, 武瑞梓, 李均利. 高效的自适应复杂网络鲁棒性优化算法[J]. 《计算机应用》唯一官方网站, 2024, 44(11): 3530-3539. |
阅读次数 | ||||||
全文 |
|
|||||
摘要 |
|
|||||