An efficient 3D palmprint recognition method was proposed by using local texture feature sets and deep learning, in order to explore the usage of 3D palmprint in biometrics recognition. Curvature feature, shape index and surface type were employed to describe the geometry characteristics of local regions in 3D palmprint data, and then take the charasteristics as the input of the deep neural network to finish 3D palmprint recognition task. Comprehensive experiments on Hong Kong Polytechnic University 3D palmprint database were further conducted by using different geometrical features and deep neural network models. The final experimental results of 3D palmprint recognition validate that the proposed method outperforms existing state-of-the-art methods in terms of recognition accuracy and runtime, showing high potential for real-time palmprint recognition applications.
Bing YANG,Wen-bo MO,Jin-liang YAO. 3D palmprint recognition by using local features and deep learning. Journal of ZheJiang University (Engineering Science), 2020, 54(3): 540-545.
Fig.2Mean curvature images and Gaussian curvature images of different 3D palms
Fig.3Shape index images of different 3D palms
曲率
G> 0
G = 0
G< 0
M< 0
峰（ST = 1）
岭（ST = 2）
鞍岭（ST = 3）
M= 0
无（ST = 4）
平坦（ST = 5）
低点（ST = 6）
M> 0
坑（ST = 7）
谷（ST = 8）
鞍谷（ST = 9）
Tab.1Three dimensional surface type definition
Fig.4Surface type images of different 3D palms
Fig.5Examples from Hong Kong Polytechnic University 3D palmprint database
三维特征
表面类型
平均曲率
高斯曲率
文献[5]
99.15
93.55
67.10
文献[15]
98.78
91.88
91.87
AlexNet模型
99.40
99.20
98.75
Tab.2Identification accuracy by using different 3D palmprint recognition methods %
方法
N=1
N=2
N=4
N=10
ST+文献[5]
90.22
94.17
97.26
99.15
ST+ AlexNet
91.37
95.83
99.05
99.40
CBR[7]
95.11
97.31
99.52
99.66
Tab.3Identification accuracy by using different training samples %
方法
tr
方法
tr
ST[15]
63 275.86
CBR[7]
858.29
MCI[15]
9 403.33
AlexNet+ST
96.562
GCI[15]
9 403.30
AlexNet+MCI
94.993
ST[5]
547.03
AlexNet+GCI
94.352
Tab.4Runtime of different 3D palmprint recognition methods ms
三维特征
AlexNet
GoogleNet
Vgg16
ResNet50
表面类型
99.40
99.55
99.25
99.45
形状指数
99.30
99.00
97.25
98.90
平均曲率
99.20
99.15
99.55
98.80
高斯曲率
98.75
96.60
97.50
96.25
Tab.5Identification accuracy by using different features and network structures of deep learning models %
三维特征
AlexNet
GoogleNet
Vgg16
ResNet50
表面类型
96.562
103.896
261.432
257.531
形状指数
95.865
103.812
261.175
256.421
平均曲率
94.993
103.461
259.813
254.594
高斯曲率
94.352
103.153
258.231
254.451
Tab.6Runtime by using different features and network structuresof deep learning models ms
[1]
HUANG D, JIA W, ZHANG D Palmprint verification based on principal lines[J]. Pattern Recognition, 2008, 41 (4): 1316- 1328
doi: 10.1016/j.patcog.2007.08.016
[2]
FEI L, XU Y, TANG W, et al Double-orientation code and nonlinear matching scheme for palmprint recognition[J]. Pattern Recognition, 2016, 49 (1): 89- 101
[3]
ZHANG L, CHENG Z, SHEN Y, et al Palmprint and palmvein recognition based on DCNN and a new large-scale contactless palmvein dataset[J]. Symmetry, 2018, 10 (4): 78
doi: 10.3390/sym10040078
[4]
白雪飞, 高楠, 张宗华, 等 基于分块ST与主成分分析的三维掌纹识别[J]. 天津大学学报: 自然科学与工程技术版, 2018, 51 (6): 631- 637 BAI Xue-fei, GAO Nan, ZHANG Zong-hua, et al Three dimensional palmprint identification based on blocked ST and PCA[J]. Journal of Tianjin University: Science and Technology, 2018, 51 (6): 631- 637
[5]
ZHANG L, SHEN Y, LI H, et al 3D palmprint identification using block-wise features and collaborative representation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015, 37 (8): 1730- 1736
doi: 10.1109/TPAMI.2014.2372764
[6]
YANG B, XIANG X, XU D, et al 3D palmprint recognition using shape index representation and fragile bits[J]. Multimedia Tools and Applications, 2017, 76 (14): 15357- 15375
doi: 10.1007/s11042-016-3832-1
[7]
FEI L, ZHANG B, XU Y, et al Precision direction and compact surface type representation for 3D palmprint identification[J]. Pattern Recognition, 2019, 87 (3): 237- 247
[8]
BAI X, GAO N, ZHANG Z, et al Person recognition using 3D palmprint data based on full-field sinusoidal fringe projection[J]. IEEE Transactions on Instrumentation and Measurement, 2019, 68 (9): 3287- 3298
doi: 10.1109/TIM.2018.2877226
[9]
KUHNEL W, HUNT B. Differential Geometry: Curves-Surfaces-Manifolds: 2nd ed [M]. New York: American Mathematical Society, 2006: 66-68.
[10]
KRIZHEVSKY A, SUTSKEVER I, HINTON G. ImageNet classification with deep convolutional neural networks [C] // Proceedings of the 25th International Conference on Neural Information Processing Systems. Lake Tahoe: Curran Associates Inc., 2012: 1097-1105.
[11]
SZEGEDY, LIU W, JIA Y, et al. Going deeper with convolutions [C] // Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR'15). Washington: IEEE Computer Society Press, 2015: 1-9.
[12]
SIMONYAN K and ZISSERMAN A. Very deep convolutional networks for large-scale image recognition [J]. arXiv Preprint, arXiv: 1409.1556, 2014.
[13]
HE K, ZHANG X, REN S, et al. Deep residual learning for image recognition [J]. arXiv Preprint, arXiv: 1512.03385, 2015.
[14]
LI W, ZHANG D, LU G, et al A novel 3D palmprint acquisition system[J]. IEEE Transactions on Systems, Man, and Cybernetics, Part A: Systems and Humans, 2012, 42 (2): 443- 452
doi: 10.1109/TSMCA.2011.2164066
[15]
ZHANG D, LU G, LI W, et al Palmprint recognition using 3-D information[J]. IEEE Transactions on Systems Man and Cybernetics Part C Applications and Reviews, 2009, 39 (5): 505- 519
doi: 10.1109/TSMCC.2009.2020790
[16]
XU Y, FEI L, WEN J, et al Discriminative and robust competitive code for palmprint recognition[J]. IEEE Transactions on Systems, Man, and Cybernetics: Systems, 2018, 48 (2): 232- 241
doi: 10.1109/TSMC.2016.2597291
