Laser & Optoelectronics Progress, Volume. 56, Issue 10, 101010(2019)
Palm Vein Classification Based on Deep Neural Network and Random Forest
Figures & Tables(11)
Fig. 4. Acquisition device and five examples of different human palm vein images collected using this device. (a) PolyU database; (b) CASIA database; (c) self-built database
Fig. 5. Misclassified images and pseudo color images. (a) Class-1 example of misclassified image; (b) class-2 example of misclassified image; (c) class-3 example of misclassified images; (d) class-4 example of misclassified images; (e) class-5 example of misclassified images; (f) class-6 example of misclassified images; (g) class-7 example of misclassified images; (h) class-8 example of misclassified images; (i) class-9 example of misclassified images; (j) class-10 example of misclassified images; (k) c
Fig. 6. Palm vein ROI map, feature map and pseudo color image. (a) ROI map; (b) pseudo color image of Fig. (a); (c) feature map extracted by method in Ref. [11]; (d) pseudo color image of Fig. (c); (e) feature map extracted from 4th convolutional layer; (f) pseudo color image of Fig. (e)
Fig. 7. Classification error of each database versus number of classification decision trees
Table 1. Classification errors of palm vein features extracted from different layers of AlexNet network on different databases%
View tableTable 1. Classification errors of palm vein features extracted from different layers of AlexNet network on different databases%
Database Error Conv1 Conv2 Conv3 Conv4 Conv5 Fc6 Fc7 Fc8 PolyU Oob error 0 0 0.01 0 0.02 0.69 5.53 31.44 Test error 1.10 0.10 0 0 0 1.90 5.50 33.40 CASIA Oob error 2.78 0 0.025 0.025 0.05 3.85 17.28 63.33 Test error 27.00 7.75 4.50 3.00 5.25 15.25 32.50 78.50 Self-built Oob error 7.55 0.38 0 0 0 0.075 0.38 5.65 Test error 9.25 4.75 1.25 0.50 0.75 1.50 4.25 16.50
Table 2. Effect of PCA on classification error of each database%
View tableTable 2. Effect of PCA on classification error of each database%
Method PolyU CASIA Self-built AlexNet+RF 0.30 6.75 1.50 AlexNet+PCA+RF 0 3.00 0.50
Table 3. Misclassification data classification errors from different methods
View tableTable 3. Misclassification data classification errors from different methods
Class Method Classification error /% 14 classes Method in Ref. [11] 3.10 Proposed method 0 214 classes Method in Ref. [11] 17.15 Proposed method 0.23
Table 4. Recognition accuracy of each method%
View tableTable 4. Recognition accuracy of each method%
Method PolyU CASIA Self-built Method in Ref. [10] 92.90 79.00 87.25 Method in Ref. [9] 99.50 91.50 97.75 AlexNet 99.30 87.75 98.00 AlexNet+PCA+SVM 99.90 91.00 99.50 VGG16+PCA+RF 99.80 90.50 98.50 Proposed method 100 97.00 99.50
Tools
Get Citation
Copy Citation Text
Lisha Yuan, Mengying Lou, Yaqin Liu, Feng Yang, Jing Huang. Palm Vein Classification Based on Deep Neural Network and Random Forest[J]. Laser & Optoelectronics Progress, 2019, 56(10): 101010
Paper Information
Category: Image Processing
Received: Oct. 30, 2018
Accepted: Dec. 12, 2018
Published Online: Jul. 4, 2019
The Author Email: Yaqin Liu (liuyq@smu.edu.cn), Jing Huang (jhuangyg@smu.edu.cn)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20