Laser & Optoelectronics Progress, Volume. 58, Issue 16, 1610012(2021)
Blurred License Plate Character Recognition Algorithm Based on Deep Learning
Figures & Tables(8)
Fig. 1. CNN configuration diagram. (a) Depthwise separable convolution bolck; (b) CNN structure
Fig. 4. Sample dataset example. (a) Samples of GAN_LP ; (b) samples of Reldv1; (c) samples of CCPDv1
Table 1. Statistics of datasets
View tableTable 1. Statistics of datasets
Dateset Training set Validation set Test set GAN-LP 12200 980 7406 Reldv1 2500 0 500 CCPDv1 3000 0 1000
Table 2. Hyper parameters setting
View tableTable 2. Hyper parameters setting
Parameter Setting Base_lr 0.001 γ 0.1 Lr_policy Multistep Active_fun ReLU Iterations 2×104 Type SGD Weight_decay 0.0005 Momentum 0.8 BN Yes Batch_size 64
Table 3. Comprehensive experimental results of improved CRNN+CTC network unit: %
View tableTable 3. Comprehensive experimental results of improved CRNN+CTC network unit: %
Dataset GAN-LP CCPDv1 Reldv1 ACRR 97.6 97.8 98.2 CCRR 95.1 93.3 95.7 fRA 96.8 96.9 97.8 fCRA 97.1 97.2 98.4
Table 4. Experimental comparison results
View tableTable 4. Experimental comparison results
Method GAN_LP CCPDv1 Reldv1 RA% CRA% T /ms RA% CRA% T /ms RA% CRA% T /ms Traditional[ 10 ]84.2 77.3 25 83.4 74.2 31 85.5 75.7 32 LPR-Net[ 15 ]94.3 92.7 62 87.4 86.6 72 88.5 86.3 46 CRNN+CTC[ 18 ]96.9 96.2 46 97.1 96.3 48 96.4 97.0 44 Ours 96.8 97.1 29 96.9 97.2 33 97.8 98.4 37
Tools
Get Citation
Copy Citation Text
Caizhen Zhang, Ying Li, binlong Kang, yuan Chang. Blurred License Plate Character Recognition Algorithm Based on Deep Learning[J]. Laser & Optoelectronics Progress, 2021, 58(16): 1610012
Paper Information
Category: Image Processing
Received: Nov. 13, 2020
Accepted: Dec. 22, 2020
Published Online: Aug. 19, 2021
The Author Email: Li Ying (873238393@qq.com)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20