Chinese Journal of Lasers, Volume. 50, Issue 2, 0211001(2023)
Spectral Classification and Characteristic Spectral Analysis of Nearshore Aquatic Plants Based on AlexNet
Figures & Tables(12)
Fig. 1. Four aquatic plants. (a) Pontederia cordata L.; (b) Thalia dealbata; (c) Typha angustifolia L.; (d) Hydrocotyle vulgaris
Fig. 2. Spectral curves of 4 species of aquatic plants. (a) Original spectra; (b) first-order derivative spectra; (c) second-order derivative spectra; (d) first-order derivative logarithm spectra; (e) second-order derivative logarithm spectra
Fig. 3. Gray scale images of 4 aquatic plants. (a) Typha angustifolia L.; (b) Pontederia cordata L.; (c) Hydrocotyle vulgaris; (d) Thalia dealbata
Fig. 5. Training process diagrams of different models. (a) Training accuracy graph; (b) training loss graph
Fig. 6. Training loss value of each model under different preprocessing methods. (a) CNN3 model; (b) VGG16 model; (c) our model
Fig. 7. Intermediate feature maps. (a) Typha angustifolia L.; (b) Pontederia cordata L.; (c) Hydrocotyle vulgaris; (d) Thalia dealbata
Fig. 8. Feature maps. (a) Typha angustifolia L.; (b) Pontederia cordata L.; (c) Hydrocotyle vulgaris; (d) Thalia dealbata
Table 1. Spectral data classification under different models
View tableTable 1. Spectral data classification under different models
Model Training speed /（s·epoch-1） Testing speed /（frame·s-1） Accuracy /% CNN3 5.56 0.030 85.06 VGG16 43.68 0.048 99.48 Our model 13.56 0.032 99.50
Table 2. Division of different training sets
View tableTable 2. Division of different training sets
Training data ratio /% Training data quantity Testing data quantity 40 1596 500 60 2394 500 80 3192 500
Table 3. Classification accuracy with different training set sizes
View tableTable 3. Classification accuracy with different training set sizes
Model Training data ratio /% Accuracy /% CNN3
40 80.83 60 84.93 80 85.06 VGG16
40 97.70 60 98.74 80 99.48 Our model
40 99.15 60 99.44 80 99.50
Table 4. Classification accuracy comparison of three kinds of models before and after sample data preprocessing
View tableTable 4. Classification accuracy comparison of three kinds of models before and after sample data preprocessing
Model Accuracy /% Original data First-order derivative Second-order derivative First-order derivative logarithm Second-order derivative logarithm CNN3 76.67 85.06 79.59 84.74 80.21 VGG16 96.45 99.48 97.99 98.41 97.20 Our model 97.80 99.50 98.56 99.10 98.95
Tools
Get Citation
Copy Citation Text
Zongsheng Zheng, Bei Liu, Peng Lu, Zhenhua Wang, Guoliang Zou, jiahui Zhao, Yunfei Li. Spectral Classification and Characteristic Spectral Analysis of Nearshore Aquatic Plants Based on AlexNet[J]. Chinese Journal of Lasers, 2023, 50(2): 0211001
Paper Information
Category: spectroscopy
Received: Mar. 9, 2022
Accepted: Apr. 25, 2022
Published Online: Feb. 7, 2023
The Author Email: Liu Bei (godbei@foxmail.com)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20