Spectroscopy and Spectral Analysis, Volume. 42, Issue 5, 1385(2022)
Identification of Pesticide Residue Types in Chinese Cabbage Based on Hyperspectral and Convolutional Neural Network
Figures & Tables(14)
Fig. 2. Schematic representation of selection of ROI on Chinese cabbage sample
Fig. 3. Spectral data preprocessing by MSC
(a): Without preprocessing; (b): With MSC preprocessing
Fig. 5. CNN hyperparameter selection
(a): TLA for different LR; (b): TLA for different BS; (c): OA for different Epochs
Fig. 7. Low frequency portions of wavelet transform based on db1 function
(a)—(f) corresponding to 1~6 layers of DWT, respectively
Fig. 9. Flow chart of hyperspectral image classification based on DWT and deep learning
Fig. 10. Modeling results based on DWT, PCA and CARS
(a)—(d) Overall accuracies of CNN, MLP, KNN and SVM models based on DWT;(e) Overall accuracies of four models based on PCA; (f) OAs of four models based on CARS
Table 1. MLP network structure
View tableView in ArticleTable 1. MLP network structure
层 神经元个数 激活函数 全连接层FC1 512 relu 全连接层FC2 256 relu 全连接层FC3 256 relu 全连接层FC4 128 relu 输出层 5 softmax
Table 2. Parameter settings of the CNN structure
View tableView in ArticleTable 2. Parameter settings of the CNN structure
层 离散小波变换层数 1~4 5~6 卷积炽C1 3×3×16 3×3×16 激活层A1 relu relu 舍弃层D1 0.3 0.3 卷积层C2 5×5×32 3×3×32 正则层L21 0.001 0.001 激活层A2 relu relu 舍弃层D2 0.5 0.5 池化层MaxPooling 2×2×2 2×2×2 全连接层FC1 256 256 全连接层FC2 256 256 正则层L22 0.001〗〗0.001 输出层 5 5 激活层A3 softmax softmax
Table 3. Overall accuracies and Kappa coefficients of different algorithms
View tableView in ArticleTable 3. Overall accuracies and Kappa coefficients of different algorithms
模型 降维算法 None CARS PCA DWT OA
/%Kappa Time
/msOA
/%Kappa Time
/msOA
/%Kappa Time
/msOA
/%Kappa Time
/msKNN 12.00 -0.10 16.98 11.20 -0.11 10.00 56.00 0.49 4.00 66.40 0.58 20.02 SVM 89.60 0.87 74.77 26.40 0.08 4.99 68.00 0.60 4.99 90.40 0.88 14.03 MLP 67.20 0.59 80.78 72.80 0.66 59.99 66.40 0.58 67.01 83.20 0.79 63.23 CNN 82.40 0.78 698.13 26.40 0.08 4.99 64.00 0.55 94.01 91.20 0.89 86.01
Table 4. Overall accuracies of the various prediction algorithms
View tableView in ArticleTable 4. Overall accuracies of the various prediction algorithms
降维
算法模型 药物种类 毒死蜱 氯氰菊酯 灭多威 乐果 无残留 None KNN 16.00 12.00 8.00 24.00 0.00 CARS 16.00 12.00 4.00 20.00 4.00 PCA 58.82 42.00 63.79 44.44 69.56 DWT 60.00 80.00 48.00 64.00 80.00 None SVM 96.00 92.00 84.00 84.00 92.00 CARS 100.00 8.00 20.00 4.00 0.00 PCA 68.62 66.00 70.69 66.66 67.39 DWT 96.00 100.00 80.00 84.00 92.00 None MLP 88.00 88.00 52.00 12.00 96.00 CARS 56.00 52.00 92.00 72.00 92.00 PCA 60.78 68.00 75.86 46.67 78.26 DWT 88.00 84.00 72.00 76.00 96.00 None CNN 76.00 84.00 76.00 80.00 96.00 CARS 76.00 80.00 76.00 80.00 100.00 PCA 60.78 74.00 63.79 44.44 76.08 DWT 88.00 88.00 84.00 96.00 100.00
Tools
Get Citation
Copy Citation Text
Rong-chang JIANG, Ming-sheng GU, Qing-he ZHAO, Xin-ran LI, Jing-xin SHEN, Zhong-bin SU. Identification of Pesticide Residue Types in Chinese Cabbage Based on Hyperspectral and Convolutional Neural Network[J]. Spectroscopy and Spectral Analysis, 2022, 42(5): 1385
Paper Information
Category: Research Articles
Received: Aug. 3, 2021
Accepted: --
Published Online: Nov. 10, 2022
The Author Email: JIANG Rong-chang (jake_jrc@qq.com)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20