• Spectroscopy and Spectral Analysis
  • Vol. 42, Issue 5, 1572 (2022)
Wen-bin SUN2、*, Rong WANG1、1; 3; 4;, Rong-hua GAO1、1; 3; *;, Qi-feng LI1、1; 3;, Hua-rui WU1、1; 3;, and Lu FENG1、1; 3;
Author Affiliations
  • 11. Beijing Research Center for Information Technology in Agriculture, Beijing 100097, China
  • 22. College of Information Engineering, Guangdong University of Technology, Guangzhou 510006, China
  • show less
    DOI: 10.3964/j.issn.1000-0593(2022)05-1572-09 Cite this Article
    Wen-bin SUN, Rong WANG, Rong-hua GAO, Qi-feng LI, Hua-rui WU, Lu FENG. Crop Disease Recognition Based on Visible Spectrum and Improved Attention Module[J]. Spectroscopy and Spectral Analysis, 2022, 42(5): 1572 Copy Citation Text show less
    Part of the disease sample in the dataset
    Fig. 1. Part of the disease sample in the dataset
    Structure diagram of SE module
    Fig. 2. Structure diagram of SE module
    The overall structure of the SMLP-Res module
    Fig. 3. The overall structure of the SMLP-Res module
    Crop disease recognition model based on improved channel attention mechanism
    Fig. 4. Crop disease recognition model based on improved channel attention mechanism
    Relationship between test error of model and epochs
    Fig. 5. Relationship between test error of model and epochs
    Relationship between test error of model and epochs
    Fig. 6. Relationship between test error of model and epochs
    Comparison results of different disease identification methods
    Fig. 7. Comparison results of different disease identification methods
    Recognition results of some disease samples
    Fig. 8. Recognition results of some disease samples
    Results of heat map analysis of early and late samples of different tomato diseases
    Fig. 9. Results of heat map analysis of early and late samples of different tomato diseases
    数据集样本图像类别作物样本疾病样本
    Plant Village54 306381424
    AI Challenger 201835 861591027
    Table 1. Basic information of two Dataset
    参数SMLP_ResNet18SMLP_ResNet50SMLP_ResNet101
    n233
    m244
    l2623
    w233
    Table 2. Parameter table of SMLP_ResNet disease model
    方法准确率
    /%
    训练时长/轮
    /min
    模型权重
    /M
    精确率
    /%
    AlexNet[8]97.82-218.0-
    AlexNet[5]99.08-218.0-
    GoogleNet22[8]98.36---
    ResNet1899.054.742.898.57
    SENet1899.195.043.199.00
    SMLP_ResNet1899.324.548.699.10
    Table 3. Recognition results of different disease models
    方法发表年份数据集图像数量/张模型准确率/%
    Too et al. [9]2019PlantVillage54 306VGG1681.83
    Gensheng et al.[10]2019茶叶病害4 980VGG1690.00
    Wang et al.[11]2017PlantVillage54 306VGG1690.40
    Agarwal M et al.[12]2020番茄叶片病害18 160VGG1693.50
    Wang et al.[11]2017PlantVillage54 306Inception-V380.00
    Gandhi et al.2018PlantVillage56 000Inception-V388.60
    Agarwal M et al.[12]2020番茄叶片病害18 160Inception-V377.50
    Elhassouny & Smarandache2019番茄叶片病害7 176Mobilenet88.40
    Gandhi et al.2018PlantVillage56 000Mobilenet92.00
    Agarwal M et al.[12]2020番茄叶片病害18 160Mobilenet82.60
    Agarwal M et al.[12]2020番茄叶片病害18 160CNN98.40
    Proposed Model2021PlantVillage54 306SMLP_ResNet1899.32
    Table 4. Accuracy comparison with pre-training models in previous studies
    方法准确率/%模型权重大小/M参数量/百万
    AlexNet83.50217.5157.02
    ResNet1883.8342.7511.21
    SENet1884.5342.9511.26
    SMLP_ResNet1886.9348.5812.73
    Table 5. Parameter table of SMLP_ResNet disease model
    Wen-bin SUN, Rong WANG, Rong-hua GAO, Qi-feng LI, Hua-rui WU, Lu FENG. Crop Disease Recognition Based on Visible Spectrum and Improved Attention Module[J]. Spectroscopy and Spectral Analysis, 2022, 42(5): 1572
    Download Citation