• Spectroscopy and Spectral Analysis
  • Vol. 41, Issue 12, 3837 (2021)
Ye-lan WU1、*, Yi-yu CHEN1、1;, Xiao-qin LIAN1、1;, Yu LIAO2、2;, Chao GAO1、1;, Hui-ning GUAN1、1;, and Chong-chong YU1、1;
Author Affiliations
  • 11. Key Laboratory of Internet and Big Data in Light Industry, Beijing Technology and Business University, Beijing 100048, China
  • 22. Institute of Agricultural Engineering, Jiangxi Academy of Agricultural Sciences, Nanchang 330200, China
  • show less
    DOI: 10.3964/j.issn.1000-0593(2021)12-3837-07 Cite this Article
    Ye-lan WU, Yi-yu CHEN, Xiao-qin LIAN, Yu LIAO, Chao GAO, Hui-ning GUAN, Chong-chong YU. Study on the Identification Method of Citrus Leaves Based on Hyperspectral Imaging Technique[J]. Spectroscopy and Spectral Analysis, 2021, 41(12): 3837 Copy Citation Text show less
    Hyperspectral imaging system
    Fig. 1. Hyperspectral imaging system
    Hyperspectral images of leaves of five citrus species
    Fig. 2. Hyperspectral images of leaves of five citrus species
    Spectral curves of 13 250 samples
    Fig. 3. Spectral curves of 13 250 samples
    Average spectral curves of five types of leaves
    Fig. 4. Average spectral curves of five types of leaves
    The first four principal component load curves of the original spectrum
    Fig. 5. The first four principal component load curves of the original spectrum
    Sample distribution scatter diagramFigure (a), (b), (c) and (d) respectively represent the scatter plots of sample distribution using original spectra and the spectra after preprocessing by 1st Der, MSC and SNV
    Fig. 6. Sample distribution scatter diagram
    Figure (a), (b), (c) and (d) respectively represent the scatter plots of sample distribution using original spectra and the spectra after preprocessing by 1st Der, MSC and SNV
    模型预处理
    方法
    识别率/%
    正常
    叶片
    溃疡病
    叶片
    除草剂
    叶片
    红蜘蛛
    叶片
    煤烟病
    叶片
    SVM原始100.0099.7890.03100.0092.08
    1st Der100.0099.7891.0599.5894.09
    MSC99.7399.7880.6997.3588.38
    SNV100.0099.7879.8098.6087.17
    RF原始99.4698.4484.9198.7493.19
    1st Der99.4699.3374.0498.7493.29
    MSC98.9197.1072.7696.2385.97
    SNV98.9197.1073.4096.3787.17
    Table 1. Modeling results of full-band data
    模型预处理
    方法
    识别率/%
    正常
    叶片
    溃疡病
    叶片
    除草剂
    叶片
    红蜘蛛
    叶片
    煤烟病
    叶片
    SVM原始100.0099.1172.5198.4681.66
    1st Der99.1899.7879.2898.4687.27
    MSC98.1098.4472.7694.9781.56
    SNV98.1098.6672.5194.8381.16
    RF原始98.6496.6574.3098.4685.97
    1st Der99.1896.6580.0598.0491.58
    MSC97.5597.3268.8093.8580.96
    SNV97.2897.3268.4193.7281.06
    Table 2. PCA characteristic wavelength modeling results
    模型评价指标FSPCA
    原始1st DerMSCSNV原始1st DerMSCSNV
    SVMOA95.23%95.98%91.30%91.03%87.53%90.82%86.50%86.32%
    Kappa0.938 50.948 20.887 90.884 40.839 20.881 60.826 00.823 6
    RFOA93.84%91.42%88.01%88.56%88.77%91.79%84.93%84.81%
    Kappa0.920 50.889 20.845 30.852 30.855 00.894 00.805 60.804 0
    Table 3. Modeling results under different pretreatment methods of all-band and PCA wavelength data
    Ye-lan WU, Yi-yu CHEN, Xiao-qin LIAN, Yu LIAO, Chao GAO, Hui-ning GUAN, Chong-chong YU. Study on the Identification Method of Citrus Leaves Based on Hyperspectral Imaging Technique[J]. Spectroscopy and Spectral Analysis, 2021, 41(12): 3837
    Download Citation