• Laser & Optoelectronics Progress
  • Vol. 60, Issue 5, 0530002 (2023)
Fusheng Li1,2,* and Xiaolong Zeng1,2
Author Affiliations
  • 1School of Automation Engineering, University of Electronic Science and Technology of China, Chengdu 611731, Sichuan, China
  • 2Yangtze Delta Region Institute, University of Electronic Science and Technology of China, Huzhou 313099, Zhejiang, China
  • show less
    DOI: 10.3788/LOP213241 Cite this Article Set citation alerts
    Fusheng Li, Xiaolong Zeng. Quantitative Analysis Method of Soil Elements Combining Sensitivity Dimensionality Reduction and Support Vector Regression[J]. Laser & Optoelectronics Progress, 2023, 60(5): 0530002 Copy Citation Text show less
    Spectra before and after background subtraction. (a) Original spectrum and estimated background; (b) original spectrum and corrected spectrum
    Fig. 1. Spectra before and after background subtraction. (a) Original spectrum and estimated background; (b) original spectrum and corrected spectrum
    Preprocessing effect after background subtraction
    Fig. 2. Preprocessing effect after background subtraction
    Flow chart of quantitative analysis method based on BOA-SVR
    Fig. 3. Flow chart of quantitative analysis method based on BOA-SVR
    Physical image of the sample and XRF spectrometer. (a) Sample; (b) XRF spectrometer
    Fig. 4. Physical image of the sample and XRF spectrometer. (a) Sample; (b) XRF spectrometer
    Sensitivity analysis result of the As element
    Fig. 5. Sensitivity analysis result of the As element
    Prediction results of the model under different feature dimensions
    Fig. 6. Prediction results of the model under different feature dimensions
    Prediction results of Cu element. (a) SVR model with feature dimension reduction; (b) SVR model with all features as inputs; (c) PLS model
    Fig. 7. Prediction results of Cu element. (a) SVR model with feature dimension reduction; (b) SVR model with all features as inputs; (c) PLS model
    Prediction results of As element. (a) SVR model with feature dimension reduction; (b) SVR model with all features as inputs; (c) PLS model
    Fig. 8. Prediction results of As element. (a) SVR model with feature dimension reduction; (b) SVR model with all features as inputs; (c) PLS model
    No.Reference valuePredictive valueRelative error
    SVR*SVRPLSSVR*SVRPLS
    225.723.9723.370.000.06720.09061.0000
    5916.0932.84908.241062.300.01840.00850.1597
    77.29.8466.270.000.36718.20471.0000
    10187.0179.84220.50195.000.03830.17910.0428
    1254.254.8874.2360.070.01250.36960.1082
    2122.610.60184.530.000.53087.16501.0000
    24118.0112.03133.18108.470.05060.12860.0807
    2745.070.77109.6382.730.57271.43620.8384
    32177.0170.59131.64184.140.03620.25630.0403
    3337.050.4856.6537.600.36420.53120.0162
    4829.029.4429.5439.150.01500.01870.3498
    52390.0457.35300.90237.890.17270.22850.3900
    5311.419.6912.830.000.72700.12591.0000
    Table 1. Prediction results of three models on verification set in Cu element verification
    ModelRCMSERC2RPMSERP2
    SVR*11.03340.997022.88030.9918
    SVR6.93560.998873.82960.9146
    PLS24.13190.985666.11330.9315
    Table 2. Cu element prediction results obtained by three models
    ModelRCMSERC2RPMSERP2
    SVR*1.12710.986311.68680.9526
    SVR0.30380.999616.52710.7534
    PLS17.09480.419237.59090.4899
    Table 3. As element prediction results obtained by three models
    Fusheng Li, Xiaolong Zeng. Quantitative Analysis Method of Soil Elements Combining Sensitivity Dimensionality Reduction and Support Vector Regression[J]. Laser & Optoelectronics Progress, 2023, 60(5): 0530002
    Download Citation