• Laser & Optoelectronics Progress
  • Vol. 57, Issue 24, 242803 (2020)
Si Ran1、2, Jianli Ding1、2、*, Xiangyu Ge1、2, Bohua Liu1、2, and Junyong Zhang1、2
Author Affiliations
  • 1College of Resources & Environmental Science, Xinjiang University, Urumqi, Xinjiang 830046, China;
  • 2Key Laboratory of Oasis Ecology, Xinjiang University, Urumqi, Xinjiang 830046, China
  • show less
    DOI: 10.3788/LOP57.242803 Cite this Article Set citation alerts
    Si Ran, Jianli Ding, Xiangyu Ge, Bohua Liu, Junyong Zhang. Estimation Method of VIS-NIR Spectroscopy for Soil Organic Matter Based on Sparse Networks[J]. Laser & Optoelectronics Progress, 2020, 57(24): 242803 Copy Citation Text show less
    AE network structure
    Fig. 1. AE network structure
    Results of combined method of PCA and Mahalanobis distance to eliminate abnormal samples
    Fig. 2. Results of combined method of PCA and Mahalanobis distance to eliminate abnormal samples
    Reflectance curves of SOM and different VIS-NIR spectra. (a) Original spectrum; (b) first-order spectrum; (c) second-order spectrum
    Fig. 3. Reflectance curves of SOM and different VIS-NIR spectra. (a) Original spectrum; (b) first-order spectrum; (c) second-order spectrum
    Correlation curves of different VIS-NIR spectra and SOM. (a) Original spectrum; (b) first-order spectrum; (c) second-order spectrum
    Fig. 4. Correlation curves of different VIS-NIR spectra and SOM. (a) Original spectrum; (b) first-order spectrum; (c) second-order spectrum
    PCA eigenvalues and contribution rate
    Fig. 5. PCA eigenvalues and contribution rate
    RMSE curve of characteristic band
    Fig. 6. RMSE curve of characteristic band
    SAE network structure and parameters
    Fig. 7. SAE network structure and parameters
    Relationship between variable selection method and R2 in SOM
    Fig. 8. Relationship between variable selection method and R2 in SOM
    DatasetCountMinMiddleMaxMeanSDCV
    Whole dataset890.4012.5645.4713.078.080.62
    Calibration dataset590.4012.5645.4712.676.980.55
    Validation dataset300.5612.7242.5413.849.990.76
    Table 1. SOM statistics
    ModelExtract variableCalibrationPrediction
    R2RMSER2RMSERPDRPIQ
    PLSRSPA0.645.020.535.561.471.70
    PCA0.674.810.615.081.612.09
    SAE0.664.900.615.231.582.07
    BPSPA0.754.360.714.431.852.21
    PCA0.803.900.764.082.012.40
    SAE0.853.380.823.532.322.78
    Table 2. Model accuracy of different feature variable selections and modeling methods
    Si Ran, Jianli Ding, Xiangyu Ge, Bohua Liu, Junyong Zhang. Estimation Method of VIS-NIR Spectroscopy for Soil Organic Matter Based on Sparse Networks[J]. Laser & Optoelectronics Progress, 2020, 57(24): 242803
    Download Citation