Fig. 1. Spatial distribution of study area
Fig. 2. Reflectance spectra with different pre-processing methods
Fig. 3. Spectral characteristics of wavelet transform
Fig. 4. Correlation analysis of wavelet transform
Fig. 5. The variables selected by SPA, CARS and VCPA
Fig. 6. Fitting diagrams of the optimal models of tea nitrogen based on (a) CWT (1 scale)-SPA-PLS, (b) CWT (1 scale)-CARS-PLS and (c) CWT (1 scale)-VCPA-PLS
模型 | 光谱处理 | 主成分 | 建模集 | 预测集 |
---|
| RMSEC | | RMSEP |
---|
| SG | 8 | 0.50 | 0.42 | 0.37 | 0.48 | | Detrending | 9 | 0.59 | 0.35 | 0.58 | 0.38 | SPA-PLSR | 1st | 7 | 0.63 | 0.37 | 0.61 | 0.35 | | SNV | 10 | 0.71 | 0.31 | 0.60 | 0.39 | | MSC | 8 | 0.62 | 0.38 | 0.68 | 0.29 | | SG | 5 | 0.62 | 0.35 | 0.55 | 0.41 | | Detrending | 10 | 0.75 | 0.30 | 0.71 | 0.29 | CARS-PLSR | 1st | 11 | 0.79 | 0.26 | 0.73 | 0.32 | | SNV | 11 | 0.71 | 0.33 | 0.68 | 0.29 | | MSC | 11 | 0.73 | 0.29 | 0.62 | 0.39 | | SG | 6 | 0.81 | 0.24 | 0.65 | 0.37 | | Detrending | 10 | 0.83 | 0.25 | 0.73 | 0.31 | VCPA-PLSR | 1st | 9 | 0.86 | 0.18 | 0.84 | 0.25 | | SNV | 7 | 0.81 | 0.25 | 0.67 | 0.35 | | MSC | 6 | 0.82 | 0.24 | 0.72 | 0.32 |
|
Table 1. Models based on different preprocessing methods and variables selection methods
模型 | 分解尺度 | 主成分 | 建模集 | 预测集 |
---|
| RMSEC | | RMSEP |
---|
SPA-PLSR | 1 | 5 | 0.83 | 0.25 | 0.72 | 0.37 | 2 | 7 | 0.72 | 0.30 | 0.68 | 0.36 | 3 | 10 | 0.74 | 0.31 | 0.69 | 0.33 | 4 | 11 | 0.78 | 0.28 | 0.67 | 0.35 | 5 | 11 | 0.79 | 0.26 | 0.60 | 0.41 | 6 | 11 | 0.65 | 0.33 | 0.60 | 0.42 | CARS-PLSR | 1 | 7 | 0.91 | 0.19 | 0.88 | 0.24 | 2 | 8 | 0.81 | 0.24 | 0.76 | 0.32 | 3 | 12 | 0.81 | 0.25 | 0.73 | 0.31 | 4 | 8 | 0.82 | 0.26 | 0.69 | 0.29 | 5 | 10 | 0.78 | 0.28 | 0.64 | 0.35 | 6 | 11 | 0.73 | 0.32 | 0.68 | 0.32 | VCPA-PLSR | 1 | 7 | 0.95 | 0.16 | 0.90 | 0.23 | 2 | 7 | 0.83 | 0.23 | 0.82 | 0.27 | 3 | 10 | 0.81 | 0.26 | 0.79 | 0.26 | 4 | 8 | 0.82 | 0.24 | 0.73 | 0.30 | 5 | 8 | 0.81 | 0.25 | 0.77 | 0.30 | 6 | 7 | 0.73 | 0.29 | 0.74 | 0.35 |
|
Table 2. Models based on different decomposition scales and variables selection methods