Table 3.
The classification accuracies (%) in SPA bands with PLS-DA and SVM models (Cal and Pre are the abbreviations of calibration dataset and prediction dataset).
| Spectral Type | Models | (A) Spectra | (B) Images | (C) Spectrum and Image Fusion | |||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| SPA Bands | GLCM | GLRLM | SPA Bands +GLCM | SPA Bands +GLRLM | |||||||
| Cal | Pre | Cal | Pre | Cal | Pre | Cal | Pre | Cal | Pre | ||
| (I) VNIR | PLS-DA | 78.1 | 75.7 | 78.3 | 77.0 | 77.3 | 79.7 | 81.9 | 86.5 | 84.7 | 86.5 |
| SVM | 78.7 | 80.8 | 76.5 | 70.1 | 79.9 | 76.6 | 80.1 | 81.9 | 83.3 | 83.1 | |
| (II) SWIR | PLS-DA | 82.2 | 79.7 | 75.5 | 78.4 | 74.9 | 74.3 | 87.0 | 86.5 | 85.4 | 83.8 |
| SVM | 86.8 | 80.9 | 68.4 | 72.3 | 76.0 | 75.9 | 81.9 | 82.3 | 87.5 | 88.4 | |
| (III) VNIR and SWIR Fusion | PLS-DA | 86.5 | 83.8 | 79.0 | 77.0 | 78.1 | 78.4 | 89.1 | 89.2 | 89.0 | 93.2 |
| SVM | 92.0 | 88.4 | 73.6 | 71.1 | 81.5 | 79.5 | 82.0 | 82.2 | 88.2 | 89.6 | |