This is an implementation of A Fast Neighborhood Grouping Method for Hyperspectral Band Selection.
Four public datasets, i.e., Indian Pines, Botswana, Pavia University, and Salinas, are employed to verify the effectiveness of the proposed FNGBS.
MATLAB, libsvm, cruve fitting tool
With respect to FNGBS algorithm, to run the code, please perform 'main.m'. As for obtained recommended bands, we need to conduct 'main_recomBand.m'. Note that the hyperparameters of each dataset is set as follows
| Dataset | k | Z |
|---|---|---|
| Indian Pines | 2 | 10% |
| Botswana | 2 | 1% |
| Pavia University | 3 | 1% |
| Salinas | 3 | 1% |
To qualitatively measure the proposed FNGBS, KNN and SVM classifiers are employed to verify the effectiveness of the algorithm.
| Dataset | E-FDPC | WaLuDi | SNNC | TOF | FNGBS (1%) | FNGBS (100%) |
|---|---|---|---|---|---|---|
| Indian Pines (6 bands) | 0.121 | 7.430 | 0.4411 | 0.4165 | 0.2542 | 0.2995 |
| Botswana (8 bands) | 0.661 | 99.281 | 3.738 | 1.843 | 0.892 | 3.442 |
| Pavia University (13 bands) | 0.282 | 27.930 | 1.201 | 0.925 | 0.336 | 1.421 |
| Salinas (6 bands) | 0.381 | 40.382 | 1.61 | 1.276 | 0.465 | 1.464 |
Please consider cite this paper if you find it helpful.
@article{wang2020afast,
title={A Fast Neighborhood Grouping Method for Hyperspectral Band Selection},
author={Q. Wang, Q. Li and X. Li},
journal={IEEE Transactions on Geoscience and Remote Sensing},
year={2020},
doi={10.1109/TGRS.2020.3011002}
}
If you has any questions, please send e-mail to [email protected].