Research Article
Composite Clustering Sampling Strategy for Multiscale Spectral-Spatial Classification of Hyperspectral Images
Table 4
Classification accuracy and test set independence rate with different sampling strategies of the Indian Pines dataset.
| Sampling strategy | 3D-CNN | 1D-3D-2D-CNN | Test set-independent sample rate (%) | OA (%) | AA (%) | Kappa | Time (min.) | OA (%) | AA (%) | Kappa | Time (min.) |
| Area | 46.81 | 42.55 | 0.3950 | 9.24 | 49.76 | 51.12 | 0.4392 | 1.95 | 90.93 | Random | 90.33 | 91.69 | 0.8894 | 9.24 | 94.36 | 92.91 | 0.9356 | 1.87 | 1.75 | | 63.55 | 62.66 | 0.5808 | 10.15 | 77.19 | 70.83 | 0.7402 | 1.98 | 81.88 | | 58.76 | 58.84 | 0.5284 | 10.11 | 73.72 | 70.68 | 0.7007 | 2.01 | 81.48 | | 65.57 | 64.35 | 0.6059 | 10.10 | 76.06 | 72.53 | 0.7278 | 2.00 | 81.52 | | 66.24 | 66.20 | 0.6136 | 10.12 | 77.01 | 70.58 | 0.7385 | 2.04 | 80.17 | | 66.73 | 68.39 | 0.6161 | 10.21 | 78.60 | 77.01 | 0.7558 | 2.13 | 77.23 | | 63.57 | 65.34 | 0.5803 | 9.37 | 79.83 | 77.50 | 0.7699 | 2.10 | 75.51 | | 67.90 | 68.93 | 0.6295 | 9.41 | 78.27 | 77.38 | 0.7501 | 2.05 | 75.15 | | 66.19 | 68.59 | 0.6104 | 9.48 | 80.23 | 79.38 | 0.7742 | 2.12 | 75.02 | | 65.49 | 68.57 | 0.6019 | 9.42 | 78.62 | 79.64 | 0.7554 | 2.29 | 74.23 | | 65.08 | 69.98 | 0.5963 | 9.50 | 82.86 | 82.08 | 0.8039 | 2.34 | 71.68 | | 69.22 | 72.12 | 0.6443 | 9.51 | 80.45 | 78.71 | 0.7769 | 2.36 | 70.20 | | 69.73 | 73.32 | 0.6517 | 9.56 | 80.30 | 80.22 | 0.7743 | 2.40 | 69.34 | | 70.94 | 71.98 | 0.6650 | 9.61 | 80.09 | 78.36 | 0.7720 | 2.46 | 67.98 | | 69.40 | 73.11 | 0.6467 | 9.69 | 83.01 | 81.13 | 0.8059 | 2.44 | 67.34 | | 70.24 | 72.06 | 0.6566 | 9.25 | 81.94 | 80.28 | 0.7932 | 2.46 | 65.81 |
|
|