Table 6. Comparison of classification accuracy by PCD deep learning models. Unit: %
Model | Platform | Down-sampling intensity | OA | Kappa | F-Score |
Pinus koraiensis | Larix kaempferi |
PointNet | TLS | 1024 | 94.9 | 87.3 | 96.4 | 90.9 |
2048 | 89.7 | 74.0 | 93.0 | 81.0 |
4096 | 89.7 | 74.0 | 93.0 | 81.0 |
8192 | 96.2 | 90.6 | 97.3 | 93.3 |
MLS | 1024 | 87.2 | 66.6 | 91.4 | 75.0 |
2048 | 87.2 | 67.5 | 91.2 | 76.2 |
4096 | 83.3 | 58.3 | 88.5 | 69.8 |
8192 | 92.3 | 81.5 | 94.5 | 87.0 |
PointNet++ | TLS | 1024 | 96.2 | 90.6 | 97.3 | 93.3 |
2048 | 93.6 | 85.1 | 95.3 | 89.8 |
4096 | 94.9 | 87.7 | 96.4 | 91.3 |
8192 | 97.4 | 93.7 | 98.2 | 95.5 |
MLS | 1024 | 96.2 | 90.6 | 97.3 | 93.3 |
2048 | 98.7 | 96.9 | 99.1 | 97.8 |
4096 | 92.3 | 81.5 | 94.5 | 87.0 |
8192 | 93.6 | 84.0 | 95.6 | 88.4 |