Table 2 Comparison of evaluation metrics of classification results between this paper’s method and nine point cloud semantic segmentation methods on the DALES dataset(%).
From: An aerial point cloud classification using point transformer via multi-feature fusion
Method | IoU | mIoU | OA | |||||||
|---|---|---|---|---|---|---|---|---|---|---|
Grounds | Building | Cars | Trucks | Poles | Power lines | Fences | Vegetation | |||
KPConv | 97.12 | 96.62 | 85.33 | 41.90 | 75.01 | 95.50 | 63.50 | 94.11 | 81.12 | 97.82 |
PointNet++ | 94.11 | 89.14 | 75.42 | 30.32 | 40.03 | 79.92 | 46.22 | 91.20 | 68.31 | 95.71 |
ConvPoint | 96.90 | 96.32 | 75.50 | 21.73 | 40.32 | 86.70 | 29.60 | 91.92 | 67.42 | 97.22 |
SuperPoint | 94.73 | 93.45 | 62.91 | 18.71 | 28.50 | 65.23 | 33.61 | 87.91 | 60.62 | 95.50 |
PointCNN | 97.50 | 95.70 | 40.65 | 4.82 | 57.62 | 26.74 | 52.60 | 91.72 | 58.40 | 97.22 |
ShellNet | 96.01 | 95.42 | 32.22 | 39.60 | 20.01 | 27.40 | 60.02 | 88.40 | 57.42 | 96.41 |
Point transformer | 93.74 | 95.90 | 82.88 | 33.32 | 67.33 | 91.14 | 57.39 | 92.46 | 76.77 | 96.05 |
Ours method | 98.96 | 97.88 | 85.34 | 48.75 | 72.15 | 95.90 | 63.75 | 94.80 | 82.18 | 98.23 |
