Research Article
Land Cover Classification Using SegNet with Slope, Aspect, and Multidirectional Shaded Relief Images Derived from Digital Surface Model
Table 3
Classification evaluation of Vaihingen dataset for training, evaluation, and test.
| Dataset | Training accuracy (%) | Evaluation accuracy (%) | Test data (%) | Overall | Building | Accuracy | IoU | Accuracy | IoU |
| IR image | 97.74 | 91.49 | 83.71 | 57.98 | 89.34 | 84.08 | DSM | 91.97 | 85.25 | 69.90 | 41.32 | 83.88 | 77.66 | Slope | 94.42 | 88.41 | 67.93 | 40.31 | 78.59 | 71.28 | Aspect | 95.87 | 89.07 | 64.55 | 35.90 | 79.01 | 70.58 | SRI | 93.50 | 89.37 | 70.00 | 42.59 | 83.74 | 75.90 | All data | 91.36 | 90.61 | 83.56 | 55.75 | 89.45 | 84.90 |
|
|