Research Article
A Deep Multiscale Fusion Method via Low-Rank Sparse Decomposition for Object Saliency Detection Based on Urban Data in Optical Remote Sensing Images
Table 1
Different indexes with different methods on different objects.
| | Object | Method | Precision | Recall | -measure | MAE |
| | Airplane1 | RA | 79.9% | 73.5% | 72.8% | 19.3% | | RB | 81.1% | 75.8% | 76.4% | 17.2% | | SC | 81.8% | 74.6% | 77.2% | 15.7% | | RAD | 87.4% | 77.4% | 79.5% | 14.6% | | SCLR | 91.7% | 75.4% | 81.8% | 12.5% | | Proposed | 95.6% | 65.3% | 82.5% | 9.8% |
| | Cloud | RA | 84.6% | 68.9% | 73.8% | 21.2% | | RB | 89.1% | 71.8% | 75.4% | 17.8% | | SC | 90.7% | 74.1% | 76.2% | 14.1% | | RAD | 91.6% | 73.7% | 78.4% | 12.6% | | SCLR | 93.6% | 72.8% | 80.9% | 11.3% | | Proposed | 97.4% | 71.5% | 83.6% | 7.6% |
| | Vehicle | RA | 89.2% | 77.4% | 78.7% | 19.5% | | RB | 91.5% | 79.9% | 80.8% | 17.6% | | SC | 93.1% | 79.3% | 82.1% | 13.8% | | RAD | 93.6% | 79.5% | 82.5% | 13.1% | | SCLR | 94.3% | 78.6% | 83.7% | 11.7% | | Proposed | 98.2% | 72.4% | 89.1% | 8.7% |
| | Playground | RA | 85.7% | 77.8% | 81.9% | 16.5% | | RB | 87.8% | 74.2% | 83.7% | 14.8% | | SC | 89.9% | 74.6% | 84.1% | 13.1% | | RAD | 91.8% | 73.3% | 84.5% | 12.5% | | SCLR | 92.4% | 71.8% | 86.7% | 10.2% | | Proposed | 97.2% | 59.6% | 89.7% | 9.4% |
| | Airplane2 | RA | 86.4% | 78.9% | 77.1% | 15.8% | | RB | 87.6% | 78.3% | 78.6% | 14.6% | | SC | 88.2% | 77.1% | 79.4% | 13.5% | | RAD | 88.3% | 76.6% | 80.8% | 12.9% | | SCLR | 91.3% | 75.8% | 81.6% | 11.9% | | Proposed | 96.3% | 62.4% | 83.9% | 8.1% |
| | Boat | RA | 85.4% | 84.1% | 72.4 | 24.5% | | RB | 88.6% | 78.2% | 74.1% | 21.2% | | SC | 89.7% | 76.1% | 76.4% | 19.4% | | RAD | 91.6% | 74.8% | 79.2% | 15.7% | | SCLR | 92.7% | 73.4% | 82.5% | 11.4% | | Proposed | 95.2% | 68.3% | 89.7% | 7.4% |
|
|
