Author Affiliations
1State Key Laboratory of Information Engineering in Surveying, Mapping and Remote Sensing, Wuhan University, Wuhan, Hubei 430079, China2Electronic Information School, Wuhan University, Wuhan, Hubei 430079, Chinashow less
Fig. 1. Flowchart of classification algorithm
Fig. 2. Tree structure of image. (a) Original image; (b) min-tree; (c) max-tree; (d) tree of shape
Fig. 3. Calculation process of node attributes. (a) Original image; (b) remove node B; (c) tree of shape before removing node B; (d) tree of shape after removing node B
Fig. 4. Real remote sensing image. (a) Sample images of Aν; (b)(c) change curves of and Aν from leaf node to root node; (d)(e) significant areas corresponding to two significant maximum points in Fig. (c)
Fig. 5. Hyperspectral images of Indian Pines dataset. (a) False color composite image; (b) survey results of feature types
Fig. 6. Hyperspectral images of Pavia university dataset. (a) False color composite image; (b) survey results of feature types
Fig. 7. Classification results of different algorithms on Indian Pines dataset. (a) SVM; (b) EMP; (c) EMAP; (d) EEP; (e) SC-MK; (f) ESP
Fig. 8. Classification results of different algorithms on Pavia university dataset. (a) SVM; (b) EMP; (c) EMAP; (d) EEP; (e) SC-MK; (f) ESP
Fig. 9. SP of different feature images on Pavia university dataset. (a) Original image; (b) SP0; (c) SP1; (d) SP2; (e) SP3; (f) SP4; (g) SP5; (h) SP6; (i) SP7; (j) SP8; (k) SP9; (l) SP10
Fig. 10. Relationship curves between hν and OA
No. | Name | Number of samples |
---|
1 | Corn-notill | 1428 | 2 | Corn-mintill | 830 | 3 | Grass-pasture | 483 | 4 | Grass-tree | 730 | 5 | Hay-windrow | 478 | 6 | Soybean-notill | 972 | 7 | Soybean-mintill | 2455 | 8 | Soybean-clean | 593 | 9 | Wood | 1265 | 10 | Building-grass-tree-drive | 386 | Total | | 9620 |
|
Table 1. Number of samples in Indian Pines dataset
No. | Name | Number of samples |
---|
1 | Asphalt | 6631 | 2 | Meadow | 18649 | 3 | Gravel | 2099 | 4 | Tree | 3064 | 5 | Painted metal sheet | 1345 | 6 | Bare soil | 5029 | 7 | Bitumen | 1330 | 8 | Self-blocking brick | 3682 | 9 | Shadow | 947 | Total | | 42776 |
|
Table 2. Number of samples in Pavia university dataset
Class | SVM | EMP | EMAP | EEP | SC-MK | ESP |
---|
1 | 66.99 | 70.29 | 76.79 | 79.30 | 81.00 | 92.54 | 2 | 67.59 | 70.92 | 78.60 | 79.28 | 87.78 | 94.88 | 3 | 91.48 | 90.81 | 91.34 | 92.24 | 95.61 | 97.71 | 4 | 93.71 | 94.81 | 96.65 | 97.40 | 97.32 | 99.94 | 5 | 99.44 | 99.39 | 99.65 | 99.51 | 99.98 | 100.00 | 6 | 74.32 | 69.92 | 82.66 | 84.39 | 85.80 | 94.08 | 7 | 55.89 | 60.64 | 72.67 | 74.02 | 77.09 | 91.86 | 8 | 72.69 | 67.66 | 74.99 | 79.96 | 88.03 | 93.08 | 9 | 84.32 | 85.88 | 92.91 | 93.79 | 94.04 | 99.10 | 10 | 73.27 | 79.58 | 92.77 | 94.88 | 94.67 | 98.87 | OA /% | 72.41 | 74.20 | 82.33 | 83.88 | 86.50 | 95.00 | AA /% | 77.97 | 78.99 | 85.90 | 87.48 | 90.13 | 96.20 | K | 68.38 | 70.40 | 79.63 | 81.42 | 84.40 | 94.19 |
|
Table 3. Classification accuracy of Indian Pines dataset
Class | SVM | EMP | EMAP | EEP | SC-MK | ESP |
---|
1 | 93.76 | 98.17 | 93.46 | 94.48 | 95.37 | 98.27 | 2 | 94.48 | 98.47 | 90.88 | 95.79 | 95.62 | 97.65 | 3 | 67.72 | 73.85 | 89.47 | 97.40 | 97.76 | 96.10 | 4 | 81.87 | 96.86 | 97.07 | 98.79 | 96.34 | 86.84 | 5 | 95.78 | 98.89 | 99.20 | 99.52 | 99.96 | 99.82 | 6 | 62.66 | 85.03 | 95.68 | 96.65 | 97.78 | 99.99 | 7 | 60.36 | 94.66 | 97.38 | 97.62 | 99.95 | 100.00 | 8 | 81.19 | 92.51 | 85.97 | 97.16 | 94.84 | 97.65 | 9 | 99.53 | 99.69 | 100.00 | 98.13 | 99.99 | 98.22 | OA /% | 84.24 | 94.43 | 92.44 | 96.76 | 96.28 | 97.33 | AA /% | 79.72 | 92.76 | 94.35 | 97.53 | 97.51 | 97.17 | K | 82.52 | 93.01 | 90.12 | 95.68 | 95.11 | 96.45 |
|
Table 4. Classification accuracy of Pavia university dataset
Feature map | Number of scale shapes |
---|
Original image | 53653 | SP0 | 2259 | SP1 | 1072 | SP2 | 645 | SP3 | 416 | SP4 | 242 | SP5 | 170 | SP6 | 120 | SP7 | 103 | SP8 | 85 | SP9 | 60 | SP10 | 41 |
|
Table 5. Comparison of number of scale shapes of different feature images