Figure 1.
The evolution process of energy functional.
Figure 1.
The evolution process of energy functional.
Figure 2.
Disjoint domain decomposition.
Figure 2.
Disjoint domain decomposition.
Figure 3.
All possible triangles in the neighborhood (a) down; (b) down and left (P); (c) mix.
Figure 3.
All possible triangles in the neighborhood (a) down; (b) down and left (P); (c) mix.
Figure 4.
Eight types of the triangular tangent planes through (a) two of the common edges from the four tangent plane (R); (b) two of the common edges from the four tangent plane (P); (c) four of the tangent planes through mixed neighbors.
Figure 4.
Eight types of the triangular tangent planes through (a) two of the common edges from the four tangent plane (R); (b) two of the common edges from the four tangent plane (P); (c) four of the tangent planes through mixed neighbors.
Figure 5.
Flow of the proposed curvature filter domain transform recursive filter (CFDTRF-LDM).
Figure 5.
Flow of the proposed curvature filter domain transform recursive filter (CFDTRF-LDM).
Figure 6.
Curvature filter (CF) and domain transform recursive filter (DTRF) comparison for Indian Pines; (a) the 10th band of spectrum; (b) the 60th band of spectrum; (c) the 130th band of spectrum; (d) the 180th band of spectrum; (e) the 10th band filtering of CF; (f) the 60th band filtering of CF; (g) the 130th band filtering of CF; (h) the 180th band filtering of CF; (i) the 10th band filtering of DTRF; (j) the 60th band filtering of DTRF; (k) the 130th band filtering of DTRF; (h) the 180th band filtering of DTRF.
Figure 6.
Curvature filter (CF) and domain transform recursive filter (DTRF) comparison for Indian Pines; (a) the 10th band of spectrum; (b) the 60th band of spectrum; (c) the 130th band of spectrum; (d) the 180th band of spectrum; (e) the 10th band filtering of CF; (f) the 60th band filtering of CF; (g) the 130th band filtering of CF; (h) the 180th band filtering of CF; (i) the 10th band filtering of DTRF; (j) the 60th band filtering of DTRF; (k) the 130th band filtering of DTRF; (h) the 180th band filtering of DTRF.
Figure 7.
Average of Moran’s I for hyperspectral images (HSI) (a) Indian Pines (b) Salinas Valley (c) Kennedy Space Center.
Figure 7.
Average of Moran’s I for hyperspectral images (HSI) (a) Indian Pines (b) Salinas Valley (c) Kennedy Space Center.
Figure 8.
Classification maps of different methods on the Indian Pines dataset (a) ground; (b) training; (c) SVM, overall accuracy (OA) = 77.47%; (d) principal component analysis (PCA)-SVM, OA = 77.81% (e) large margin distribution machine (LDM), OA = 79.85%; (f) PCA-LDM, OA = 78.49%; (g) edge-preserving filter (EPF), OA = 90.35%; (h) image fusion with multiple subsets of adjacent bands and recursive filter (IFRF), OA = 90.64%; (i) PCA-EPFs, OA = 91.62%; (j) LDM-FL, OA = 93.31%; (k) CF-SVM, OA = 88.09%; (l) CF-LDM, OA = 88.54%; (m) DTRF-SVM, OA = 92.29%; (n) DTRF-LDM, OA = 94.60%; (o) CFDTRFF-SVM, OA = 94.13%; (p) CFDTRF-LDM, OA = 96.64%.
Figure 8.
Classification maps of different methods on the Indian Pines dataset (a) ground; (b) training; (c) SVM, overall accuracy (OA) = 77.47%; (d) principal component analysis (PCA)-SVM, OA = 77.81% (e) large margin distribution machine (LDM), OA = 79.85%; (f) PCA-LDM, OA = 78.49%; (g) edge-preserving filter (EPF), OA = 90.35%; (h) image fusion with multiple subsets of adjacent bands and recursive filter (IFRF), OA = 90.64%; (i) PCA-EPFs, OA = 91.62%; (j) LDM-FL, OA = 93.31%; (k) CF-SVM, OA = 88.09%; (l) CF-LDM, OA = 88.54%; (m) DTRF-SVM, OA = 92.29%; (n) DTRF-LDM, OA = 94.60%; (o) CFDTRFF-SVM, OA = 94.13%; (p) CFDTRF-LDM, OA = 96.64%.
Figure 9.
Classification maps of different methods on the Salinas Valley dataset (a) ground; (b) training; (c) SVM, OA = 87.99%; (d) PCA-SVM, OA = 87.45%; (e) LDM, OA = 88.96%; (f) PCA-LDM, OA = 89.19%; (g) EPF, OA = 91.37%; (h) IFRF, OA = 97.52%; (i) PCA-EPFs, OA = 98.68%; (j) LDM-FL, OA = 98.76%; (k) CF-SVM, OA = 89.20%; (l) CF-LDM, OA = 90.56%; (m) DTRF-SVM, OA = 96.71%; (n) DTRF-LDM, OA = 98.52%; (o) CFDTRFF-SVM, OA = 97.93%; (p) CFDTRF-LDM, OA = 99.16%.
Figure 9.
Classification maps of different methods on the Salinas Valley dataset (a) ground; (b) training; (c) SVM, OA = 87.99%; (d) PCA-SVM, OA = 87.45%; (e) LDM, OA = 88.96%; (f) PCA-LDM, OA = 89.19%; (g) EPF, OA = 91.37%; (h) IFRF, OA = 97.52%; (i) PCA-EPFs, OA = 98.68%; (j) LDM-FL, OA = 98.76%; (k) CF-SVM, OA = 89.20%; (l) CF-LDM, OA = 90.56%; (m) DTRF-SVM, OA = 96.71%; (n) DTRF-LDM, OA = 98.52%; (o) CFDTRFF-SVM, OA = 97.93%; (p) CFDTRF-LDM, OA = 99.16%.
Figure 10.
Classification maps of different methods on the Salinas Valley dataset (a) Ground; (b) Training; (c) SVM, OA = 82.45%; (d) PCA-SVM, OA = 79.49%; (e) LDM, OA = 85.11%; (f) PCA-LDM, OA = 80.65%; (g) EPF, OA = 89.03%; (h) IFRF, OA = 86.21%; (i) PCA-EPFs, OA = 94.12%; (j) LDM-FL, OA = 90.17%; (k) CF-SVM, OA = 90.07%; (l) CF-LDM, OA = 91.02%; (m) DTRF-SVM, OA = 92.62%; (n) DTRF-LDM, OA = 95.24%; (o) CFDTRFF-SVM, OA = 95.89%; (p) CFDTRF-LDM, OA = 97.33%.
Figure 10.
Classification maps of different methods on the Salinas Valley dataset (a) Ground; (b) Training; (c) SVM, OA = 82.45%; (d) PCA-SVM, OA = 79.49%; (e) LDM, OA = 85.11%; (f) PCA-LDM, OA = 80.65%; (g) EPF, OA = 89.03%; (h) IFRF, OA = 86.21%; (i) PCA-EPFs, OA = 94.12%; (j) LDM-FL, OA = 90.17%; (k) CF-SVM, OA = 90.07%; (l) CF-LDM, OA = 91.02%; (m) DTRF-SVM, OA = 92.62%; (n) DTRF-LDM, OA = 95.24%; (o) CFDTRFF-SVM, OA = 95.89%; (p) CFDTRF-LDM, OA = 97.33%.
Figure 11.
Comparison of SVM, LDM, PCA-SVM and PCA-LDM on three datasets.
Figure 11.
Comparison of SVM, LDM, PCA-SVM and PCA-LDM on three datasets.
Figure 12.
Comparison of SVM, CF-SVM, LDM and CF-LDM on three datasets.
Figure 12.
Comparison of SVM, CF-SVM, LDM and CF-LDM on three datasets.
Figure 13.
Comparison of SVM, DTRF-SVM, LDM and DTRF-LDM on three datasets.
Figure 13.
Comparison of SVM, DTRF-SVM, LDM and DTRF-LDM on three datasets.
Figure 14.
Comparison of EPF, PCA-EPFs, IFRF, LDM-FL and CFDTRF-LDM on three datasets.
Figure 14.
Comparison of EPF, PCA-EPFs, IFRF, LDM-FL and CFDTRF-LDM on three datasets.
Figure 15.
Effect of different training ratios on classification performance (a) Indian Pines (b) Salinas Valley (c) Kennedy Space Center.
Figure 15.
Effect of different training ratios on classification performance (a) Indian Pines (b) Salinas Valley (c) Kennedy Space Center.
Table 1.
Comparison of classification accuracies (in percent) provided by seven methods for Indian Pines (part A).
Table 1.
Comparison of classification accuracies (in percent) provided by seven methods for Indian Pines (part A).
Ground | Sum | Train | SVM | PCA-SVM | LDM | PCA-LDM | EPF | IFRF | PCA-EPFs |
---|
Alfalfa | 54 | 11 | 45.59 | 73.20 | 90.08 | 83.42 | 55.88 | 89.51 | 83.28 |
Corn-no-till | 1434 | 72 | 64.06 | 67.16 | 72.90 | 74.16 | 84.55 | 89.87 | 86.18 |
Corn-min-till | 834 | 42 | 72.28 | 71.24 | 67.41 | 57.38 | 88.49 | 78.09 | 91.12 |
Corn | 234 | 12 | 15.73 | 34.51 | 56.38 | 60.08 | 19.11 | 69.25 | 78.80 |
Grass-pasture | 497 | 25 | 87.22 | 84.36 | 88.90 | 91.85 | 91.56 | 92.72 | 91.49 |
Grass-trees | 747 | 37 | 94.11 | 95.83 | 94.98 | 94.35 | 99.89 | 97.97 | 93.53 |
Grass-pasture-mowed | 26 | 5 | 45.46 | 73.57 | 83.57 | 69.11 | 43.86 | 64.18 | 60.59 |
Hay-windrowed | 489 | 24 | 98.26 | 96.96 | 96.46 | 94.62 | 100.00 | 99.51 | 99.83 |
Oats | 20 | 4 | 29.46 | 26.68 | 73.53 | 87.85 | 18.30 | 41.29 | 42.32 |
Soybeans-no-till | 968 | 48 | 65.89 | 61.67 | 67.49 | 72.83 | 86.69 | 84.51 | 87.86 |
Soybeans-min-till | 2468 | 123 | 82.38 | 82.87 | 79.34 | 73.30 | 97.83 | 94.58 | 96.35 |
Soybeans-clean-till | 614 | 31 | 76.40 | 76.03 | 80.35 | 73.29 | 95.47 | 89.31 | 88.21 |
Wheat | 212 | 11 | 95.66 | 98.28 | 99.51 | 99.01 | 99.88 | 99.16 | 76.38 |
Woods | 1294 | 65 | 95.64 | 97.25 | 92.65 | 91.81 | 99.57 | 98.55 | 98.36 |
Bldg-grass-tree | 380 | 19 | 41.31 | 33.53 | 61.07 | 56.65 | 53.77 | 76.75 | 91.20 |
Stone-steel-towers | 95 | 5 | 81.22 | 57.49 | 86.38 | 76.38 | 93.87 | 67.63 | 58.59 |
OA/% | - | | 77.47 | 77.81 | 79.85 | 78.49 | 90.35 | 90.64 | 91.62 |
AA/% | - | | 68.17 | 70.66 | 80.69 | 78.51 | 76.80 | 83.31 | 82.76 |
Kappa/% | - | | 74.12 | 74.52 | 77.03 | 77.36 | 88.92 | 89.30 | 90.43 |
Table 2.
Comparison of classification accuracies (in percent) provided by seven methods for Indian Pines (part B).
Table 2.
Comparison of classification accuracies (in percent) provided by seven methods for Indian Pines (part B).
Ground | Sum | Train | LDM-FL | CF-SVM | CF-LDM | DTRF-SVM | DTRF-LDM | CFDTRFF-SVM | CFDTRF-LDM |
---|
Alfalfa | 54 | 11 | 93.13 | 86.69 | 89.41 | 92.38 | 98.73 | 76.32 | 93.62 |
Corn-no-till | 1434 | 72 | 92.17 | 84.60 | 86.26 | 87.98 | 91.07 | 94.74 | 96.85 |
Corn-min-till | 834 | 42 | 89.89 | 83.40 | 81.41 | 92.51 | 92.21 | 92.29 | 94.71 |
Corn | 234 | 12 | 77.60 | 69.97 | 73.33 | 78.79 | 94.72 | 73.56 | 88.17 |
Grass-pasture | 497 | 25 | 93.43 | 94.71 | 94.67 | 89.58 | 96.36 | 92.68 | 92.29 |
Grass-trees | 747 | 37 | 96.56 | 97.39 | 98.51 | 94.76 | 97.01 | 97.32 | 99.01 |
Grass-pasture-mowed | 26 | 5 | 100.0 | 74.11 | 97.50 | 27.94 | 100.0 | 98.61 | 93.45 |
Hay-windrowed | 489 | 24 | 100.00 | 98.65 | 98.80 | 99.78 | 100.00 | 99.40 | 99.78 |
Oats | 20 | 4 | 93.74 | 52.35 | 98.33 | 5.88 | 93.20 | 70.31 | 100.00 |
Soybeans-no-till | 968 | 48 | 91.57 | 78.69 | 85.60 | 86.39 | 92.45 | 87.69 | 95.21 |
Soybeans-min-till | 2468 | 123 | 92.56 | 91.60 | 86.97 | 95.57 | 95.02 | 96.80 | 96.97 |
Soybeans-clean-till | 614 | 31 | 91.17 | 87.21 | 86.19 | 89.96 | 90.69 | 91.53 | 92.06 |
Wheat | 212 | 11 | 99.14 | 99.12 | 99.26 | 97.24 | 94.09 | 99.25 | 99.75 |
Woods | 1294 | 65 | 98.98 | 98.34 | 96.48 | 98.74 | 99.67 | 98.82 | 99.96 |
Bldg-grass-tree | 380 | 19 | 90.20 | 48.87 | 68.97 | 92.66 | 93.68 | 91.12 | 99.16 |
Stone-steel-towers | 95 | 5 | 92.09 | 84.00 | 88.38 | 67.51 | 81.45 | 70.04 | 95.67 |
OA/% | - | | 93.31 | 88.09 | 88.54 | 92.29 | 94.60 | 94.13 | 96.64 |
AA/% | - | | 93.26 | 83.11 | 89.38 | 81.10 | 94.40 | 89.41 | 96.04 |
Kappa/% | - | | 92.39 | 86.36 | 86.94 | 91.20 | 93.84 | 93.29 | 96.16 |
Table 3.
Comparison of classification accuracies (in percent) provided by seven methods for Salinas Valley (part A).
Table 3.
Comparison of classification accuracies (in percent) provided by seven methods for Salinas Valley (part A).
Ground | Sum | Training | SVM | PCA-SVM | LDM | PCA-LDM | EPF | IFRF | PCA-EPFs |
---|
Broccoli-green-weeds-1 | 2009 | 16 | 96.68 | 98.91 | 99.06 | 99.32 | 99.84 | 99.93 | 99.86 |
Broccoli green-weeds-2 | 3726 | 30 | 99.03 | 98.59 | 99.08 | 99.03 | 100.00 | 98.88 | 99.43 |
Fallow | 1976 | 16 | 95.91 | 85.30 | 94.30 | 98.04 | 86.58 | 99.96 | 99.66 |
Fallow-rough-plough | 1394 | 11 | 96.34 | 94.50 | 99.06 | 99.35 | 99.87 | 95.11 | 92.39 |
Fallow-smooth | 2678 | 21 | 90.60 | 97.33 | 95.97 | 96.51 | 99.50 | 96.28 | 98.29 |
Stubble | 3959 | 32 | 99.57 | 99.50 | 99.84 | 99.76 | 100.00 | 99.62 | 99.84 |
Celery | 3579 | 29 | 99.35 | 99.34 | 99.61 | 99.46 | 100.00 | 99.16 | 99.18 |
Grapes-untrained | 11271 | 90 | 89.94 | 86.19 | 78.05 | 76.14 | 95.01 | 96.04 | 98.45 |
Soil-vineyard-develop | 6203 | 50 | 98.22 | 99.00 | 99.35 | 99.60 | 99.96 | 100.00 | 100.00 |
Corn-senesced-green weeds | 3278 | 26 | 89.33 | 83.14 | 92.17 | 93.11 | 92.23 | 99.02 | 99.41 |
Lettuce-romaine-4wk | 1068 | 9 | 70.95 | 66.80 | 92.34 | 91.60 | 97.76 | 90.80 | 88.48 |
Lettuce-romaine-5wk | 1927 | 15 | 98.29 | 93.45 | 99.69 | 99.62 | 100.00 | 98.08 | 98.45 |
Lettuce-romaine-6wk | 916 | 7 | 98.43 | 50.42 | 97.66 | 97.99 | 100.00 | 83.46 | 96.41 |
Lettuce-romaine-7wk | 1070 | 9 | 88.09 | 94.49 | 94.12 | 95.50 | 99.60 | 95.31 | 96.89 |
Vineyard-untrained | 7268 | 58 | 51.00 | 61.05 | 63.39 | 65.96 | 53.39 | 97.89 | 99.86 |
Vineyard-vertical-trellis | 1807 | 14 | 81.60 | 86.77 | 96.47 | 97.13 | 91.75 | 93.45 | 95.86 |
OA/% | - | | 87.99 | 87.45 | 88.96 | 89.19 | 91.37 | 97.52 | 98.68 |
AA/% | - | | 90.21 | 87.17 | 93.76 | 94.26 | 94.72 | 96.44 | 97.65 |
Kappa/% | - | | 86.57 | 85.96 | 87.70 | 87.97 | 90.34 | 97.23 | 98.53 |
Table 4.
Comparison of classification accuracies (in percent) provided by seven methods for Salinas Valley (part B).
Table 4.
Comparison of classification accuracies (in percent) provided by seven methods for Salinas Valley (part B).
Ground | Sum | Training | LDM-FL | CF-SVM | CF-LDM | DTRF-SVM | DTRF-LDM | CFDTRFF-SVM | CFDTRF-LDM |
---|
Broccoli-green-weeds-1 | 2009 | 16 | 99.99 | 99.91 | 99.95 | 99.96 | 100.00 | 100.00 | 100.00 |
Broccoli green-weeds-2 | 3726 | 30 | 99.75 | 97.44 | 99.57 | 99.36 | 99.81 | 99.80 | 99.98 |
Fallow | 1976 | 16 | 99.96 | 92.66 | 99.95 | 97.74 | 98.31 | 97.05 | 100.00 |
Fallow-rough-plough | 1394 | 11 | 96.41 | 98.59 | 98.88 | 89.95 | 91.38 | 99.15 | 98.46 |
Fallow-smooth | 2678 | 21 | 99.03 | 96.73 | 99.04 | 94.93 | 94.79 | 98.07 | 98.73 |
Stubble | 3959 | 32 | 99.55 | 99.36 | 99.76 | 97.70 | 98.84 | 99.58 | 99.90 |
Celery | 3579 | 29 | 99.85 | 99.57 | 99.76 | 99.87 | 99.83 | 99.73 | 99.72 |
Grapes-untrained | 11271 | 90 | 98.50 | 87.98 | 83.92 | 97.82 | 99.24 | 97.54 | 99.18 |
Soil-vineyard-develop | 6203 | 50 | 100.00 | 99.87 | 99.67 | 100.00 | 100.00 | 99.63 | 100.00 |
Corn-senesced-green weeds | 3278 | 26 | 99.32 | 86.16 | 93.23 | 96.27 | 96.53 | 95.97 | 98.19 |
Lettuce-romaine-4wk | 1068 | 9 | 91.08 | 56.40 | 94.38 | 64.07 | 94.06 | 96.65 | 96.84 |
Lettuce-romaine-5wk | 1927 | 15 | 98.37 | 86.48 | 100.00 | 98.52 | 99.01 | 99.95 | 100.00 |
Lettuce-romaine-6wk | 916 | 7 | 93.09 | 97.61 | 97.04 | 93.30 | 94.46 | 93.36 | 98.26 |
Lettuce-romaine-7wk | 1070 | 9 | 95.06 | 89.91 | 97.41 | 74.42 | 94.66 | 91.38 | 93.20 |
Vineyard-untrained | 7268 | 58 | 99.16 | 64.93 | 76.36 | 98.21 | 99.40 | 96.37 | 99.03 |
Vineyard-vertical-trellis | 1807 | 14 | 96.24 | 88.78 | 97.80 | 95.21 | 99.43 | 95.69 | 97.82 |
OA/% | - | | 98.76 | 89.20 | 92.60 | 96.71 | 98.52 | 97.93 | 99.16 |
AA/% | - | | 97.84 | 90.15 | 96.04 | 93.58 | 97.48 | 97.49 | 98.71 |
Kappa/% | - | | 98.62 | 87.92 | 91.76 | 96.33 | 98.35 | 97.70 | 99.06 |
Table 5.
Comparison of classification accuracies (in percent) provided by seven methods for Kennedy Space Center (part A).
Table 5.
Comparison of classification accuracies (in percent) provided by seven methods for Kennedy Space Center (part A).
Ground | Sum | Training | SVM | PCA-SVM | LDM | PCA-LDM | EPF | IFRF | PCA-EPFs |
---|
Scrub | 761 | 30 | 97.66 | 87.95 | 89.54 | 98.43 | 100.00 | 95.91 | 99.00 |
Swamp willow | 243 | 10 | 85.42 | 70.41 | 84.78 | 68.98 | 81.12 | 35.46 | 59.96 |
Cabbage palm hammock | 256 | 10 | 84.99 | 70.54 | 88.39 | 78.69 | 94.85 | 92.98 | 97.27 |
Cabbage palm/oak | 252 | 10 | 49.68 | 46.28 | 55.33 | 44.40 | 80.40 | 57.85 | 97.32 |
Slash pine | 161 | 6 | 8.28 | 38.67 | 52.71 | 26.32 | 12.83 | 63.77 | 64.70 |
Oak/broadleaf hammock | 229 | 9 | 25.68 | 35.82 | 56.79 | 0.23 | 21.37 | 60.51 | 77.86 |
Hardwood swamp | 105 | 4 | 38.89 | 57.99 | 66.96 | 41.09 | 48.51 | 77.87 | 81.76 |
Graminoid marsh | 431 | 17 | 75.44 | 63.40 | 72.97 | 71.72 | 93.46 | 91.28 | 99.71 |
Spartina marsh | 520 | 21 | 94.09 | 92.51 | 94.58 | 96.59 | 100.00 | 87.57 | 100.00 |
Cattail marsh | 404 | 16 | 90.81 | 92.34 | 94.56 | 92.54 | 95.80 | 97.04 | 92.78 |
Salt marsh | 419 | 17 | 93.85 | 92.56 | 90.71 | 88.33 | 98.33 | 88.03 | 99.83 |
Muld flats | 503 | 20 | 78.52 | 73.93 | 83.61 | 84.78 | 93.53 | 96.80 | 94.18 |
Water | 927 | 37 | 99.94 | 98.54 | 98.18 | 98.08 | 100.00 | 100.00 | 100.00 |
OA/% | - | | 82.45 | 79.49 | 85.11 | 80.65 | 89.03 | 87.16 | 94.12 |
AA/% | - | | 71.02 | 70.84 | 79.16 | 68.47 | 78.48 | 80.39 | 89.57 |
Kappa/% | - | | 80.33 | 77.14 | 83.42 | 78.32 | 87.72 | 85.63 | 93.42 |
Table 6.
Comparison of classification accuracies (in percent) provided by seven methods for Kennedy Space Center (part B).
Table 6.
Comparison of classification accuracies (in percent) provided by seven methods for Kennedy Space Center (part B).
Ground | Sum | Training | LDM-FL | CF-SVM | CF-LDM | DTRF-SVM | DTRF-LDM | CFDTRFF-SVM | CFDTRF-LDM |
---|
Scrub | 761 | 30 | 93.73 | 98.42 | 94.80 | 99.41 | 99.01 | 99.76 | 98.08 |
Swamp willow | 243 | 10 | 74.77 | 83.89 | 87.59 | 65.14 | 85.82 | 98.07 | 82.52 |
Cabbage palm hammock | 256 | 10 | 84.72 | 86.24 | 88.07 | 82.80 | 100.00 | 96.34 | 98.80 |
Cabbage palm/oak | 252 | 10 | 78.55 | 66.98 | 75.04 | 63.97 | 93.04 | 89.69 | 94.77 |
Slash pine | 161 | 6 | 72.59 | 31.36 | 63.05 | 65.57 | 86.10 | 91.74 | 84.26 |
Oak/broadleaf hammock | 229 | 9 | 96.68 | 70.39 | 64.02 | 94.79 | 100.00 | 75.75 | 99.09 |
Hardwood swamp | 105 | 4 | 100.00 | 48.10 | 68.19 | 97.54 | 100.00 | 46.56 | 100.00 |
Graminoid marsh | 431 | 17 | 90.59 | 90.15 | 90.89 | 98.25 | 93.98 | 97.06 | 97.41 |
Spartina marsh | 520 | 21 | 100.00 | 99.50 | 96.84 | 100.00 | 100.00 | 100.00 | 100.00 |
Cattail marsh | 404 | 16 | 70.08 | 96.72 | 96.79 | 89.76 | 75.69 | 98.08 | 100.00 |
Salt marsh | 419 | 17 | 100.00 | 97.06 | 95.95 | 96.87 | 99.94 | 96.76 | 99.75 |
Muld flats | 503 | 20 | 83.57 | 91.74 | 91.50 | 95.26 | 92.08 | 98.38 | 94.99 |
Water | 927 | 37 | 98.57 | 100.00 | 100.00 | 99.89 | 99.89 | 99.78 | 100.00 |
OA/% | - | | 90.17 | 90.07 | 91.02 | 92.62 | 95.24 | 95.89 | 97.33 |
AA/% | - | | 87.99 | 81.58 | 85.60 | 88.40 | 94.27 | 91.38 | 96.13 |
Kappa/% | - | | 89.05 | 88.92 | 89.99 | 91.77 | 94.70 | 95.42 | 97.03 |