Acta Photonica Sinica, Volume. 54, Issue 4, 0410002(2025)
Hyperspectral Image Classification Method Based on Dynamic Graph-spectral Feature Extraction
Figures & Tables(22)
Fig. 6. The structure of dynamic graph construction and feature extraction network
Fig. 7. The structure of region-global spectral feature extraction network
Fig. 11. University of Pavia data set category and ground-truth map
Fig. 13. Classification map of different methods on the Indian Pines
Fig. 14. Classification map of different methods on the University of Pavia
Table 1. Classification result of ablation experiments on the Indian Pines
View tableView in ArticleTable 1. Classification result of ablation experiments on the Indian Pines
Class DGCFN/% RGSFN/% DGSFEC/% 1 98.54±1.95 96.09±7.25 99.51±0.97 2 93.80±0.50 95.26±1.21 94.79±0.54 3 99.22±0.45 99.15±0.68 98.96±0.69 4 98.49±0.96 97.41±4.38 98.59±1.95 5 96.80±0.59 97.91±1.33 98.14±1.03 6 99.45±0.27 98.87±0.80 99.42±0.60 7 88.00±9.30 99.60±1.20 100.00±0 8 99.95±0.09 99.69±0.76 99.98±0.06 9 72.22±16.10 84.44±18.22 83.89±10.95 10 96.71±0.90 98.22±0.63 98.83±0.76 11 98.39±0.34 99.39±0.30 99.58±0.25 12 95.97±0.63 97.82±1.35 95.34±1.08 13 98.70±0.77 96.86±0.99 99.30±0.68 14 99.58±0.29 99.97±0.04 99.95±0.07 15 97.35±1.11 96.85±4.91 97.52±2.42 16 83.33±5.02 80.83±13.35 92.62±4.21 OA 97.49±0.14 98.17±0.77 98.33±0.21 AA 94.78±1.11 96.15±2.61 97.28±0.65 Kappa 97.14±0.17 97.91±0.88 98.10±0.24
Table 2. Classification result of ablation experiments on the University of Pavia
View tableView in ArticleTable 2. Classification result of ablation experiments on the University of Pavia
Class DGCFN/% RGSFN/% DGSFEC/% 1 87.11±1.77 86.74±19.93 95.45±2.67 2 98.98±0.14 99.32±0.60 99.83±0.15 3 92.70±1.77 71.32±24.46 84.54±7.32 4 70.08±2.44 83.16±14.51 95.04±5.05 5 92.49±1.61 88.99±6.41 94.69±2.06 6 99.39±0.22 86.42±29.89 99.35±0.82 7 93.62±0.49 65.41±17.90 92.15±7.59 8 87.20±1.65 83.07±27.27 89.89±1.17 9 76.50±3.67 79.89±11.13 85.35±10.18 OA 93.14±0.33 90.12±11.76 96.26±0.79 AA 88.65±0.52 82.70±16.01 92.64±1.76 Kappa 90.90±0.44 85.56±19.45 95.05±1.06
Table 3. Classification result of ablation experiments on the Salinas
View tableView in ArticleTable 3. Classification result of ablation experiments on the Salinas
Class DGCFN / %RGSFN / %DGSFEC/% 1 99.75±0.46 98.95±1.55 99.80±0.27 2 99.99±0.03 98.87±3.35 100.00±0 3 99.50±0.97 87.08±29.32 100.00±0 4 97.79±1.38 96.86±4.68 97.39±0.31 5 97.60±1.43 98.11±1.77 98.49±0.62 6 99.59±0.36 99.59±0.48 100.00±0 7 99.28±0.32 99.40±0.49 100.00±0 8 99.41±0.65 91.83±15.33 99.20±0.99 9 99.57±1.29 99.93±0.12 100.00±0 10 99.51±0.79 98.27±2.54 99.91±0.14 11 99.02±2.04 92.98±11.83 99.05±0.41 12 99.26±1.11 98.76±3.41 98.95±1.01 13 92.22±5.68 95.49±2.99 97.68±4.96 14 98.58±1.56 95.82±9.52 99.91±0.29 15 98.41±2.03 90.25±15.23 98.26±0.35 16 99.86±0.21 99.98±0.03 100.00±0 OA 99.09±0.46 95.71±4.86 99.34±0.36 AA 98.71±0.68 96.38±4.02 99.29±0.61 Kappa 98.99±0.51 95.22±5.39 99.27±0.41
Table 4. Classification result of different methods on the Indian Pines
View tableView in ArticleTable 4. Classification result of different methods on the Indian Pines
Class 3DCNN/% HybridSN/% MiniGCN/% MDGCN/% AMGCFN/% Hybrid Former/% Spectral Former/% SSBC/% DGSFEC/% 1 75.99 96.10 71.66 93.75 68.23 97.21 74.92 90.91 99.51 2 88.10 94.79 84.92 92.63 89.44 97.31 84.05 93.00 94.79 3 78.34 98.86 75.98 93.12 92.34 98.24 91.84 93.66 98.96 4 78.51 93.99 76.51 96.14 91.46 96.98 93.73 100.00 98.59 5 89.16 97.68 88.88 96.03 90.37 87.84 86.51 99.13 98.14 6 97.10 98.69 90.26 97.43 94.59 99.19 94.98 99.28 99.42 7 68.38 100.00 85.44 69.23 81.63 98.42 87.69 100.00 100.00 8 96.93 99.95 95.10 97.99 99.26 99.65 68.73 100.00 99.98 9 64.36 83.89 79.00 100.00 94.73 85.71 65.07 89.47 83.89 10 81.13 98.80 83.30 84.39 92.19 96.04 99.38 93.28 98.83 11 87.56 99.32 86.17 94.93 96.11 98.32 93.48 98.20 99.58 12 80.52 93.48 87.64 90.05 84.23 96.91 84.24 95.91 95.34 13 98.52 99.24 84.75 100.00 96.08 99.39 100.00 97.95 99.30 14 96.42 99.70 96.28 99.35 97.60 99.70 76.92 98.42 99.95 15 73.42 98.56 80.56 98.88 94.97 98.67 100.00 99.73 97.52 16 89.01 93.45 93.60 98.41 86.34 95.52 100.00 98.86 92.62 OA 86.60 97.97 86.48 94.34 93.36 98.12 80.97 96.80 98.33 AA 81.08 96.65 85.06 93.90 90.60 96.79 87.60 96.74 97.28 Kappa 84.70 97.68 84.56 93.51 92.44 97.86 78.45 96.35 98.10
Table 5. Classification result of different methods on the University of Pavia
View tableView in ArticleTable 5. Classification result of different methods on the University of Pavia
Class 3DCNN/% HybridSN/% MiniGCN/% MDGCN/% AMGCFN/% Hybrid Former/% Spectral Former/% SSBC/% DGSFEC/% 1 89.59 95.27 96.99 59.37 95.04 89.70 85.39 89.00 95.45 2 91.97 99.29 94.50 78.61 98.96 96.41 89.64 99.17 99.83 3 75.88 91.77 84.27 77.48 95.28 78.34 81.10 85.37 84.54 4 94.55 85.38 85.89 73.47 83.71 97.41 94.16 93.34 95.04 5 96.91 97.35 99.73 95.67 41.84 99.54 99.37 97.82 94.69 6 84.12 95.17 92.04 80.76 96.62 90.66 76.85 92.33 99.35 7 82.69 97.96 74.61 78.92 87.44 94.62 87.05 84.43 92.15 8 84.42 87.54 69.78 43.78 98.07 77.19 90.52 85.90 89.89 9 91.55 77.74 94.70 67.94 93.48 97.79 99.87 95.84 85.35 OA 87.51 95.13 90.21 72.77 94.45 92.09 87.94 93.98 96.26 AA 85.51 91.62 88.06 72.89 87.83 90.84 89.33 91.47 92.64 Kappa 83.63 93.53 86.77 65.36 92.65 89.57 83.88 92.00 95.05
Table 6. Classification result of different methods on the Salinas
View tableView in ArticleTable 6. Classification result of different methods on the Salinas
Class 3DCNN/% HybridSN/% MiniGCN/% MDGCN/% AMGCFN/% HybridFormer/% SpectralFormer/% SSBC/% DGSFEC/% 1 96.73 99.97 99.80 88.83 99.79 99.31 99.34 98.04 99.80 2 99.05 100.00 99.79 95.83 97.66 99.58 86.60 99.81 100.00 3 97.47 99.97 99.33 92.70 99.84 99.29 43.14 97.65 100.00 4 98.58 96.38 97.60 94.35 94.26 98.50 99.71 99.49 97.39 5 98.66 98.12 88.66 63.67 81.78 99.31 97.31 97.96 98.49 6 98.26 99.68 99.95 93.08 92.69 100.00 99.36 99.87 100.00 7 98.47 99.96 99.21 80.73 99.20 99.92 98.92 98.31 100.00 8 86.77 98.80 78.08 69.21 99.32 93.84 67.57 90.12 99.20 9 98.99 99.99 99.49 88.74 99.44 99.54 97.49 99.12 100.00 10 93.15 99.16 88.29 88.98 98.70 97.32 80.48 95.69 99.91 11 94.30 96.77 82.50 87.28 86.52 97.93 78.21 94.80 99.05 12 97.73 97.86 96.67 92.30 96.32 99.77 99.10 99.74 98.95 13 97.30 98.53 93.61 77.77 43.59 99.79 97.34 95.70 97.68 14 96.89 99.11 99.51 80.48 91.31 98.81 90.93 99.81 99.91 15 78.33 96.32 71.57 81.86 97.88 90.04 50.06 97.89 98.26 16 76.45 99.94 99.29 96.45 99.10 99.84 95.02 98.38 100.00 OA 91.48 98.81 89.54 83.08 96.07 96.99 80.93 96.65 99.34 AA 93.67 98.79 93.33 85.77 92.34 98.52 86.29 97.65 99.29 Kappa 90.53 98.67 88.36 81.30 95.62 96.65 78.81 96.27 99.27
Table 7. Running time of different algorithms
View tableView in ArticleTable 7. Running time of different algorithms
Data set Method 3DCNN/s HybridSN/s MiniGCN/s MDGCN/s AMGCFN/s Hybrid Former/s Spectral Former/s SSBC/s DGSFEC/s Indian Pines Training time 109.48 27.49 131.53 269.88 26.42 312.56 224.59 100.56 125.19 Test time 0.99 0.55 1.32 1.11 0.71 4.72 2.20 2.96 4.96 Total time 110.47 28.04 132.85 270.99 27.13 317.28 226.79 103.52 130.15 University of Pavia Training time 69.84 72.27 579 1205.24 185.28 84.37 464.45 161.01 92.98 Test time 5.89 2.63 45.29 2.94 7.55 25.38 3.32 22.21 6.12 Total time 75.73 74.9 624.29 1208.18 192.83 109.75 467.77 183.22 99.10 Salinas Training time 159.85 20.64 300.32 242.96 88.26 171.51 705.88 566.59 116.61 Test time 5.48 3.09 47.5 1.39 7.6 22.97 10.08 14.46 7.67 Total time 165.33 23.73 347.82 244.35 95.86 194.48 715.96 581.05 124.28
Tools
Get Citation
Copy Citation Text
Chenjie XU, Dan LI, Fanqiang KONG. Hyperspectral Image Classification Method Based on Dynamic Graph-spectral Feature Extraction[J]. Acta Photonica Sinica, 2025, 54(4): 0410002
Paper Information
Category:
Received: Oct. 16, 2024
Accepted: Dec. 16, 2024
Published Online: May. 15, 2025
The Author Email: Dan LI (danli@nuaa.edu.cn)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20