Optics and Precision Engineering, Volume. 32, Issue 9, 1395(2024)
Active learning-clustering-group convolutions network for hyperspectral images classification
Figures & Tables(15)
Fig. 5. Ground truth map of IP dataset and classification maps of each classification method on IP dataset
Fig. 6. Ground truth map of BW dataset and the classification maps of each classification method on BW dataset
Fig. 7. Ground truth map of HT dataset and the classification maps of each classification method on HT dataset
Table 1. Classification results on IP dataset
View tableView in ArticleTable 1. Classification results on IP dataset
Method OA±SD/% Kappa Time/s ClusterCNN 95.84±0.43 0.952 5 186.18 SSRN 93.07±0.92 0.920 5 57.50 HybridSN 94.90±0.52 0.941 7 257.70 FC-3DCNN 70.13±1.23 0.655 4 65.00 3D-DenseNet 85.04±1.76 0.828 4 336.10 CGNet 98.28±0.17 0.980 3 172.66
Table 2. Classification results on BW dataset
View tableView in ArticleTable 2. Classification results on BW dataset
Method OA±SD/% Kappa Time/s ClusterCNN 97.12±0.97 0.968 7 56.32 SSRN 96.44±0.81 0.961 4 17.80 HybridSN 94.32±1.67 0.938 4 77.30 FC-3DCNN 94.15±1.91 0.936 6 21.11 3D-DenseNet 92.25±1.38 0.916 0 111 CGNet 98.20±0.68 0.980 5 54.34
Table 3. Classification results on HT dataset
View tableView in ArticleTable 3. Classification results on HT dataset
Method OA±SD/% Kappa Time/s ClusterCNN 96.33±0.60 0.960 3 260.50 SSRN 96.99±0.33 0.967 5 91.40 HybridSN 95.99±1.11 0.956 7 380.2 FC-3DCNN 92.51±0.84 0.919 0 90.82 3D-DenseNet 93.62±0.97 0.931 0 486.9 CGNet 96.91±0.28 0.966 5 255.89
Table 4. Number of trainable parameters and FLOPs on IP， BW， and HT datasets
View tableView in ArticleTable 4. Number of trainable parameters and FLOPs on IP， BW， and HT datasets
Method Parameters（IP/BW/HT） FLOPs（IP/BW/HT） ClusterCNN 201，488/201，358/201，423 26，196，864/26，196，608/26，196，736 SSRN 63，520/63，470/63，495 1，754，920/1，754，824/1，754，872 HybridSN 4，845，696/4，845，438/4，845，567 51，346，384/51，345，872/51，346，128 FC-3DCNN 995，456/995，198/995，327 4，220，000/4，219，488/4，219，744 3D-DenseNet 2，549，584/2，548，878/2，549，231 9，690，464/9，689，056/9，689，760 CGNet 58，176/57，982/58，111 5，524，480/5，508，736/5，524，352
Table 5. Classification results of each method on IP dataset
View tableView in ArticleTable 5. Classification results of each method on IP dataset
Method OA±SD/% Kappa Time/s Training data ClusterCNN 95.84±0.43 0.952 5 186.18 6% CGNet 98.28±0.17 0.980 3 172.66 6% AL-ClusterCNN 98.51±0.35 0.982 9 186.04 6% 97.12±0.43 0.967 1 148.31 5% AL-CGNet 99.57±0.12 0.995 0 176.84 6% 98.90±0.35 0.987 4 145.68 5%
Table 6. Classification results of each method on BW dataset
View tableView in ArticleTable 6. Classification results of each method on BW dataset
Method OA±SD/% Kappa Time/s Training data ClusterCNN 97.12±0.97 0.968 7 56.32 6% CGNet 98.20±0.68 0.980 5 54.34 6% AL-ClusterCNN 98.61±0.55 0.984 9 56.47 6% 97.68±0.60 0.974 9 48.14 5% AL-CGNet 99.23±0.47 0.991 7 58.16 6% 98.81±0.58 0.987 1 47.33 5%
Table 7. Classification results of each method on HT dataset
View tableView in ArticleTable 7. Classification results of each method on HT dataset
Method OA±SD/% Kappa Time/s Training data ClusterCNN 96.33±0.60 0.960 3 260.50 6% CGNet 96.91±0.28 0.966 5 255.89 6% AL-ClusterCNN 98.50±0.53 0.983 8 274.39 6% 97.81±0.58 0.976 3 223.69 5% AL-CGNet 98.82±0.16 0.987 2 256.77 6% 97.88±0.37 0.977 1 214.88 5%
Table 8. Classification results of AL-CGNet method on IP/BW/HT dataset using fewer training samples
View tableView in ArticleTable 8. Classification results of AL-CGNet method on IP/BW/HT dataset using fewer training samples
Index 4% 3% OA±SD/% Kappa Time/s OA±SD/% Kappa Time/s IP 98.48±0.35 0.982 7 110.63 96.42±0.82 0.959 1 81.93 BW 98.24±1.18 0.981 0 37.25 97.22±1.43 0.969 9 28.79 HT 96.92±0.50 0.966 7 170.58 95.81±0.63 0.954 7 121.31
Tools
Get Citation
Copy Citation Text
Jing LIU, Yinqiao LI, Yi LIU. Active learning-clustering-group convolutions network for hyperspectral images classification[J]. Optics and Precision Engineering, 2024, 32(9): 1395
Paper Information
Category:
Received: Nov. 2, 2023
Accepted: --
Published Online: Jun. 2, 2024
The Author Email: LIU Jing (zyhalj1975@163.com)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20