Laser & Optoelectronics Progress, Volume. 60, Issue 12, 1228002(2023)
Hyperspectral Remote-Sensing Classification Combining Transformer and Multiscale Residual Mechanisms
Fig. 1. Network structure of SMSaNet
Fig. 2. Multiscale spectral enhancement residual fusion module
Fig. 3. Spectral attention module
Fig. 4. Swin Transformer feature extraction module
Fig. 5. Swin Transformer block
Fig. 6. MSA and W-MSA
Fig. 7. W-MSA and SW-MSA
Fig. 8. Classification result chart on India dataset
Fig. 9. Classification result chart on PU dataset
Fig. 10. Class activation mapping (CAM). (a) CAM of multiscale spectral enhanced residual fusion module; (b) CAM of spectral attention module
Fig. 11. OA values corresponding to different ratios of training samples. (a) Inida dataset; (b) PU dataset
|
|
|
|
|
|
|
|
Get Citation
Copy Citation Text
Yuhan Chen, Bo Wang, Qingyun Yan, Bingjie Huang, Tong Jia, Bin Xue. Hyperspectral Remote-Sensing Classification Combining Transformer and Multiscale Residual Mechanisms[J]. Laser & Optoelectronics Progress, 2023, 60(12): 1228002
Category: Remote Sensing and Sensors
Received: Mar. 8, 2022
Accepted: Jun. 13, 2022
Published Online: Jun. 1, 2023
The Author Email: Wang Bo (wangbo@nuist.edu.cn)