.png){kind=icon}
Return
Transformer-Based Multi-Scale Feature Remote Sensing Image Classification Model
.png){kind=icon}
OA
AI
DOI：10.1109/ACCESS.2024.3520253.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
To address the challenge of poor classification accuracy due to complex backgrounds, large intra-scale variations, and high inter-scale similarity in remote sensing scene classification (RSSC), we propose a new remote sensing scene classification model called multi-scale dual-branch classification network (MDBC-Net). The model is composed of a Trans-branch and CNN-branch in parallel, which can fully utilize the local attention of the CNN-branch structure and the global attention mechanism of the Trans-branch structure, thereby improving the model's ability to focus on features of different scales. Due to the complexity of backgrounds in RSSC, we require features at different scales to obtain richer scene information. Thus we design a down-sampling module in the model to obtain multi-scale features. Finally, we adopt the polynomial form of cross entropy for the trained loss function to improve the generalization performance and robustness of the model. Experiments have shown that the model achieves advanced performance on three datasets: NWPU-RESISC45, AID, and UC Served.
Keywords:
Feature extraction
Transformers
Remote sensing
Biological system modeling
Merging
Scene classification
Convolutional neural networks
Sun
Computational modeling
Training
scene classification
multi-scale features
local attention
global attention
polynomial form
Journal
IF:
3.6
Papers:
9.8W
Citations:
29.4W
Organization
Cited Papers
Matrix Effects in the Detection of Pb and Ba in Soils Using Laser-Induced Breakdown Spectroscopy
PREAI
ShareSaveShareSaveEnhancing Multiscale Representations With Transformer for Remote Sensing Image Semantic Segmentation
PREAI
ShareSave
FSwin Transformer: Feature-Space Window Attention Vision Transformer for Image Classification.png)
IEEE ACCESSOAAI
IEEE ACCESSIF3.6
OAAIShareSave
View more