Laser & Optoelectronics Progress, Volume. 62, Issue 4, 0428006(2025)
Remote Sensing Scene Classification Method Based on Multi-Scale Graph Convolution Context Feature Aggregation
Figures & Tables(10)
Table 1. Experience results on AID dataset
View tableTable 1. Experience results on AID dataset
Type Method 20% Training ratio 50% Training ratio CNN-based GoogleNet［ 28 ］83.44 86.39 VGG16［ 6 ］86.59 89.64 ResNet50［ 7 ］92.57 95.96 ConvNeXt-T［ 22 ］94.43 96.57 ACNet［ 29 ］94.38 96.76 LSMNet［ 30 ］94.31 96.78 Transformer-based ViT-B［ 13 ］93.74 95.84 Swin-T［ 14 ］94.56 96.92 T2T-ViT-l2［ 31 ］94.39 96.29 PVT-V2-B0［ 32 ］93.52 96.27 EMTCAL［ 15 ］94.69 96.41 GCN-based DFAGCN［ 19 ］‒ 94.88 ViG-S［ 21 ］93.22 96.34 MCAGCN （proposed） 95.64 97.58
Table 2. Experience results on NWPU-RESISC45 dataset
View tableTable 2. Experience results on NWPU-RESISC45 dataset
Type Method 10% Training Ratio 20% Training Ratio CNN-based GoogleNet［ 28 ］76.19 78.48 VGG16［ 6 ］76.47 79.79 ResNet50［ 7 ］88.48 91.86 ConvNeXt-T［ 22 ］91.12 93.34 ACNet［ 29 ］91.09 92.42 LSMNet［ 30 ］90.80 93.16 Transformer-based ViT-B［ 13 ］90.05 92.61 Swin-T［ 14 ］90.84 93.18 T2T-ViT-l2［ 31 ］90.62 93.19 PVT-V2-B0［ 32 ］89.72 92.95 EMTCAL［ 15 ］91.63 93.65 GCN-based DFAGCN［ 19 ］‒ 89.29 ViG-S［ 21 ］91.61 93.52 MCAGCN （proposed） 92.24 94.37
Table 3. Results of ablation experiments for modules
View tableTable 3. Results of ablation experiments for modules
Model Module OA /% Parameter /Mbit FLOPs /Gbit Net 1 IFEM 96.57 28.77 4.46 Net 2 IFEM+DAM 96.75 31.50 4.63 Net 3 IFEM+MFAM 97.08 34.60 4.98 Net 4 IFEM+CIEM+DAM 97.14 37.79 7.41 Net 5 IFEM+CIEM+MFAM 97.38 40.89 7.76 Net 6 IFEM+CIEM+MFAM+LSLoss 97.58 40.89 7.76
Table 4. Results of ablation experiments for backbone
View tableTable 4. Results of ablation experiments for backbone
Model OA /% Parameters /Mbit FLOPs /Gbit ResNet18 95.51 11.19 1.82 MCAGCN （ResNet18） 96.43 16.31 3.29 Swin-T 96.92 28.26 4.37 MCAGCN （Swin-T） 97.35 39.82 7.67 ConvNeXt-T 96.57 28.77 4.46 MCAGCN （ConvNeXt-T） 97.58 40.89 7.76
Tools
Get Citation
Copy Citation Text
Baolan Chen, Huawang Li, Yinxiao Wang. Remote Sensing Scene Classification Method Based on Multi-Scale Graph Convolution Context Feature Aggregation[J]. Laser & Optoelectronics Progress, 2025, 62(4): 0428006
Paper Information
Category: Remote Sensing and Sensors
Received: Jun. 12, 2024
Accepted: Jul. 25, 2024
Published Online: Feb. 18, 2025
The Author Email:
CSTR:32186.14.LOP241466
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20