Remote Sensing Technology and Application, Volume. 40, Issue 1, 25(2025)
Multi-scale Scene Classification and Non-agricultural Application of Cultivated Land High-resolution Image
Figures & Tables(19)
Fig. 3. Multi-scale scene classification and non-agricultural detection method of cultivated land high-resolution images
Fig. 7. Results of cultivated land multi-scale scene classification
Fig. 8. Display of misclassification of cultivated land in small-scale scene classification
Fig. 9. Comparison of non-agricultural detection results by different methods
Fig. 10. Comparison of cultivated land classification effect by different methods
Fig. 11. Comparison of effects before and after boundary constraints
Table 1.
Sample information statistics View tableView in ArticleTable 1.
Sample information statistics 样本尺度 分辨率/m 耕地样本数量 非耕地样本数量 训练样本数量 验证样本数量 128×128 0.8 360 545 724 181 200×200 2 200 2 200 1 920 480 256×256 0.8 360 550 728 182 512×512 0.8 360 545 724 181
Table 2.
Binary mapping of classification results at different scales View tableView in ArticleTable 2.
Binary mapping of classification results at different scales 尺度 耕地（P） 非耕地（N） L 2 1 M 2 1 S 2 -2
Table 3.
Eigenvalue meaning and mapping relationship View tableView in ArticleTable 3.
Eigenvalue meaning and mapping relationship 特征值 特征值含义 映射值 0 LN+MN+SN 0 1 LN+MP+SN或LP+MN+SN 0 2 LP+MP+SN 0 4 LN+MN+SP 0 5 LP+MN+SP或LN+MP+SP 1 6 LP+MP+SP 1
Table 4.
Comparison of recognition accuracy of cultivated land scene classification model View tableView in ArticleTable 4.
Comparison of recognition accuracy of cultivated land scene classification model Model A P R F1 Minute ViT 0.993 2 0.984 4 0.988 3 0.986 3 34 Swin Transformer 0.995 1 0.988 3 0.992 2 0.990 3 66 Swin Transformer v2 0.993 2 0.980 7 0.992 2 0.986 4 104 ResNet101 0.991 2 0.966 0 1.000 0 0.982 7 43 Inception_v3 0.990 2 0.969 5 0.992 2 0.972 9 32 MobileNet_v2 0.991 2 0.976 8 0.988 3 0.982 5 28 MobileNet_v3 0.989 3 0.973 0 0.984 4 0.978 6 23 Xception 0.994 1 0.992 1 0.984 4 0.988 2 34 Xception_CBAM 0.996 1 0.984 6 1.000 0 0.992 2 36
Table 5.
Comparison of recall rates of non-agricultural detection methods View tableView in ArticleTable 5.
Comparison of recall rates of non-agricultural detection methods 非农化检测方法 检测正确的图斑数 参考图斑数 查全率 PSPNet 288 489 0.588 9 Unet 294 489 0.601 2 DeepLabV3+ 298 489 0.609 4 MSC-Xception 333 489 0.681 0
Table 6.
Comparison of classification accuracy of different algorithms View tableView in ArticleTable 6.
Comparison of classification accuracy of different algorithms 算法 A P R F1 MSC-Xception 0.887 0 0.958 6 0.809 0 0.877 4 DeeplabV3+ 0.824 5 0.987 0 0.657 7 0.789 4 Unet 0.819 3 0.991 1 0.644 4 0.781 0 PSPNet 0.813 4 0.982 0 0.638 5 0.773 8 OBIA-RF 0.802 5 0.810 0 0.790 5 0.800 1 OBIA-SVM 0.778 5 0.830 4 0.725 1 0.774 2
Table 7.
Experimental accuracy of different model transfer View tableView in ArticleTable 7.
Experimental accuracy of different model transfer 影像 算法 A P R F1 一 MSC-Xception 0.928 8 0.983 0 0.885 4 0.931 6 DeeplabV3+ 0.833 8 0.976 6 0.713 8 0.824 8 PSPNet 0.681 4 0.936 4 0.448 6 0.606 6 Unet 0.806 3 0.981 2 0.659 0 0.788 4 二 MSC-Xception 0.910 4 0.933 3 0.903 2 0.918 0 DeeplabV3+ 0.833 1 0.915 2 0.770 9 0.836 9 PSPNet 0.651 1 0.885 4 0.426 8 0.576 0 Unet 0.952 1 0.477 7 0.636 2
Tools
Get Citation
Copy Citation Text
Wei CHEN, Hao LI, Qihua ZHANG, Yanlan HE, Shengli WANG. Multi-scale Scene Classification and Non-agricultural Application of Cultivated Land High-resolution Image[J]. Remote Sensing Technology and Application, 2025, 40(1): 25
Paper Information
Category:
Received: Oct. 26, 2023
Accepted: --
Published Online: May. 22, 2025
The Author Email: Hao LI (lihao@hhu.edu.cn)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20