Semantic Segmentation of Rice Fields in Sub-Meter Satellite Imagery Using an HRNet-CA-Enhanced DeepLabV3+ Framework
Remote Sensing · 11.07.2025 · 17 (14) · 2404
DOI: 10.3390/rs17142404 ↗Kurz gefasst
Fachartikel in der Zeitschrift Remote Sensing, online erschienen am 11.07.2025. Verfasst von Yifan Shao und 6 weiteren Autor:innen. Die Arbeit wurde bisher 7-mal zitiert (Crossref) und nennt 39 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Special Funding for Science and Technology of Sanya Yazhou Bay Science and Technology City, National Key R&D Program, Natural Science Foundation of Hainan Province.
Abstract (Originalsprache)
Accurate monitoring of rice-planting areas underpins food security and evidence-based farm management. Recent work has advanced along three complementary lines—multi-source data fusion (to mitigate cloud and spectral confusion), temporal feature extraction (to exploit phenology), and deep-network architecture optimization. However, even the best fusion- and time-series-based approaches still struggle to preserve fine spatial details in sub-meter scenes. Targeting this gap, we propose an HRNet-CA-enhanced DeepLabV3+ that retains the original model’s strengths while resolving its two key weaknesses: (i) detail loss caused by repeated down-sampling and feature-pyramid compression and (ii) boundary blurring due to insufficient multi-scale information fusion. The Xception backbone is replaced with a High-Resolution Network (HRNet) to maintain full-resolution feature streams through multi-resolution parallel convolutions and cross-scale interactions. A coordinate attention (CA) block is embedded in the decoder to strengthen spatially explicit context and sharpen class boundaries. The rice dataset consisted of 23,295 images (11,295 rice + 12,000 non-rice) via preprocessing and manual labeling and benchmarked the proposed model against classical segmentation networks. Our approach boosts boundary segmentation accuracy to 92.28% MIOU and raises texture-level discrimination to 95.93% F1, without extra inference latency. Although this study focuses on architecture optimization, the HRNet-CA backbone is readily compatible with future multi-source fusion and time-series modules, offering a unified path toward operational paddy mapping in fragmented sub-meter landscapes.
Förderung
- Special Funding for Science and Technology of Sanya Yazhou Bay Science and Technology City
- National Key R&D Program
- Natural Science Foundation of Hainan Province
- Special Funding for Southern Propagation of the National Institute of Southern Propagation Research, Chinese Academy of Agricultural Sciences, Sanya
- Special Funding for Basic Scientific Research Operations of Central-level Public Welfare Research Institutes
- Science and Technology Innovation Project of the Chinese Academy of Agricultural Sciences
- Hainan Provincial Graduate Student Innovative Research Project
Zitieren
APA
Shao, Y., Pan, P., Zhao, H., Li, J., Yu, G., Zhou, G., & Zhang, J. (2025). Semantic Segmentation of Rice Fields in Sub-Meter Satellite Imagery Using an HRNet-CA-Enhanced DeepLabV3+ Framework. Remote Sensing, 17(14), 2404. https://doi.org/10.3390/rs17142404
Vancouver
Shao Y, Pan P, Zhao H, Li J, Yu G, Zhou G, et al. Semantic Segmentation of Rice Fields in Sub-Meter Satellite Imagery Using an HRNet-CA-Enhanced DeepLabV3+ Framework. Remote Sensing. 2025;17(14):2404. doi:10.3390/rs17142404
BibTeX
@article{Shao2025,
author = {Shao, Yifan and Pan, Pan and Zhao, Hongxin and Li, Jiale and Yu, Guoping and Zhou, Guomin and Zhang, Jianhua},
title = {Semantic Segmentation of Rice Fields in Sub-Meter Satellite Imagery Using an HRNet-CA-Enhanced DeepLabV3+ Framework},
journal = {Remote Sensing},
year = {2025},
volume = {17},
number = {14},
pages = {2404},
doi = {10.3390/rs17142404}
}
EUVIMED