A Knowledge-Guided refinement framework for Late-Season Rice–Weed segmentation using UAV imagery under canopy closure
• A knowledge-guided refinement framework is proposed for late-season rice–weed mapping under canopy closure. • The WeedGraphNet outperformed appearance-only U-Net baselines on late-season UAV imagery • High segmentation accuracy is achieved on held-out tiles, with an overall accuracy of 0.974 and a mean IoU of 0.883. • Strong class-wise performance is obtained, with IoUs of 0.970 (rice), 0.820 (weed), and weed F1-score 0.901.
Authors
- Muhammad Nasar Ahmad (ORCID: https://orcid.org/0000-0002-4468-8105)
- Xinyu Miao
- Xiaohui Yuan
- Lichuan Gu
- Shuai Yang
Institutions
- University of North Texas (US)
- Anhui Agricultural University (CN)
- Anhui University (CN)
Publication Details
- Journal
- International Journal of Applied Earth Observation and Geoinformation
- Published
- 2026-05-29
- DOI
- https://doi.org/10.1016/j.jag.2026.105361
- Primary Topic
- Smart Agriculture and AI
- Type
- article
- Field-Weighted Citation Impact
- 0.00
Funders
- National Natural Science Foundation of China