Scholarlink

<p>Agricultural parcels (APs) are fundamental spatial units for precision agriculture and effective agricultural management. However, large-scale AP mapping remains limited, particularly at national scales. To address this gap, this study proposes a national-scale AP mapping framework based on 10-m resolution Sentinel-2 imagery by enhancing the Hierarchical Boundary-Guided multi-task Network (HBGNet). Specifically, we developed HBGNet 2.0, an upgraded version of HBGNet, which introduces a novel self-adaptive weighting loss function, thereby balancing the learning of different tasks and improving AP delineation. Using HBGNet 2.0, we generated a 10-m resolution national-scale AP dataset for Ukraine in 2023. HBGNet 2.0 demonstrates the best overall performance among all compared models (ResUNet-a, BFINet, BsiNet, SEANet, and HBGNet), achieving the highest OA (92.86%), F1 (91.03%), and IoU (85.70%) metrics. Compared with three publicly available national AP datasets for Ukraine, our dataset exhibits better spatial detail and higher delineation accuracy. This high-precision AP dataset has the potential to serve as a critical resource for Ukrainian agricultural monitoring and for assessing conflict-related crop damage.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A 10-m National Agricultural Parcel Dataset of Ukraine Generated with an Enhanced Multi-Task Network and Sentinel-2 Imagery

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月10日

查看全文
  • Junbin Li,
  • Hongwei Zeng,
  • Hang Zhao,
  • Miao Zhang,
  • Yang Chen,
  • Qiangyi Yu,
  • Wenbin Wu,
  • Yuanwei Qin,
  • Bingfang Wu,
  • Shaohua Wang

摘要

Agricultural parcels (APs) are fundamental spatial units for precision agriculture and effective agricultural management. However, large-scale AP mapping remains limited, particularly at national scales. To address this gap, this study proposes a national-scale AP mapping framework based on 10-m resolution Sentinel-2 imagery by enhancing the Hierarchical Boundary-Guided multi-task Network (HBGNet). Specifically, we developed HBGNet 2.0, an upgraded version of HBGNet, which introduces a novel self-adaptive weighting loss function, thereby balancing the learning of different tasks and improving AP delineation. Using HBGNet 2.0, we generated a 10-m resolution national-scale AP dataset for Ukraine in 2023. HBGNet 2.0 demonstrates the best overall performance among all compared models (ResUNet-a, BFINet, BsiNet, SEANet, and HBGNet), achieving the highest OA (92.86%), F1 (91.03%), and IoU (85.70%) metrics. Compared with three publicly available national AP datasets for Ukraine, our dataset exhibits better spatial detail and higher delineation accuracy. This high-precision AP dataset has the potential to serve as a critical resource for Ukrainian agricultural monitoring and for assessing conflict-related crop damage.