Chen et al. (2026) Extracting Summer-Harvested Crops in the Baojixia Irrigation District Using CycleGAN and Transfer Learning

⚠️ Warning: This summary was generated from the abstract only, as the full text was not available.

Identification

• Journal: Remote Sensing
• Year: 2026
• Date: 2026-09-14
• Authors: Zili Chen, Zhilong Gao, Zefeng Jia, Pengjie Pan, Wen Gao, Jun Zhang, Zijie Niu, Dongyan Zhang
• DOI: 10.3390/rs18183155

Research Groups

Not specified

Short Summary

The study proposes a cross-scale collaborative extraction framework using CycleGAN and transfer learning to map summer harvest crops in the Baojixia Irrigation District, effectively bridging the resolution gap between UAV and satellite imagery.

Objective

• To develop a cost-effective and high-accuracy method for mapping fragmented crop planting structures (winter wheat and rapeseed) by transferring high-resolution spatial knowledge from UAV imagery to satellite imagery.

Study Configuration

• Spatial Scale: Baojixia Irrigation District
• Temporal Scale: 2023 (Summer harvest period)

Methodology and Data

• Models used: CycleGAN (for style translation), U-Net (selected as the optimal segmentation backbone), DeepLabv3+, SegFormer, and HRNet (evaluated for comparison).
• Data sources: UAV imagery and satellite imagery.

Main Results

• The proposed framework significantly outperformed the baseline U-Net model, achieving the following metrics:
  • mIoU: 85.09% (increase of 2.98%)
  • mPA (Recall): 91.63% (increase of 2.00%)
  • Precision: 91.97% (increase of 1.66%)
  • Accuracy: 93.57% (increase of 1.52%)
  • F1-Score: 91.80% (increase of 1.83%)
• The use of CycleGAN for UAV-to-satellite style translation accelerated model convergence and reduced the need for manual annotation.

Contributions

• Provides a solution to the trade-off between the high spatial resolution of UAVs and the wide coverage of satellites.
• Demonstrates a successful method for transferring high-resolution prior knowledge into the satellite feature space to improve crop identification in complex, fragmented agricultural landscapes.

Funding

Not specified

Citation

@article{Chen2026Extracting,
  author = {Chen, Zili and Gao, Zhilong and Jia, Zefeng and Pan, Pengjie and Gao, Wen and Zhang, Jun and Niu, Zijie and Zhang, Dongyan},
  title = {Extracting Summer-Harvested Crops in the Baojixia Irrigation District Using CycleGAN and Transfer Learning},
  journal = {Remote Sensing},
  year = {2026},
  doi = {10.3390/rs18183155},
  url = {https://doi.org/10.3390/rs18183155}
}