A UAV-satellite collaborative framework for multi-scale farmland area extraction and calculation to enable precision irrigation

Accurately calculating farmland area at multiple scales is key to realizing the precise allocation and dispatch of water resources for intelligent irrigation and is thereby of great importance to the scientific allocation of agricultural resources. Based on the previous research, this paper proposes a farmland area calculation method and its application strategy, leveraging the collaboration between Unmanned Aerial Vehicle (UAV) and satellite remote sensing images. Firstly, combining the spatial resolution of the image and the enhanced interactive segmentation results based on Segment Anything Model (SAM), the farmland area extraction and area calculation in the satellite image are realized; furthermore, according to the different requirements of irrigation accuracy for different crop types, the area classification extraction and calculation strategy of UAV and satellite cooperation was designed. Finally, the feasibility and accuracy of the proposed method are verified by using the self-built satellite image data set and UAV image data set. Experimental results show that the SAM-based enhanced interactive segmentation performs excellently in the farmland area extraction task from satellite remote sensing image (zero-shot setting), with the test set yielding a Dice Similarity Coefficient (DSC) of 94.08 ± 1.13% and an Intersection over Union (IoU) of 95.79 ± 1.12%. The farmland area calculation method based on satellite remote sensing images has a large calculation error (R_E > 10%) in small farmland area (4000–20,000 m2), and a good performance (R_E < 10%) in large farmland area (above 20,000 m2). The collaborative strategy proposed in this study effectively synergizes the extensive coverage of satellite with the high precision of UAV data: large-area assessments rely on satellite-based rapid estimation, while critical localized areas are supplemented with UAV-enabled refined measurements. Accuracy validation demonstrates that the computational errors of the proposed collaborative strategy remain below 4% and 10% for UAV and satellite imagery, respectively, meeting the precision requirements for agricultural engineering applications.