Researchers at the University of Aberdeen have developed SAGRNet, an advanced AI model designed to significantly improve land cover mapping accuracy, especially for vegetation. This innovative model employs deep learning to analyze entire landscape objects, enhancing efficiency and precision compared to traditional pixel-by-pixel methods. The study detailing SAGRNet's development was published in the ISPRS Journal of Photogrammetry and Remote Sensing.
SAGRNet was trained using satellite images of diverse landscapes in north-east Scotland, and further tested across five globally distributed urban fringe areas. These areas, including Guangzhou, Durban, Sydney, New York City, and Porto Alegre, were selected to represent diverse ecological backgrounds. This approach ensures the model's robustness and transferability across various environments.
The model's open availability allows decision-makers to quickly assess the impact of events like floods and droughts on large areas of land. SAGRNet can also monitor crop growth, aiding in harvest predictions and sustainable land use decisions. This technology aligns with ongoing advancements in AI and remote sensing, as highlighted by the upcoming IEEE IGARSS 2025 symposium and Esri's recent land cover map update.
SAGRNet's ability to rapidly and accurately assess landscape changes is crucial for understanding climate change impacts. The model's versatility makes it suitable for large-scale applications such as land resource surveys and ecological monitoring. The development of SAGRNet represents a significant step forward in environmental monitoring and sustainable land management, offering valuable tools for researchers and policymakers globally.