GeoAI research

Our team is dedicated to frontier research in Geospatial Intelligence (Geo-AI), building a comprehensive technical framework centered on spatiotemporal foundation models. By integrating multi-source heterogeneous data—including point clouds, imagery, and text—we have overcome critical bottlenecks in cross-modal alignment and spatiotemporal big data mining. We focus on developing AI-driven perception algorithms and structural modeling techniques, achieving a seamless loop from low-level data processing to high-level semantic understanding, providing a robust digital foundation for modeling complex geographic environments.

Focusing on core scenarios in precision forestry and smart power grids, we have developed a series of high-efficiency perception solutions covering individual tree segmentation, carbon accounting, and risk monitoring for power facilities. Building upon this, we further explore the evolution of Geospatial Agents. These agents leverage natural language interaction to enable autonomous feature extraction, dynamic change monitoring, and the automated construction of spatial analysis models. This transition from “automated processing” to “decision-making intelligence” significantly enhances the efficiency and intelligent reasoning capabilities of geospatial information systems.