Collaborative Research: III: Toward Robust and Private Cross-view Geo-localization

Cross-view geo-localization seeks to determine where a photograph was taken by connecting information from different viewpoints, such as ground-level photographs, aerial imagery, and satellite images. As geospatial data become increasingly available from vehicles, mobile devices, aerial platforms, and satellites, the ability to accurately understand locations from multiple viewpoints is becoming increasingly important. However, current artificial intelligence systems often struggle to generalize across geographic regions, adapt to changing environmental conditions, explain their decisions, and protect sensitive information. These limitations restrict the broader use of geospatial artificial intelligence in transportation, mapping, environmental monitoring, emergency response, and public safety. This project addresses these challenges by advancing the scientific foundations of cross-view geospatial understanding through scalable, trustworthy, and privacy-preserving methods. The project will produce openly available datasets, models, and software resources while providing research and training opportunities for graduate and undergraduate students. The resulting advances are expected to accelerate innovation in geospatial artificial intelligence and strengthen the nation's capabilities in managing and understanding large-scale geospatial information. To meet these goals, this project develops new datasets, models, and algorithms for cross-view geo-localization using multimodal geospatial data. Research activities include creating large-scale datasets that combine ground-level imagery, aerial imagery, satellite imagery, geographic information system data, segmentation annotations, and natural language descriptions collected from diverse regions worldwide. The project develops foundation models that learn shared representations across multiple views and modalities, investigates reasoning-based approaches that provide interpretable explanations for localization decisions, and designs privacy-preserving techniques that protect sensitive geospatial information during model training and deployment. The project also establishes benchmarks for evaluating scalability, reliability, interpretability, and privacy across diverse geographic environments. Together, these activities advance the scientific foundations of geospatial artificial intelligence and enable more robust and trustworthy cross-view geo-localization systems. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.