Developing AI tools for Geospatial Intelligence
Developing AI tools for Geospatial Intelligence
Through advances in artificial intelligence, Oak Ridge National Laboratory is transforming geospatial analytic methods in support of information exploitation and dissemination into geospatial intelligence (GEOINT) foundational data layers. We apply GeoAI methods to address several national security applications including infrastructure mapping, change detection, damage assessment, socioeconomic analysis, population dynamics, geomatics, maritime safety, geospatial data privacy and trustworthiness of AI systems.
With core strengths in machine learning for high-performance computing, geoassurance, probabilistic reasoning, and spatio-temporal analytics, we are forming alliances and collaborating to support geospatial science and human security missions.
GeoAI Workshops and Events
Trillion Pixel Challenge
The annual Trillion Pixel Challenge workshop is an opportunity to foster collaboration, share insights and best practices, and ensure scientific robustness and operational utility. The 2024 workshop, co-hosted with NASA, focused on the intersection of AI foundation models and geospatial digital twins, which are virtual replicas of the physical world. Experts gathered to discuss the integration and challenges of employing AI in digital twins in topics such as data integration, enhancing analytical and predictive analysis for future scenarios, and improving decision-making.
GeoAI at ACMSigspatial
This annual, one-day workshop at the ACM Sigspatial Conference includes keynotes, individual presentations, and paper presentation awards. The event focuses on AI for geographic knowledge discovery, impacting decision-making in areas such as humanitarian mapping, intelligent transport systems, urban expansion analysis, health data analysis and epidemiology, handling natural disasters, and the general monitoring of the Earth’s surface.
CV4EO at WACV
We organize the Computer Vision for Earth Observation workshop series at the annual IEEE/CVF Winter Conference on Applications of Computer Vision, fostering application-oriented discussions between the computer vision community and experts from geoscience domains, remote sensing data providers, governmental agencies, and other organizations utilizing computer vision-enabled EO data analysis for decision-making in disaster response, national security, environmental protection, and other application areas.
We're a multidisciplinary team
We're a multidisciplinary team
To deliver our mission, we need dedicated, mission-driven researchers with expertise in electrical engineering, applied mathematics, computer engineering, statistics, computer science, data science, and more!
In addition to these academic backgrounds, our team has expertise in Artificial Intelligence, Probabilistic Reasoning, Computer Vision, Risk Analysis, Visual Analytics, High-Performance Computing, Geographic Information Science, Ontology Reasoning, and Remote Sensing.
Exploring complex geospatial analytic workflows
Settlement Characterization
Settlement Characterization
With over two decades of experience in creating foundational data layers, ORNL continues to exploit overhead imagery toward developing scalable ways to characterize a wide variety of socioeconomic neighborhoods by using artificial intelligence methods and high performance computing resources. The foundational datasets are being used in applications that include identification of zones for economic stimulus, mapping of unstructured settlements, studying spatial similarities between cities, and supporting population distribution studies.
AI-Driven Data Flows
AI-Driven Data Flows
Automated feature extraction from geospatial data requires constant retraining of machine-learning algorithms on the best available data. This method can cause loss of accuracy due to low-level noise and varying regional conditions. The Remote Sensing Flow (RESFlow) approach partitions images into groups based on similarities and localizes context into buckets to expedite analysis by orders of magnitude while maintaining accuracy.
Mapping Building Footprints, Road Networks, and More
Mapping Building Footprints, Road Networks, and More
Automatic feature extraction at scale enables streamlined processing of vast amounts of remote sensing data at high spatial and temporal resolution with minimal human involvement. Oak Ridge National Laboratory employs machine learning, computer vision, and high-performance computing resources toward automated feature extraction under multimodal sensing sources, scarce training data and distribution shifts, to create foundational data layers for building footprints, road networks and other critical infrastructures.
Gravity Mapping
Gravity Mapping
Using methods developed by ORNL scientists, Geomatics researchers are adapting artificial intelligence techniques to uncover and better understand complex relationships between geological and topological density patterns. BY incorporating AI into the techniques, researchers are demonstrating the possibility of producing high-resolution gravity maps on a global scale with consistently reliable quality at reduced cost and effort. The resulting Gravity maps are aimed at supporting navigation systems, early detection of potential earthquakes, and measurement of changes in water patterns.
Keeping people safe
Disaster Response
Our deep-learning methods underlie a critical geographic dataset used during hurricane response efforts
Critical Infrastructure Resilience
ORNL tools help ensure energy supply
AI on the Edge
ORNL researchers developed a mobile system to quickly detect objects and analyze images following disasters
