Invention Reference Number
This technology introduces a scalable method for identifying features of the built environment, such as mobile home locations, using advanced data modeling techniques. By leveraging a transfer learning framework, the approach enables reliable inferences across multiple footprint datasets—overcoming limitations of source-specific models. This innovation enhances the ability to extract actionable insights from spatial data to support applications in urban planning, emergency response, and demographic analysis.
Description
Traditional machine learning models for analyzing the built environment are constrained by dependence on a single data source, which restricts scalability across regions or dataset types. This invention resolves that challenge through a morphology-driven learning approach that generalizes across unique building footprint datasets. The model identifies meaningful physical and spatial patterns that characterize different building types and applies these learned relationships to new datasets with high accuracy. By enabling consistent performance regardless of source variation, this technique expands analytical capabilities for understanding building-level characteristics, including type, age, and relative height. The resulting framework provides a transferable foundation for large-scale, data-driven assessments of urban and rural environments without being limited to a specific data provider.
Benefits
- Scalable analysis across diverse geographic datasets
- Improved accuracy in identifying built environment targets
- Enables expansion of existing models to previously inaccessible regions
Applications and Industries
- Urban planning and infrastructure assessment
- Emergency management and disaster response
- Population and demographic studies
- Real estate and environmental analytics
Contact
To learn more about this technology, email partnerships@ornl.gov or call 865-574-1051.
