Vivaldi: A Machine Learning Approach to Inferring Building Height

Invention Reference Number

202505875

West Shinjuku, Tokyo, Japan financial district cityscape over residential apartments. Image from Envato

Oak Ridge National Laboratory has developed a novel data-driven framework to enhance the accuracy of building height estimation at a building-by-building level, particularly for high-rise structures. Traditional methods often struggle to estimate the height of taller buildings due to data imbalance and classification inconsistencies. This �鶹Ӱ�� improves precision by segmenting the built environment into distinct categories, enabling tailored analysis that supports applications in energy modeling, disaster response, and urban planning.

Description

This technology introduces a new framework for inferring building height at the building-by-building level by leveraging a segmented modeling approach. Unlike conventional single-model methods, this solution divides the built environment into multiple structural categories based on inherent distribution patterns, allowing tailored estimation strategies for each. The framework uses unique combinations of building morphology and contextual data, such as footprint complexity and spatial relationships, to drive specialist model selection and improve predictive performance. This segmentation is guided by a customized classification mechanism, which determines the appropriate predictive model without relying on direct height input. The ensemble modeling approach significantly enhances estimation accuracy, particularly for structures exceeding a certain threshold height, which are typically underrepresented in existing datasets. This improvement has important implications for sectors where accurate urban morphology modeling is critical.