Date of Award
December 2025
Document Type
Dissertation
Degree Name
Doctor of Philosophy (PhD)
Department
Energy Engineering
First Advisor
Olusegun Tomomewo
Abstract
Buildings contribute significantly to worldwide energy use and greenhouse gas emissions, underscoring the importance of precise evaluation of building geometry and envelope performance for efficient energy management and climate mitigation initiatives. Recent advances in high-resolution satellite remote sensing, uncrewed aircraft systems (UAS), and deep-learning-based data analysis have produced a growing body of scholarly work focused on three-dimensional building extraction and thermal loss detection. However, these methods are often evaluated in isolation or under controlled conditions, limiting their demonstrated applicability to real-world, cold-region urban environments where seasonal variability and heterogeneous building stocks present unique challenges. This dissertation addresses this gap through a structured synthesis and applied evaluation of remote sensing-based approaches for 3D building geometry extraction and thermal performance assessment.
This research develops and applies an integrated remote sensing framework that combines GaoFen-7 (GF-7) high-resolution stereo satellite imagery, UAS-based thermal photogrammetry, and deep-learning-driven multi-feature data fusion methods drawn from the contemporary literature. Rather than proposing a single novel algorithm, the study critically reviews and contextualizes three anchor journal methodologies corresponding to building geometry extraction, multi-feature fusion networks, and thermal photogrammetric analysis. These approaches are systematically applied to residential and commercial buildings at the University of North Dakota, across the City of Grand Forks, and within representative building stocks in the State of North Dakota. The framework integrates spectral, geometric, and thermal information to generate three-dimensional building models and spatially resolved indicators of building-envelope performance suitable for energy analysis and decision support.
The results demonstrate that literature-established remote sensing and deep-learning techniques can be cohesively applied across sensing platforms and spatial scales to produce reliable 3D building representations in cold-region environments. Multi-feature fusion approaches show increased robustness under challenging conditions such as snow cover, low solar angles, and mixed land use, while the integration of UAS thermal photogrammetry with three-dimensional geometry enhances the identification and interpretation of insulation defects, air leakage, and thermal inefficiencies. The findings indicate that an integrated, literature-grounded remote sensing framework can support scalable energy efficiency assessment, retrofit prioritization, and energy management planning for institutional, municipal, and state-level applications. This dissertation contributes to the advancement of applied geospatial energy analysis by bridging established remote sensing research with practical implementation in cold-region built environments.
Recommended Citation
Dunlevy, Matt, "A Review Of Remote Sensing-Based Approaches For 3D Building Geometry And Thermal Loss Detection In Urban Energy Efficiency Applications" (2025). Theses and Dissertations. 8218.
https://commons.und.edu/theses/8218