Date of Graduation
Summer 2022
Degree
Master of Science in Geospatial Sciences
Department
Geography, Geology, and Planning
Committee Chair
Xin Miao
Abstract
Three-dimensional city (3D) models are very useful in supporting natural disaster preparation and response. LiDAR surveying is currently the main method by which 3D city models are created; however, LiDAR data on a local scale is hard to obtain for developing countries. This project sought to test whether or not urban feature height data obtained using the photogrammetric sun-angle shadow method is a viable alternative to LiDAR-derived 3D city models. A core element of this work was the development of a toolset to be shared freely to the public to promote crowdsourcing of 3D building data. Prior works were reviewed and a shadow-overlapping method for estimating building heights was selected and implemented in the Python programming language. Shadow detection methods in the literature were also reviewed and eight were modified into a simple command-line Python tool to batch process shadow detection using multiple algorithms. The Missouri State University, Springfield, MO, campus was selected as the study site for testing the shadow-overlapping process and LiDAR/DEM data was used to create building footprints with ground truth height values. For the shadow detection methods that produced the most accurate results, roughly 60% of building height estimates were within 10 feet (or one floor) of the true height, and buildings whose heights were between 37-50 feet consistently had the lowest margins of error. A systematic finding, however, was that shorter buildings’ heights were overestimated and taller buildings were underestimated. A similar pattern was identified with building size/square footage with smaller buildings being overestimated and larger buildings being underestimated. In the end, the results suggested that the shadow-overlapping method is likely not reliable enough to produce height estimations comparable to LiDAR-derived methods and that these height estimates are not suitable for downstream calculations. However, for a simple/generalized 3D cartographic representation of an area, it appeared that this low-cost method could produce adequate results.
Keywords
3D city models, building height estimation, GIS, photogrammetry, Python, shadows
Subject Categories
Geographic Information Sciences | Remote Sensing | Spatial Science
Copyright
© Lonnie Lee Byrnside III
Recommended Citation
Byrnside, Lonnie Lee III, "Shadow-Based Automatic Building Height Estimation From High Spatial Resolution Satellite Imagery" (2022). MSU Graduate Theses/Dissertations. 3768.
https://bearworks.missouristate.edu/theses/3768
Open Access
Included in
Geographic Information Sciences Commons, Remote Sensing Commons, Spatial Science Commons
