Creating 3D city models from satellite imagery for integrated assessment and forecasting of solar energy

Mohammed Kadhim, Nada 2018. Creating 3D city models from satellite imagery for integrated assessment and forecasting of solar energy. PhD Thesis, Cardiff University. Item availability restricted.

Preview	PDF - Accepted Post-Print Version Download (19MB) | Preview
	PDF - Supplemental Material Restricted to Repository staff only Download (166kB)

Abstract

Buildings are the most prominent component in the urban environment. The geometric identification of urban buildings plays an important role in a range of urban applications, including 3D representations of buildings, energy consumption analysis, sustainable development, urban planning, risk assessment, and change detection. In particular, 3D building models can provide a comprehensive assessment of surfaces exposed to solar radiation. However, the identification of the available surfaces on urban structures and the actual locations which receive a sufficient amount of sunlight to increase installed power capacity (e.g. Photovoltaic systems) are crucial considerations for solar energy supply efficiency. Although considerable research has been devoted to detecting the rooftops of buildings, less attention has been paid to creating and completing 3D models of urban buildings. Therefore, there is a need to increase our understanding of the solar energy potential of the surfaces of building envelopes so we can formulate future adaptive energy policies for improving the sustainability of cities. The goal of this thesis was to develop a new approach to automatically model existing buildings for the exploitation of solar energy potential within an urban environment. By investigating building footprints and heights based on shadow information derived from satellite images, 3D city models were generated. Footprints were detected using a two level segmentation process: (1) the iterative graph cuts approach for determining building regions and (2) the active contour method and the adjusted-geometry parameters method for modifying the edges and shapes of the extracted building footprints. Building heights were estimated based on the simulation of artificial shadow regions using identified building footprints and solar information in the image metadata at pre-defined height increments. The difference between the actual and simulated shadow regions at every height increment was computed using the Jaccard similarity coefficient. The 3D models at the first level of detail were then obtained by extruding the building footprints based on their heights by creating image voxels and using the marching cube approach. In conclusion, 3D models of buildings can be generated solely from 2D data of the buildings’attributes in any selected urban area. The approach outperforms the past attempts, and mean error is reduced by at least 21%. Qualitative evaluations of the study illustrate that it is possible to achieve 3D building models based on satellite images with a mean error of less than 5 m. This comprehensive study allows for 3D city models to be generated in the absence of elevation attributes and additional data. Experiments revealed that this novel, automated method can be useful in a number of spatial analyses and urban sustainability applications.

Item Type:	Thesis (PhD)
Status:	Unpublished
Schools:	Engineering
Uncontrolled Keywords:	3D City Models, Building Detection, Height Estimation, GrabCut partitioning, Shadow Anlaysis, Very High Resolution (VHR)Satellite Imagery
Date of First Compliant Deposit:	19 February 2018
Last Modified:	12 Apr 2021 15:27
URI:	https://orca.cardiff.ac.uk/id/eprint/109232

Actions (repository staff only)

Edit Item