Developing a data-fusing method for mapping fine-scale urban three-dimensional building structure

Abstract

Understanding urban morphology is essential for various urban management studies and local environmental issues and guiding sustainable city development. Existing studies mainly focus on analyzing urban morphology from the horizontal aspect, while the urban vertical structure has rarely been discussed due to the scarcity of reliable and fine-scale urban three-dimensional (3-D) building data. This study develops an effective data-fusing methodology to estimate the heights of individual buildings at a city scale. We examined a machine-learning regression model by collecting public materials, including multi-source remote sensing-(RS)-based products, building-derived features, and relevant data to verify its performance in building height estimation. By applying the model in Shenzhen City, a dense city in the Guangdong-Hong Kong-Macao Greater Bay Area, results demonstrated that integrating rich multi-source explanatory variables could achieve high-accuracy building height retrieval. Using multiple building morphological metrics derived by building height data as proxy measures, the urban 3-D form patterns were further analyzed to understand current heterogeneous urban morphological structures. The proposed methodology can be conveniently applied to worldwide cities for urban 3-D morphology retrieval. Also, the available building height information is useful for planners to design morphological control for cities and thus contributes to sustainable and smart city development.