Quantification and automatic computation of 3D window views in high-rise, high-density cities based on city information models and machine learning

Abstract

High-quality window views, e.g., with high proportions of greenery, sky, and sea, and a great openness benefit urban dwellers’ physical and mental health and life satisfaction, especially in high-rise, high-density cities, e.g., Hong Kong. Towards healthy and sustainable high-rise, high-density urban development, high-quality window views have shown significant socioeconomic impacts on landscape management, urban planning and design, and real estate valuation and transaction. Thus, quantification and automatic computation of window views are significant in examining the disparity of window-level exposures, e.g., to nature and openness for multiple urban applications and analytics, e.g., multi-level urban environment improvement, precise housing valuation, and analytics for urban health and environmental justice. However, quantification and automatic computation of window views in high-rise, high-density cities are challenging without a systematic assessment framework and mature window view collection and assessment methods. To bridge the research gap, the study proposes an urban computing approach to 3D window views in high-rise, high-density cities making full use of the up-to-date City Information Models (CIMs) and machine learning techniques. The thesis first theoretically constructs a Hierarchy of Window View Characteristics (HoWVC) to systematically quantify window views for landscape management, urban planning and design, and real estate valuation and transaction. The HoWVC comprises basic, compound, and perceived window view characteristics and identifies attributes and descriptors at pixel, patch, and graph levels. Then, the thesis presents an urban computing paradigm for HoWVC on CIMs by developing a representative window view index, identifying four procedures to compute the index, and formulating two urban computing norms. Last, three automatic computation methods for example basic, compound, and perceived window view characteristics, i.e., view feature, openness, and perceived openness are implemented based on the developed urban computing paradigm. Multi-scale experimental results in high-rise, high-density urban areas of Hong Kong confirmed the accuracy and efficiency of the proposed urban computing approach to the systematic and automatic quantification of 3D window views. The thesis’s contributions are three-fold. Theoretically, the proposed HoWVC extends the traditionally separated characterizations of window view content into a systematic representation. The defined hierarchical window view characteristics push the boundary of current window view studies for comprehensively quantifying window views for landscape management, urban planning and design, and real estate valuation and transaction. The identified attributes and descriptors extend the knowledge of current window view studies for automatic computation. Methodologically, the urban computing paradigm and three automatic computation methods turn traditional small-scale quantifications into urban computing. Besides, the paradigm and methods complement existing studies on overhead and street-level urban views. Urban computing of window views enables a new angle of sensing for urban environments, especially in high-rise, high-density cities. For agencies and practitioners in landscape management, urban planning and design, and real estate valuation and transaction, the proposed approach enables systematic and automatic assessments of window views instead of previous qualitative judgments. The urban-scale assessment results unlock a plethora of window view-based applications, e.g., multi-level urban environment improvement and precise housing valuation, and intra-city and inter-city analytics for urban health, environmental justice, and urban sustainability.