Urban fabric decoded: High-precision building material identification via deep learning and remote sensing

Area of Science:

• Urban Planning and Sustainability
• Artificial Intelligence in Environmental Science
• Remote Sensing and Geospatial Analysis

Background:

• Accurate building material data is crucial for embodied carbon reduction, retrofitting, and urban circularity.
• Current material databases are often project-specific or geographically limited, providing only approximate data.
• Inadequate records and cost constraints hinder the acquisition of large-scale, precise material information.

Purpose of the Study:

• To introduce a novel automated framework for identifying building materials using remote sensing and Google Street View imagery.
• To demonstrate the scalability and adaptability of the framework across diverse Danish urban landscapes.
• To provide high-resolution insights into material distribution for sustainable urban development.

Main Methods:

• Utilized a deep learning model trained on sensing technology and Google Street View data.
• Initial model training and validation were performed using a comprehensive dataset from Odense.
• The framework was extended to characterize building materials in Copenhagen, Aarhus, and Aalborg.

Main Results:

• Successfully identified roof and facade materials across multiple Danish cities with high resolution.
• Demonstrated the model's scalability and adaptability to varied geographic contexts and architectural styles.
• Provided detailed insights into the distribution of building materials across different urban settings.

Conclusions:

• The automated framework offers a scalable and adaptable solution for precise building material identification.
• Findings are pivotal for informing sustainable urban planning and revising building codes for emission reduction.
• Optimized retrofitting efforts can be achieved by leveraging these high-resolution material distribution insights.