Journal Article

Abstract

Rapid urbanization demands new approaches for the continuous monitoring of urban conditions to ensure smooth operation and liveable cities – key goals of successful Urban Facility Management (FM). Traditional inspection methods offer limited scalability and low temporal resolution. This paper explores how computer vision can transform urban condition monitoring through the analysis of street-level imagery, and p ositions these methods as a technical foundation for dynamic digital twins that continuously capture and analyze urban conditions rather than serving merely as static, formal scale models. Based on a systematic review of recent studies, it highlights data-driven methods – particularly image segmentation and deep learning models such as DeepLabv3+ – used to assess indicators such as greenery, cleanliness, or accessibility. The findings demonstrate computer vision's potential to generate high-frequency data that can feed dynamic digital twins, ena bling continuous monitoring and immediate response strategies. Integrating these techniques into Urban FM supports more data-driven decision-making and adaptive management of the built environment.