Augmenting urban planning with computer vision: A review of the state-of-the-art

• Computer vision (CV) uncovers hidden patterns, advancing evidence-based urban planning decisions • CV aids urban planning visualisation and automated urban planning and design processes • Machine learning algorithms complement CV, improving classification and modelling tasks • Convolutional neural networks enable CV to excel in classifying diverse urban environment data • CV facilitates more informed decision-making, and advancing urban planning research and practice Computer Vision (CV) offers powerful methods for analysing built environments, yet its integration into planning workflows remains fragmented. This study conducts a systematic review of CV applications in urban planning using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) protocol, synthesising prevailing models, methods, and application domains. The review identifies five dominant CV research clusters in urban contexts (built environment analysis, urban sensing and data acquisition, smart mobility, methodological advances, and 3D urban modelling) and synthesises that CV is primarily used to extract physical and perceptual urban features, support diagnostic analytics, and enable emerging visualisation and generative design workflows. Advances in deep learning, convolutional neural networks, and emerging foundation models (including vision transformers and diffusion-based generative systems) are expanding capability across diverse CV tasks and lowering barriers to adoption. Building on an inductive synthesis of application patterns across the reviewed studies, the paper proposes an Urban Visual Intelligence (UVI) hierarchy that clarifies how imagery is translated into progressively more planning-relevant outputs, from observation and diagnosis to associative explanation, forecasting, and intervention prototyping. The review also highlights constraints related to data representativeness, transferability, interpretability, privacy and generative unreliability, underscoring the need for stronger validation and governance as outputs become more decision relevant. Overall, the review consolidates how CV is reshaping urban analytics and offers a structured basis for responsible integration into planning practice.