A multi-dimensional urban spatial perception framework for urban diagnosis in Wuhan driven by multi-source data

To address the evaluation bias in conventional urban spatial design and governance that prioritizes functional provision over lived experience, this study proposes an implementable multi-source data-driven framework for multi-dimensional urban spatial perception, supporting urban diagnosis and grid-based governance. The framework integrates three categories of information within a unified spatial unit: (1) extracting interpretable environmental attributes from street-view images and predicting six dimensions of affective perception, including beauty, liveliness, and perceived safety; (2) characterizing supply-agglomeration patterns and supply intensity using spatiotemporal trajectories and mobile phone signaling data; and (3) identifying functional rhythms-daily, periodic, and occasional-by leveraging POI semantics and temporal periodicity. Building on these components, we construct a three-dimensional matrix of "affective perception-agglomeration intensity-regional function," classify Wuhan's urban space into 18 types of multi-dimensional perception units, and identify key areas requiring targeted, category-specific governance. Furthermore, the typology is linked to governance-oriented references for different spatial types, including feasible street- and community-level actions, corresponding leverage variables, and potential indicators for future monitoring or before-after assessment. Rather than constituting a complete closed-loop health-check system, the framework provides a diagnostic basis for urban problem identification, intervention prioritization, and refined governance.