Assessing Divergent Perception–Function Coupling Across Old and New Urban Fabrics: A Nonlinear Spatial Analysis of Xi’an, China

Although visual assessments of the built environment have progressed, existing studies frequently evaluate metropolitan spaces as uniform entities. This overlooks the interplay between subjective spatial perception and functional infrastructure across distinct architectural morphologies. This research synthesizes over 133,000 street-level images and 731,000 point-of-interest records to categorize the spatial alignment of perceived street quality and functional density in Xi’an, China, using perception scores generated by a pretrained model. Generalized Additive Models are employed to investigate their complex, nonlinear associations. Results reveal that while historic districts rely on dense commercial areas, which are associated with high urban vibrancy, newly developed zones maintain vibrant atmospheres despite dispersed infrastructure. Regarding pedestrian comfort, modern workplace-dominated environments consistently score high, whereas historic areas with similar office densities show lower comfort. Nonlinear modeling uncovers stark regional contrasts: the positive association between lifestyle amenities and vibrancy plateaus faster in historic sectors than in modern developments. Additionally, spatial comfort demonstrates an inverted-U trajectory relative to workplace density in older neighborhoods, with an estimated turning point at approximately 113 units/km2, beyond which comfort declines, whereas a comparable decline is not observed in newer districts. These outcomes highlight that equivalent visual perception metrics can obscure vastly different structural realities. Consequently, future architectural interventions must integrate localized functional density with visual evaluations to formulate context-aware design strategies.