Perceived and actual pedestrian crash risk on local streets: identifying actionable urban design qualities

Street view imagery and deep learning map perceived road safety at city scale, yet identify where perception fails, not which design decision to revisit. This paper compares the weight pedestrians assign to five urban design qualities, elicited from a survey of 871 respondents, with the weight the same qualities carry in a crash model that accounts for pedestrian exposure, across 582 local street units in Seoul. Sidewalk presence dominates perceptual judgement yet accompanies higher crash risk (IRR = 1.77), rising with width; linkage is protective but discounted. Expressing misperception as design attributes turns a documented mismatch into targeted, buildable guidance.