Perceived built environment and non-motorist crashes: An exploration with street view imagery

Walking, cycling, and other forms of non-motorized travel are widely promoted in urban and transportation planning for their evident health and environmental benefits. However, ensuring travel safety for non-motorists remains a significant challenge. Existing risk assessments of non-motorist safety issues primarily focus on objectively measured traffic conditions (e.g., land use, road width, and the presence of sidewalks), while overlooking individuals' perceptions of the built environment. In this paper, we explore how the perceived built environment can impact traffic crashes involving non-motorists by employing Google Street View (GSV) data. Specifically, we have quantified six perceptual attributes (i.e., beautiful, boring, depressing, lively, safe, and wealthy) around non-motorist crash locations using GSV images and machine learning models trained by the MIT Place Pulse 2.0 dataset. Negative binomial regression models are developed to examine the associations between these perceptual attributes and a nine-year non-motorist traffic crash dataset in Hartford, Connecticut. We also apply different buffer sizes to test the sensitivity of the associations, revealing the buffer size that is the most relevant to travel safety. Our results indicate that the perceived beauty, safety, and wealthiness of the built environment are negatively associated with non-motorist crash risk, whereas the perceived liveliness, depression, and boredom of the environment have positive correlations. The findings can shed insights into the mechanistic intersection of environmental perceptions and traffic crashes involving non-motorists. By incorporating the perceptual dimensions into crash analysis, stakeholders in the planning and transportation sectors can develop targeted, street-level interventions to enhance road safety.