2:45 PM - 3:00 PM
[078] Effects of Street Qualities and Destination Attractiveness on Bike Ridership Using Street Imagery and Deep Learning
Keywords:Bike-Sharing Program, Visual Bikeability, Destination attraction, Deep Learning
There is a growing consensus among planners and policymakers that active transit such as walking and bicycling provides various socioeconomic and environmental benefits that enhance urban sustainability. In this active mobility paradigm, the Bike-Sharing Program (BSP) has attracted great attention as an effective traffic policy response to catalyze cycling for the last decade. While many empirical studies have revealed the relationship between BSP patterns and built environments regarding urban form and design at the neighborhood-level, limited understanding has remained of whether street qualities and destination attractiveness encourage bike use. This study fills this gap by examining the correlation between BSP usage and neighborhood- and eye-level environmental characteristics around bike stations in Seoul, Korea. This study notably estimates cyclist eye-level visual bikeability using a semantic segmentation technique, a type of computer vision task, based on a deep learning network. In addition, our study extends the previous literature by exploring the complex mechanism between destination attractions and visual environmental features to estimate the effects of streetscape quality on BSP patterns, and whether this mechanism varies across time factors. Based on these estimations, our study employed Heckmans’ two-step model to identify how built environments at different levels affect BSP usage, especially for returns, even after accounting for selection bias. The findings of this study showed that patterns of bike returns were associated with population density, intersection density, and residential land use, and its correlations were varied by weekdays and weekends. Furthermore, visual greenery and outdoor openness were critical determinants in ameliorating or promoting correlations between attractions and bike returns. This study not only informs planners, traffic engineers, and policymakers of the importance of urban space quality to promote bike trips but also helps to provide empirical evidence to maintain support for investment in cycling infrastructure.