Respondent Dynamic Attention to Streetscape Composition in Nanjing, China

Abstract

Scholars are interested in understanding human responses and perceptions concerning the configuration of streetscape environments that serve multiple functions. However, drivers’ visual attention to the streetscape has seldom been studied dynamically in multi-modal settings. By employing eye-tracking and semantic segmentation, visual attention partitions and objects and patterns are inspected in a per-second count along three typical roadways in Nanjing, China. In our study of 28 participants, it was found that people are likely to focus on the frame center (p-value < 0.005) in all methods of transportation. Roads and buildings are constantly observed along the roadway (p-value < 0.005), while smaller transportation objects across multi-modal conditions are noticed more in per-area counts (p-value < 0.025). Besides, vehicles are focused on more in a higher-speed driving lane (p-values < 0.005), while greenery and humans attract more attention in a slower lane (p-values < 0.005). The results indicate that the previous visual engagement results should be reconsidered on several points, and that the risk of distractions from non-traffic-related elements could be overestimated. The potential of the road surface in integrating safety and information-providing has been ignored in current studies. This study showed that greenery and other functional elements will not distract users in driving lanes; decreasing the calculation burden to two-ninth is possible in smart driving. These results could be helpful in future sustainable cities.