Exploring the Multidimensional Visual Perception of Urban Riverfront Street Environments: A Framework Using Street View Images, Deep Learning and Eye-Tracking

Urban waterfront areas (UWAs), which are essential natural resources and highly perceived public areas in cities, play a crucial role in improving the quality of the urban environment. While numerous studies have delved into the visual perception of urban environments, little attention has been paid to understanding how the visual perception of urban riverfront streets (URSs) differs with various aspects within their unique spatial environment. This study took the Gusu District in Suzhou, China, as a case study, applying deep learning to street-view images to identify urban riverside landscape elements and evaluate their visual attention, aesthetic preference, and distinctiveness through eye-tracking technology and questionnaires. Subsequently, a multidimensional assessment was conducted to analyze how landscape elements influence visual perception in the urban riverfront street. This study concludes that (1) riverfront streets in the Gusu District present balanced visual attention, with high aesthetic preference but limited distinctiveness, and only a few roads in the ancient city score highly for distinctiveness. (2) Greenery, traditional-style buildings, water, and riverfronts positively impact visual perception, while buildings have a negative impact, and backgrounds such as the sky and roads exhibit minimal influence. This study validated the scientific accuracy, appropriateness, and precision of assessments of visual attention, aesthetics, and distinctiveness to quantitatively evaluate the multidimensional human perception of URSs.