How do streetscape visual components affect public perception in post-renewed neighborhoods: a case study in Chengdu

Streets are fundamental public spaces within urban environments, profoundly impacting public well-being through their visual quality. This study endeavors to construct a multiple linear regression model to explore the relationship between visual components and public perception in selected post-renewal neighborhoods. Moreover, it innovative employs a mediating effect test to delve into the underlying mechanisms through which streetscape visual elements influence perception of beauty. By quantifying streetscape visual components, we identified 7 key elements by categorizing 150 semantic categories obtained from the Deeplab V3+ model and ADE20K dataset. Furthermore, we utilized the Shannon Wiener index and Pielou's evenness index to characterize the distribution of these visual components. Subsequently, these 9 components and questionnaire data were employed as variables in the multiple regression model and mediating effect test. The results unveil a significant correlation between visual components and public perception. Particularly noteworthy, 5 visual components demonstrate full mediation in influencing perception of beauty. This implies that public perception of streetscape is indirectly influenced by various factors. The research findings are poised to serve as a valuable reference for regional urban planning, design, and management.