Visual comfort impact assessment for walking spaces of urban historic district in China based on semantic segmentation algorithm

Abstract

Urban historic districts (UHDs) are primarily walkable areas in built environments that preserve the traditional and local characteristics of a specific historical period. However, the impacts of the visual perception elements (VPEs) in UHD walking spaces on visual comfort related to the health-benefits have not been studied in depth. In contrast to previous studies, this study develops a comprehensive methodology framework for assessing visual comfort of UHD walking spaces to provide evidence for health-benefits based on Semantic Segmentation (Semseg) algorithm, integrating street view image data mining, Semseg computing, K-means cluster analysis, visual perception analysis and multiple linear regression analysis. A total of 11,633 images from Baidu Maps-Panoramic Static Imagery (BM-PSI) API were collected from 17 typical UHD walking spaces in China, forming 17 datasets of visual sequence images (VSIs). Three types of VPEs (Natural-VPE, Artificial-VPE and Cultural-VPE) were identified and quantified by Semseg Computing to calculate Visual Pixel Ratio (VPR) indicators. 15 representative VSIs were filtered by K-means cluster analysis, which were subsequently evaluated through web-based questionnaires involving 308 respondents with diverse demographic characteristics to derive General Visual Comfort Ratings (G-VCR) and Sub-Visual Comfort Ratings (S-VCR) indicators. Finally, multiple linear regression analysis was conducted to reveal the relationships between VPR and both G-VCR and S-VCR indicators respectively. Overall, the results reveal that Natural-VPE and Cultural-VPE have a significant positive impact on visual comfort, while Artificial-VPE exerts a opposite effect. Furthermore, demographic characteristics affect visual comfort differently: female respondents showed stronger preference for Natural-VPE, while male respondents were more sensitive to Artificial-VPE; middle-aged and elderly respondents favored environments with more Natural-VPE and Cultural-VPE, whereas young respondents were more affected by Artificial-VPE; higher-educated respondents exhibited greater appreciation for Natural-VPE and Cultural-VPE compared to non-higher educated respondents. The findings can provide valuable insights for promoting sustainable renewal of UHD walking spaces aligned with the Healthy City concept.