How Blue–Green coastal landscapes shape runners' Perceptions: A deep-learning street-view study of Qingdao's coastal greenway

With rapid urbanization, understanding how coastal blue-green landscapes support health-oriented physical activity has become increasingly important. Using the 21.25 km coastal greenway in Qingdao as a case study, this research examines how the visual characteristics of coastal open spaces influence runners' perceptions of security, comfort, resilience, and attractiveness. A runner-eye-level street-view image dataset was constructed by integrating field video recordings with Baidu Map street views sampled at 50 m intervals, resulting in 417 sampling points. Based on this dataset, a SegFormer semantic segmentation model pretrained on ADE20K was applied to quantify eight visual indicators identified in the literature and image analysis, including green and blue view indices, sky view factor, spatial enclosure, walkability, crowding, public transport accessibility, and sports-facility density. Perception data were collected through an online questionnaire in which 137 respondents rated ten randomly presented street-view images, resulting in 1370 valid ratings. K-means clustering was then applied to classify the coastal open spaces into three spatial types—high sky-view open spaces, high green-view semi-enclosed spaces, and low green-view enclosed spaces—followed by Spearman correlation analysis and multiple linear regression to assess how visual features and spatial types jointly shape runners' perceptions. The results show that high sky-view open spaces achieve the highest scores across all four perception dimensions, whereas low green-view enclosed spaces perform the worst, particularly in comfort and attractiveness. The blue view index and walkability consistently and significantly enhance comfort, resilience, and attractiveness, while excessive spatial enclosure weakens perceptions of security and resilience. These findings highlight the differentiated roles of blue–green visual exposure and spatial form in shaping runners’ experiences and provide AI-supported evidence to inform the design of health-supportive coastal greenways in high-density urban contexts. • Integrates runner videos, Baidu SV and SegFormer to quantify eight blue–green indicators along a 21.25 km greenway. • K-means reveals three spatial types with distinct patterns of runners’ security, comfort and attractiveness. • Blue view and walkability improve comfort and attractiveness; enclosure reduces security and resilience.