Assessment of Elderly-Oriented Spatial Visual Quality in Pocket Parks: A New Method Based on Panoramic Semantic Segmentation

With the intensification of global population aging, pocket parks in high-density urban areas have emerged as key venues for the daily activities of older adults. However, existing research predominantly relies on street view images (SVI) captured from vehicle lanes, which inadequately captures the spatial heterogeneity and micro-scale visual features within these parks. This study aims to assess the age-friendly spatial visual quality (SVQ) of pocket parks from a panoramic perspective. This study examines 29 pocket parks on Xiamen Island. An index for Older-Adult Friendly Visual Quality Index (OAVQ-I) was developed. Panoramic imagery combined with deep learning-based semantic segmentation was employed to quantitatively analyze six indicators: Green View Rate (GVR), Natural Elements Richness (NER), Sky Openness (SO), Enclosure (Encl), Rest Facilities Visibility (RFV) and Walkable Space Ratio (WSR). Regression analysis was subsequently applied to determine indicator weights. The results indicate that RFV, Encl, and SO are the decisive factors, collectively accounting for 87% of quality variation. Notably, a high green view ratio does not necessarily indicate high quality; excessively dense vegetation may create an "ecological trap" by obstructing facilities and limiting light penetration. Through comparative case analysis, three spatial patterns were identified: the High-quality friendly type, the Moderately friendly type, and the Improvement needed - low quality type. This study advances the evaluation of urban green spaces from macroscopic statistics toward microscopic functional diagnosis. The findings suggest that urban renewal should transition from increasing greenery to enhancing quality, with particular emphasis on pruning mid-level vegetation to improve facility visibility and lighting conditions, thereby creating more supportive environments for older adults.