A Visual-Attention-Driven Framework for Quantifying Wayfinding Cue Strength for Pedestrians in Urban Scenes: A Pilot Study of Older Adults in Hong Kong

In urban environments, wayfinding is a fundamental task through which individuals access essential services. Quantifying the strength of perceived cues that facilitate wayfinding can inform urban design interventions aimed at reducing potential wayfinding difficulties. This study develops a visual-attention-driven computational framework to measure such cue strength from street-view images. In implementation, the framework adopts an image-inpainting-based strategy and incorporates a task-specific eye-tracking fine-tuning procedure to improve the strength of visual saliency extraction. A POI-based cognitive weighting scheme is then introduced to integrate the extracted visual values with cognitive information into a composite measure of wayfinding cue strength. Due to the lack of available data, a self-collected dataset integrating eye-tracking data from older adults, cognitive data, and ground-truth annotations for 30 intersections in Hong Kong was constructed to evaluate the framework and provide a data foundation for future work. The results revealed a significant positive correlation between the computed scores and the mean human ratings on a 1–5 Likert scale, while the linear regression model yielded an R2 of 0.62, providing preliminary evidence for the effectiveness of the proposed computational framework. The findings suggest that the cognitive attributes of spatially extensive environmental entities, such as neighborhood parks, could be considered in future site-selection processes, as they may enhance pedestrians’ spatial understanding and thereby facilitate wayfinding.