Resumen
Walking exercise is a prevalent physical activity in urban areas, with streetscapes playing a significant role in shaping preferences. Understanding this influence is essential for creating urban environments conducive to walking exercise and improving residents? quality of life. In this study, we utilize scenic beauty estimation and deep learning methods, leveraging street view images and walking exercise trajectories to analyze this influence from a human-centric perspective. We begin by generating sampling points along streets covered by trajectories and acquiring street view images. Subsequently, we apply a deep learning model to segment the images, yielding six visual indicators. Additionally, we use scenic beauty estimation to derive the seventh visual indicator. Finally, we match these indicators with trajectory data to implement preference analysis. The main findings are: (1) preferences for walking and running exercises differ on multiple indicators; (2) there are gender distinctions, with males preferring openness and females prioritizing enclosed spaces; (3) age plays a role, with those aged 30?40 preferring openness and those aged 40?50 preferring enclosed spaces; (4) preferences for different indicators vary over time and across different locations. These insights can inform policymakers in tailoring urban planning and design to specific population segments and promoting sustainable residential landscapes.