A study on the coupling mechanism between the urban environment and depression perception based on deep learning and street view image.

With the accelerated urbanization process, the impact of the urban environment on the mental health of residents has attracted extensive attention. However, previous research has primarily focused on single environmental factors, lacking a systematic quantitative framework based on multi-dimensional urban elements, especially at the refined scale, resulting in dual bottlenecks in methodology and data. In this work, a novel deep learning (DL) approach, driven by SVI, is developed to quantify depression perception (DP) between residents in Wuhan and the urban environment. Specifically, a database containing 133,114 SVIs is constructed, extracting six visual elements of green view index (GVI), sky view factor (SVF), degree of enclosure (DOE), degree of motorization (DOM), degree of non-motorization (DON), and rate of sidewalk (ROS). Stepwise regression is applied to elucidate which urban elements exacerbate or mitigate DP. Results indicate that GVI, SVF, and DON are negatively correlated with DP. Moreover, the risk of negative emotions is increased due to the high DOE and DOM, presenting a clear coupling relationship. Meanwhile, spatially, the city exhibits a pattern of high DP along transport-industrial corridors and low DP across blue-green and cultural corridors. Additionally, historical districts such as Tanhualin display a cultural buffering effect of "high-enclosure/low-DP". This study establishes a deep connection between street-level urban environments and DP, expands the "emotion-space" mechanism framework in the design of healthy cities theoretically, and provides quantitative support and an emotion-oriented strategy for urban renewal, street development, and the planning of green infrastructure in practice.