Mapping approach for emotional response to urban visual environments based on street view images and EEG signals

Urban visual environment plays a critical role in city planning and public health by influencing residents’ emotional responses. However, existing studies rely on subjective assessments of small-scale areas, lacking objective, large-scale emotional maps. This study introduces a novel electroencephalogram (EEG)-based framework to quantify emotional responses to urban visual environments. First, urban visual indicators are extracted from 352,112 street view images (SVI) in Guangzhou and 680,280 in Shenzhen using a deep learning-based semantic segmentation. Then, EEG experiments are conducted with 24 participants exposed to four groups of SVI visual stimuli, employing machine learning models to quantify relationships between visual indicators and emotions. Finally, a series of models are integrated to generate city-wide visual emotional maps across four emotion dimensions and valence-arousal space. Results show that the models established between the visual environment and emotional responses display R² values in the range of 0.39 to 0.69, enabling visual emotional mapping. Correlation analysis further shows a higher green view index and color entropy significantly correlate with positive emotions (p < 0.01). Conversely, elevated building density and openness are linked to negative emotions (p < 0.01). In terms of HAPV (high-arousal, positive-valence) dimensions, Shenzhen had higher emotion scores than Guangzhou, with mean values of 0.237 and 0.226, respectively. These differences correspond to colorful urban center landscapes in Shenzhen and predominantly green landscapes in Guangzhou as a result of their different urban planning strategy. This study contributes to the optimization and transformation of urban landscapes for improved visual comfort.