The Impact of Visual Character on Perceived Stress Levels: An Intelligent Approach Applied to University Campus Design

Various types of streetscapes have been the subject of past research, with university campus planning being identified as one example where a psychological impact has been observed. However, due to the visual complexity of campus streetscapes, little or no clear approach is available to quantitively assess their potential impacts. Furthermore, collecting empirical data about environmental stress levels in urban spaces remains a significant challenge. In response, this chapter presents an “intelligent” approach to estimating the relationship between street view imagery (SVI) properties and perceived stress in seven university campuses in China. Specifically, an automatic, semantic segmentation method is used to measure the visual properties of 6056 SVIs—the visual element proportions (VEPs) of design elements and visual features. Then, a human–machine adversarial model using a random forest is applied to predict the perceived stress scores (PSSs) of SVIs. Through this combination of a computer vision technique and machine learning, this research identifies the various impacts of visual elements on PSSs. The research also tests the significance of three visual features holistically contributing to lower stress levels in campus design. This chapter concludes with a discussion of the findings and a contribution to architectural and urban studies.