Extracting human perceptions from street view images for better assessing urban renewal potential

Abstract

Accurate and efficient assessment of large-scale urban renewal potential is an indispensable prerequisite for managing and facilitating projects. However, few studies consider the built environment when assessing urban renewal potential because it is difficult to measure. Street view images can show the physical setting of a place for humans to perceive the built environment. Hence, we separately extracted emotional and visual perceptions from street view images to construct a new comprehensive indicator set to assess multi-class urban renewal potentials. To establish the assessment model, we applied a backpropagation neural network based on the presence and background learning (PBL-BPNN). The renewal potential assessment based on the proposed indicator set can reach the highest accuracy. Emotional perceptions contribute more to assessing renewal potential than visual perceptions because they are more consistent in portraying the blighted built environment. Emotionally, the ratings of safety, boring, depression, and lively are stable in the blighted built environment. Visually, greenness and imageability often remain at lower values, highlighting the importance of greenspace and urban furniture in determining urban renewal. Furthermore, multi-class renewal potentials can be used for scenario analysis by assuming different renewal intentions. The results can support governments and planners in making efficient urban renewal decisions.