Using google street view to reveal environmental justice: Assessing public perceived walkability in macroscale city

Abstract

Walkability is an important issue in urban planning equity, which is primarily influenced by the objective environment and subjective perception. However, assessing the objective environment of the city on a large scale or obtaining the general public’s perceptual evaluation of the urban environment with less cost is challenging. This research adopted two-stage studies to identify the relationship between streetscape elements and perceived walkability using Google Street View images with machine learning in Taipei City. In Study I, the Zhongzheng District in Taipei was selected as the sample area and successfully developed a walkability prediction model that integrates street elements segmented by semantic segmentation technique, with the perceived walkability. Roads and terrains were identified as key predictors that affect perceived walkability in this random forest regression model. Based on this prediction model, we expanded the walkability assessment citywide with semantic segmentation in Study II, and the citywide walkability maps were produced. Accordingly, the socio-spatial equality of walkability was further audited. We further adopted spatial linear regression analysis to examine the relationship between neighborhood socioeconomic indicators (individual income, elderly %, and less educated %) and perceived walkability. Through the geographically weighted regression analysis, the results indicated that situated in peripheral areas were more sensitive to local socioeconomic indicators. This highlights the significant presence of social-spatial vulnerability in the walkability of Taipei City. Our research demonstrated the feasibility of using machine learning to audit urban socio-spatial justice from urban micro-to-macro scales.