The integration of UAV, deep learning, and GIS in the assessment of a new neighborhood concept

Abstract

Built environments of a neighborhood are significant for promoting physical activities and preventing diseases. In quantitative research for built environments at the neighborhood scale, the operational unit neighborhoods should be clearly predefined. Most of the time, census geography is the surrogate neighborhood unit in health research. However, neighborhood boundaries based on census geography can hardly respect the social and spatial dimensions of the residents of a neighborhood. A recently proposed concept, sidewalk-homogenous neighborhoods, attempts to move beyond census geography and operate as a measurable neighborhood unit that captures the economic and behavioral components of residents for community health research. This paper evaluated this newly proposed neighborhood concept by assessing built environments among different residential communities using the integration of unmanned aerial vehicle (UAV) images, deep learning, and Geographic Information Systems (GIS). The applicability of the neighborhood concept was tested in four residential areas at different economic levels in Northeast Ohio of the United States. The study addresses that the sidewalk-homogenous neighborhoods concept can help identify the spatial disparity of built environments between neighborhoods at different economic levels. The study also reveals the inequality in UAV research opportunities between economically advantaged and disadvantaged neighborhoods.