Decoding the built environment-behavior nexus at residential boundaries: A multi-dimensional analysis of 85 street segments in Dalian

Abstract

Boundaries in communities often become sites of conflict and “Not In My Backyard” (NIMBY) effects, reducing the efficiency of public space utilization, especially in older residential areas. However, by re-envisioning boundaries as potential public spaces and exploring their interactions with various stakeholders, these negative effects can be mitigated. This study analyzes multi-source data from 85 street segments in Dalian, China, to examine these interactions with the urban environment. A VIM-VDM model was developed to quantify interaction intensity, using two behavioral patterns—traversal and stationary activities—as dependent variables. Through machine learning-based semantic segmentation of street view images, 24 environmental factors across five dimensions were incorporated into a bivariate LOESS regression model to assess their impact on three boundary interaction types. Using Python's OpenCV library, population trajectory maps were generated and targeted strategies for two boundary types—H-type and I-type—were proposed. Results show that increasing entrance density and sidewalk width significantly enhance interactions in Fenced Type boundaries, while shop density, transparency, and seating improve interactions in Building-service Type boundaries. Factors like street length and height-width ratio suppress interactions in Building-life Type boundaries, whereas convex space ratio and build-to-line rate promote them. These findings suggest shifting urban renewal focus from “isolated centers” to “interactive boundaries” to foster shared spaces and resources.