Measuring and understanding changes in the physical built environment of cities with street view images

被引：0

作者：

Yang Zhou ^{[1
]}

Jean-Claude Thill ^{[2
]}

Xingjian Liu ^{[3
]}

Chen Zhong ^{[4
]}

Wei Tu ^{[5
]}

机构：

[1] Central China Normal University,Key Laboratory for Geographical Process Analysis and Simulation of Hubei Province, College of Urban and Environmental Sciences

[2] University of North Carolina at Charlotte,Department of Geography and Earth Sciences

[3] University of Hong Kong,Department of Urban Planning and Design

[4] The Bartlett Centre for Advanced Spatial Analysis (CASA),Key Laboratory for Geo

[5] University College London,Environmental Monitoring of Great Bay Area, School of Architecture and Urban Planning

[6] Shenzhen University,Guangdong Key Laboratory for Urban Informatics

[7] Shenzhen University,Shenzhen Key Laboratory of Spatial Information Smart Sensing and Services, and Department of Urban Informatics, School of Architecture and Urban Planning

[8] State Key Laboratory of Subtropical Building and Urban Science,undefined

[9] Shenzhen University,undefined

来源：

Urban Informatics | / 4卷 / 1期

关键词：

Built environment; Street view images; Urban changes; Street space;

D O I：

10.1007/s44212-025-00069-9

中图分类号：

学科分类号：

摘要：

Street view images (SVIs) may provide information on near-surface urban changes which are not necessarily captured by spaceborne remote sensing data. The application of SVIs in assessing diverse built environment changes at the street level and over time remains challenging. This paper presents a stepwise rule-based method to identify key types of urban built environment changes using multi-year SVIs. In particular, physical/built environment changes along streets are evaluated based on proposed Street Units of Analysis (SUA) that account for both street layouts and street view features. The approach employs sharp variations of visual attributes derived from deep learning segmentation model DeepLabv3+. A stepwise rule-based algorithm classifies SUAs. Using panoramic SVIs from 2015–2019 in Wuhan, China, we identify critical types of changes such as those related to highway bridges, sidewalk increases, building increases, road losses, greenness increases, and mixed changes. Identified changes take place on over 50% of roads in the study area. In addition, the robustness of proposed approach is assessed based on results produced by manual labeling and by a fuzzy rough sets analysis. The approach is found to be robust and effective by having an 81.7% agreement with manually labeled analysis and an 80.5% agreement with fuzzy rough sets analysis. Overall, this study contributes to the development of a cost effective and efficient method for detecting physical changes on SUAs, which can be further utilized in studies that link urban changes, space use, and policy interventions.

引用