| 摘要: |
| 街道空间作为城市空间基本单元，
其更新效能是国家治理现代化的重要维度。
为量化城市街道空间质量与活力的匹配关
系，支撑街道精细化治理，以武汉市中心城
区为例，构建“量化评价—动态匹配—多维
解耦”技术框架。结合街景图像与社交媒体
数据测度街道空间质量及城市活力，并以此
为基础测算两者之间的匹配关系，利用多分
类Logistic 回归模型解析影响因素。结果表
明：武汉市中心城区72.1%的街道存在空间
质量与城市活力的错配现象。交叉口密度、
区域平均房价、夜间灯光指数、POI 混合度
等指标，可有效抑制匹配关系的不良转向。
研究从微观尺度揭示街道空间“供给—需
求”的结构性矛盾，为街道规划从“追求效
益”的单一空间扩展向“承载价值”的多维
提升转型提供策略支持。 |
| 关键词: 街道空间质量 城市活力 街景图
像 影响因素 深度学习 |
| DOI：10.13791/j.cnki.hsfwest.20250403002 |
| 分类号: |
| 基金项目:国家自然科学基金面上项目（52278062）；国家自然科学基金青年科学基金项目（C类）（51708233） |
|
| Research on the matching relationship between street space quality and urban vitality andinfluencing factors: Taking Wuhan City center as an example |
|
SHAN Zhuoran,ZHANG Xinyue,YUAN Man,CHEN Yuedi
|
| Abstract: |
| In face of the process of rapid urbanization and global pursuit of sustainable development,
renewal and quality enhancement of urban street space becomes increasingly critical. Internationally,
studies by scholars such as Jane Jacobs and Jan Gehl have long emphasized the role of streets as vital
public realms that foster social interaction, economic vitality, and cultural expression. In China, aligned
with the national strategic vision for urban regeneration and high-quality development, street space
transformation has become an essential component of modern urban governance. However, a
noticeable paradox persists across global cities: substantial investments in physical upgrades often fail
to translate into sustained urban vitality. This disjunction between spatial supply and user demand
reflects deeper structural and contextual imbalances that necessitate systematic analysis and contextsensitive
policy responses. Taking the central urban area of Wuhan as a case study, this research
constructs a technical framework entitled “quantitative evaluation-dynamic matchingmultidimensional
decoupling”. A comprehensive set of multi-source big data was employed. Baidu
Street View imagery was used to quantify street space quality, while Sina Weibo check-in data served
to measure urban vitality levels. These datasets were integrated into a geospatial modeling platform to
support fine-grained analysis at the street segment level. To extract the multidimensional perceptual
indicators from street view images, Deep Learning (DL) was employed with a pre-trained
DeepLabV3+ model trained on the Cityscapes data set. Perceptual indicators within the view included
5 main dimensions including walkability (sidewalk width, crossing facilities), visual appeal (green
view index,) sky openness, spatial comfort (visual enclosure, building interface continuity), traffic
inclusiveness (pedestrian-vehicle separation), and interface activities (street wall continuity, groundfloor
transparency) were used as indexes of street space quality. Furthermore, urban vitality was
measured in terms of the density and distribution of geotagged social media check-ins. Such data
proxies represent human activities and street-level behavior to varying extents. To examine the
correspondence between spatial supply and behavioral demand, a matching index was developed based
on normalized rankings of street space quality and urban vitality. This index enabled the classification
of street segments into four distinct matching types: high-quality – high-vitality (H-H), low-quality –
high-vitality (L-H), high-quality–low-vitality (H-L), and low-quality–low-vitality (L-L). Furthermore,
a multinomial logistic regression model was utilized to identify key factors influencing these matching
types from a multidimensional perspective that incorporated built environment, economic
development, and social context variables. Specific metrics included building density, intersection
density, regional housing price, nighttime light intensity, and density of convenience facilities. The
results reveal a significant proportion of mismatch (72.1%) between street space quality and urban
vitality in Wuhan’s city center, indicating widespread disconnection between spatial supply and user
demand. H-H streets (27.9%) were predominantly located in mixed-use central areas such as JianghanRoad Pedestrian Street and Zhongnan Road-Wuluo Road commercial districts, where high functional diversity and accessibility supported synergistic qualityvitality
integration. L-H types (22.26%) clustered in historic inner-city neighborhoods like Wuchang Ancient City and old concession areas, as well as urban
villages such as Hua’anli, where cultural heritage, social networks, and economic informality compensated for physical deficiencies. H-L streets were frequently
found in newly developed waterfront zones and high-green-coverage suburbs, where superior physical conditions were undermined by monofunctional land use,
inadequate daily facilities, and poor connectivity. L-L segments typically appeared in peripheral industrial zones and urban-rural fringes characterized by poor
accessibility, resource deprivation, and functional isolation. The regression analysis identified several factors that effectively inhibit undesirable matching
outcomes. Higher intersection density, average housing price, nighttime light intensity, POI mix, and density of convenience facilities. These factors collectively
enhance street connectivity, economic vitality, functional diversity, and service accessibility, thereby promoting better alignment between space quality and usage
intensity. Conversely, and somewhat counterintuitively, greater accessibility to public green spaces was found to promote mismatch. This suggests that large,
poorly integrated green patches may disrupt street connectivity, reduce commercial visibility, and hinder vibrancy, highlighting the importance of strategic
landscape integration rather than mere proximity. This paper contributes to literature through 3 innovative parts: 1) introducing street view image segmentation
(via SVI and DL) to extract 3D visual indicators to evaluate quality of the street space; 2) A ranking-difference matching index to quantify space-vitality
relationships, identify mismatches, and break linear “high vitality-high quality” reliance; 3) Integrating built environment/economic/social factors, using
multinomial logistic regression to reveal mismatch drivers and inform multidimensional indicator coordination in stock renewal. |
| Key words: street space quality urban vitality street view images influencing factors deep learning |
