【作者简介】张延吉，福州大学社会学系，游永熠，华南理工大学建筑学院

【文章来源】《社会学研究》2026年第1期

【内容提要】本研究以2013—2022年中国296个地级及以上城市为分析对象，采用目标检测、图像回归等深度学习方法测量街景图像中的环境维护与富裕感知水平，以反映社区的社会经济水平并量化城市社会空间的全域及邻域分异。研究表明，在空间分布上，全域分异在中部地区最突出，邻域分异自东向西、自发达城市向欠发达城市依次降低；在发展趋势上，全域分异在沿海发达城市有所减弱，发达城市的邻域分异在高基数水平上持续上升。市场主体和政府政策的宏观环境与微观空间特征均对社会空间分异具有关键作用。

【关键词】社会空间分异；计算机视觉；基尼指数；空间基尼指数；双主体—双视角解释框架

【全文链接】https://shxyj.ajcass.com/Magazine/show/?id=121911

Current Status, Evolutionary Trends and Influencing Factors of Socio spatial Differentiation in Urban China: An Analysis Based on Street View Images Using Deep Learning

Abstract: This study uses object detection and image regression to measure environmental maintenance and perceived wealth across street view images from 296 prefecture-level (and above) cities in China between 2013 and 2022. These measurements reflect the socioeconomic status of each community and quantify both global and local socio spatial differentiation within each city. The findings indicate that in terms of spatial distributions, global differentiation is most pronounced in the central region, whereas local differentiation decreases from east to west and from developed to less developed cities. Interms of evolving trends, there has been a decline in global differentiation in some developed coastal cities,while local differentl iation has increased within developed cities from a high baseline level. Both the macro- environ mental and micro-spation characteristics of market actors and government policies play a crucial role in determining urban socio spatial differentiation.