广州泮塘社区街道空间品质感知评价研究OACHSSCD
Perception and Evaluation of Street Space Quality in Puntoon Community,Guangzhou
在"老城市新活力"与"人民城市"理念驱动下,以广州泮塘社区为对象,探究使用者对社区街道空间偏好及其影响因素,以期为老旧社区的街道空间更新提供科学化、人本化的测度方法,研究综合图像识别技术与可视化SP法,构建"物理特征识别-偏好模型构建-效用优化"的闭环框架,为社区街道空间品质提升提供量化路径.采用人视角拍摄泮塘社区21条街道的368张街景图像,借助PSPNet模型与ADE20K数据集进行语义分割,通过K-means聚类量化视觉特征,并划定八项评价要素的水平等级.基于此设计可视化SP问卷,运用混合Logit模型构建街道空间效用模型,发现色彩丰富度(RI=17.16%)与界面透明度(RI=17.13%)是影响偏好的核心视觉要素;社交活跃度、空间舒适性等次之,而设施便利性影响最小.同时亚组分析表明,居民更关注社交功能,游客则对设施便利性与空间舒适性更敏感.据此针对T1-T3三类街道样本提出优化策略,提高街道样本的效用,为老旧社区的街道空间更新提供了可行的量化路径和实践参考.
Guided by the concepts of"New Vitality for Old Cities"and"People's Cities"in Guangzhou,this study found that renewing old communities requires a balance between preserving the cultural context and meeting the demands of modern life.The Puntoon Community in Guangzhou,as a micro-renovation pilot project,faces many issues.These include redefining the boundaries between public and private spaces in the street areas during the transformation process.This is a typical case for research as the study integrates image-recognition technology with visualized stated preference to construct a research framework of"physical feature recognition preference model construction utility optimization",providing a quantitative approach to enhancing the quality of community street spaces. The study collected a total of 368 street-view images of 21 streets in the Puntoon Community,taken from the perspective of pedestrians.Semantic segmentation was conducted by combining the PSPNet model and the ADE20K dataset.Visual features were quantified using the K-means clustering algorithm,and eight evaluation elements,including walking safety and color richness,were categorized into distinct levels.Based on this,a visual SP questionnaire was designed,and 327 sample data were analyzed using the mixed logit(MXL)model.The findings revealed that color richness(RI=17.16%)and interface transparency(RI=17.13%)were the core visual elements that influence preferences,followed by social activity and spatial comfort;facility convenience had the least impact.Subgroup analysis indicated that residents were more interested in social functions,and tourists were more sensitive to facility convenience and spatial comfort.The study classified streets into three types through utility evaluation:T1,T2,and T3.T1 streets have insufficient pedestrian safety and attractiveness.T2 streets lack facilities and greenery.T3 streets need to improve interface transparency and social vitality.Hence,some targeted improvement measures are proposed,including implementing traffic diversions and adding social spaces in T1 streets,optimizing the signage and greenery in T2 streets,and enhancing social scene creation in T3 streets.After the renovation,the space utility of the streets significantly improved. The subjective and objective integrated evaluation systems in this study provide a quantified basis,based on user preferences,for renewing streets in old communities.In the future,it can further promote the sustainable renewal of blocks by combining residents'emotional needs with community cohesion.
向科;俞星瑶
华南理工大学建筑学院、亚热带建筑与城市科学全国重点实验室华南理工大学建筑学院、亚热带建筑与城市科学全国重点实验室
建筑与水利
街道感知评价图像识别可视化SP广州泮塘社区
streetperception evaluationimage recognitionvisualized stated preferencePuntoon Community,Guangzhou
《南方建筑》 2026 (3)
50-61,12
广东省哲学社会科学规划项目(GD24CYS39):人民城市理念下主观感知导向的岭南特色城市风貌研究广东省自然科学基金项目(2024A1515012576):低碳导向的亚热带建筑典型过渡空间"行为-气候"耦合的空间表征参数研究.
评论