Evaluation of Street Space Renovation in Historic Areas Using Deep Learning Based on Street View Imagery in the Human Visual Field  

作　　者：Zhu Xiaotong Bai Mei Bai Yuxin Li Min(Translated) Liu Jiayan(Proofread) 

机构地区：[1]School of Architecture and Art,Hebei University of Engineering,Handan,P.R.China [2]不详

出　　处：《China City Planning Review》2024年第4期25-34,共10页城市规划（英文版）

摘　　要：Regular evaluation of street space is essential for achieving sustainable development and dynamic maintenance of historic areas.Although quantitative evaluations using street view imagery are precise and efficient,they often fall short in capturing pedestrians’visual experience,largely because images are collected from vehicles.Accordingly,this paper acquires street view imagery in the human visual field before and after the street space renovation by adjusting relevant parameters,and performs image semantic segmentation.From a pedestrian’s viewpoint,the paper develops street space evaluation indicators across four dimensions:comfort,identity,diversity,and walkability.The mean square deviation method is applied to assign weights to these indicators,enabling a comprehensive evaluation of street space in historic areas.In addition to evaluating the renovation results,it proposes improvement suggestions that may provide insights into the evaluation practices of street space renovations in historic areas and contribute to improving street space quality.

关 键 词：street space human visual field street view imagery historic areas deep learning 

分 类 号：TU984.113[建筑科学—城市规划与设计]