基于人机对抗模型的街景特征与可步行性关系研究  

Exploring the Relationship between Streetscape Characteristics and Walkability Based on Human-Machine Confrontation Model

作　　者：黄敏虹 谢伊文 陈嘉林 刘京一 HUANG Minhong;XIE Yiwen;CHEN Jialin;LIU Jingyi

机构地区：[1]华南农业大学,广州510642

出　　处：《广东园林》2024年第1期33-40,共8页Guangdong Landscape Architecture

基　　金：广州市科技局基础与应用基础研究项目(编号:2023A04J2007)。

摘　　要：基于涉及街道景观要素与街道可步行性关系的研究还存在一定缺失,以广州市天河区代表性综合街区华师-体育中心片区为对象,对城市街区的可步行性进行研究。综合马斯洛需求层次理论和“步行需求层次”模型,选取安全性、舒适性、社交性、美观性为指标建立街景特征评价指标体系。通过Insta360 ONE X2全景相机采集、图像语义切割、人机对抗模型评分、SPSS分析(Pearson相关性分析、多元线性回归分析)等多种方法,研究街景特征与可步行性关系。结果表明,4类街景特征与可步行性均具有一定相关性,其中,舒适性方面的绿视率对街道可步行性的影响最为显著。基于分析结果,街道空间可以从安全性、舒适性、社交性和美观性的街景特征层面进行步行性优化提升,如在安全性方面,建议增设护栏或绿化带,增设、整合标识牌等。There still needs more research on the relationship between street landscape elements and street walkability.Combining Maslow's hierarchy of needs theory and the'walkability hierarchy'model,safety,comfort,sociability,and aesthetics were selected as indicators to establish a street scene feature evaluation index.Through methods such as collecting data with the Insta360 ONE X2 panoramic camera,image semantic segmentation,adversarial model scoring,and SPSS analysis(Pearson correlation analysis,multiple linear regression analysis),the relationship between street scene features and walkability was studied.The results show that the four types of street scene features are all correlated with walkability,with the green view rate in terms of comfort having the most significant impact on street walkability.Based on the results of the analysis,street spaces can be optimized for walkability from the perspective of safety,comfort,sociability,and aesthetics of street scene features,such as suggesting the addition of fences or green belts,and integrating signage.

关 键 词：街景图像 语义分割 人机对抗 街景特征评价 

分 类 号：TU986[建筑科学—城市规划与设计]