基于街景数据的城市街道感知特征空间异质性研究——以西安市为例

Spatial Heterogeneity of Urban Street Perception Characteristics Based on Street View Data:A Case Study of Xi′an,China

理解城市街道环境感知特征对城市街道规划和更新具有重要参考意义,目前多利用街景数据研究城市街道环境特征,但缺少基于多维街道感知特征的城市街道分类分析。该文以西安市为例,基于百度街景影像语义分割结果,从绿色度、开放性、包围性、步行友好性、成像性5个维度定量研究街道的感知特征,分析各个维度感知变量的空间异质性特征;以5个维度的感知变量构建描述街道环境的感知向量,利用K-Means聚类算法对街道感知向量进行聚类,将具有相似感知特征的街道分类。结果表明:①西安市街道感知特征具有空间分异性,其中绿色感知热、冷点区域交错分布,开放性和成像性空间分布均具有渐变趋势,包围性和步行友好性呈分散团簇状分布;②基于感知特征向量可将街道分成4类,其中包围性和步行友好性类别间差异较小,其余感知变量类别间差异显著,C1类街道绿色度感知最强,C2类街道开放性感知最强,C3类街道成像性感知较强,C4类街道各感知变量差异较小。研究结果有助于理解城市街道感知的空间分异特征,为城市街道规划与更新提供参考依据。

Understanding the characteristics of urban street environment perception has important reference significance for both urban planning and renewal.Street view data has become an important data source for studying urban street environment features at present.Previous researches mainly evaluate street environment features(e.g.,greening,walkability)in a certain dimension by semantic segmentation of street view images.However,it lacks the analysis of urban street classification based on multidimensional street perception characteristics.This study takes Xi′an as an example,based on the semantic segmentation results of Baidu Street View images,quantifies the perceptual characteristics of streets from five dimensions,namely greenness,openness,enclosure,walkability and imageability,and analyzes the spatial heterogeneity characteristics of the perceptual variables in each dimension.On this basis,the perceptual vectors describing the street environment are constructed with the perceptual variables of the five dimensions.K-Means clustering algorithm is used to cluster the street perceptual vectors and categorize the streets with similar perceptual characteristics.The results are shown as follows.①Street perception in Xi′an presents spatial differentiation,in which the hot spots and cold spots of greenness perception are staggered,the spatial distribution of openness and imageability is gradual,and the enclosure and walkability are distributed in scattered clusters.②Streets can be divided into four categories based on perception characteristics,in which the difference between the categories of enclosure and walkability is small,while the difference between the other categories of perception variables is significant.The first category of streets(C1)has the strongest greenness perception,the second category of streets(C2)has the strongest openness perception,the third category of streets(C3)has strong imaging perception,and the difference between the perception elements of the fourth category of streets(C4)is small.The results of this study are helpful to understand the spatial heterogeneity of urban street perception and provide reference for urban street renewal and planning.