基于街景数据的街道空间品质评价思路优化及实证研究——以兰州市城关区为例

The Optimized Idea of Street Space Quality Evaluation and Empirical Studies Based on Street View Data——Taking Lanzhou City Chengguan District as an Example

为寻求利用街景数据从人本尺度进行街道空间品质评价的思路方法,研究通过构建步行安全性、街道舒适性、空间交往性三大评价指标,利用基于深度学习的全卷积神经网络进行街景图片语义分割,以兰州市城关区街道空间为例进行实证研究.结果表明:1)街道客观空间品质方面:城关区街道空间绿视率以较低水平及低水平分布为主,整体上街道绿视率低;大部分开敞度高的街道绿视率低;街道空间开敞度、绿视率与城市功能相关,以市民活动为主要功能的街道开敞度、绿视率更高.2)街道空间人群感知效应方面:市民经常活动的地方街道空间步行较为安全且具有更高的空间舒适性;街道综合品质评价结果的空间分布与空间交往性、街道舒适性具有高度一致性;兰州市城关区街道空间综合品质较低,具有较大提升空间;兰州市火车站街道空间品质水平居中,具有较好的提升潜力.

In order to find a way to evaluate the quality of street space from a human scale by using streetscape data,the study uses three evaluation indexes:pedestrian safety,street comfort and spatial interaction,a full convolutional neural network based on deep learning for semantic segmentation of streetscape images is adopted,and an empirical study on the street space of Chengguan District,Lanzhou City is conducted.The results show that:1)the objective spatial quality of the streets:the green view rate of the street space in Chengguan District is mainly distributed at a low level,and the overall green view rate of the streets is low;the green view rate of most streets with high openness is low;The degree of street space openness and green visibility are related to urban functions,those streets with civic activities as the main function have higher degree of openness and green visibility.2)Perception effect of crowd in street space:the street space in places where people frequently move is safer on foot and has higher spatial comfort;the spatial distribution of the evaluation results of the comprehensive street quality is highly consistent with spatial interaction and street comfort;The comprehensive quality of street space in the Chengguan District of Lanzhou is low and has much room for improvement;Lanzhou East Railway Station neighbourhood has a medium level of spatial quality,with good potential for improvement.