基于街景图像的旅游地街道空间视觉品质评价方法  

Street Space Visual Quality Evaluation Method of Tourism Sites Based on Street View Images

作　　者：黄竞雄 梁嘉祺 杨盟盛 李渊[1,3] HUANG Jingxiong;LIANG Jiaqi;YANG Mengsheng;LI Yuan(School of Architecture and Civil Engineering,Xiamen University,Xiamen 361005,China;School of Architecture,Tsinghua University,Beijing 100084,China;Xiamen Key Laboratory of Integrated Application of Intelligent Technology for Architectural Heritage Protection,Xiamen 361005,China)

机构地区：[1]厦门大学建筑与土木工程学院,厦门361005 [2]清华大学建筑学院,北京100084 [3]厦门市建筑遗产保护智能技术集成应用重点实验室,厦门361005

出　　处：《地球信息科学学报》2024年第2期352-366,共15页Journal of Geo-information Science

基　　金：国家自然科学基金项目(42171219、52078444);福建省自然科学基金项目(2020J01011)。

摘　　要：街道空间是旅游者在旅游地的主要接触空间,良好的街道视觉品质对旅游产业的可持续发展至关重要,而如何建构视觉品质评价方法则需要进一步探索。本文以厦门市鼓浪屿作为研究区域,首先建立街道空间视觉品质量化模型,采用人工模拟旅游者行进的方式采集街景数据,进行成像参数校正和图像编码;其次,采用全卷积网络进行图像语义分割和视觉要素提取,获得街道环境视觉参数;最后,根据量化模型将视觉参数聚合到地理要素中,建立地理信息数据库进行街道空间视觉品质评价。结果表明:(1)鼓浪屿街道空间视觉品质存在明显的空间分异性;(2)建筑密度、街道宽度与植被小品是塑造街道空间视觉品质的基础视觉要素;(3)植物类公园、主要交通枢纽、主要商业设施的分布对街道空间视觉品质产生显著影响。本文为街道视觉品质评价工作提供了可供迁移应用的数据采集方式和研究方法,基于全卷积网络的街景图像语义分割精度较高,能够为街景图像以及其他类型的图像数据分析提供参考。在管理实践上,本文可以为旅游地街道空间的管理与规划、资源整合与配置、人流引导与调控等工作提供有价值的借鉴。Street space is the main space that affects tourists' experience of tourism sites.The visual quality of street space is crucial to the development of tourism.However,the evaluation method of visual quality needs further exploration.This paper selected Gulangyu,the famous tourism site in Xiamen,as a study case.First,we established a quantitative model of visual quality combining the existing research on street space and the visual elements of tourism sites.Then,we collected street view data of each intersection by traveling like tourists,corrected imaging parameters,and encoded the street view images.Second,based on the deep learning method(Fully Convolutional Networks,FCN),we segmented the collected street view images semantically and extracted the visual elements in street view data.Finally,by combining with GIS,we set up a geographic information database to analyze the visual and spatial characteristics of each sampling point's visual elements.This database was aimed at providing a basis for further evaluation of the visual quality of street space.It was aggregated using the street line as the smallest unit.In our study,we calculated the visual quality indicators to evaluate the street space in Gulangyu.The results show that:(1) There is obvious spatial differentiation in the visual elements of street space in Gulangyu;(2) Building density,street width,and vegetation sketches are the basic visual elements that shape the visual quality of street space;(3) The distribution of botanical parks,major docks,and commercial facilities significantly impacts the street space's visual quality.In detail,green plants,buildings,roads,sky,and street facilities show the differences between a center and a roundabout.While the distribution of pedestrians shows differences between the east and the west.The green view rate,enclosure,sky openness,and diversity of street space also have obvious center-roundabout spatial differentiation.Moreover,there is an obvious spatial agglomeration effect in the green view rate,crowding degree,a

关 键 词：街景图像 全卷积网络 街道空间 视觉要素 视觉品质 评价方法 地理信息系统 鼓浪屿 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TU984.113[自动化与计算机技术—计算机科学与技术]