摘要
城市街道景观是城市整体景观的重要组成部分,直接影响人们对城市环境的感知与评价。以天津市鼓楼街道为例,利用街景图像数据和LaDeco景观视觉评估模型,深入分析城市街道景观的视觉要素,包括自然和人工景观的比例、主要视觉要素、形态特征及景观特征的粒度信息。结果表明,鼓楼街道以人工景观为主,自然景观不稳定且不连贯;主要视觉要素包括街道、天空和建筑,LaDeco模型能有效识别这些要素;道路、天空和建筑是街道景观的主要形态特征,其中道路占比较高,而建筑和天空影响着城市环境的审美感知;景观的布局、比例和相对位置共同构建了整体视觉印象,对街道形象和城市环境产生重要影响。提供了一种有效的评估方式,能快速、客观地计算图像中视觉要素的比例,为城市街道景观的研究、规划和评估提供新的思路和方法。
Urban streetscape holds significant importance within the broader cityscape,exerting a direct influence on individuals’perception and evaluation of the urban environment.The paper centers on Gulou Street in Tianjin,employing the streetscape image data and the LaDeco landscape visual evaluation model for an in-depth analysis of the visual components of urban streetscape,including the proportion between natural and artificial landscapes,major visual components,morphology features and granular details of landscape feature.Findings reveal that Gulou Street is predominantly characterized by artificial landscapes,with an unstable and disconnected expression of natural landscapes.Major visual components encompass streets,the sky and buildings,effectively discerned by the LaDeco model.Roads,the sky and buildings constitute the principal morphology feature,with roads holding a higher proportion.The sky and buildings significantly influence the aesthetic perception of the urban environment.The configuration,scale and relative positioning of landscape collectively shape the overall visual impression,profoundly influencing the street identity and the urban environment.The paper introduces an efficient assessment method for rapidly and objectively quantifying the proportion of visual components within an image,presenting novel ideas and methodologies for the research,planning and evaluation of urban streetscape.
作者
邹尚恩
林朝
杨瀚文
胡一可
TSOU Shangen;LIN Zhao;YANG Hanwen;HU Yike
出处
《景观设计》
2024年第1期22-25,共4页
Landscape Design
基金
国家自然科学基金重点项目“基于中华语境‘建筑-人-环境’融贯机制的当代营建体系重构研究”(52038007)。
关键词
城市街道
景观评估
街景图像
深度学习
景观视觉要素
urban street
landscape assessment
streetscape image
deep learning
landscape visual component
