图像语义分割技术下园林要素识别与视觉感知关联性研究  

Relationship Between Garden Element Recognition and Visual Perception Under Image Semantic Segmentation Technolog

作　　者：刘文洁 黎欢 刘文雪 孙乔昀 Liu Wenjie;Li Huan;Liu Wenxue;Sun Qiaoyun

机构地区：[1]深圳大学建筑与城市规划学院 [2]深圳大学美丽中国研究院 [3]澳门科技大学计算机科学与工程学院,中国澳门 [4]福建农林大学风景园林与艺术学院

出　　处：《华中建筑》2024年第11期138-142,共5页Huazhong Architecture

基　　金：广东省基础与应用基础研究基金“国家公园景观视觉感知与游憩行为关联机制研究”(编号:2023A1515110856);广东省哲学社会科学规划项目“国家公园导览地图表达方法与绘制标准研究”(编号:GD22YYS09)。

摘　　要：园林环境的物理特性与公众视觉感知体验密切相关,深入分析园林构成要素对视觉感知的微观机制,是理解游客视觉感知体验的关键途径。研究以9座园林为案例,运用图像语义分割技术、语义差异分析(SD)等量化方法,探究16项感知指标与园林要素间的关联。研究发现,不同地区的园林在视觉感知体验上展现出独特的地域风格,尤其体现在自然与人工设计相融合的园林要素对视觉美感的显著提升作用。然而,园林内部的主路径规划与低层植被区域在视觉感知评估中得分较低,主要归因于其视觉特征的单一性和辨识度不足等。此研究通过量化分析,揭示了园林内部不同构成要素在视觉感知评价上的异质性效应,进一步深化对园林感知机制的理解。The physical characteristics of garden environments are closely related to the public's visual perception experience.A nalyzing the micro-mechanisms of how garden elements affect visual perception is a key approach to understanding the core elements of visitors'visual perception experiences.Taking nine gardens as case studies,we employ quantitative methods such as image semantic segmentation and semantic differential analysis(SD)to explore the relationship between 16 perception indicators and garden elements.It is found that gardens in different regions exhibit significant visual characteristics,particularly in how the integration of natural and artificial design elements significantly enhances visual aesthetics.However,the main path planning and low-layer vegetation areas within the gardens scored lower in visual perception evaluations,mainly due to their monotonous visual features and lack of distinctiveness.Through quantitative analysis,we reveal the heterogeneous effects of different garden elements on visual perception evaluation,further deepening the understanding of garden perception mechanisms.

关 键 词：园林环境 视觉感知 图像语义分割 关联性 评价 

分 类 号：TU984.119[建筑科学—城市规划与设计]