This study systematically analyzes the security perception mechanism of underground public spaces by comparing the visual environmental elements of new and old subway stations in Milan (gradual urbanization) and Beijing (rapid expansion). Using semantic segmentation technology combined with 360-degree image acquisition, ADE20K and YOLOv8, combined with volunteer scores and XGBoost-SHAP machine learning model, the study found that artificial lighting, ground openness and population density are common core elements to enhance the sense of security, while wall enclosure and infrastructure aging (such as exposed pipelines in old Beijing stations and damaged facilities in old Milan stations) significantly reduce the sense of security. In terms of urban differences, the platform doors of Milan's new station have become a key positive factor due to technological innovation. Although the new station in Beijing is modernized, hidden dangers are exposed due to uneven lighting; the renovation of old stations requires targeted strategies (such as the addition of commercial facilities in Milan and hidden pipelines in Beijing). Research and innovatively propose the "visual-space-perception" ternary analysis framework, providing cross-cultural methodological support for underground space design and urban transportation system optimization, emphasizing the dynamic adaptation needs of security perception under the differences in urbanization modes.
研究内容
Research contents
聚焦于地下公共交通空间的安全感知差异，以米兰和北京的地铁站为案例，系统探讨了新旧地铁站中视觉环境要素对乘客安全感的影响机制，并揭示不同城市化模式（渐进式发展 vs. 快速扩张）对地铁系统设计与安全感知的深层次关联。具体研究内容包括：
新旧地铁站的视觉环境对比
时间维度：米兰地铁系统自1960年代逐步扩展，旧站多建于2000年前，新站集中于21世纪后（如2013年开通的5号线和2022年的4号线）；北京地铁在2000年前仅2条线路，2008年奥运会后进入快速建设期，新站多建于2014年后。
设计特征：旧站普遍存在基础设施老化（如米兰的陈旧照明、北京裸露管道）、空间封闭性强等问题；新站则注重现代化设计（如米兰的站台门、北京的高开放性地板）。
视觉要素的分类与影响分析
功能性要素：人工照明、地面开放度（地板面积占比）、站台门等对安全感有显著正向作用，但北京新站部分换乘通道因照明不均仍存在隐患。
社会性要素：人群密度、商业设施（如自动售货机）在米兰旧站中提升安全感，而在北京旧站因设施维护不足反而引发负面感知。
秩序性要素：墙壁占比高（限制视野）、裸露管道（北京旧站）、散落垃圾和破损设施（米兰旧站）均显著降低安全感。
城市化模式的差异化影响
米兰：长期渐进发展使旧站改造更注重局部更新（如增加商业设施），而新站通过技术创新（如站台门）提升安全体验。
北京：快速扩张导致新旧站差异悬殊，旧站遗留问题（如管道外露）难以通过简单翻新解决，新站虽现代化但仍存在设计一致性不足的问题（如照明标准不统一）。
跨文化安全感知的异同
共性：开放空间、充足照明和人群存在是两地乘客共同关注的安全要素。
差异：米兰乘客对站台门等新设施的敏感性更高（因历史缺位），而北京乘客更易受物理环境无序性（如管道外露）的影响，反映出文化背景对安全认知的塑造作用。
理论框架构建
研究提出“视觉-空间-感知”三元分析模型，强调地下空间的视觉可达性、社会监控度与环境秩序是安全感知的核心维度，为未来地下公共空间设计提供了理论依据。
Focusing on the differences in safety perception in underground public transportation spaces, taking subway stations in Milan and Beijing as cases, the mechanism of influence of visual environmental elements on passengers' sense of security in old and new subway stations is systematically explored, and the deep correlation between different urbanization models (gradual development vs. rapid expansion) on subway system design and security perception. Specific research contents include:
Comparison of visual environment between old and new subway stations
Time dimension: The Milan subway system has gradually expanded since the 1960s. Most of the old stations were built before 2000, and the new stations were concentrated after the 21st century (such as Line 5 opened in 2013 and Line 4 opened in 2022); Beijing Metro only had two lines before 2000, and entered a period of rapid construction after the 2008 Olympics. Most of the new stations were built after 2014.
Design characteristics: Old stations generally have problems such as aging infrastructure (such as Milan's old lighting, exposed pipelines in Beijing), and strong space enclosure; new stations focus on modern design (such as Milan's platform doors and high open floors in Beijing).
Classification and impact analysis of visual elements
Functional elements: artificial lighting, floor openness (floor area proportion), platform doors, etc. have significant positive effects on the sense of security, but some transfer channels at Beijing New Station still have hidden dangers due to uneven lighting.
Social factors: crowd density and commercial facilities (such as vending machines) enhance the sense of security in the old Milan station, while in the old Beijing station, negative perceptions arise due to insufficient maintenance of facilities.
Elements of order: high wall proportion (limited field of view), exposed pipes (old Beijing station), scattered garbage and damaged facilities (old Milan station) all significantly reduce the sense of security.
The differentiated impact of urbanization models
Milan: Long-term and gradual development has made the renovation of old stations pay more attention to local renewal (such as adding commercial facilities), while the new stations improve the safety experience through technological innovation (such as platform doors).
Beijing: Rapid expansion leads to huge differences between old and new stations. Legacy problems (such as exposed pipelines) of old stations are difficult to solve through simple renovation. Although the new station is modern, there are still problems with insufficient design consistency (such as inconsistent lighting standards).
The similarities and similarities of cross-cultural security perception
Common features: Open space, sufficient lighting and crowded existence are safety elements that passengers in both places are concerned about.
Difference: Milan passengers are more sensitive to new facilities such as platform doors (due to historical absence), while Beijing passengers are more susceptible to disorder in the physical environment (such as exposure to pipelines), reflecting the role of cultural background in shaping safety perceptions.
Theoretical framework construction
The research proposes a ternary analysis model of "visual-space-perception", emphasizing that the visual accessibility, social monitoring and environmental order of underground space are the core dimensions of security perception, and provides a theoretical basis for future underground public space design.
研究方法
Research methods
数据采集：使用360度全景相机拍摄地铁站关键区域（如站台、安检口），并将图像转换为多视角照片。
语义分割：结合ADE20K数据集和YOLOv8模型，分割图像中的视觉要素（如灯光、人群、管道等），计算各要素的面积占比。
感知评分：招募两地长期居民作为志愿者，通过图形界面为图像的安全感评分（0-100分）。
建模与分析：采用XGBoost模型构建回归模型，分析视觉要素与安全感的关系，并通过SHAP值解释各要素的影响
Data acquisition: Use a 360-degree panoramic camera to capture key areas of the subway station (such as platforms and security checkpoints) and convert
the images into multi-view photos.
Semantic segmentation: Combining the ADE20K dataset and the YOLOv8 model, visual elements in the image (such as lights, crowds, pipelines, etc.) are
segmented, and the area proportion of each element is calculated.
Perception score: Recruiting long-term residents of the two places as volunteers, and rating the security of the images through a graphical interface (0-100
points).
Modeling and analysis: The XGBoost model is used to construct a regression model, analyze the relationship between visual elements and sense of security,
and explain the influence of each element through SHAP values.
研究结果
Research conclusions
共性因素：人工照明、地面开放程度和人群存在对安全感有显著正向影响，而墙壁、垃圾和破损设施则负向影响。
城市差异：
米兰：新站的站台门显著提升安全感，旧站的商业设施（如自动售货机）对安全感有积极作用。
北京：旧站的裸露管道是主要负面因素，新站部分区域因照明不足仍存在安全隐患。
建议：
旧站需改善照明、修复破损设施（米兰）或隐藏管道（北京）。
新站应优化空间开放性（如减少立柱遮挡）和照明一致性。
方法论贡献：提出了适用于地下公共空间的视觉分析与机器学习框架，为未来研究提供新工具。
Common factors: artificial lighting, the degree of ground openness and the presence of people have a significant positive impact on the sense of security,
while walls, garbage and damaged facilities have negative impacts.
City Differences:
Milan: The platform doors of the new station significantly enhance the sense of security, and the commercial facilities of the old station (such as vending
machines) have a positive effect on the sense of security.
Beijing: The exposed pipes in the old station are the main negative factors, and some areas of the new station still have safety hazards due to insufficient
lighting.
suggestion:
Old stations need to improve lighting, repair damaged facilities (Milan) or hide pipelines (Beijing).
New stations should optimize spatial openness (such as reducing column occlusion) and lighting consistency.
Methodological contribution: A visual analysis and machine learning framework suitable for underground public spaces is proposed to provide new tools
for future research.
