Existing walkability assessment systems mainly rely on objective measurements and ignore subjective assessments and the diverse effects of urban spatial elements. A new comprehensive assessment framework is proposed that combines macro-indicators, micro-indicators and walking preference indicators. Leverage big data and deep learning techniques such as semantic segmentation and instance segmentation for efficient assessment of walking environments. Taking the area within the Fifth Ring Road in Beijing as the research object, a large amount of data was collected for empirical analysis. The study found that macroscopic indicators more accurately reflect walking ability in high-density cities, that there are significant differences and similarities between subjective and objective indicators, and that the uneven distribution of walking environments may be related to economic policies and social welfare inequality.
研究内容
Research contents
现有评估系统的局限性：指出了当前步行能力评估系统主要依赖客观测量，忽略了主观评估和不同城市空间元素对步行需求的多样影响。
综合评估框架的构建：提出了一个新的综合评估框架，该框架结合了宏观指标（Macro-Scale Index, MAI）、微观指标（Micro-Scale Index, MII）和步行偏好指标（Street Walking Preferences Index, SPI）。
大数据与深度学习的应用：利用大数据和深度学习技术，如语义分割、实例分割和离散选择模型，对步行环境进行高效评估。
实证研究：以北京市五环路以内的区域为研究对象，收集了大量道路网络数据、街景图像数据、兴趣点数据和调查数据，进行了实证分析和评估。
Limitations of existing assessment systems: It is pointed out that the current walkability assessment system mainly relies on objective measurements, ignoring subjective
assessments and the diverse impacts of different urban spatial elements on walking demand.
Construction of a comprehensive assessment framework: A new comprehensive assessment framework is proposed, which combines macro indicators (Macro-Scale Index,
MAI), micro indicators (Micro-Scale Index, MII) and walking preference indicators (Street Walking Preferences Index, SPI).
Application of big data and deep learning: Utilize big data and deep learning techniques, such as semantic segmentation, instance segmentation and discrete choice
models, to conduct efficient assessment of walking environments.
Empirical research: Taking the area within the Fifth Ring Road in Beijing as the research object, a large amount of road network data, street view image data, point of
interest data and survey data were collected, and empirical analysis and evaluation were conducted.
研究方法
Research methods
数据收集：通过OpenStreetMap、百度地图API和高德开放平台API等渠道，收集了道路网络数据、街景图像数据和兴趣点数据；通过问卷调查收集了行人的步行偏好数据。
模型构建：
宏观指标（MAI）：基于设施分类标准和步行衰减曲线，构建了计算基本步行能力指数并进行校正的模型。
微观指标（MII）：使用语义分割和实例分割模型量化街道元素，如绿视率、界面透明度、相对人行道宽度等。
步行偏好指标（SPI）：基于Place Pulse 2.0数据集，构建了深度学习模型来预测街道景观的感知分数。
统计分析：利用数学统计方法分析街道元素与三个指标之间的相关性，通过多变量回归分析探索主观和客观测量结果之间的差异和相似性。
综合指数计算：使用归一化加权公式将宏观指标、微观指标和步行偏好指标合成综合指数，以全面反映步行能力。
Data collection: Road network data, street view image data and point-of-interest data were collected through OpenStreetMap, Baidu Map API and Amap Open Platform API; pedestrian walking preference data were collected through questionnaire surveys.
Model building:
Macroscopic index (MAI): Based on facility classification standards and walking attenuation curves, a model is constructed to calculate and correct the basic walking ability index.
Micro-indicators (MII): Use semantic segmentation and instance segmentation models to quantify street elements such as green visibility, interface transparency, relative sidewalk width, etc.
Walking Preference Index (SPI): Based on the Place Pulse 2.0 dataset, a deep learning model was constructed to predict the perception score of streetscape.
Statistical analysis: Use mathematical statistics to analyze the correlation between street elements and the three indicators, and explore the differences and similarities between subjective and objective measurements through multivariable regression analysis.
Comprehensive index calculation: Use a normalized weighted formula to combine macro indicators, micro indicators and walking preference indicators into a comprehensive index to comprehensively reflect walking ability.
研究结果
Research conclusions
宏观指标（MAI）：显示北京市五环路以内东北部区域的步行能力较高，而西南部较低。不同设施类型的分布也呈现出显著的空间差异。
微观指标（MII）：外围地区的评分较高，主路通常比相邻的本地道路评价更好。街道元素如绿视率、界面透明度、相对人行道宽度等对步行能力有显著影响。
步行偏好指标（SPI）：揭示了行人在不同维度上的感知差异，如“安全”、“活跃”和“富有”等正面感知维度在城市中心到外围地区呈现先升后降的趋势，而“美丽”维度则在外围地区评分较高。
综合指数评估：通过综合指数映射到土地利用分类图上，发现高步行能力区域主要集中在商业中心、居住区和公共设施周边，而工业区则评分较低。这表明土地利用类型与步行能力之间存在显著相关性。
研究发现：
宏观指标在高密度城市中更准确地反映步行能力，特别是在环形交通发展模式下。
主客观指标之间存在显著差异和相似性，反映了不同区域行人的实际步行需求和偏好。
步行环境的不均衡分布可能与经济政策和社会福利不平等相关，需要进一步研究。
Macroscopic indicator (MAI): It shows that the walking ability in the northeastern area within the Fifth Ring Road of Beijing is higher, while that in the southwest is lower. The distribution of different facility types also shows significant spatial differences.
Micro-indicators (MII): Peripheral areas are rated higher, and main roads are generally rated better than adjacent local roads. Street elements such as green visibility, interface transparency, relative sidewalk width, etc. have a significant impact on walkability.
Walking Preference Index (SPI): Reveals pedestrians' perceived differences in different dimensions. Positive perception dimensions such as "safe", "active" and "rich" show a trend of first increasing and then decreasing from the city center to peripheral areas, while " The "Beauty" dimension scores higher in peripheral areas.
Comprehensive index evaluation: By mapping the comprehensive index onto the land use classification map, it is found that areas with high walkability are mainly concentrated around commercial centers, residential areas and public facilities, while industrial areas have lower scores. This indicates a significant correlation between land use type and walkability.
Research found:
Macro-level indicators more accurately reflect walkability in high-density cities, especially in ring-shaped development patterns.
There are significant differences and similarities between subjective and objective indicators, reflecting the actual walking needs and preferences of pedestrians in different areas.
The uneven distribution of walkable environments may be related to economic policies and social welfare inequalities and requires further research.
