基于机器学习的历史街区周边生活性街道感知测度及空间分布特征研究

Research on Perception Measure and Spatial Distribution Characteristics of Living Streets aroundHistorical Blocks Based on Machine Learning

关键词:  风景园林  机器学习  生活性街道  街景图像  历史街区  量化评价

Key words:  landscape architecture  machine learning  living street  street view image  historic block  quantitative evaluation

摘要:

历史街区周边的生活性街道具有生活空间、文化资源、社交场所等多维属性。随着中国城镇化转型与城市更新的推进，如何平衡社会生活与旅游开发 是历史街区周边生活性街道需要重点关注的问题，对其开展感知测度研究，具有传承传统居住文化、活化历史街区资源和提升城市整体风貌的重要意义。以南京 市高淳老街历史街区周边11个生活性街道为研究对象，基于16 632次街景图像分析，应用机器学习与深度学习方法，筛选出街道空间关键感知指标，包括绿视 率、可步行性和意象化等，进而构建街道空间感知模型。研究结果显示，绿视率和可步行性对人的主观感知具有较为显著的积极作用，而街道历史文化主题元素 对意象化感知具有较大影响。研究结果为历史街区周边生活性街道更新提供了可借鉴的定量评估方法与策略建议，并可推广应用于更大范围历史街区周边街道的 感知评价研究。

Abstract:

Historical districts, as important carriers of a city's historical and cultural heritage, carry rich cultural heritage and unique spatial features, and are one of the key elements for the sustainable development of a city. In recent years, with the advancement of urban renewal, how to organically integrate historical districts with surrounding communities while protecting them has become an important issue in urban planning and design. The relevant policies issued by the Ministry of Housing and Urban-Rural Development emphasize that urban renewal should respect the will of the people, adhere to the principle of "retention, renovation, and demolition" simultaneously, and pay attention to the livability and cultural inheritance of historical districts. However, the research on the spatial perception of the living streets around historical districts, which serve as an important link between historical and modern cities, is relatively weak. Living streets not only serve the function of transportation, but also carry out multiple social functions such as neighborhood interaction, daily shopping, and leisure fitness. Compared with ordinary residential streets, the residential streets around historical districts have unique features and cultural values. Their spatial perception is influenced by various factors such as historical buildings, the scale of streets and alleys, and the green environment. At present, with the emergence of new data environments, data sources such as street view images, mobile phone signaling, and POI have provided new technical support for street space research. The development of machine learning technology has also made it possible to conduct quantitative analysis of street space perception, enabling more efficient processing of large-scale street scene data and revealing the relationship between street space perception and environmental elements. This study takes 11 living streets around the historical block of Gaochun Old Street in Nanjing City as the research objects. Based on the analysis of 16,632 street scene images, machine learning and deep learning methods are applied to screen out key perception indicators of street space, such as green view rate, walkability, and imagery, and construct a street space perception model. The results show that the transparency, imagery, positivity, and coordination of the living streets around Gaochun Old Street present obvious spatial heterogeneity. The permeability and coordination degree show a distribution characteristic of "high in the northwest and low in the northeast". The high values of imagery are concentrated around historical districts. The overall level of enthusiasm is relatively even, but it is higher in areas close to historical districts. Meanwhile, the green vision rate is positively correlated with the coordination degree and walkability. The sky rate is negatively correlated with positivity. The degree of enclosure is positively correlated with the degree of positivity and imagery. The ranking of the importance of street scene elements to perception indicators shows that elements such as buildings, sidewalks, and trees contribute the most to imagery, transparency, and coordination. The research results propose renewal principles such as the overall integration of historical districts and surrounding community Spaces, the network-like connection of historical and cultural Spaces with social life, and the behavior-oriented reshaping of a sense of belonging in residential Spaces. Specific strategies include optimizing historical features, enhancing the landscape environment, and improving slow traffic Spaces, providing quantitative assessment methods and strategic suggestions for the renewal of living streets around historical districts that can be referred to, and can be promoted and applied to the perception evaluation research of streets around historical districts on a larger scale.