[1]李 渊,王耀梅,杜亚男,等.基于机器学习的旅游街道空间安全感知评价及其影响因素研究——以厦门鼓浪屿为例[J].中国名城,2025,39(5):78-88.[doi:10.19924/j.cnki.1674-4144.2025.005.011]
　LI Yuan,WANG Yaomei,DU Yanan,et al.Study on the Evaluation of Security Perception in Tourist Street Space and its Influencing Factors Based on Machine Learning: Taking Gulangyu Island in Xiamen as an Example[J].China Ancient City,2025,39(5):78-88.[doi:10.19924/j.cnki.1674-4144.2025.005.011]
点击复制

基于机器学习的旅游街道空间安全感知评价及其影响因素研究——以厦门鼓浪屿为例()
分享到：

中国名城[ISSN:1674-4144/CN:32-1793/GO]

卷:
39
期数:
2025年5期
页码:
78-88
栏目:
出版日期:
2025-05-05

文章信息/Info

Title:
Study on the Evaluation of Security Perception in Tourist Street Space and its Influencing Factors Based on Machine Learning: Taking Gulangyu Island in Xiamen as an Example
文章编号:
1674-4144（2025）005-0078-11
作者:
李 渊; 王耀梅; 杜亚男; 杨盟盛; 张 娜
李 渊，厦门大学建筑与土木工程学院教授；王耀梅，厦门大学建筑与土木工程学院硕士研究生；杜亚男，厦门大学建筑与土木工程学院博士研究生；杨盟盛，厦门大学建筑与土木工程学院博士研究生；张 娜，厦门大学建筑与土木工程学院博士研究生。
Author(s):
LI Yuan; WANG Yaomei; DU Yanan; YANG Mengsheng; ZHANG Na
关键词:
旅游街道; 图像分割; 安全感知; XGBoost; SHAP
Keywords:
tourist streets; street view images; security perception; XGBoost; SHAP
分类号:
TU984；F59
DOI:
10.19924/j.cnki.1674-4144.2025.005.011
文献标志码:
A
摘要:
街道空间是旅游目的地的主要游览空间，准确识别影响街道安全感知的视觉要素对改善旅游者的体验至关重要。然而，旅游者的主观感知与客观环境之间的差异难以进行定量研究。以鼓浪屿为研究地，通过全岛交叉口街景图像构建数据集，运用图像语义分割技术提取要素占比，并结合主客观评价采用 XGBoost 算法与SHAP模型分析各街景要素对街道空间安全感知的具体影响。结果显示，鼓浪屿的安全感知在空间分布上呈现出“由岛屿中心向外逐渐递增”的趋势。人行道、树木、建筑、道路、天空、墙壁的图像占比值是影响力最大的6类视觉要素。其中人行道、树木、道路、天空的图像占比值与安全感知呈现出线性递增的关系；建筑和墙壁的图像占比值与安全感知呈现线性递减的关系。此外，各视觉要素之间交互影响。研究通过量化分析不仅能为旅游地的安全感知研究提供理论参考和实践依据，而且可以为街道空间的设计与优化提供科学指导，从而提升旅游者的安全感和体验质量。
Abstract:
Street space is the main space that tourists travel in tourist destinations, and accurately identifying the visual elements that affect the perception of street safety is crucial to the improvement of tourist experience. However, the subjective and objective perceptions of tourists often differ, making it difficult to conduct extensive quantitative studies. In this study, Gulangyu Island, a tourist destination, is selected as a case study, and the dataset is constructed based on the street images of intersections throughout the island, the percentage of each element in the image extracted through image segmentation techniques, and the subjective ratings combimed with an objective machine-learning evaluation system to contruct a street security perception model. The study adopts XGBoost algorithm and SHAP interpretation framework to analyse the specific influence of each streetscape element on the perception of street space safety. The results show that the overall spatial distribution of security perception in Gulangyu Island shows the trend of ’gradually increasing from the centre of the island to the outside’. The images of pavements, trees, buildings and roads, sky, and walls are the six most influential visual elements. Among them, the percentage of images of pavements, trees, roads and sky shows a linear increasing relationship with the perception of street safety, while the percentage of images of buildings and walls shows a linear decreasing relationship with the perception of street safety. In addition to this, there is an interaction effect among visual elements. This study not only provides methodological reference and theoretical basis for the study of security perception in tourist places, but also aims to improve the safety level of tourist places through more accurate analysis results.
更新日期/Last Update: 2025-05-14