Research On Safety Perception And Comfort Of Urban Street Based On Street View Images And Machine Learning

With the rapid progress of urbanization,various urban diseases are becoming increasingly prominent,and cities are becoming more crowded,chaotic and bloated.In the planning and design process,urban roads are often designed to take into account factors such as traffic efficiency and the maximum traffic flow that can be sustained,while pedestrian space is only considered in terms of greening rate and pavement width,and less from the perspective of pedestrians.The traditional method of planning from the pedestrian’s point of view is to conduct a large number of questionnaires,which can be used to get the most intuitive information,but it is often time-consuming and labour-intensive.If the street environment is organised in a rational way to form a pedestrian space plan with an open view and a certain order and structure,it will be more conducive to enhancing the safety and comfort of pedestrians.Accurate identification of visual factors affecting safety perception and comfort on the street is an important support for improving urban traffic environment and enhancing pedestrian travel safety.However,it is difficult for established studies to quantify street safety perception and comfort in complex scenes on a large scale.Based on this,the study conducts an empirical study on the gathering place of universities in An Ning District of Lanzhou.Firstly,visual perception factors are extracted from street view images by using image semantic segmentation and object detection,then the street safety perception and comfort dataset is constructed by manual scoring combined with deep learning,and finally a street safety perception and comfort model is established based on a lightweight gradient boosting machine and SHAP interpretation framework to identify the visual factors affecting street safety perception and comfort.The results show that:(1)The safety perception scores of colleges and commercial streets were higher,while those of urban roads were generally lower,among which,Lanzhou City College had the highest mean score(M=6.94,SD=0.88)and Gansu Agricultural University had the largest swing in evaluation(M=6.59,SD=1.13);the overall situation of street comfort was medium,and the ratings of the study area showed a decreasing trend from north to south,among which,Lanzhou City University(M=6.95,SD=1.33)and Lanzhou Jiaotong University(M=6.37,SD=0.71)had the highest ratings,Zhixing College of Northwest Normal University(M=5.97,SD=0.42)had the next highest ratings and Gansu Agricultural University(M=3.34,SD=1.42)had the lowest ratings;in addition,the street comfort ratings of the immediately adjacent universities were all higher.(2)The image shared values of sky,pavement,road and tree were the four most influential factors on the perception of street safety,where the image shared values of sky were linear,the image shared values of pavement and tree approximate logarithmic functions,and the image shared values of road resemble quadratic functions;for street comfort,wall,tree,road,sky,building and pavement were the six most influential visual perception factors.For street comfort,walls,trees,roads,sky,buildings and pavements were the six most influential visual perception factors.For street comfort,walls,trees and pavements were facilitators,skies were inhibitors,and roads and buildings had a non-linear relationship with street comfort.Therefore,the planning and design needs to determine the appropriate width of the pavement and increase the number of street trees,while taking into account factors such as tree species and canopy,to create a safe and comfortable walking environment.(3)The two-factor effect of the proportion and number of visual factors indicates that a reasonable layout of factors not only creates good spatial views,but also helps to establish the right visual richness,which in turn contributes to the perceived safety and comfort of the street.Therefore,in the street planning and design,the street environment should be organised in a reasonable way to form a certain sense of spatial order,which helps people to identify the street environment in the process of turning many into one,so as to get the safety and comfort of moving in the street space.There was also a synergistic effect between the number of visual factors and the visual factors ratio.When the number of people and motor vehicles was low,lower values of pedestrian and motor vehicle image ratios lead to higher perceptions of safety and comfort,providing further evidence that human behavioural activity also affects street safety perception and comfort.