Research On The Spatial Distribution Characteristics Of City Image Based On Street View Image

City image is an effective way to reflect the characteristics and identifiability of each city,and it is also the memory of the city in the mind of residents.However,with the continuous expansion of the city scale,more and more cities are increasingly experiencing the convergence of culture,characteristics and buildings.Therefore,it is very necessary to create a good city image.The traditional research methods usually need to cooperate with the interviewees by means of questionnaire,cognitive map and other ways.This way has the problems of few data samples,time-consuming and high cost,which can not fully meet the needs of the rapid development of the current city,and there is no unified standard for the evaluation of the city image.In view of the above problems,this paper selects POI data and street network data from Internet big data in recent years to identify and recognize the five elements of city image structure proposed by Kevin Lynch;and classifies POI data into four categories: landmark buildings,scenic spots,public life and cultural education,and studies their spatial distribution in the city and each category composition of elements;finally,an image evaluation model is put forward,which includes six indexes of road feasibility,walking safety,space comfort,social contact,color richness and imageability.The recognition and segmentation of street view image is made by using deep learning,and the proportion of urban elements in each image is obtained.The index value of each coordinate point is calculated according to the formula.In order to solve the problem of different importance of indicators,the AHP is used to assign weight to each indicator and calculate the total image score,which can be visualized in the map and comprehensively evaluate the city image.This paper chooses Xicheng District of Beijing as an example to carry out the above experiments and points out the existing image problems in the study area.The experimental results show that:(1)In this paper,the POI thermal map generated by nuclear density is used to identify the image area;the boundary is analyzed with reference to the base map;the paths,nodes and signs are identified by superimposing the road network,POI and administrative base map;(2)The image type of Xicheng District is the most prominent,almost all of which are distributed in the image area;the symbol buildings are mainly distributed in the center of the research area;the scenic spots are concentrated in the north and Southeast;the cultural and educational categories are evenly distributed in the whole research area;(3)The overall image level of Xicheng District is good,and the image difference in the south and north is large,showing the trend of "the north is more important than the south".The Zhanlanguan and Deshengmen area in the north are the areas with the highest image value,while the Guang'anmen gate and Baizhifang areas in the south are the areas with the lowest image level.(4)As an old urban area,the most distinctive historical and cultural buildings in Xicheng District are Hutong,Siheyuan,ancient city wall,etc.during the urban construction,we should focus on protecting these buildings,developing more street culture,and making the urban characteristics more distinctive.