The Study Of City Image Based On UAV Photography Data In Xi'an

With the coming of fourth industrial revolution,the technological innovation brought by artificial intelligence and machine learning has greatly enhanced the speed and accuracy of recognition and analysis of images,video,voice,text and other information.At the same time,the rapid development of manufacturing has led to the rapid expansion of the civilian UAV market.The number of users using consumer UAV aerial photography has gradually increased.More and more city photos taken by UAVs from "God's perspective".This new perspective changes the traditional ground perspective that used to be taken by cameras or mobile phones.In this situation,based on the photos and related data of UAVs photos of Xi'an obtained from the social platform "Skypixel",this paper explores Xi'an's city image through Microsoft Azure vision analysis,sentiment analysis,social network analysis and GIS spatial analysis.This study explores the spatial and temporal distribution characteristics of aerial photos taken by UAVs,the visual characteristics of city images,the classification of city images elements and the evaluation levels,and makes an exploratory analysis of the current city image.Finally,a comprehensive formation mechanism for city image from UAVs is explored,and suggestions are put forward for the construction of city images of Xi'an and the management of UAVs.This research aims to strengthen urban design and improve the quality of urban space through the study of city image.The main conclusions are as follows:(1)The number of UAVs photos show a single peak in annual changes,and 2016 is the peak year.In the monthly changes,the winter photos are more than the spring,summer and autumn seasons.In terms of spatial distribution,UAVs photos distribution is concentrated in the main urban areas such as Yanta District,Beilin District,Weiyang District,Lianhu District,Baqiao District and Xincheng District,and the number of photographs accounts for 88.76 percent of the total.(2)Through computer vision analysis,the photo content can be divided into 6 categories: transportation,human landscape,natural landscape,people and activities,adjectives,and colors.Social network feature maps show divergent forms with dense kernels and sparse surroundings.In the color analysis of image photos,there are a total of 12 different foreground and background colors,of which black and gray occupy a larger proportion in visual color matching.(3)The selection of the content attributes of city image indicates that the number of photos in the characteristic category of the ancient city is the largest,accounting for more than half.The number of photos in the modern city category is second,accounting for about 30%.The number of photos of single ancient buildings,ancient buildings,antique buildings,transportation facilities and night scenes is more than that of daytime photos.More than 90% of the images do not include people,and more than 84% of the imagery photos are taken from the UAV at an oblique angle.(4)Through the classification of city image elements,the city image of Xi'an is divided into 9 path images,3 edge images,16 node images,5 district images,and 14 landmark images.(5)From the cognitive analysis of city image evaluations,the number of positive evaluations with a score value greater than 0.50 accounted for 58.52%,which was greater than the sum of negative evaluations and neutral evaluations.The evaluations distribution shows the characteristics of single-peak normal distribution,which reaches its peak in the 0.51-0.59 group.The city image presents the overall positive evaluations,some neutral evaluations and a small number of negative evaluations status.(6)This paper explores and constructs the comprehensive formation mechanism of Xi'an's city images from the perspective of UAV.By analyzing the city image construction in Xi'an,then put forward suggestions on city images construction and UAV using policy.