Research On The Method Of Street Digital Portrait Based On Big Datas Of Street View Picture

In the past 40 years of reform and opening up,with the rapid development of urbanization,cities have taken on a new look.Similarly,based on such a big background of The Times,the phenomenon of a lot of cities of china are similar and the annihilation of urban characteristics has also become a hot topic.However,most researches on the phenomenon of "one side of a thousand cities" are qualitative studies from top to bottom,the specific expression of "one side of a thousand cities" and how cities resemble each other have been neglected and difficult to solve.Similarly,the purpose of urban construction is to provide a better living environment for people.When dealing with the phenomenon of one side of a thousand cities,we should start from a humanistic perspective and consider the specific perception and experience of people in urban space and their recognition of the city.Therefore,starting from the most commonly used public space--street space,this paper studies the uniqueness of urban landscape by digitizing the three scenes in the street scene,including the foreground,middle and back,through the street boundary,node and direction,and forming a digital portrait of the street.At the same time,it is intended to use digital Street portraits as a "measure" for comparative study among multiple cities,so as to study the degree of similarity between cities,and quantitatively analyze the specific representation of the "one side" in the "one side of a thousand cities".However due to reasons such as the outbreak,construct the streets digital portrait but no more comparative study of the city.With the gradual improvement of artificial intelligence,machine learning and other technologies,emerging technologies such as image recognition,image classification and image generation provide a good support for the study of cities.In this background,through integrating machine learning technology,Res Ne T residual network framework and Mobile Nets modified network,this paper systematically identifies and classifies the foreground,middle view,back view,category of street digital portraits and road information of street scene,and analyzes the characteristics of Nanjing Street digital portraits after detailed analysis of logarithmic and spatial attributes.The organization and structure of this article are as follows,with five chapters in total:The first chapter is the introduction part,combing the context and thinking of the article.Specifically,this paper elaborates the theoretical background,research purpose and significance of this research,as well as a specific introduction to the research content,research methods and research framework.The second chapter,as the theoretical part,introduces the relationship among streets,people and cities from the perspective of urban image,and summarizes the two basic concepts of streets and digital portraits,which defines the boundary of this study.From the value-oriented evolution of street development,the evolution of subject background and the evolution of technical methods,the previous studies were reviewed.On the basis of discussing the feasibility of digital Street portraits,this paper constructs the theory of digital Street portraits in cities,including the micro scale and city scale involved in the theory,the composition of foreground,middle view and back view,as well as the label system of label and feature label,which provides theoretical support and parameter standard for the following methods and experiments.The third chapter is the method data part,according to the above theory of digital portrait of urban streets,a complete label index system is constructed.The classification system of street digital portraits is formed by subdividing the types of street digital portraits through the interaction of characteristic labels and human activity categories in the index system.It also introduces the selected research objects,data,technical methods and experimental process.The fourth chapter is the analysis result part,which is divided into two categories and four parts,namely data feature analysis and spatial feature analysis.It includes the analysis and classification of the overall data features of Nanjing street digital portrait,the analysis and classification of the overall spatial features of Nanjing street digital portrait.On the basis of this combined with the conclusion to put forward the corresponding characteristics of the street modeling recommendations.The fifth chapter is the summary and prospect part,which summarizes the main conclusions and summarizes the viewpoints,expounds the innovation of this paper and the shortcomings of the reflection paper,so as to provide reference basis and research direction for the follow-up research.