Research On The Classification Of Urban Architecture Style And Features Based On Technology Of Machine Learning With Street View Data

With the implementation of industrial construction and modernist architecture in Chinese cities,Chinese cities are faced with the problem of disappearing features.However,as an important part of urban landscape,the architectural landscape has always been difficult to be comprehensively and quantitatively analyzed on a large scale.However,the current artificial intelligence technology is gradually mature,and the application of machine learning in image recognition and image classification is gradually improved.It makes it possible to quantify landscape studies.Therefore,in this context,machine learning technology is used to identify and classify the architectural features in urban street view images.The architectural style of urban street view is the collection of building facades,functions and other features that are most easily perceived at the street level.It is a component part of the physical style of a city,including buildings,monomer buildings,building components,skin symbols and other major elements.This paper adopts baidu street view platform to collect street view data of the whole city of nanjing.By integrating Info GAN framework of semisupervised classification and Res Net framework of supervised classification in machine learning technology,it constructs the classification technology method of street view architectural style.Firstly,this paper studies the ontology and classification theory of urban architectural style,and summarizes the classification standards of current urban architectural style under the guidance of different fields.Combined with the research results of scholars from three different disciplines on urban architectural style,the classification standards of urban architectural style or the value orientation of classification are summarized,and the three classification orientations of location type,function type and form type are constructed as the theoretical basis for subsequent classification.Then through data cleaning,Mobile Nets semantic segmentation,mask operation,Info GAN model preliminary classification,as well as the Res Net framework supervised classification steps,the case analysis of nanjing street view data,get the classification results of machine learning.Then,it is verified from two perspectives of spatial location and functional fit to discuss the feasibility of machine learning architectural style classification technology and the possibility of research on urban style.The specific organization structure of this paper is as follows:The first chapter is the introduction of this paper,which mainly introduces the research background,the research progress of the style ontology,the concept analysis and the research method and significance of this paper.The second chapter is the theoretical construction part,which establishes the value orientation of the classification of urban landscape from the perspectives of urban planning,architecture and database construction,and provides basic theoretical parameters and standards for the subsequent data experiments.The third chapter is the technical method flow of the concrete example.This paper expounds the technical process of street view classification taking nanjing city as an example and the various problems and solutions encountered therein.The fourth chapter is the case study results and discussion,showing the classification results of the case study,as well as the verification results,and discussing the phenomenon reflected in the results,evaluating the technical methods,and considering the possibility of landscape research scenario.The fifth chapter is the summary part of this paper,which summarizes the main conclusions of this paper,condenses the innovation points,and provides some reference and direction for the follow-up research.