Research On The Influence Of Built Environment On Housing Price Involving Street View And Remote Sensing Images

Housing price is closely related to the life of urban residents.Housing price is not only related to the structural characteristics of the real estate itself,but also has an important relationship with the main research object of this paper,namely the built environment.It is of great help to identify and understand the built environment for analyzing the spatial differentiation of urban housing prices.However,it is difficult to quantify the built environment,and the traditional manual review method is timeconsuming and labor-consuming.With the development of deep learning and the popularization of street view products and remote sensing products,a new technical route has been brought to the quantification of built environment.In this paper,the deep learning model Seg Net and UNet are used to obtain the semantic segmentation objects of street view and remote sensing images respectively,in order to extract the characteristics of built environment.Street view and remote sensing images show the current situation of urban built environment from the perspective of the ground and the sky respectively,and the fusion of these two kinds of data can more truly reflect the built environment.In order to achieve this goal,this paper designs a data fusion method of remote sensing and street view semantic segmentation objects,which considers the difference of scale and perspective.In this method,Bayesian empirical Kriging interpolation is applied to the street scene semantic segmentation objects,and continuous distribution street view features which are consistent with the remote sensing data are generated.Then,from the perspective of three-dimensional,Mix Tree index and Mix Building index are generated.The experimental results show that the integration of street view and remote sensing data can bring higher fitting accuracy to the housing price regression model.At the same time,the characteristics of Mix Tree and Mix Building obtained by the data fusion method in this paper match with the actual situation of the green and building in Shenzhen,which proves the scientificity of the fusion method from the side.In addition,this paper adopts three models,namely HPM in the economic field,GWR in geography field and XGBoost in the computer field,to regress the house price respectively.It is found that the accuracy of GWR is the highest,XGBoost is the second,and HPM is the worst.These three models explore the relationship between variables and housing prices from the perspective of global,local and data fitting respectively.Through the comparison of the regression results,this paper has a more comprehensive understanding of the impact of the characteristics of built environment and non-built environment on housing prices.Specifically,this paper finds that in the built environment indicators,Mix Tree,Sky and Trans indicators shows a more obvious positive correlation with housing prices,while Volume,on the contrary,has a negative impact on the surrounding real estate.In particular,the influence of Mix Building indicators on housing prices is obviously polarized.In the non-built environment indicators,the economic-related indicators –Dfinanical,Dindustry and Dbussiness have the most important impact on the housing price;in addition,Dcity Center and Dmid School are also the key factors affecting the housing price.The street view and remote sensing images data fusion methods designed in this paper provides a data fusion framework for other field,for instance public health,air pollution and urban microclimate,which is highly related to the built environment.At the same time,this paper explores the universal laws between the characteristics of different built environment and housing prices,in order to get universal domain knowledge to guide built environment designing and urban planning.