Prediction and effectiveness evaluation of urban economic development on fine scale of streetscape
With the acceleration of urbanization,traditional large-scale remote sensing image and economic statistical methods cannot accurately depict the dynamics of urban economic development from a fine scale.Streetscape images can reflect the material spatial features of the urban built environment.And based on this,economic development indicator(EDI)can be predicted at a finer scale.The study proposes an improved Deeplabv3_MEP semantic segmentation model for streetscape images to extract the percentage of streetscape elements.Then,graph neural network(GCN)and convolutional neural network(CNN)are used separately with streetscape factor index and streetscape images as inputs to predict EDI.And the XGBoost model is used to analyze the driving factors of EDI.The carbon sinks are calculated and a Lasso regression model is constructed to evaluate the effectiveness of green economic development in administrative regions.The results show that:(1)At the city-level division scale,the economic indicator of Jinan city shows a trend of high concentration towards the city center and a gradual decrease towards the outskirts.(2)At the district-level division scale,Lixia district has the highest level of economic development.And the development level in sporadic areas in the east is very high,while that in other areas is lower.(3)At the street-level division scale,the closer the street is to the district or county center,the higher the average income level of residents,and the closer the street is to the city center,the higher the average income level of residents.(4)Driving factors such as wall,sky,road,and car,contribute more to EDI,while factors such as pole and motorcycle contribute less,with bus being the lowest.(5)There is a phenomenon of mismatch between the green economic development index and the level of economic development.
streetscape imageDeepLabv3_MEP modelGCN modelCNN modelPrediction of economic development indicatorsJinan
NETL
NSTL
万方数据
