Impervious Surface Extraction Based On Deep Learning And Spatio-temporal Evolution Analysis

In urban studies,impervious surface is generally defined as the artificial surface where water cannot quickly penetrate into the surface,such as buildings,parking lots,streets,highways,etc.,and it is an important indicator for studying urban expansion and measuring urban ecological environment.Hangzhou is an important central city in the Yangtze River Delta,and its urban expansion can be used as a reference for urban development in China.Therefore,it is of great significance to carry out researches on impervious surface for monitoring urban dynamics and guiding urban planning and construction.Due to the complexity of urban land use and the diversity of impervious surface materials,it is challenging to directly extract impervious surface from high-resolution remote sensing images.Aiming at the demand of extracting urban impervious surface from high-resolution remote sensing images,this paper proposes an impervious surface extraction method based on deep learning,which makes use of deep learning's outstanding ability in analysis,calculation and feature extraction to achieve accurate extraction of urban impervious surface.The method proposed in this paper can improve the extraction accuracy and detail accuracy of impervious surface,and has certain generalization and migration ability.The main research contents of this paper are as follows:(1)Construction of remote sensing impervious surface data set.Based on the high-resolution Google Earth remote sensing image,the impervious surface was labeled.After data cropping,data screening,data enhancement and other operations,a high-resolution remote sensing image impervious surface data set with 12,000 samples was constructed.It provides data support for the subsequent extraction of remote sensing impervious surface based on deep learning.(2)A deep learning impenetrable surface extraction method combining Feature Pyramid Network(FPN)and U-Net Network was proposed.By introducing the characteristics of the pyramid network,realizes the remote sensing image multi-scale feature fusion,to build a tight network model of the water extract made the model of the improved training plan,improve the generalization ability and applicability of the model,the formation is of high precision through the water extraction method,and through experimental verification method is feasible.At the same time,a comprehensive comparative experiment of the extraction method of impervious surface was designed to comprehensively evaluate the extraction results of impervious surface from the aspects of visual effect and extraction accuracy,etc.,to verify the superiority of the proposed method.(3)Automatic extraction and spatiotemporal evolution analysis of long time series impervious surface.The improved multi-scale U-NET impervious surface extraction model proposed in this paper is used to realize the automatic extraction of impervious surface with a time span of 30 years in the study area,and the advantages of the proposed method in the extraction efficiency and automation of impervious surface are verified.Based on the extraction results,the temporal and spatial variation characteristics and rules of impervious surface in the main urban area of Hangzhou were analyzed,and the urban expansion trend and driving factors of impervious surface changes were discussed,so as to provide decision-making reference for guiding urban planning and management.In this paper,based on the deep learning method can accurately extract the city information opaque surface,avoid the traditional methods cannot extract through the surface of a priori knowledge dependence is stronger,the problem such as extracting feature selection is relatively complex and inefficient,by improving the network structure,improves to detail and ability to opaque on the edge of the water extraction,the extraction precision is improved further,and has a certain generalization and migration ability.The research results of this paper provide a new data-driven idea for the extraction of impervious surface,which has certain scientific and practical significance.