Urban Core Elements Extraction And Spatial-temporal Change Analysis Based On Multi-source Data Fusion

The rapid development of urbanization and the growing demand of urban function lead to the diversity and variability of species in urban interior.Water body,green space,buildings and roads,as the core elements of a city,play an important role in the process of sustainable urban development.The research of urban core elements extraction by remote sensing image has become a hot topic of concern by many scholars.In the study of urban core elements extraction from remote sensing images,one of the most typical problems is foreign bodies with the same spectrum and different spectra of the same object.In the process of element extraction,different elements face different category confusion problems according to their spectral characteristics.Therefore,it is the focus and key of this paper to classify the problems encountered in the extraction of different urban elements and select the best features for fusion extraction.This paper takes Shenzhen city as the research area and Sentinel-2 image as the main data source to extract urban core elements based on multi-source data fusion.Urban core elements include water element,green space element,building element and road element.On this basis,the change detection of Shenzhen's urban core elements from 2017 to 2021 is carried out to analyze the change characteristics of Shenzhen's urban core elements from 2017 to 2021.The main research contents and achievements of this paper are as follows:(1)Extraction of urban core elements based on multi-source data fusion.The preprocessed Sentinel-2 image was transformed by principal component to achieve the purpose of data dimension reduction,and the four core elements of water body,green space,buildings and roads were extracted by integrating spectral features and texture features using random forest algorithm.According to the characteristics of different elements,different data and features are fused to optimize the elements.Water elements in the study area are confused with some mountain shadows due to the influence of terrain,and the slope and aspect features extracted from terrain data are integrated to achieve the optimization effect.The green space factor was optimized by combining the Sentinel-2 image's unique red-edge band and the greenness component of K-T transform.The extraction of building elements is easy to be confused with bare land with high reflectivity.The optimization effect is achieved by integrating luminance value of VIIRS and POI kernel density characteristics.The road elements are optimized by combining shape characteristics such as shape index and aspect ratio.The optimized extraction results were processed by mathematical morphology method and some noise points were removed.The overall accuracy of water body,green space,building and road elements were 96.8134%,99.5134%,90.0954% and 87.9009%,respectively,and the Kappa coefficients were 0.8254,0.9867,0.8265 and 0.7839,respectively.(2)Integration of urban core elements.The extraction results of urban core elements based on multi-source data fusion were initially merged with the method of superposition analysis,and the overlapping phenomenon,misclassification and missing classification were improved according to the neighborhood voting method.In order to verify the feasibility of the proposed method,the maximum likelihood method and support vector machine(SVM)are used for multi-classification,and the results of the three methods are compared.The overall accuracy of maximum likelihood method,support vector machine method and the proposed method are 88.5678%,90.6745% and92.7890%,respectively.The Kappa coefficients are 0.7567,0.7899 and 0.8064,respectively.The results show that the accuracy of the proposed method is slightly higher.(3)Spatial-Temporal change analysis of urban core elements.Based on the above work,the urban core elements extraction results of Shenzhen in 2017 and 2021 are obtained respectively.The transformation process and spatial distribution of urban core elements can be expressed more directly from land use transfer matrix and spatialtemporal change analysis chart.The results show that buildings,roads and part of green space are the main transfer factors in Shenzhen.It can be concluded that with the passage of time,the building and road elements of Shenzhen are still gradually expanding,which is consistent with the high urbanization process of Shenzhen.In addition,while accelerating the pace of urbanization,Shenzhen also takes into account the urban greening work.