Two-stage Fine Extraction Of Urban Villages Based On Building-Green Index And Hierarchical Constraints

In the process of rapid urban development,large cities have surrounded rural areas to realize urban modernization,forming a special urban village(UV)type of urban settlement.Urban village communities are characterized by high building density,low accessibility and mixed functional areas,which makes it very difficult to monitor and update the information of urban villages,and accurate and effective monitoring and identification of them can help achieve coordinated urban-rural development and optimize urban-rural ecological environment.Traditional urban villages are extracted and analyzed mainly from manual research,field surveys,and current land use maps,which often have problems of lagging information,high update cost,and missing data,and the extracted urban village data cannot meet the rhythm of rapid development in the era of smart cities.In addition,urban villages show obvious spatial characteristics due to their special functional positioning and development history,and previous studies on urban village identification focus on the extraction of basic geographic features,ignoring the analysis of spatial relationships,and the extraction accuracy cannot meet the needs of urban village analysis and research in the era of big data.Currently,the most common method for urban villages extraction is object-based extraction method,object-oriented method contains complete semantic information,which is conducive to extract richer information,but object-oriented segmentation method also divides the non-homogeneous areas near urban village area together into urban village,which often cannot do fine identification,and the current common deep learning fine identification method needs a large amount of label data,urban village identification of The open dataset is small and the acquisition cost of label data is high.To address the above problems,this paper proposes a two-stage urban village extraction method based on Building-Green index and hierarchical constraints using high-resolution remote sensing data,built-up area data,Open Street Map data,etc.,constructs urban village extraction units to identify urban villages and analyzes the distribution characteristics of urban villages in the study area.The main research contributions of this paper are as follows:(1)Construction of built-up area zoning grid and urban villages based on object extraction unitThe spatial pattern of urban villages is related to the location of urban villages In order to improve the accuracy of identification,this paper divides the study area into five regions based on the built-up area data according to the density of built-up areas,and different regions are trained and tested separately.Based on the simplified OSM data to assist image segmentation,small segmentation units are integrated to form the basic units for urban village extraction.(2)A high-resolution urban village extraction method based on Building-Green index is studiedThe high-resolution remote sensing images contain rich features of urban natural attributes,including spectral features,texture features,structural features,etc.,but lack spatial relationship features.In this paper,we first extract buildings by using building morphology index,then extract building features including density,area,minimum building spacing and other building features by combining with the characteristics of urban villages,then extract texture and spectral features based on remote sensing images,extract NDVI data based on GEE platform,and count the number of NDVI pixels within the basic unit of urban villages by NDVI threshold segmentation to form green space features.(3)A hierarchical constrained urban village fine extraction method is studiedThe object-oriented extraction method contains redundant contextual information and cannot achieve fine extraction of urban villages.In this paper,we propose a hierarchically constrained method for fine extraction of urban villages.Firstly,we further divide the fine grid cells in the urban village cells,and extract texture,spectrum,structure and other features in the grid cells to classify the grid cells initially.Then the preliminary classification results are constrained by using super-pixel segmentation cells,and the grid cells that obviously do not match the spatial distribution of urban villages are corrected to obtain the fine extraction results of urban villages.(4)The distribution characteristics of urban villages in the study area are analyzedThis paper analyzes the characteristics of urban villages distribution in the study area with the extraction results,and explores the relationship between urban village distribution and urban economic industry and geographic location,so as to provide a reference basis for targeted urban village transformation.