| Surface water is an indispensable strategic resource for human survival and social development.High-precision extraction and classification of surface water is of great significance for ecosystem services and sustainable economic and social development.However,surface water extraction is easily affected by background environment.For example,water extraction in urban areas is affected by ground objects such as building shadows and asphalt roads,while water extraction in non-urban areas is interfered by ground objects such as mountain shadows and vegetation.The existing single water index method is difficult to achieve high-precision surface water extraction in complex geographical environments.At present,scholars at home and abroad have publicly published remote sensing data sets related to surface water,but such data sets only regard surface water as a single type of land cover,lacking more detailed classification.Therefore,this study constructs a Water Extraction method based on Background Information(WEBI)to extract surface water in complex geography with high precision.A Water Classification based on Shape features and Flooding frequency(WCSF)method was proposed to identify the fine type of surface water.The urban agglomeration in the middle reaches of the Yangtze River is selected as the research area for the experiment.The main contents and conclusions are as follows:(1)Extraction method of surface water bodies based on local background information.This paper proposes a surface water extraction method,WEBI,which takes into account local background information,to extract large-scale surface water with high precision.Firstly,the background similarity is calculated based on the characteristic information representing the environmental background to determine whether the scene is an urban or non-urban area.Then,according to the spectral characteristics of ground objects in different scenes,the water index applicable to urban and non-urban background was constructed respectively.Finally,the corresponding index is used for calculation in different scenarios,and the calculation results of the two indexes are integrated to obtain the surface water extraction results of the whole region.(2)Classification method of surface water bodies based on shape and flooding frequency characteristics.In this paper,a refined classification method WCSF based on shape characteristics and inundation frequency characteristics of surface water is proposed.Firstly,the classification system of surface water was constructed according to the formation mode,shape characteristics and inundation frequency characteristics of water bodies,including seasonal wetlands,seasonal rivers,seasonal reservoirs,rice fields,nonseasonal lakes,non-seasonal rivers,non-seasonal reservoirs,ponds and ditches.Secondly,the image gray threshold segmentation method and multi-scale segmentation method are combined to segment the image,and the water shape characteristics and the water inundated frequency characteristics are calculated.Finally,a random forest classifier based on water shape characteristics and water inungement frequency characteristics was constructed to achieve fine identification of surface water types with a spatial resolution of 10 meters in the study area.(3)Fine classification of surface water bodies in the middle reaches of the Yangtze River.In this paper,the urban agglomeration in the middle reaches of the Yangtze River was selected as the study area.Based on Sentinel-2 remote sensing data in 2021,surface water extraction method WEBI,which takes into account local background information,and surface water classification method WCSF,which is based on shape and water inundating frequency characteristics,were used to extract and classify surface water in the study area.The results show that the proposed method has a good effect on surface water extraction.Compared with the seven water indexes widely used at present,the overall accuracy is increased by more than 3%,and the error of misclassification and missing classification reaches the lowest,which are 6.68% and 5.87% respectively.This method can effectively overcome the interference of low reflectance surface objects and maintain the integrity of extraction of small water bodies and regular water bodies.The surface water type identification method proposed in this paper divided the surface water in the middle reaches of the Yangtze River into nine types,extracting a total of 30114.54 square kilometers of water in the study area.The overall classification accuracy was 89.87%,and the Kappa coefficient was 0.84.Compared with the existing remote sensing data sets,the classification was more complete and accurate.The dynamic realization of fine remote sensing monitoring of surface water type has important theoretical significance and application value for fine management of surface water resources and accurate monitoring of ecological environment. |
