Watershed And Fishery Cages Extraction Of Water Source Protected Areas Based On GF Series Image

Water resources is an important resource related to the people's livelihood.Although China is vast,but the relative lack of water resources.Surface drinking water sources as an important source of fresh water resources and storage sites,water resources in China is a vital part of the system.Remote sensing,as a technology that can acquire and monitor surface conditions for a long time,can provide data support for the monitoring and management of surface drinking water sources.In this paper,the domestic high score satellite image as the data source,for the surface drinking water source for the watershed extraction and protection area mapping,water source cage extraction of two research.In the first part,this paper designs the decision tree classification algorithm to extract the waters of the water source area.The algorithm mainly uses the multi-band feature of the multi-spectral remote sensing image.Through NDVI,HSV transform based on Munsell,PCA main Component analysis results and the combination between the band,the design of a three-tier classifier,and gradually filter non-water objects.At the same time,this method has better effect in extracting the turbid water with higher sediment concentration than other traditional water extraction methods.Then,based on the relevant national standard and the actual situation of remote sensing image,the process of plotting the water and land area protection area in the remote sensing area is designed.In the second part,this paper will extract the cage materials of the fishery which may cause the water pollution of the drinking water source,so that the relevant departments can grasp the distribution of the groundwater in the waters of the water source.Firstly,this paper analyzes the spectral characteristics,spatial characteristics and texture features of fishery cages in high-frequency remote sensing images.Then,based on the analysis results,the spectral features and texture features are used to combine Class objects to extract.Firstly,the spatial filtering of the remote sensing image is enhanced,and then the feature of the image texture is obtained by the gray level co-occurrence matrix.By comparing the different texture features under different window sizes,two features are more suitable for the extraction of the cages.(Entropy)and Second Moment(second moment),and use these two measures to build the NDES index,better highlight the difference between the fishery cages and water bodies.At the same time,considering the directionality of the texture features,the multi-directional texture feature combination method is used to filter the area of the misty water body,and the general range of the cage material is obtained.Then,by narrowing the range of extraction method,the object-oriented segmentation method is used to extract the details of the cage material in the approximate range of the acquired cage.In the end,the experiment of water area extraction,select the Danjiangkou reservoir,Baotou part of Yellow River,Yarkand River,the Zhangjiashan part of Jing River.The results are compared with traditional water extraction methods such as NDWI,NDVI,PCA single band threshold and HSV transformation method In comparison,the generalization of the method and the advantages of extracting the turbid water with high sediment concentration were proved from the water extraction effect and precision.Select Baotou part of Yellow River and Nierji reservoir to get the protection part reigon,The vector data of the protected area of the water source area were obtained successfully.The experiment was carried out to extract the fishery cages from the water source of Danjiangkou Reservoir.The accuracy of the extraction of the cage and the accuracy of the cage area with GF-2 image was proved.