Research On Seawater Extraction Algorithm Based On GEE Platform

Water extraction in remote sensing big data cloud platform has always been a fast,accurate and batch process.There are many signs that researchers at home and abroad have made great progress in the research of extracting water body according to water body index all over the world,but they have not yet used the fusion algorithm on GEE to automatically extract seawater body according to regional image features.Therefore,according to the multiple index water indicators in the existing research,the appropriate water index can be selected according to local conditions,combined with elevation and vegetation index,and then through the adaptive threshold cutting algorithm Otsu,so as to be more adaptive in different sea areas.The main researches done are listed as followings:(1)Based on sentinel-2 MSI medium and high-resolution optical images,an effective method is used to combine the water body indexes MNDWI,AWEI and TCW calculated from sentinel-2image waveband with Otsu threshold of maximum inter class variance method to form an automatic detection method for extracting seawater.The results show that this method is more accurate and flexible than the single index method.Compared with the published surface water verification set,this method can extract the seawater of natural and artificial shoreline coastal areas more accurately.At the same time,this paper uses the same method to screen the water index of multi band landsat8 optical satellite,and the results are also applicable,which shows the robustness of this method in extracting seawater from optical remote sensing images.(2)Based on the Sentinel-1 GRD synthetic aperture radar image in 2020,the adaptive seawater thresholding algorithm for SAR radar satellite is obtained by combining multiple polarization expressions of SAR and combining the features of sea surface noise for synthesis with the Otsu adaptive threshold cut method.Compared with the discriminative extraction of inland surface water,the seawater surface has its own characteristics and difficulties,and the results of radar images are very similar to the validation set of seawater presentation features,and the land seawater boundary is smoother and smaller features can be extracted.(3)The overall accuracy of the experimental result data of both satellite functions reaches more than 91%,kappa coefficient reaches more than 83%,consumer accuracy and producer accuracy reaches more than 92% and optical remote sensing has stronger discriminative performance for independent or small features,followed by SAR algorithm.Accordingly,further,the seawater mask function with adaptive threshold as the kernel for presenting the selected image at the selected moment of click was designed on the Java Script interactive window to visualize and reflect the flexibility of the algorithm.