Research on Inland Water Body Recognition Method Based on SVM Model for Spaceborne GNSS-R
Inland water body detection is one of the important research directions in the field of remote sensing,covering river morphology change monitoring,real-time flood monitoring,surface water change analysis and other fields.Support Vector Machine(SVM)performs well for resolving nonlinear classification problems and can get a low error rate.A method of inland water body recognition based on Global Navigation Satellite System Reflectometry(GNSS-R)signals is proposed,which uses satellite-borne GNSS-R Signal to Noise Ratio(SNR)as pixels to construct remote sensing images,and applies the Synthetic Minority Oversampling Technique(SMOTE)to perform data processing.On this basis,the water body information is extracted by the SVM model,which is verified by the Cyclone Global Navigation Satellite System(CYGNSS)data in the Congo Basin.The results show that the classification accuracy without SMOTE algorithm processing is 65.1%and the water body classification accuracy is 5.39%,while after SVM model and SMOTE algorithm processing,the data classification accuracy is improved to 96.49%and the water body classification accuracy is improved to 96.32%.The accuracy and water body precision rates are improved,which shows the effectiveness of the spaceborne GNSS-R inland water body identification method based on SVM.
国家科技期刊平台
NSTL
万方数据
维普
