基于小波变换和DNN算法的GNSS-IR 土壤湿度反演OA

To effectively improve the accuracy of Global Navigation Satellite System Interferometric Reflectometry(GNSS-IR)soil moisture retrieval,a soil moisture retrieval method combining digital signal analysis and Deep Neural Network(DNN)is proposed.This method utilizes Wavelet Transform(WT)instead of the traditional polynomial fitting method to reduce noise,thereby effectively enhancing the extraction accuracy of the reflected signal.The Hilbert Transform(HT)is employed to obtain the average instantaneous properties of the observation signal,including the average instantaneous amplitude,average instantaneous frequency,and average instantaneous phase for each observation period.The Deep Neural Network(DNN)algorithm is used to establish a nonlinear mapping relationship between these three attributes and soil moisture,enabling the inversion of soil moisture.The model is established and evaluated using GNSS observation data collected in 2015 and 2016 at the PBO P037 station near Chattaffy County,Colorado,USA.The results demonstrate a Root Mean Square Error(RMSE)of 0.009 5 cm3/cm3 for this method,which represents a significant improvement compared to the traditional linear regression model.Consequently,the proposed approach effectively enhances the accuracy of GNSS-IR soil moisture retrieval.