SVRM辅助的北斗GEO卫星反射信号土壤湿度反演方法

提出了一种支持向量回归机(SVRM)辅助的北斗地球静止轨道(GEO)卫星反射信号土壤湿度反演方法。使用全球导航卫星系统反射信号(GNSS-R)右旋圆极化(RHCP)天线和左旋圆极化(LHCP)天线接收体制进行了地基实验,采集了北斗GEO卫星直射、反射信号原始数据,并从中提取直射、反射信号的相关功率,结合北斗GEO卫星的高度角与方位角信息作为输入,烘干称重法获取的土壤湿度作为输出对使用径向基(RBF)核函数的ε-SVRM进行了训练。独立测试集上的结果表明,SVRM辅助的北斗GEO卫星反射信号土壤湿度反演方法获取的土壤湿度结果与烘干称重法获取的土壤湿度参考值误差控制在3%以内,线性回归方程决定系数为0.897 9,均方根误差RMSE为1.492 6%,证明了该方法具有良好的泛化特性,实际应用中效果良好。

SVRM-assisted soil moisture retrieval method using reflected signal from BeiDou GEO satellites

We propose a support vector regression machine (SVRM)-assisted soil moisture retrieval method using the reflected signal from BeiDou geosynchronous orbit (GEO) satellites. This method uses a right hand circular polarization (RHCP) antenna and a left hand circular polarization (LHCP) antenna to gain the direct and reflected signal's power data from the BeiDou GEO satellites, respectively. Furthermore, it uses the direct and reflected signal power, BeiDou GEO satellites' elevation angle and azimuth angle as the input features and uses the soil moisture data which is obtained by oven-drying method as the output target of the ε-SVRM which uses a radial basis function (RBF) kernel function. The collected data is separated into two sets randomly:one as training set and the other as test set. The test results show that the error between retrieval model's prediction and the value of oven-drying method is less than 3%; the regression coefficient of determination is 0.897 9; the root mean square error (RMSE) is 1.492 6%, which proves that this method has good generalization ability and the practical results meet the application requirement.