基于GA-SVM的GNSS-IR土壤湿度反演方法

针对提高大范围土壤湿度测量精度的问题，研究了土壤湿度的全球卫星导航系统干涉测量法（GNSS-IR），提出了一种基于支持向量机（SVM）的土壤湿度反演模型，利用遗传算法（GA）的自动寻优功能寻找SVM的最佳参数。结果表明，GA-SVM模型在测试集上得到的土壤湿度反演值与实测值的平均绝对百分比误差（MAPE）仅为0.69%，最大相对误差（MRE）为1.22%，线性回归方程决定系数达到了0.956 9。进一步与统计回归、粒子群优化的SVM模型（PSO-SVM）及反向传播（BP）神经网络方法进行对比，结果说明:在样本数目有限的情况下，GA-SVM方法更适用于土壤湿度的GNSS-IR技术反演，且反演精度较高，泛化性能良好。 

GNSS-IR soil moisture inversion method based on GA-SVM

In order to improve the precision of soil moisture measurement in a wide range，in this paper, the global navigation satellite system interferometry and reflectometry (GNSS-IR) for soil moisture was studied and a soil moisture inversion model based on support vector machine (SVM) was proposed. In this model, the automatic optimizing function of genetic algorithm (GA) was applied to optimize the parameters of SVM. The results show that the mean absolute percentage error (MAPE), the maximum relative error (MRE) and the coefficient of determination for equation of linear regression are 0.69%, 1.22% and 0.9569 respectively between the soil moisture inverted by the proposed GA-SVM model and the ground measured values. In addition, the performance of GA-SVM model was also compared with the statistical regression, particle swarm optimization SVM model (PSO-SVM) and back propagation (BP) neural network. The comparison results show that the GA-SVM method is more suitable for the GNSS-IR soil moisture inversion than other machine learning algorithms in small training set scenario, and it has higher inversion precision and better generalization performance.