GA辅助NLS的GNSS-IR土壤湿度反演方法北大核心CSCD

采用具有阻尼因子的函数模型,使用遗传算法(genetic algorithm,GA)辅助非线性最小二乘(nonlinear least squares,NLS)方法对相位参数进行求解。结果表明:1)相较于标准余弦函数模型,该方法的反演相位与土壤湿度的相关系数有较为明显的提升,反演结果也更加稳定,在5°~15°、5°~20°、5°~25°三个高度角范围内的相关系数均大于0.68,不同高度角之间的相关系数差值小于0.07;2)反演精度有不同程度提高,R 2提高5.72%~76.06%,RMSE减小6.12%~24.24%,MAE减小2.7%~28.3%,将该方案所求相位用于多星线性回归模型后平均RMSE减小10%。

The Method of GNSS-IR Soil Moisture Inversion Using GA-Assisted NLS

We adopt the function model with damping factor and use the genetic algorithm(GA) assisted nonlinear least squares(NLS) to solve the phase parameters. The results show that:1) Compared with the standard cosine function model, the correlation coefficient between the inversion phase and soil moisture is significantly improved, and the inversion results are more stable. In the three elevation angle ranges of 5°-15°, 5°-20°, and 5°-25°, the correlation coefficients can all be greater than 0.68. The difference between the correlation coefficients of different elevation angles is less than 0.07. 2) The inversion accuracy has been improved to varying degrees, with R² increased by 5.72%-76.06%, RMSE decreased by 6.12%-24.24%, MAE decreased by 2.7%-28.3%. By applying the inversion results of this method in multi-satellite linear regression model, the average RMSE decreased by 10%.