一种稳健最小二乘支持向量机GNSS-IR土壤湿度反演方法
A Robust Least Squares Support Vector Machine GNSS-IR Soil Moisture Inversion Method
王式太 1蒋威 2杨可心 2马岳 2姜新伟2
作者信息
- 1. 桂林理工大学测绘地理信息学院,广西桂林 541006;广西空间信息与测绘重点实验室,广西桂林 541006
- 2. 桂林理工大学测绘地理信息学院,广西桂林 541006
- 折叠
摘要
全球卫星导航系统干涉测量(global navigation satellite system interferometric reflectometry,GNSS-IR)是一种新型的遥感技术,可利用多径信噪比序列的延迟相位值反演土壤湿度值,其延迟相位求解通常使用信赖域算法,该算法一定程度依赖初值设定.文章先使用遗传算法求解出延迟相位粗略值,再将该数值作为信赖域的初值用于迭代计算,提升了部分卫星延迟相位的求解精度及稳定性.此外,针对多径信噪比序列易受环境因素影响引入粗差,进而影响模型反演精度,文章采用稳健最小二乘支持向量机作为反演模型,同时又考虑到多星融合的时空尺度优势,将该模型分别做了单星反演至五星融合反演,并与最小二乘支持向量机模型做对比.分析结果表明,当三星融合时该模型提升精度最为明显,MAE最高可降低15.6％,RMSE最高可降低12.0％.
Abstract
Global navigation satellite system interferometric reflection is a new remote sensing technology that can use the delay phase values of multipath signal-to-noise ratio sequences to invert soil moisture values.The delay phase solution is usually obtained using a trust region algorithm,which relies to some extent on initial value settings.So this article first uses genetic algorithm to solve for the rough value of the delay phase,and then uses this value as the initial value of the trust region for iterative calculation,which improves the accuracy and stability of solving some satellite delay phases.In addition,for multipath signal-to-noise ratio sequences,which are susceptible to environmental factors and introduce gross errors,which can affect the inversion accuracy of the model,this paper adopts a robust least squares support vector machine as the inversion model.At the same time,taking into account the spatiotemporal scale advantage of multi-satellite fusion,the model is respectively inverted from single star to five star fusion,and compared with the least squares support vector machine model.The results show that the model has the most significant improvement in accuracy when integrating with Samsung,with a maximum MAE reduction of 15.6％and a maximum RMSE reduction of 12.0％.
关键词
GNSS-IR/土壤湿度/遗传算法/多卫星融合/稳健最小二乘支持向量机Key words
GNSS-IR/soil moisture/genetic algorithm/multi satellite fusion/robust least squares support vector machine引用本文复制引用
基金项目
广西空间信息与测绘重点实验室项目(19-050-11-27)
广西壮族自治区高等学校中青年教师科研基础能力提升项目(2022KY1163)
出版年
2024
NETL
NSTL
万方数据