支持向量机在雪地环境中的参数反演

Inversion of Snowfield Parameters Based on SVM

针对传统雪地环境参数反演方法中模型复杂、未知参量多、网络拓扑结构难确定等问题,提出了一种基于最小二乘支持向量回归技术的雪地环境土壤湿度和雪层厚度的反演方法。首先根据雪地环境建立了分层粗糙介质微波散射模型,利用微扰法得出其后向散射系数;然后进行数据敏感性分析,选取雷达频率为1.27GHz(L波段)、双入射角(42°和53°),设计了两种反演方案,分别为单极化和双极化的方式作为微波信号样本信息,经过适当的训练,利用最小二乘支持向量机反演土壤湿度和雪层厚度。结果表明:用多入射角、双极化时,反演结果具有较高的精度;在加噪的情况下,用多入射角、双极化时,该方法的反演结果保持了较好的抗噪能力,为雪地环境中土壤湿度和雪层厚度的反演研究提供了一种可行的方法。

For inversion methods of traditional snowfield environment with some problems including complicated model, a lot of sealed parameters,and difficultly confirmed network topology, there is an inversive way based on the least squares support vector machine （LS-SVM） technique to retrieve the soil moisture and snow thickness of snowfield environment. Firstly, the layered geometry model of medium is constructed by the environment of snowfield, in order to obtain the back-scattering coefficient with small perturbation method （SPM）. After data sensitivity analysis,with the 1.27GHz radar frequency （L-band） and dual angle of incidence （42degree/53degree） selected, two types of inversion scheme are designed, hence the single polarization and the dual polarization are selected as the microwave signal sample information. Via appropriate training, the inversion of soil moisture and snow thickness is researched depending on the least squares support vector machine technique. The research results indicate that the inversion result has high accuracy using the dual polarization and incidence angle. Meanwhile,in the having noise situation, the good anti-noise ability is showed by the inversion result with the way of dual polarization and incidence angle,thus providing a feasible approach for the retrieval of soil moisture and snow thickness in the snowfield environment.