基于支持向量机的粗糙海面风速及海表盐度反演研究

Inversion study of the rough sea surface wind speed and sea surface salinity based on the support vector machine

将支持向量机(Support Vector Machine,SVM)回归技术应用到海况参数(如海表盐度、海面风速等)反演研究.利用双尺度模型(Two-Scale Model,TSM)作为前向电磁算法,数值模拟不同雷达参数下风驱粗糙海面微波后向散射系数,经过敏感性分析,选取L波段(1.4GHz)、C波段(6.8GHz)及其合适的入射角作为雷达参数,并设计多种反演方案,分别以单频率双极化双角度、双频率双极化双角度及双极化后向散射系数的比值作为SVM的训练样本数据信息,经过适当的训练,利用SVM回归技术对海洋表面风速和盐度进行了反演研究.研究结果表明,针对于海面风速的反演,C波段的反演精度最高,针对于海表盐度的反演,L波段同极化散射系数比值作为SVM输入的反演精度较高.最后,检验了SVM反演方法的抗噪声性能,表明文中提出的SVM方法能较好地应用于实际海况参数反演问题.

In this paper,the support vector machine（SVM）regression techniques are applied to the inversion of sea state parameters （e.g.salinity and wind speed of the sea surface）.The two scale model （TSM）is used to simulate backscattering coefficients of the rough sea surface with different radar parame-ters.After the sensitivity analysis,the L band （1.4 GHz）and the C band （6.8 GHz）are selected with ap-propriate angles as radar parameters.Then a variety of schemes of inversion are designed,in which single-frequency dual-polarization double angle,dual-frequency dual-polarization double angle and the ratio be-tween the VV and HH polarization backscattering coefficients are chosen respectively as the samples infor-mation.After appropriate training,the SVM forecasting model is applied to inverse the salinity and wind speed of the sea surface.As shown by the results,at the C band,the inversion of the sea surface wind speed bears the highest accuracy,whereas at the L band,the inversion of the sea surface salinity demon-strates the highest accuracy when the ratio between backscattering coefficients is chosen as the samples in-formation.The anti-noise performance of the SVM model is also examined,and the results show that the SVM model performs favorably in the sea state parameter inversion problem.