一种裸露土壤湿度反演方法

Inversion of bare soil moisture based on least squares support vector machine

针对目前土壤湿度反演方法研究较少且缺少实时性的现状,该文提出一种土壤湿度反演方法——最小二乘支持向量机技术。以积分方程模型为正向算法,数值模拟不同雷达参数(频率、入射角及极化)下后向散射系数随土壤含水量和地表粗糙度的变化情况。经过数据敏感性分析,选取C-波段和X-波段、小入射角下的同极化后向散射系数作为支持向量回归的训练样本信息;经过适当的训练,利用支持向量回归技术对土壤含水量进行了反演研究;并考虑通过多频率、多极化、多入射角数据的组合,消除地表粗糙度的影响,提高反演精度。模拟结果表明,该方法反演土壤湿度具有较高的精度和较好的实时性;同时,与人工神经网络方法的结果比较,证明了该方法的有效性,为土壤湿度的反演研究提供了一种方法。

According to the fact that there is no enough approach and real-time of soil moisture inversion,an inversion method for bare soil moisture was presented by using least squares support vector machine（LS-SVM）technique.Based on the integral equation model（IEM）,the backscatter coefficient with the change of soil moisture content and surface roughness were numerical simulated in different radar parameters（frequency,incidence angle and polarization）.After data sensitivity analysis,backscattering coefficient of C-band and X-band,with small incidence angle were selected as the support vector regression training sample information.After appropriate training,least squares support vector regression techniques was adopted to provide estimation of soil moisture under different inversion scheme.In order to eliminate the influence of surface roughness and improve the inversion accuracy,the combination of multiple frequency,multi-polarization,and incident angle data were considered.The results of simulation indicated that the approach could inverse soil moisture with higher accuracy and better real-time.The effectiveness of proposed method was proved to be useful for building soil moisture inversion research through comparing with the results of the artificial neural network method.