一种改进的局部线性回归估计器及其在雷达高度计海况偏差估计中的应用

Improved Local Linear Regression Estimator and Its Application to Estimation for Radar Altimeter Sea State Bias

在建立雷达高度计海况偏差(Sea State Bias,SSB)非参数模型时,通常会用到局部线性回归(Local Linear Regression,LLR)估计器,而传统的局部线性回归估计器涉及高维矩阵运算,当建模的数据量较大时,估计海况偏差需要大量的时间,从而使得非参数估计方法很难用于高维海况偏差模型。该文提出一种改进的局部线性回归(Improved Local Linear Regression,ILLR)估计器,可以避免传统的LLR估计器所需的高维矩阵运算,在不影响海况偏差估计结果的条件下,将局部线性回归估计器获取加权函数的时间复杂度由O(N)2降低为O(N),从而大幅地降低估计海况偏差所需的时间,为实现高维非参数海况偏差模型的实时运算奠定了基础。

The Local Linear Regression （LLR） estimator is usually used when developing a nonparametric model for radar altimeter Sea State Bias （SSB）. However, the conventional LLR estimator contains matrices with high dimension. When a large number of data are used to develop the SSB model, the SSB estimation costs too much time. Therefore, the nonparametric estimation method can hardly be used to develop high-dimensional SSB models This paper presents an Improved LLR （ILLR） estimator, complexity from O（N2） to O（N） which can avoid high-dimensional matrix operations. The improved LLR estimator can greatly reduce the time for SSB estimation without affecting the estimated accuracy. So the improved LLR estimator can laid the foundation for high-dimensional SSB models.