正则化总体最小二乘用于光学线阵遥感影像定位

发布时间：2019-03-04 13:35

【摘要】：针对光学线阵遥感影像几何定位中,系统误差改正参数对应系数矩阵存在误差的问题,提出了一种基于正则化总体最小二乘的光学线阵遥感影像定位方法。首先依据有理函数模型的定义,构建共线条件方程,利用线性化构建光学线阵影像定位方法和系统误差改正方法,然后依据EIV模型的定义和性质构建相应的优化目标函数,并引入正则化项,依据Lagrange条件极值原理推导基于正则化总体最小二乘的系统误差参数迭代估计方法。实验结果表明:与经典的最小二乘平差算法相比,该方法的总体定位精度提高了11.61%,且比Tikhonov正则化法的定位精度平均提高了6.06%。本文提出方法在不增加任何额外控制信息的情况下,是提高光学线阵影像定位精度的有效途径。
[Abstract]:Aiming at the error of the coefficient matrix corresponding to the system error correction parameters in the geometric positioning of optical linear remote sensing images, an optical linear array remote sensing image localization method based on the regularization population least squares is proposed in this paper. Firstly, according to the definition of rational function model, the collinear conditional equation is constructed, and the linear array image localization method and the system error correction method are constructed. Then, according to the definition and properties of the EIV model, the corresponding optimization objective function is constructed. The regularization term is introduced and the iterative estimation method of system error parameters based on regularization population least squares is derived based on the Lagrange conditional extremum principle. The experimental results show that, compared with the classical least square adjustment algorithm, the overall positioning accuracy of this method is increased by 11.61% and 6.06% higher than that of the Tikhonov regularization method on average. The method proposed in this paper is an effective way to improve the positioning accuracy of optical linear array images without adding any additional control information.
【作者单位】：信息工程大学地理空间信息学院;地理信息工程国家重点实验室;
【基金】：国家自然科学基金资助项目(No.41471387;No.41301526) 地理信息工程国家重点实验室开放基金资助项目(No.SKLGIE2015-M-3-1)
【分类号】：TP751

【相似文献】