【摘要】 正则化的目的是在求解过程中对方程的解进行约束,得到稳定可用的解。遥感影像外定向是确定外方位元素,它是正射纠正和三维信息提取的基础。在计算卫星遥感影像外方位元素时存在病态问题。本文使用两种正则化方法求解这一病态问题:Tikhonov正则化和两步解法。这两种方法与主成分估计法和岭-压组合估计法进行了比较,实验结果表明,基于GCV参数选取方法的Tikhonv正则化方法和主成分估计及两步解法求解结果较优。由不同精度的控制点进行外定向的检验精度表明,除了应用较优的算法,精确的控制点对外定向也很重要。更多还原

【Abstract】 The purpose of regularization is to incorporate further information in order to restrain the problem and to get a useful and stable solution.Remote sensing image exterior orientation,which is to compute the exterior orientation elements,is fundamental work of orthorectification and three-dimensional information extraction.The ill-posed problem is often confronted when computing the satellite remote sensing image orientation elements.In this paper,two regularization methods are used to solve this ill-posed problem: Tikhonov regularization and two steps method.These two methods are compared with the methods of principal component estimator and combined ridge-stein estimator.The experiments indicate that the Tikhonov regularization based on GCV,principal component estimator and two steps method are better.The accuracy of the exterior orientation computed by control points with different accuracies indicates that accurate control points are also important to exterior orientation besides better algorithm.更多还原