正则化算法在遥感影像外定向中的应用  

Application of Regularization Methods to Remote Sensing Image Exterior Orientation

作　　者：贾秀鹏[1] 焦伟利[1] 王振杰[2] 

机构地区：[1]中国科学院中国遥感卫星地面站重点实验室 [2]山东理工大学建筑工程学院,淄博255049

出　　处：《遥感信息》2006年第4期49-52,共4页Remote Sensing Information

基　　金：国家自然科学基金项目(60272032)"卫星遥感信息智能处理的信息论方法研究"资助

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。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.

关 键 词：正则化 外定向 L-曲线法 两步解法 病态问题 

分 类 号：P237.9[天文地球—摄影测量与遥感]