基于Cook距离的协方差函数拟合点优选  

Optimal selection of covariance function fitting points based on Cook distance

作　　者：李成仁[1] 陆游 袁豹[1] 

机构地区：[1]河海大学地球科学与工程学院,江苏南京210098 [2]广东天信电力工程检测有限公司,广东广州510600

出　　处：《测绘工程》2014年第5期21-23,共3页Engineering of Surveying and Mapping

摘　　要：进行最小二乘配置中的协方差函数拟合时,选取的拟合点应尽可能反映该区域的异常位的实际情况。提出基于Cook距离的协方差函数拟合点优选,计算所有拟合点的Cook距离,剔除拟合模型中的强扰动点,再进行协方差函数拟合。结合GPS高程拟合实例分析,文中方法可以有效提高高程拟合精度,并且通过Cook距离可以对拟合点进行精简和检出拟合点中误差较大点或粗差点。When fitting covariance function of least squares collocation, fitting points should be selected as far as possible to reflect the actual situation in the region exception. Optimal selection of covariance function fitting points based on cook distance is proposed, by calculating Cook distance of all fitting points, and excluding the strong disturbances in the fitting model to fitting covariance function. By the example of GPS height fitting, the results show that this method can effectively improve the fitting accuracy, and that by Cook distance the fitting points can be streamlined, and points with large errors can be checked out.

关 键 词：COOK距离 优选 最小二乘配置 协方差函数 高程拟合 

分 类 号：P207[天文地球—测绘科学与技术]