一种IGrubbs-LWLR的区域高程异常拟合方法  

A Regional Elevation Anomaly Fitting Method of IGrubbs-LWLR

作　　者：张炎 刘立龙 蒙金龙 徐勇 何广焕 胡鹏程 ZHANG Yan;LIU Lilong;MENG Jinlong;XU Yong;HE Guanghuan;HU Pengcheng(College of Geomatics and Geoinformation,Guilin University of Technology,Guilin 541006,China;Guangxi Key Laboratory of Spatial Information and Geomatics,Guilin 541006,China;School of Municipal Construction and Transportation,Guangxi Polytechnic of Construction,Nanning 530007,China)

机构地区：[1]桂林理工大学测绘地理信息学院,广西桂林541006 [2]广西空间信息与测绘重点实验室,广西桂林541006 [3]广西建设职业技术学院市政与交通学院,广西南宁530007

出　　处：《无线电工程》2023年第10期2345-2351,共7页Radio Engineering

基　　金：国家自然科学基金(42064002);广西自然科学基金(2020GXNSFBA297160,2021GXNSFBA220046);广西中青年教师基础能力提升项目(2021KY0268)。

摘　　要：针对测量数据中粗差干扰及高程异常拟合方法选择较为困难的问题,结合格拉布斯(Grubbs)法判别粗差的原理,提出一种改进格拉布斯(Improved Grubbs,IGrubbs)结合局部加权线性回归(Local Weighted Linear Regression,LWLR)的拟合模型构建法。在原Grubbs法则的基础上,引入自适应迭代,在训练数据中,对粗差进行识别,并设定粗差剔除完成的指标参数,从而降低原方法中发生误判或漏判的概率,并利用局部加权线性回归法通过预处理后的训练样本数据来建立区域高程异常拟合模型。实验结果表明,相较于传统Grubbs法则,改进后的Grubbs法对于高程异常数据中的粗差剔除更为快速有效,且利用局部加权线性回归法所构建的区域高程异常拟合模型的预测精度及稳定性也得到一定程度的提高,对今后工程中的测高工作具备一定的参考意义。To overcome the difficulty in choosing a fitting method for elevation anomalies and disturbance of gross error in the measurement data,combined with the principle of Grubbs method for determining gross error,a fitting model construction method for Improved Grubbs(IGrubbs)combined with Local Weighted Linear Regression(LWLR)is proposed.On the basis of the original Grubbs criterion,the adaptive iteration is introduced,and the gross error is identified and the indicator parameters for the gross error elimination are set in the training data,so as to reduce the probability of misjudgment or missed judgment in the original method.The regional elevation anomaly fitting model is established by LWLR in the pre-processed training sample data.Experimental results show that compared with the traditional Grubbs criterion,the improved Grubbs is faster and more effective for the gross error elimination in elevation anomaly data,and the prediction accuracy and stability of the regional elevation anomaly fitting model constructed by using LWLR are also improved to a certain extent,which has certain reference significance for the altimetry work in future engineering.

关 键 词：改进格拉布斯法则 局部加权线性回归法 粗差剔除 区域高程异常拟合 

分 类 号：P228[天文地球—大地测量学与测量工程]