董洲洋,徐卫明,庄昊,孟浩

• 1:海军大连舰艇学院军事海洋与测绘系
• 2:32023部队

摘要(Abstract)：

针对数学拟合法在进行全球定位系统(GPS)水准拟合时,因受自身模型限制,导致GPS水准拟合精度不高的问题,该文提出了一种基于EGM2008模型和深度学习的GPS水准拟合法。首先使用深度学习中的分段线性整流函数(ReLU)作为神经元激活函数加快网络的收敛速度,然后利用自适应矩估计函数(Adam)作为优化函数加速获取最优解,并采用正则化丢弃法(Dropout)增强深度学习网络的泛化能力。通过实测数据计算表明:该文方法相比常用的多项式拟合法,丘陵地区外符合精度提高了约65%,达到1.7 cm;高差变化较大的山地外符合精度提高了约90%,达到1.2 cm。

关键词(KeyWords)： EGM2008模型;GPS水准拟合;“移去-恢复”法;深度学习

Abstract:

Keywords:

基金项目(Foundation): 国家自然科学基金项目(61071006)

作者(Author): 董洲洋,徐卫明,庄昊,孟浩

DOI: 10.16251/j.cnki.1009-2307.2021.04.009

参考文献(References)：

• [1] 李建成，陈俊勇，宁津生，等.地球重力场逼近理论与中国2000似大地水准面的确定[M].武汉：武汉大学出版社，2003.(LI Jiancheng,CHEN Junyong,NING Jinsheng,et al.Approximation theory of Earth gravity field and determination of 2000 quasi geoid in China[M].Wuhan:Wuhan University Press,2003.)
• [2] 赫林，李建成，褚永海.1985国家高程基准与全球高程基准之间的垂直偏差[J].测绘学报，2016,45(7):768-774.(HE Lin,LI Jiancheng,CHU Yonghai.Vertical deviation between 1985 national elevation datum and global elevation datum[J].Acta Geodaetica et Cartographica Sinica,2016,45(7):768-774.)
• [3] 刘斌，郭际明，史俊波，等.利用EGM2008模型与地形改正进行GPS高程拟合[J].武汉大学学报(信息科学版),2016,41(4):554-558.(LIU Bin,GUO Jiming,SHI Junbo,et al.GPS height fitting using EGM2008 model and terrain correction[J].Geomatics and Information Science of Wuhan University,2016,41(4):554-558.)
• [4] 吴富梅，魏子卿.利用GNSS和EGM2008模型进行跨海高程传递[J].武汉大学学报(信息科学版),2016,41(5):698-703.(WU Fumei,WEI Ziqing.Sea crossing height transfer using GNSS and EGM2008 models[J].Geomatics and Information Science of Wuhan University,2016,41(5):698-703.)
• [5] 贾雪，徐炜，刘超，等.顾及地球重力场模型的GNSS高程转换方法[J].测绘科学，2019,44(5):14-20.(JIA Xue,XU Wei,LIU Chao,et al.GNSS elevation conversion method considering gravity field model[J].Science of Surveying and Mapping,2019,44(5):14-20.)
• [6] 梁伟，徐新禹，李建成，等.联合EGM2008模型重力异常和GOCE观测数据构建超高阶地球重力场模型SGG-UGM-1[J].测绘学报，2018,47(4):425-434.(LIANG Wei,XU Xinyu,LI Jiancheng,et al.Combining EGM2008 model gravity anomaly and GOCE observation data to build super-high-order earth gravity field model SGG-UGM-1[J].Acta Geodaetica et Cartographica Sinica,2018,47(4):425-434.)
• [7] 王乐洋，吴飞，吴良才.GPS高程转换的总体最小二乘拟合推估模型[J].武汉大学学报(信息科学版),2016,41(9):1259-1264.(WANG Leyang,WU Fei,WU Liangcai.Global least square fitting estimation model of GPS height conversion[J].Geomatics and Information Science of Wuhan University,2016,41(9):1259-1264.)
• [8] ZHU X X,TUIA D,MOU L C,et al.Deep learning in remote sensing:a comprehensive review and list of resources[J].IEEE Geoscience & Remote Sensing Magazine,2017,5(4):8-36.
• [9] 周志华.机器学习[M].北京：清华大学出版社，2018.(ZHOU Zhihua.Machine learning[M].Beijing:Tsinghua University Press,2018.)
• [10] 何海清，黄声享，陈婷.一种顾及粗差的径向神经网络高程曲面拟合法[J].武汉大学学报(信息科学版),2015,40(4):547-551.(HE Haiqing,HUANG Shengxiang,CHEN Ting.A radial neural network height surface fitting method considering gross errors[J].Geomatics and Information Science of Wuhan University,2015,40(4):547-551.)
• [11] ZAKI A,MANSI A H,SELIM M,et al.Comparison of satellite altimetric gravity and global geopotential models with shipborne gravity in the Red Sea[J].Marine Geodesy,2018,41(3):258-269.
• [12] 李建成，褚永海，徐新禹.区域与全球高程基准差异的确定[J].测绘学报，2017,46(10):1262-1273.(LI Jiancheng,CHU Yonghai,XU Xinyu.Determination of regional and global elevation datum differences[J].Acta Geodaetica et Cartographica Sinica,2017,46(10):1262-1273.)
• [13] 李建成.我国现代高程测定关键技术若干问题的研究及进展[J].武汉大学学报(信息科学版),2007,32(11):980-987.(LI Jiancheng.Research and progress on some key technologies of modern height determination in China[J].Geomatics and Information Science of Wuhan University,2007,32(11):980-987.)
• [14] MAKAN A,MEHDI A.Application of Molodensky’s method for precise determination of geoid in Iran[J].Journal of Geodetic Science,2011,1(3):259-270.
• [15] LI Jiancheng,JIANG Weiping.Long distance transference of height datum across seas[J].Geo-Spatial Information Science,2002,5(3):1-5.
• [16] 张兴福，刘成.综合EGM2008模型和SRTM/DTM2006.0剩余地形模型的GPS高程转换方法[J].测绘学报，2012,41(1):25-32.(ZHANG Xingfu,LIU Cheng.GPS height conversion method of integrated EGM2008 model and SRTM/DTM2006.0 residual terrain model[J].Acta Geodaetica et Cartographica Sinica,2012,41(1):25-32.)
• [17] FILMER M S,FEATHERSTONE W E,KUHN M.Erratum to:the effect of EGM2008-based normal,normal-orthometric and Helmert orthometric height systems on the Australian levelling network[J].Journal of Geodesy,2013,84(8):501-513.
• [18] 章传银，郭春喜，陈俊勇，等.EGM2008地球重力场模型在中国大陆适用性分析[J].测绘学报，2009,38(4):283-289.(ZHANG Chuanyin,GUO Chunxi,CHEN Junyong,et al.Applicability analysis of EGM2008 Earth gravity field model in China's mainland[J].Acta Geodaetica et Cartographica Sinica,2009,38(4):283-289.)
• [19] 蒋昂波，王维.ReLU激活函数优化研究[J].传感器与微系统，2018,37(2):50-52.(JIANG Angbo,WANG Wei.Research on the optimization of ReLU activation function[J].Transducer and Microsystem Technologies,2018,37(2):50-52.)
• [20] 陈振宏，兰艳艳，郭嘉丰，等.基于差异合并的分布式随机梯度下降算法[J].计算机学报，2015,38(10):2054-2063.(CHEN Zhenhong,LAN Yanyan,GUO Jiafeng,et al.Distributed random gradient descent algorithm based on difference merging[J].Chinese Journal of Computers,2015,38(10):2054-2063.)
• [21] 魏贞原.深度学习[M].北京：电子工业出版社，2018.(WEI Zhenyuan.Deep learning[M].Beijing:Publishing House of Electronics Industry,2018.)