利用RBF实现线状工程中GNSS高程转换的精度分析
|
摘要
高程异常的转换精度是限制GNSS实现快速高程测量的关键因素之一,因此构建合理的高程异常转换模型具有重要意义。本文以美国GSVS2011项目的 GNSS/水准数据为例,对利用RBF实现GNSS高程转换方法展开研究,详细分析了Thinplate(T)、Cubic(C)、Gaussian(G)、Multiquadrics(M)、Linear(L) 5种RBF核函数的高程转换精度。试验结果表明:(1)在点间距较小的情况下,5种核函数均能获得较高的拟合精度;(2)T、C、M这3种核函数拟合效果较优,而且超过95.3%的拟合高差满足四等水准测量的限差要求;(3)在高程异常变化较为平坦的测区,用RBF拟合时建议点间距不宜大于30 km。
The application of GNSS rapid height measurement is mainly affected by accuracy of height anomaly transformation.Therefore,it is rather important to construct a reasonable height anomaly transformation model.This paper discusses the RBF method of GNSS height transformation with GSVS2011 GNSS/leveling data in the United States,the accuracy of height transformation based on five RBF kernel functions such as Thinplate(T),Cubic(C),Gaussian(G),Multiquadrics(M),Linear(L) is analyzed in details. The experimental results show that(1)When the distance between GNSS/leveling points is small,the five kernel functions can also obtain higher fitting accuracy.(2)Three kinds of kernel functions(T、C、M) have a better fitting effect,and more than 95.3% of the fitted height difference meets fourth-order level tolerance.(3)It is suggested that the point interval should not be greater than 30 km in the survey area where the height anomaly changes relatively flatly when fitting with RBF.
The application of GNSS rapid height measurement is mainly affected by accuracy of height anomaly transformation.Therefore,it is rather important to construct a reasonable height anomaly transformation model.This paper discusses the RBF method of GNSS height transformation with GSVS2011 GNSS/leveling data in the United States,the accuracy of height transformation based on five RBF kernel functions such as Thinplate(T),Cubic(C),Gaussian(G),Multiquadrics(M),Linear(L) is analyzed in details. The experimental results show that(1)When the distance between GNSS/leveling points is small,the five kernel functions can also obtain higher fitting accuracy.(2)Three kinds of kernel functions(T、C、M) have a better fitting effect,and more than 95.3% of the fitted height difference meets fourth-order level tolerance.(3)It is suggested that the point interval should not be greater than 30 km in the survey area where the height anomaly changes relatively flatly when fitting with RBF.
引文
[1]孙佳龙,焦明连,郭淑艳.基于双调和样条内插和高斯曲率极值的多面函数拟合高程异常方法[J].测绘通报,2014(6):14-16.
[2]王珍,程垒.GPS高程拟合方法的研究和应用[J].测绘通报,2015(S1):99-102.
[3]周长志,孙佳龙,郭淑艳.基于移动-多面函数的高程异常拟合方法[J].测绘通报,2016(12):25-27.
[4]王小辉,王琪洁,丁元兰,等.基于二次曲面和BP神经网络组合模型的GPS高程异常拟合[J].大地测量与地球动力学,2012,32(6):103-105.
[5]赵建虎,王真祥,王胜平,等.大区域似大地水准面模型的建立方法研究[J].武汉大学学报(信息科学版),2008,33(1):68-71.
[6]马鸿炜,郭金运,李国伟,等.一种区域似大地水准面模型拟合方法分析[J].海洋测绘,2014,34(1):8-11.
[7]张兴福,刘成,王国辉,等.基于EGM2008重力场模型的GPS高程转换方法及精度分析[J].地球物理学进展,2012,27(1):38-44.
[8]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.
[9]罗陶荣,王中元,梁宁,等.利用EGM2008模型与加权组合模型进行高程异常拟合[J].测绘通报,2018(1):28-32.
[10]邱斌,朱建军.全球定位系统高程转换的RCR算法[J].测绘通报,2004(7):16-18.
[11]刘晓刚,刘雁雨,曹纪东,等.GPS水准采用移去恢复技术拟合大地水准面方法的研究[J].测绘工程,2008,17(3):70-73.
[12]管真,陈剑杰,尉伯虎,等.基于EGM2008的GPS水准拟合在复杂地形中的应用研究[J].大地测量与地球动力学,2012,32(4):120-124.
[13]LEE C M,KO C N.Time Series Prediction Using RBFNeural Networks with a Nonlinear Time-varying Evolution PSO Algorithm[J].Neurocomputing,2009,73(1):449-460.
[14]刘永,张立毅.BP和RBF神经网络的实现及其性能比较[J].电子测量技术,2007,30(4):77-80.
[15]苏美娟.径向基函数神经网络学习算法研究[D].苏州:苏州大学,2007.
[16]潘琪.径向基神经网络基函数中心确定方法改进研究[D].哈尔滨:东北农业大学,2017.
[2]王珍,程垒.GPS高程拟合方法的研究和应用[J].测绘通报,2015(S1):99-102.
[3]周长志,孙佳龙,郭淑艳.基于移动-多面函数的高程异常拟合方法[J].测绘通报,2016(12):25-27.
[4]王小辉,王琪洁,丁元兰,等.基于二次曲面和BP神经网络组合模型的GPS高程异常拟合[J].大地测量与地球动力学,2012,32(6):103-105.
[5]赵建虎,王真祥,王胜平,等.大区域似大地水准面模型的建立方法研究[J].武汉大学学报(信息科学版),2008,33(1):68-71.
[6]马鸿炜,郭金运,李国伟,等.一种区域似大地水准面模型拟合方法分析[J].海洋测绘,2014,34(1):8-11.
[7]张兴福,刘成,王国辉,等.基于EGM2008重力场模型的GPS高程转换方法及精度分析[J].地球物理学进展,2012,27(1):38-44.
[8]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.
[9]罗陶荣,王中元,梁宁,等.利用EGM2008模型与加权组合模型进行高程异常拟合[J].测绘通报,2018(1):28-32.
[10]邱斌,朱建军.全球定位系统高程转换的RCR算法[J].测绘通报,2004(7):16-18.
[11]刘晓刚,刘雁雨,曹纪东,等.GPS水准采用移去恢复技术拟合大地水准面方法的研究[J].测绘工程,2008,17(3):70-73.
[12]管真,陈剑杰,尉伯虎,等.基于EGM2008的GPS水准拟合在复杂地形中的应用研究[J].大地测量与地球动力学,2012,32(4):120-124.
[13]LEE C M,KO C N.Time Series Prediction Using RBFNeural Networks with a Nonlinear Time-varying Evolution PSO Algorithm[J].Neurocomputing,2009,73(1):449-460.
[14]刘永,张立毅.BP和RBF神经网络的实现及其性能比较[J].电子测量技术,2007,30(4):77-80.
[15]苏美娟.径向基函数神经网络学习算法研究[D].苏州:苏州大学,2007.
[16]潘琪.径向基神经网络基函数中心确定方法改进研究[D].哈尔滨:东北农业大学,2017.