基于地球重力场模型和BP神经网络的GNSS高程异常拟合
GNSS Height Anomaly Fitting Based on Earth Gravity Field Models and BP Neural Network
范思思 1范喆宇2
作者信息
- 1. 四川省地质调查研究院测绘地理信息中心,四川成都 610072
- 2. 四川英徕测绘技术有限公司,四川成都 610031
- 折叠
摘要
GNSS高程异常的高精度确定在精密工程测量中具有重要应用价值.为系统比较不同地球重力场模型和拟合方法在GNSS高程异常拟合中的精度情况,通过某一工程算例,比较了 EGM2008、XGM2019e、EIGEN-6C4和SGG-UGM-2四种高阶次地球重力场模型计算该区域模型高程异常的精度,同时对比了传统的三次曲面拟合和BP神经网络拟合的精度.实验表明,在该计算区域中,并不是重力场模型阶次越高精度就越好,EIGEN-6C4模型计算的高程异常精度最高,RMS能达到35.2 cm;BP神经网络的拟合结果相对于三次曲面拟合亦有显著提升,外符合精度最优可达到1.28 cm.因此,结合高精度地球重力场模型与BP神经网络方法有望得到较高精度的高程异常,可用于其他工程算例的研究.
Abstract
The high-precision determination of GNSS height anomalies has important application value in precise engineering surveying.In order to systematically compare the accuracy of different Earth gravity field models and fitting methods in GNSS height anomaly fitting,this paper chooses to compare the accuracy of four high-degree/order Earth gravity field models,EGM2008,XGM2019e,EIGEN-6C4,and SGG-UGM-2,in calculating the height anomaly of the regional model through a certain engineering example.At the same time,the accuracy of traditional cubic surface fitting and BP neural network fitting is compared.Experiments have shown that in this calculation area,the higher the order of the gravity field model,but not the better the accuracy.The EIGEN-6C4 model has the highest accuracy in calculating height anomalies,with an RMS of 35.2 cm.The fitting results of BP neural network have also significantly improved compared to cubic surface fitting,and the optimal external fitting accuracy can reach 1.28 cm.Therefore,combining high-precision Earth gravity field models with BP neural network methods is expected to obtain high-precision height anomalies for the study of other engineering cases.
关键词
地球重力场模型/BP神经网络/GNSS高程异常/拟合Key words
Earth gravity field model/BP neural network/GNSS height anomaly/fitting引用本文复制引用
出版年
2024
NETL
NSTL
万方数据