基于神经网络的似大地水准面模型精化方法研究

论文总字数：39239字

摘 要

利用似大地水准面精化模型，可以精确的将GPS测定大地高转换为正常高，进而满足高程测量的需求。本文基于二阶多项式模型和BP神经网络模型，利用GPS测量的大地高和水准测量的正常高，进行似大地水准面的精化，并利用软件实现两种模型的计算。创新性将EGM2008信息纳入到模型中，分别建立了六种模型：二阶多项式拟合模型（A）、顾及EGM2008信息的二阶多项式模型（B）、BP神经网络模型（C）、顾及EGM2008信息的BP神经网络模型（D）、二阶多项式与BP神经网络的融合模型（E）、顾及EGM2008信息的二阶多项式与BP神经网络的融合模型（F）。利用阿联酋地区的数据，分析各种精化模型的精度。其中模型A精度为11.60cm，模型B精度3.81cm，比模型A提高了67.2%，模型F精度为3.42cm，比模型A提高了70.5%。在六种模型中，顾及EGM2008信息的二阶多项式模型与BP神经网络的融合模型的效果最好。

关键词：似大地水准面精化，高程异常，EGM2008重力场模型，二阶多项式拟

合法，BP神经网络

Abstract

High-precision, high-resolution Quasi-geoid model, which can be taken as

reference to the normal height measurement framework, service for engineering

construction. In this paper,by using various mathematical model basing on GPS

leveling results to detail the Quasi-geoid,including second-order polynomial model and BP neural network model,and using the software to calculate the model.Combining with the abnormal gravity data from EGM2008, established six models:general second-order polynomial fitting model,second-order polynomial fitting

mode combined with the abnormal gravity data from EGM2008 gravity field model,general BP neural network model,BP neural network model combined with

the gravity abnormaldata from EGM2008 gravity field model,BP neural networkmodel combined with the second-order polynomial fitting result,BP neural network model combined with the gravity abnormaldata from EGM2008 gravity field model and second-order polynomial fitting result.The paper combined the engineering examples to detail the Quasi-geoid model.In the case,general second-order polynomial fitting model’s precision is 11.60 cm, second-order polynomial

fitting mode combined with the abnormal gravity data from EGM2008 gravity field model’s precision is 3.81 cm, which increased by 67.2%. The BP neural

network model combined with the gravity abnormaldata from EGM2008 gravityfield model and second-order polynomial fitting result’s accuracy is 3.42 cm,

which increased by 70.5%. In the six models,BP neural network model combined with the gravity abnormaldata from EGM2008 gravity field model and second-order polynomial fitting result works best.

.Keywords:Quasi-geoid;abnormal height;EGM2008 gravity field model;

second-order polynomial fitting;BP neural Network

目 录

摘 要................ ...................................I

Abstract.......... ......................................II

目 录............... ..................................III

图 索 引............... ...................................V

表 索 引............... ..................................VI

第一章 绪论........ .....................................1

1.1 研究背景与意义..............................................1

1.2 国内外似大地水准面精化研究动态..............................1

1.2.1 国外研究动态..........................................2

1.2.2 国内研究动态..........................................2

1.3 本文主要研究的内容..........................................3

第二章 似大地水准面精化的理论与方法..........................4

2.1 概述........................................................4

2.2 似大地水准面精化理论........................................4

2.2.1 Stokes理论与大地水准面................................5

2.2.2 Molodensky理论与似大地水准面..........................6

2.3 统计方法确定似大地水准面....................................7

2.3.1 最小二乘配置法........................................7

2.3.2 移去-恢复法...........................................8

2.4 几何方法精化似大地水准面....................................8

2.4.1 GPS/水准测量法........................................8

2.4.2 天文大地水准测量法....................................9

2.4.3 卫星无线电测高法......................................9

2.5 GPS/水准数据拟合方法........................................9

2.5.1 函数模型法............................................9

2.5.2 BP神经网络方法.......................................11

第三章 似大地水准面精化模型............ ......................13

3.1 似大地水准面精化基本技术规定...............................13

3.1.1 外业观测技术要求.....................................13

3.1.2 数据处理技术要求.....................................13

3.1.3 似大地水准面精化精度与分辨率要求.....................14

3.2 传统似大地水准面精化模型...................................14

3.2.1 二阶多项式拟合模型A................................ .15

3.2.2 BP神经网络模型C......................................16

3.2.3 方法A与BP神经网络的融合模型E.......................17

3.3 顾及EGM2008信息的似大地水准面模型............ .............18

3.3.1 EGM2008重力场模型简介................................18

3.3.2 顾及EGM2008重力场模型的二阶多项式模型B..............19

3.3.3 顾及EGM2008重力场模型的BP神经网络模型D.............20

3.3.4 方法B与BP神经网络的融合模型F.......................20

第四章 工程实例分析........................................ ...22

4.1 工程实例概况...............................................22

4.2 数据处理...................................................24

4.2.1 数据质量分析.........................................24

4.2.2 样本确定.............................................25

4.3 传统似大地水准面精化模型精度分析...........................26

4.3.1 二阶多项式拟合模型A............................. ....26

4.3.2 BP神经网络模型C......................................27

4.3.3 方法A与BP神经网络的融合模型E.......................28

4.4 顾及EGM2008信息的似大地水准面模型精度分析.................30

4.4.1 顾及EGM2008重力场模型重力异常的二阶多项式模型B......30

剩余内容已隐藏，请支付后下载全文，论文总字数：39239字