基于BP神经网络似大地水准面精化精度分析  

Precision Analysis of Quasi Geoid Refinement Based on BP Neural Network

作　　者：房新玉[1,2] 解静 刘盾 崔嘉 FANG Xinyu;XIE Jing;LIU Dun;CUI Jia(Tianjin Research Institute forWater Transport Engineering,Ministry of Transport,Tianjin 300456,China;Tianjin Key Laboratory of Surveying and Mapping forWaterway Transport Engineering,Tianjin Survey and Design Institute forWater Transport Engineering Co.,Ltd.,Tianjin 300456,China)

机构地区：[1]交通运输部天津水运工程科学研究所,天津300456 [2]天津水运工程勘察设计院有限公司天津市水运工程测绘技术重点实验室,天津300456

出　　处：《地理空间信息》2023年第11期73-75,共3页Geospatial Information

基　　金：中央级公益性科研院所基本科研业务费专项基金资助项目(TKS20220302);天津水运工程勘察设计院有限公司科研创新基金(发展基金)资助项目(SJY202106)。

摘　　要：针对BP神经网络似大地水准面精化在山区和平原等地形区域的差异,采用BP神经网络+残余高程异常的方法,建立残余高程异常与坐标的非线性函数关系,并以南方某山区和华北某平原为研究对象,对比分析了BP神经网络似大地水准面精化不同地形区域的精度。结果表明,在平原和山区均能达到较高的精度(±5 cm以内),且平原优于山区,因此BP神经网络可应用于大面积的不同地形区域的似大地水准面精化,精度可满足1∶500~1∶2000大比例尺地形图测量的需求。Aiming at the differences of quasi geoid refinement based on BP neural network in mountainous and plains,we used a method combin-ing BP neural network with residual elevation anomaly to establish the nonlinear function relationship between residual elevation anomaly and coordinates.Then,taking a southern mountain area and a plain in north China as the research objects,we compared and analyzed the precision of quasi geoid refinement based on BP neural network in different terrain regions.The results show that quasi geoid refinement based on BP neural network can reach a high precision within±5 cm both in a mountain area and in a plain,and the precision in a plain is higher than that in a moun-tain area.Therefore,BP neural network can be further used for the quasi geoid refinement in different kinds of terrain,and its precision can meet the needs of 1∶500~1∶2000 large-scale topographic map survey.

关 键 词：BP神经网络 似大地水准面精化 EGM2008模型 高程异常 

分 类 号：P223[天文地球—大地测量学与测量工程]