BP神经网络在似大地水准面精化中的应用  

Application of BP Neural Network in the Quasi-geoid Refi nement

作　　者：陈淋[1] 邵曦[1] 

机构地区：[1]四川省冶金地质勘查局测绘工程大队,四川成都610212

出　　处：《地理空间信息》2014年第1期117-118,122,共3页Geospatial Information

基　　金：国家自然科学基金资助项目(41071294);广西自然科学基金资助项目(桂科自0640178)

摘　　要：利用分辨率为5'×5'的EGM2008模型计算高程异常,从水准联测点上移走模型中的高精度中长波部分后,通过BP神经网络模拟余下的短波部分,最后再对数据进行恢复,使实验达到较高的精度。通过实例对确定高程异常的多种方案进行比较分析,得出了一些有益的结论。The EGM2008 model with five point resolutions and the method of neural networks used for transforming GPS height were introduced briel y in this paper. The processes took full advantages of the two methods and were divided into three steps. Firstly, the height abnormal values on the known points were removed by EGM2008.Secondly, the remained values were calculated by the method of neural networks. Finally, the height values were resumed. A few practicable conclusions were received from the research. Analysis of the examples of the variety of programs to determine the Height anomaly, we obtained some useful conclusions and cognition.

关 键 词：EGM2008模型 移去-恢复技术 BP神经网络 F-R共轭梯度法 

分 类 号：P223.0[天文地球—大地测量学与测量工程]