用地球位模型和BP神经网络转换GPS高程  

Conversion of GPS Height Using Geopotential Model and BP Neural Networks

作　　者：张杰[1] 

机构地区：[1]武汉大学测绘遥感信息工程国家重点实验室,湖北武汉430079

出　　处：《测绘科学技术学报》2009年第6期407-409,413,共4页Journal of Geomatics Science and Technology

基　　金：国家自然科学基金资助项目(40674005);国家863计划资助项目(2009AA12Z318)

摘　　要：研究了转换GPS高程的地球位模型和BP神经网络的拟合方法。用已知GPS水准点的高程异常移去地球位模型高程异常,然后对剩余高程异常通过BP神经网络拟合和内插,在内插点上恢复地球位模型高程异常,从而得到该点的高程异常。通过实测GPS水准数据将该方法与基于地球位模型和二次曲面的拟合方法进行了比较。试验结果表明,该方法转换GPS高程的精度优于基于地球位模型和二次曲面的拟合方法,能够满足一定的工程应用需求。The method by remove restore technology for transforming GPS height was studied which took full advantage of virtues of the geopotential model and BP neural networks. The height abnormal values on the known points of GPS leveling were removed first by the geopotential model, the remained values were used for training BP neural networks, then the height abnormal value of any point was obtained by the trained BP neural networks and the geopotential model. By GPS leveling data, the methods of BP neural networks and conicoid fitting using geopotential model were analyzed. It showed that the method of BP neural networks using for GPS height conversion was superior to that of conicoid fitting using geopotential model.

关 键 词：地球位模型 神经网络 BP算法 GPS高程 高程转换 

分 类 号：P228[天文地球—大地测量学与测量工程]