基于Matlab和RBF网络的GPS高程转换程序开发及应用

Development and application of GPS elevation conversion program based on Matlab and RBF network

由于GPS高程是以WGS-84参考椭球面为基准面的大地高,使用不方便,因此GPS高程转换模型和方法得到广泛的研究和应用。由于径向基神经网络(简称RBF网络)在函数逼近方面具有全局逼近和最佳逼近的性质,根据其基本原理,采用最近邻聚类分析算法,增加神经网络的输入神经元,综合水准测量获得的正常高和GPS测量获得的平面坐标和大地高程,开发基于RBF网络的GPS高程转换程序,分析和解决了程序设计和开发中的一些关键问题,提高了最邻近聚类算法的精度。

GPS elevation is a geodetic elevation based on WGS-84 reference ellipsoid,which is inconvenient to use.Therefore,GPS elevation transformation model and method have been widely studied and applied.Because RBF neural network(RBF network)with the properties of global approximation and the optimal approximation in terms of function approximation,according to its basic principle,using the nearest neighbor clustering analysis algorithm,increase of the neural network input neurons,comprehensive level for normal high and GPS measurement plane coordinates and geodetic height,GPS elevation conversion program is developed based on RBF network.Some key problems in program design and development are analyzed and solved,and the precision of nearest neighbor clustering algorithm is improved.