RBF神经网络在GPS高程拟合中的应用

Application of RBF Neural Networks to GPS Leveling Elevation Fitting

通过对离散GPS/水准点观测数据进行拟合从而获得区域内任意一点的高程异常是工程实践中经常遇到的问题。利用RBF神经网络方法进行了GPS水准高程拟合实验,并将得到的高程异常结果与采用BP神经网络方法和二次曲面拟合法得到的结果进行了分析比较;通过3种方式的分析比较,证明利用RBF神经网络进行GPS高程拟合的可行性以及相比其他方法所具备的优势。

To determine an orthometric height using GPS,it is necessary to know the geoid/quasi-geoid undulation.This paper had a detail introduction of the structure and basic principle of RBF neural network,and on this basis,mainly analyzed RBF neural network's algorithms for GPS elevation fitting,including basis function centers and width parameters determined.Based on the MATLAB,selecting some GPS datas from a region,this paper mainly focused on the use of RBF neural network for GPS elevation fitting,the result of which was compared with BP neural network and fitting method,and the accuracy of the convertion of GPS elevation was analysed last.Through the analysis and comparison of this three ways,this paper tried to prove that using RBF neural network for GPS elevation fitting was of feasibility and had some advantages compared to other methods.