The Discussion on BP Neural Network for GPS Height Transformation

摘要

BP neural network is a new method to convert GPS height in recent year. This method has higher precision, which is superior to quadric surface fitting in some cases. It has certain practical value for the normal height obtained from the method and can meet various large-scale-map requirements. However, there are some problems, such as how to determine the number of the hidden layers for a network and the unit for each layer and the affect of accuracy bringing about by the quality of the samples to study etc. Through analyzing the problems combined with the example, some useful conclusion is obtained in this paper.