基于广义回归神经网络的GPS高程转换

摘要

为提高GPS高程转换的精度,采用广义回归神经网络(GRNN)进行拟合.将控制点的X、y坐标作为网络输入,高程异常作为网络输出,采用实验数据训练网络,训练完成的网络作为模型进行高程异常预测.结果表明,GRNN方法具有较高的GPS转换精度.%To improve the accuracy of GPS height transform from geodetic height to normal height, General Regression Neural Network ( GRNN) was used for fitting. The X and Y coordinates of the control points were employed as the inputs of GRNN, and the elevation anomaly were the outputs of the neural network. We adopted experimental data for training the network, then, took the trained network as a model to complete the abnormal height prediction. The results show that the GRNN method is feasible and has the high accuracy of the GPS height transform.