融合重力信息的似大地水准面精化模型

Quasi-geoid refinement model fused with gravity information

以二次多项式模型为基础,提出融合重力信息的多项式模型和融合重力信息的误差反向传播(error back propagation,BP)神经网络模型2种似大地水准面精化模型。通过试验验证,3种模型的检验点拟合中误差分别为±11.7 cm、±4.3 cm和±4.0 cm,重力信息可提高似大地水准面模型的拟合效果,BP神经网络模型的拟合精度最高,与二次多项式模型相比提高了65%。

Based on the quadratic polynomial model,two kinds of Quasi-geoid refinement models are proposed:a polynomial model fused with gravity information,and a BP neural network methodfused with gravity information.Through the test verification,the checkpoint fitting error of the three models is±11.7 cm,±4.3 cm and±4.0 cm respectively,the gravity information can improve the fitting effect of the Quasi-geoid model,and the fitting accuracy of the BP neural network model is the highest,which is 65%higher than that of the quadratic polynomial model.