基于改进的二次曲面与BP神经网络组合模型的GNSS高程异常拟合  

GNSS Height Abnormal Fitting Based on Modified Combined Model of Quadric Surface and BP Neural Network

作　　者：廖勋 

机构地区：[1]湖南省第一测绘院,湖南 长沙

出　　处：《应用数学进展》2022年第8期5285-5292,共8页Advances in Applied Mathematics

摘　　要：针对GNSS高程异常拟合模型未能很好拟合高程异常,本文在传统二次曲面模型中,引入一个附加高程的趋势项,将其与BP神经网络进行组合建立组合模型,并应用于高程异常拟合计算实例中。通过实例,将二次曲面拟合模型、改进的二次曲面拟合模型、半参数平差模型以及BP神经网络模型与文中提出的组合模型进行比较分析。结果表明组合模型推估外部点精度最高。In view of the fact that the GNSS elevation anomaly fitting model fails to fit the elevation anomaly well, this paper introduces an additional elevation trend term in the traditional quadratic surface model, combines it with the BP neural network to establish a combined model, and applies it to the elevation anomaly fitting calculation example. Through examples, the quadratic surface fitting model, the improved quadratic surface fitting model, the semi-parametric adjustment model and the BP neural network model are compared and analyzed with the combined model proposed in this paper. The results show that the combined model has the highest accuracy in estimating ex-ternal points.

关 键 词：改进的二次曲面高程异常 BP神经网络 组合模型 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]