基于BP神经网络组合模型的GPS高程拟合  

GPS Height Fitting Based on BP Neural Network Combination Model

作　　者：王杰 盛君 孙晨辉 徐有威 洪年祥 WANG Jie;SHENG Jun;SUN Chenhui;XU Youwei;HONG Nianxiang(Zhejiang Institute of Surveying and Mapping Science and Technology,Hangzhou 310000,China;Hangzhou Tiantu Geographic Information Technology Company,Hangzhou 310012,China)

机构地区：[1]浙江省测绘科学技术研究院,浙江杭州310000 [2]杭州天图地理信息技术公司,浙江杭州310012

出　　处：《测绘与空间地理信息》2024年第7期89-92,95,共5页Geomatics & Spatial Information Technology

摘　　要：为了提高BP(backpropagation)神经网络的高程拟合精度,本文在BP神经网络高程拟合方法的基础上引入了模拟退火算法(SimulatedAnnealing,SA),组成新的SA-BP高程拟合方法。该组合方法充分发挥了SA算法的全局寻优的优势,优化了BP神经网络拟合方法的初始值以及权值与阈值。将本文提出的组合高程拟合方法应用于平坦测区及复杂测区实测GPS水准点高程数据中,实验结果表明,本文提出的组合高程拟合方法可对实验高程数据进行有效拟合,拟合精度较传统的曲面拟合方法与BP神经网络方法更高,验证了本文提出方法的可靠性、优越性以及针对不同地势条件的良好适应性。In order to improve the height fitting accuracy of BP(back propagation)neural network,this paper introduces simulated annealing(SA)algorithm based on BP neural network height fitting method to form a new SA-BP height fitting method.The combina-tion method gives full play to the advantage of SA algorithm in global optimization,and optimizes the initial value,weight and thresh-old of BP neural network fiting method.The combined height ftting method proposed in this paper is applied to the measured GPS benchmark elevation data in flat survey area and complex survey area.The experimental results show that the combined height fiting method proposed in this paper can effectively fit the experimental elevation data,and the fitting accuracy is higher than the traditional surface ftting method and BP neural network method,which verifies the reliability,superiority and good adaptability to different ter-rain conditions of the method proposed in this paper.

关 键 词：GPS高程拟合 BP神经网络 组合模型 精度分析 

分 类 号：P228[天文地球—大地测量学与测量工程]