摘要
基于BP神经网络法和最小二乘支持向量机(LS—SVW),提出一种新的针对道路控制测量的GPS高程拟合方法。选择高斯函数作为其核函数,将其应用至京哈高速过境段的控制点测量中,并将高程拟合结果与BP神经网络、平面拟合、最小二乘法、GA—GRNN、LS—SVM二次曲面和三次样条曲线拟合法等高程拟合方法对比。结果表明,该模型具有拟合精度高、所需样本小、泛化能力强等特点,成功地解决了高维数、非线性、小样本等问题,是一种较适合于公路控制测量的GPS高程拟合方法,具有较高的推广价值。
Based on BP neural network and least squares support vector machine（LS — SVW）,a new GPS elevation fitting method for road control measurement is proposed. The Gaussian function was chosen as its kernel function and applied to the measurement of control points in Jingha high-speed transit area. The results of elevation fitting were compared with BP neural network,plane fitting,least square method,GA — GRNN,LS — SVM Sub-surface,cubic spline curve fitting methods such as elevation comparison. The results show that this model has the advantages of high fitting precision,small sample size and strong generalization ability. It solves the problems of high dimension,nonlinearity and small sample successfully. It is a more suitable GPS for highway control survey Elevation fitting method,with a high promotional value.
作者
陈凯
CHEN Kai(School of Roads and Bridges,Zhejiang Institute of Communications,Hangzhou,Zhejiang 311112,China;School of Geosciences and Into-Physics,Central South University,Changsha,Hunan 410083,China)
出处
《公路工程》
北大核心
2018年第3期106-109,130,共5页
Highway Engineering
基金
浙江省自然基金项目(LQ16E080008)
