不同组合方法对GPS高程拟合的影响

针对平面拟合、二次曲面拟合和GA-BP神经网络3种模型的各自特点和适用范围,为综合各模型优点、提高高程拟合的精度与可靠性,对比分析了不同非线性组合和线性组合方法,即RBF神经网络组合、加权最小二乘支持向量机(WLSSVM)组合和最优加权组合、最优非负变权组合等对GPS高程拟合精度的影响。理论分析和算例结果表明,不同组合方法对GPS高程拟合精度的影响不同,WLSSVM组合和最优非负变权组合的拟合效果较好,可靠性较强;最优非负变权组合能较好地控制残差极值,有效减小误差区间,且转换精度较高。

Influence of Different Combination Methods of GPS Elevation Fitting

Based on the characteristic and applicable scope of plane fitting, quadratic surface fitting and GA-BP neural networkmodels, nonlinear and linear combined methods are proposed to integrateadvantages of each model and improve the accuracy and reliability of height fitting. We consider an RBF neural networkcombination, a weighted least squares support vector machine (WLSSVM) portfolio, an optimal weighted combination and an optimal non-negative variable weight combination. The consequences of these different combined methodson GPS height fitting precisionare compared and analyzed. The results showthat different combination methods generate different accuracy of GPS Height Fitting. The WLSSVM and optimal non-negative variable weight combinations are superior to the others: they have stronger reliability, can better control the residual extremes, effectively shorten the error range, and havehigher conversion accuracy.