基于PSO-LSSVM的GNSS高程拟合

GNSS elevation interpolation based on PSO LSSVM

粒子群算法通过迭代搜索最优解,简单且易于实现,具有很好的全局优化能力。最小二乘支持向量机应用于GNSS高程拟合,需要选择模型参数。本文采用粒子群算法优化最小二乘支持向量机参数,对粒子的速度设定限值,建立高程拟合模型,最后通过实例采用学习样本、测试样本和总样本均方误差3种适应度函数进行评价,分析比较其拟合效果。结果表明,采用本文方法所建的高程拟合模型在GNSS高程拟合中能得到满意的拟合效果,拟合精度达到四等水准测量的精度,能满足一般的工程测量的要求。

Particle swarm optimization algorithm is simple and easy to implement through iterative search for the optimal solution,and has good global optimization ability.When least squares support vector machine is applied to GNSS elevation interpolation,model parameters are needed to be selected.This paper uses particle swarm optimization algorithm to optimize the parameters of least squares support vector machine,sets the speed limit of particles,establishes an elevation interpolation model,uses three fitness functions of learning sample,test sample and total sample mean square error to evaluate through examples,and analyzes and compares their fitting effects.The results show that the elevation interpolation model built by the method in this paper can obtain satisfactory fitting effect in GNSS elevation interpolation,and the fitting accuracy can reach the accuracy of fourth grade leveling survey and meet the requirements of general engineering survey.