基于不同核函数的LSSVM在GPS高程拟合中的应用

Application of LSSVM based on different kernel functions to GPS elevation interpolation

GPS技术因其诸多优点在工程测量中被广泛应用,最小二乘支持向量机应用于GPS高程拟合有着一定的优势。最小二乘支持向量机在处理非线性问题时,采用核函数代替内积计算,巧妙地解决了高维计算问题。文中综述了最小二乘支持向量机常用的几种核函数,用实验进行对比分析,结果表明,采用几种核函数都取得较好的效果,采用径向基核函数时效果最优。

GPS technology is widely used in the engineering measurement because of many advantages. The least squares support vector machine （LSSVM） has certain advantages in GPS elevation interpolation application. When dealing with nonlinear problems, LSSVM uses kernel function instead of the inner product computation, and cleverly solves the calculation problem of high dimension. Several commonly- used kernel functions of LSSVM are introduced, compared and analyzed with experiment. The results show some of kernel functions obtain satisfactory results, one of which is optimal when using the RBF kernel function.