摘要
当控制点多和网格稠密时,基于薄板样条(TPS)插值的近地表模型重构往往很耗时,影响了静校正中近地表建模的效率.针对此问题,采用一种TPS插值和三次样条插值相结合的混合插值法重构近地表模型.首先利用矩阵递归LU分解及GPU加速的LU分解算法求解大型线性方程组,建立TPS插值函数;然后在X和Y方向上使用适当的步长对网格进行抽稀,运用TPS插值函数计算稀疏网格点的值,再通过稀疏网格点建立三次样条插值函数并计算剩余网格点的值;最后用OpenGL实现近地表模型的三维可视化.实验结果表明,文中算法提高了近地表模型重构的速度,其精度接近TPS插值精度.
When both the number of control points and the size of grid become large,TPS based reconstruction of near-surface model is often very time-consuming.Therefore,such method affects the efficiency of building near-surface model in the static correction.To address this problem,a hybrid interpolation combining TPS with cubic spline interpolation is used for reconstructing near surface model.Firstly,not only recursive LU decomposition but also GPU based LU decomposition is used to solve large linear system of equations.And then,TPS interpolation function is created.Secondly,the grid is rarefied with appropriate steps in the X and Y directions and then,TPS interpolation function is evaluated on the sparse grid.Based on which cubic spline interpolation function is created and then,it calculates the value of the other points.Finally,OpenGL visualizes the 3D near-surface model.The experimental results show that this algorithm speeds up the reconstruction of near-surface model and approximates the TPS interpolation in accuracy.
出处
《计算机辅助设计与图形学学报》
EI
CSCD
北大核心
2012年第4期466-470,共5页
Journal of Computer-Aided Design & Computer Graphics
基金
中央高校基本科研业务费专项资金(2-9-2011-0187)
关键词
薄板样条
近地表模型
静校正
三次样条
递归LU分解
thin-plate spline
near-surface model
static correction
cubic spline
recursive LU decomposition
