基于GPU的并行克里格及其在储量估算中的应用

GPU-based parallel Kriging and its application in reserve estimation

克里格法是空间信息统计学中最主要和最基本的一种局部估计方法,利用区域化变量的空间分布特征实现对未知区域的估计.针对克里格算法高时间复杂度问题,提出一种基于GPU的并行克里格算法,实现对克里格插值算法的改进,在精度不降低的情况下很好地解决了克里格算法高时间复杂度的问题.西藏甲码铜资源量估算结果表明,并行克里格方法具有良好的加速比与并行计算效率,验证了该方法的可实践性,与纯CPU计算的对比实验验证了GPU并行计算结果的正确性与可信度.

Kriging is one of the most important and basic local estimation methods in spatial information statistics,which uses the spatial distribution of the regionalized variables to estimate unknown regions.To solve the problem of high time complexity of Kriging algorithm,we propose a GPU-based parallel Kriging algorithm which improves Kriging interpolation algorithm.This method is a good solution to high time complexity of Kriging with no reduction in accuracy.The estimation result of copper resource reserve in Tibet Jiama indicates that parallel Kriging method has good speed-up and efficiency of parallel computing,which verifies the practicability of the method.Comparative experiments with pure CPU computing show that GPUbased parallel computing results are correct and reliable.