摘要
空间插值是对现有实际观测点运用插值算法来为未知点估值的一种方法,它是环境污染监控仿真系统的一个重要支撑算法。随着环境污染监控仿真系统的实时性要求越来越高,计算复杂度高的传统空间插值算法弊端日渐凸显,难以满足实际需求。结合克里金插值[1,2]高精度特点,采用改进空间查询策略、优化数据导入与导出、自适应数据分割、自适应并行模型选择等并行分析技术,使插值算法效率大大提高,同时,采用自适应异常检测与处理技术对系统进行跟踪监测,有效地提高了程序的鲁棒性。通过仿真实验验证得出,基于自适应的并行插值算法效率较串行插值算法及前人经典并行插值算法在计算效率和稳定性方面都有显著优势。特别地,当待插值数据量或搜索半径达到一定规模后,算法可达接近线性加速比。
Spatial interpolation refers to making use of a few known points to estimate the other unknown points to get a continuous image. As the number of the spatial data increases and the demands of getting a real-time interpolation system, traditional interpolation methods couldn’t meet the needs. Based on kriging, MPI was made use of to design a self-adaptive algorithm based on parallel computing. According to the time-consuming of each part of the serial program, using new method to improve spatial queries, self-adaptive data partition method, and self-adaptive parallel mode method selected a high efficiency parallel interpolation method was obtained. At the same time, designing an effective method to read and write file, and using self-adaptive anomaly detection, the program’s robustness was improved. Experiment results show that the algorithm could achieve higher efficiency than both serial methods and traditional parallel methods. It could get near-linear speedup, when the amount of data or the size of search radius is large enough. The methods have great applicability and extensity and could play an important role in massive spatial data’s real-time interpolation.
出处
《系统仿真学报》
CAS
CSCD
北大核心
2014年第4期761-768,共8页
Journal of System Simulation
基金
国家高技术研究发展计划(863计划)(009AA12Z226)
关键词
系统仿真
克里金插值
自适应
并行编程
数据分割
system simulation
kriging
self-adaptive
parallel programming
data partition
