DEM与地形分析的并行计算

Parallel Computing of the Digital Elevation Model and Digital Terrain Analysis

总结了数字高程模型构建、特征提取等并行算法的研究进展,概述了不同并行算法的主要内容;探讨了DTA并行技术在海量地形数据可视化和高性能地学计算的应用,随着DEM的需求日益增大,高精度、高分辨率DEM产品及其附加服务也逐步产品化。最后,通过分析并行计算发展的关键问题,提出DTA并行技术的研究趋势及研究意义,合适的数据划分和结果融合策略、通用并行算法、容错机制和负载均衡策略的设计是今后研究的重要内容,尤其是如何在多种计算模式共同发展的背景下利用并行计算解决地学难题,从而得到更接近现实世界地理环境的模拟,并扩大数字地形分析的应用范围。

Theory and methodology of Digital Terrain Analysis（DTA） have played key roles in GIS in recent years,and the appearance of parallel computing provides new propositions for DTA.After a detailed analysis of previous researches of many scholars,this paper reviews the evolvement of some major parallel algorithms such as the generalization of Digital Elevation Models（DEMs）,calculation of land surface parameters.Meanwhile,the main contents of each parallel algorithm have been summarized.Secondly,the applications of parallel technology of DTA in large-scale data visualization and High Performance Geo-Computation（HPGC） are discussed.With more and more demand of DEM in each area,high-precision and high-resolution DEM productions are becoming industrial.Besides the support of ordinary DEM productions,the added value of DEM service is also needed.At the end,by analyzing the focused issues of the development of parallel computing,this paper proposes the tendency and the significance of future researches：1） Make DTA conform to different computing environment by selecting and devising proper partition strategy of spatial data,and parallelism is restricted to pre-processing and post-processing of large-scale terrain data in serial mode;2） Merge the distributed computing results;3） Design universal parallel algorithm so that it can adapt to various architecture.Furthermore,appropriate fault-tolerant scheme should be considered,which can accelerate the recovery of faults efficiently and reduce the cost of fault-tolerant.To a great extent,dynamic load balancing（DLB） can enhance the efficiency of dynamic and irregular issues in order to resolve a series of problems of terrain analysis;4） Appropriate fault-tolerance and load balancing strategies will be another research topics of this domain.