【摘要】 研究空间拓扑相交关系计算的并行化,可以缩短处理大规模地理空间数据的时间,对于高效地应用GIS空间数据有着重要的现实意义.本文以开源软件GRASS GIS为实验平台,在集群环境下引入MPI并行库,采用不同的数据划分策略对空间拓扑相交关系计算算法进行并行研究与实现.首先分析了串行算法的特点及数据结构,验证了基于几何对象的数据划分策略在该算法上的可行性;其次针对基于几何对象的数据划分策略存在的问题,即较少考虑空间几何实体对象的数据量均衡性,提出基于弧段的数据划分策略;最后通过加速比指标,对两种策略划分方式进行对比分析,验证基于弧段的划分策略的正确性和高效性.经过实验可知,相比基于几何对象的数据划分,基于弧段的数据划分能得到更好的加速比.更多还原

【Abstract】 Studying the parallel computing of spatial topological intersection relations can shorten the time to deal with largescale geospatial data and it has practical significance for the application of GIS spatial data efficiently.In this paper, we use the open source software GIS GRASS as the experimental platform, and introduce the MPI parallel library in the cluster environment, and use different data partition strategies to carry on the parallel research and implementation of the spatial topological intersection relation computing algorithm. Firstly, we analyze the characteristics and data structure of the serial algorithm, and verify the feasibility of the data partition strategy based on geometric object on the algorithm. Secondly, we focus on the problem that the geometric object-based data partition strategy consider a few of the spatial entity object equilibrium,and propose the data partition strategy based on the arc section. At last, by comparing the two methods on the speedup ratio, we verify the correctness and efficiency of the partition strategy based on the arc section. The experiment shows that compared to the data partition based on the geometric object, the data partition based on the arc section can get a better speedup ratio.更多还原