几何部件缓冲区域合并的Buffer算法及其并行优化方法

摘 要：本文在介绍一种基于几何部件缓冲区域合并的矢量数据缓冲区生成算法的基础上，采用数据并行思想和MPI编程模型对缓冲区算法的并行化实现和优化方法开展研究。实验结果显示，与ArcGIS Buffer工具相比，（1）当缓冲区结果多边形不合并时，虽然串行缓冲区算法的时间开销较高，但可轻易通过并行方式实现加速。（2）当缓冲区结果合并时，本文算法要明显优于ArcGIS Buffer工具，并且经过优化的并行缓冲区算法表现出了更高的计算效率和更大规模的数据处理能力。因此，基于几何部件缓冲区域合并的Buffer算法具备一定的实用价值，本文提出的按结点数量的任务分解方法和进程间结果“树状”归并策略是对缓冲区算法进行并行优化的有效途径，对GIS中其他矢量分析算法的并行化及相关优化工作也具有一定的借鉴意义。

Research on a Buffer Algorithm Based on Region-merging of Buffered Geometry Components and Its Paral lel Optimization

Abstract: The double-sided parallel line method and the geometry rasterization-based dilation method have been extensively used for buffer generation in spatial analysis. The former involves a series of complex numerical operations and may not be suitable for parallelized computation; the latter inevitably introduces precision problems and computation complexity. In this paper, we proposed a parallel buffer construction algorithm based on region-merging of buffered geometry components in association with the message passing interface (MPI) parallel programming model. Several optimization strategies were studied for the parallel buffer algorithm. The performances of both serial and parallel buffer algorithms were comparably analyzed. Three performance bottlenecks which significantly impact the algorithm efficiency were identified: area merging operation, task load balance strategy and MPI inter-process results merging methods. Corresponding optimization approaches involving tree-like area and inter-process results merging and the parallel task partition based on vertex number oriented parallel task partition strategy were suggested to overcome these bottlenecks. Several experiments were carried out to examine the performance efficiency of the optimized parallel algorithm. The estimation results suggested that our method could provide high performance and processing ability for buffer construction in a parallel environment. Our method could provide insights into the parallelization of spatial analysis algorithm.