【作者】 邱强； 曹磊； 卢亮； 方金云；

【Author】 Qiu Qiang;Cao Lei;Lu Liang;Fang Jinyun;Integration Application Center,Institute of Computing Technology,C.A.S.;University of Chinese Academy of Science;

【机构】 中国科学院计算技术研究所,计算机应用研究中心； 中国科学院大学,信息工程学院；

【摘要】 矢量地图叠加分析是GIS空间分析中的重要组成部分。传统的矢量地图叠加分析算法主要采用扫描线算法求得交点,然后通过交点构建叠加结果。随着计算机硬件的发展以及用户需求的变化,传统的扫描线算法不能发挥多核与集群的并行优势,已经不能满足快速、精确地叠加分析要求。本文提出一种求交算法的并行策略。在新型硬件架构环境下,本文采用消息传递编程模型,利用各个计算节点之间的数据通讯,实现了大规模数据在集群环境下的快速求交运算。实验表明,该方法在多个计算节点的计算中取得了良好的加速比。更多还原

【Abstract】 Vector map overlay is an important part of spatial analysis in GIS.The traditional algorithm of vector map overlay mainly uses plane sweep algorithm to compute the point of intersection.Then it structures the result of overlay through the points.However,with the developing of computer hardware and transformation of user requirement,the traditional plane sweep algorithm has not been able to express the preponderance of multi-core system.It cannot content the accurate and fast requirement of overlay.In this paper,we provide a parallel strategy of plane sweep algorithm.In the environment of new hardware framework,we achieved the fast intersection computing of large-scale data in multi-system,.which uses message passing interface and data communications between each evaluate nodes.For experiments show that this method made good speed-up in multi-core evaluate nodes.更多还原