机构地区: 浙江大学地球科学系浙江省资源与环境信息系统重点研究实验室
出 处: 《浙江大学学报(理学版)》 2012年第2期225-230,共6页
摘 要: 高性能计算的基础是集群体系下的大规模并行计算.遥感图像处理效率的提高,依赖于并行计算技术的运用.在分析了已有网格计算环境下分布式任务分配方法的基础上,针对远海遥感影像目标物数量相对较少的特点,从软件角度利用四叉树结构对目标区域进行划分,同时采用动态负载均衡的任务分配策略与并行计算的思想,提出了对影像进行并行处理的集群体系任务分配算法模型,实验表明该集群体系下任务分配模型能有效提高图像并行处理的速度. Large-scale parallel computing under cluster architecture is the basis of HPC(high performance computing).To enhance the efficiency of image processing for remote sensing depends on the application of technology of parallel computing.Towards the characteristic about the scarce quantity of objects on some images,this paper proposed an algorithm model for task allocating and image processing under distributed computing environment.Firstly,we split the area of image which contains the objects using quad-tree policy,and dispatched the data to computing node by using the idea on dynamic load balance to perform this computing.Experiment shows that the strategy on task allocation can improve the overall computing performance effectively.
领 域: [自动化与计算机技术] [自动化与计算机技术]