异构计算环境下遥感图像并行变化检测关键技术研究

项目名称： 异构计算环境下遥感图像并行变化检测关键技术研究

项目编号： No.61303032

项目类型： 青年科学基金项目

立项/批准年度： 2014

项目学科： 自动化技术、计算机技术

项目作者： 朱虎明

作者单位： 西安电子科技大学

项目金额： 23万元

中文摘要： 随着遥感技术的不断发展，如何高效处理空-天-地一体化对地观测系统获取的遥感大数据成为热点。本课题针对应急任务下大规模遥感图像变化检测算法运算量大难以实时应用的难题，在分析变化检测模型和MapReduce并行计算思想基础上，提出并行变化检测框架。为提高并行变化检测框架中每个计算节点的计算效率，根据CPU擅长处理复杂逻辑功能和GPU擅长处理计算密集型任务的特点，提出异构计算环境下基于MapReduce的并行遥感图像变化检测算法。基于进化计算和memetic算法中不同学习机制设计异构计算环境下的并行任务调度策略，以最大化利用并行计算资源达到缩短并行变化检测算法运行时间的目的。在异构计算集群上设计并建立基于MapReduce的遥感图像并行变化检测验证软件，采用存储优化等方法优化软件性能，在大规模真实遥感图像上进行仿真，并分析算法的加速比和并行可扩展。

中文关键词： 并行计算；SAR 图像变化检测；大数据；异构计算；Mapreduce

英文摘要： With the continuous development of remote sensing technology, how to efficiently handle remote sensing data acquired by sensors,carried on airborne, spaceborne or landborne platform,has become a hot topic. But it is time-consuming for the existing change detection algorithms to process large-scale remote sensing image in many emerging applications. Therefore, through the analysis of change detection model and MapReduce paradigm, a parallel change detection framework is proposed. In order to further improve the computational efficiency of each computing node, according to the characteristics of CPU and GPU，we propose MapReduce-based parallel remote sensing image change detection algorithm in heterogeneous computing environments. Parallel task scheduling strategy based on evolutionary computation combined with different memetic mechanism in heterogeneous computing environment is designed, which can maximize utilize parallel computing resources to minimum running time of parallel change detection algorithm. On CPU/GPU cluster, the software of MapReduce-based parallel remote sensing image change detection will be implemented, and it efficiently utilizes the GPU memory hierarchy for performance enhancement. Finally, we verify the performance of the algorithm on large-scale real remote sensing images and analyze algor

英文关键词： parallel computing；SAR image change detection；big data；heterogeneous computing；Mapreduce