海量遥感数据处理架构与并行算法研究

摘要

In order to effectively improve the efficiency of massive remote sensing data processing,based on analysing the key factors inclu-ding IO,task coordination loss and multi-core multi-CPU which affect the performance of parallel computation,we propose a three-tier parallel computing system architecture (OssSinvoParallel)that consists of seven functional modules such as single calculation cell,task assemble man-ager and so on.In the paper,we present the parallel computing algorithm on macro level,and analyse the timeliness of the algorithm.Test re-sults shows that this architecture can effectually speed up the parallel processing,and in current situation that the demand of remote sensing data processing grows pressingly,it can effectively resolve the bottleneck problem of timeliness in remote sensing image processing.%为了有效提高海量遥感数据处理效率，在对 IO、任务协调损耗和多核多 CPU 等影响并行计算性能等关键因素分析的基础上，提出计算单体、任务分配管理器等七大模块构成的三层架构的并行计算体系结构（OssSinvoParallel）。从宏观层面给出并行计算算法，对算法时效性进行分析。试验结果证明该体系架构可有效提高并行处理速度，在当前遥感数据处理需求迫切增长的形势下，可有效解决遥感影像处理时效性瓶颈问题。