基于聚类算法的多源信息融合并行处理研究
|
摘要
数据耦合性强,处理逻辑序贯性高,处理时机存在随机性是数据融合并行的难点。本文分析了数据融合的处理逻辑以及并行处理的难点,提出了一种基于凝聚层次聚类算法的并行融合架构,通过在空间域上对融合数据的划分处理,将融合处理分解成多个独立子任务,结合融合算法和空情复杂性评估处理计算量,并依据计算量将子任务均衡地分发至各计算节点并行处理。使用MPI和OpenMP相结合的编程模型实现了该并行架构。试验结果表明,该并行架构具有较高的加速比,实时性高,处理容量大,可扩展性好。
Strongly data coupling,highly Sequential processing logic,randomly process time are the key difficulties of computing parallelizing of data fusion. This paper analyzes data fusion process logic and parallel difficulties,presents the parallel fusion architecture based on agglomerative hierarchical clustering which decomposes the fusion task into independent child-tasks by breaking up fusion data in spatial region,evaluates the computational complexity by the fusion algorithm and air situation complexity and distributes the task for autonomy computation on some computing nodes according to the computational complexity. MPI and Open MP programming model are used to parallel implementation. Experiments show that the parallel architecture is higher ratio of speedup,it has high real-time performance,and has a good expansibility.
Strongly data coupling,highly Sequential processing logic,randomly process time are the key difficulties of computing parallelizing of data fusion. This paper analyzes data fusion process logic and parallel difficulties,presents the parallel fusion architecture based on agglomerative hierarchical clustering which decomposes the fusion task into independent child-tasks by breaking up fusion data in spatial region,evaluates the computational complexity by the fusion algorithm and air situation complexity and distributes the task for autonomy computation on some computing nodes according to the computational complexity. MPI and Open MP programming model are used to parallel implementation. Experiments show that the parallel architecture is higher ratio of speedup,it has high real-time performance,and has a good expansibility.
引文
[1] Bar-Shalom Y. Multitarget-Multisensor Tracking:Applications and advances. Volume III[J]. Norwood,MA,Artech House,Inc.,2000,2000.
[2]张武生,李建江. MPI并行程序设计实例教程[M].北京:清华大学出版社,2009,(1):34-57.
[3]罗秋明. Open MP编译原理及实现技术[M].北京:清华大学出版社,2009,(1):43-46.
[4] Chapman B,Jost G,Van Der Pas R. Using Open MP:portable shared memory parallel programming[M]. MIT press,2008.
[5] A.费利那,斯塔德.雷达数据处理[M].北京:国防工业出版社,1992,(1):47-49.
[6]何勇,张必银.多传感器数据融合并行处理方法研究[J].舰船电子工程,2011,31(8):56-59.
[2]张武生,李建江. MPI并行程序设计实例教程[M].北京:清华大学出版社,2009,(1):34-57.
[3]罗秋明. Open MP编译原理及实现技术[M].北京:清华大学出版社,2009,(1):43-46.
[4] Chapman B,Jost G,Van Der Pas R. Using Open MP:portable shared memory parallel programming[M]. MIT press,2008.
[5] A.费利那,斯塔德.雷达数据处理[M].北京:国防工业出版社,1992,(1):47-49.
[6]何勇,张必银.多传感器数据融合并行处理方法研究[J].舰船电子工程,2011,31(8):56-59.