论文部分内容阅读
目的空间位置检索是遥感影像检索中的关键步骤,为进一步提高海量遥感影像编目数据定位检索效率,降低误检率,提出一种基于MPI和Open MP混合编程模型对射线法进行多层次并行化实现。方法首先完善传统射线法处理点在多边形边上以及射线与边的端点相交的情况;其次采用MPI实现基于程序层面多机并行,Open MP实现算法层面单机多线程并行,通过开启多个线程同时处理多边形的各个点,判断它们是否在另一个多边形的内部。结果当系统中所有节点开启线程数之和等于主节点的最佳线程数时,全局计算速度达到最佳。混合并行算法相比串行算法检索时间减少50%以上,效率更高。结论 MPI+Open MP混合并行比普通的串行执行、单纯MPI并行或单纯Open MP并行执行空间定位检索算法效率显著提高,这种并行方案普遍适用于集群环境下的并行程序,并且可以进一步拓展到其他图像处理算法领域。
The purpose of spatial location retrieval is a key step in remote sensing image retrieval. In order to further improve the positioning retrieval efficiency and reduce the false detection rate of mass remote sensing image data, a multilevel parallelization method based on MPI and Open MP hybrid programming model is proposed . The method first completes the situation that the traditional ray processing point intersects with the edge of the polygon and the ray intersects with the edge. Secondly, MPI is used to realize multi-machine parallelism based on program level. Open MP realizes the parallelism of single machine multi-thread at the algorithm level, Each point of the polygon, to determine whether they are within the other polygons. Results When the sum of the number of threads on all nodes in the system is equal to the optimal number of threads on the primary node, the global calculation speed is optimized. Compared with the serial algorithm, the hybrid parallel algorithm reduces the retrieval time by more than 50% and is more efficient. Conclusion MPI + Open MP hybrid parallel implementation than ordinary serial execution, pure MPI parallel or pure Open MP parallel execution of space location retrieval algorithm significantly improves the efficiency of this parallel program is generally applicable to cluster environment parallel program, and can be further extended to Other image processing algorithms.