利用MapReduce进行批量遥感影像瓦片金字塔构建

Parallel Batch-Building Remote Sensing Images Tile Pyramid with MapReduce

在分析面向瓦片金字塔并行构建任务分解的基础上,提出了一种利用MapReduce进行批量遥感影像瓦片金字塔并行构建的方法。实验结果表明,该方法不仅在集群上快速、有效地解决了单机上难以解决的大规模批量遥感影像瓦片金字塔的构建操作,而且具有良好的可扩展性。同时,该算法可作为大规模遥感影像并行处理的基础框架,非常容易扩展到高效能影像特征提取、遥感影像融合以及影像增量计算等其他海量遥感影像处理任务中。

A batch-parallel pyramid-building algorithm based on the MapReduce framework is proposed. The formal description of the pyramid-building task as well as the decomposition algorithm is given in the first place. Then, the details of the processing steps in the map and reduce phases are depicted. The experimental results show the feasibility, efficiency and scalability of the proposed approach. The tile pyramid building on massive remote-sensing data, which is too complicated to be done on a commodity computer, can be efficiently ac- complished on a server cluster. Furthermore, the proposed algorithm can be used as the basic framework for processing massive remote-sensing images, and be applied to promote the efficiencies of some other remote-sensing processing algorithms, e.g. the feature extraction, the image coaddition, the tile image delta,etc.