论文部分内容阅读
随着遥感影像数据量的增加,传统非监督分类迭代自组织分析(ISODATA)算法的运算将十分耗时,应用并行计算技术能够有效解决该性能瓶颈。针对现有基于并行计算模型MapReduce的遥感迭代自组织分析并行算法存在的局限性,提出一种可扩展的基于MapReduce的迭代自组织分析并行处理算法。该算法通过其包含的全局子采样算法、聚类中心点集合过滤算法以及聚类映射算法,有效克服了现有并行算法中存在的不足。实验结果表明,在同等规模遥感计算中,该算法效率高于现有并行处理算法,具有良好的加速比,且在处理更大的影像块时具有更高的精度。
With the increase of the amount of remote sensing image data, the operation of the traditional unsupervised classification iterative self-organizing analysis (ISODATA) algorithm will be very time-consuming, and application of parallel computing technology can effectively solve this performance bottleneck. In view of the existing limitations of parallel algorithms based on parallel computing model MapReduce for remote sensing iterative self-organizing analysis, an extensible MapReduce-based iterative self-organizing analysis parallel processing algorithm is proposed. This algorithm effectively overcomes the deficiencies existing in the existing parallel algorithms through its global sub-sampling algorithm, clustering center point set filtering algorithm and clustering mapping algorithm. The experimental results show that the proposed algorithm is more efficient than the existing parallel processing algorithms in the same-scale remote sensing calculation, has a good speedup and has higher accuracy when dealing with larger image blocks.