面向林火监测的气象数据分块策略研究

Research on Meteorological Data Partitioning Strategy for Forest Fire Monitoring

摘要: 随着各式各样遥感仪器空间分辨率、光谱分辨率的不断提高，监测森林状况的遥感影像数据量会随着时间的积累而急剧増长。传统的遥感影像数据存储系统一般都采用SAN(存储区域网络)架构，面对当前日益増长的遥感影像数据,系统在存储方面存在存储效率较低，扩展性较差，扩展成本较高等问题。要解决这些问题，最好的方式就是将遥感影像数据的集中式管理模式转变为分布式管理模式。本文基于上述问题，以Hadoop平台为基础，以气象数据的分布式计算为目的，研究气象遥感数据的分块策略。实验采用对HDF数据按行分割的分块策略进行实验，并存储与HDFS之上，并通过实验结果得到数据分块后的存储效率更高的结论。

Abstract: With the improvement of every kind of remote sensing instrument spatial resolution and spectral resolution, amount of the remote sensing image data for monitoring forest condition will accumulate with time and rapid growth. The traditional remote sensing image data storage systems generally use the SAN (storage area network) architecture. Faced with the increasingly extended remote sensing data amount, stored in SAN architecture has low storage efficiency, poor scalability, high expansion cost. To solve these problems, the best way is to transform the centralized management model of remote sensing image data into a distributed management model. On the basis of Hadoop platform, this paper studies the partitioning strategy of meteorological remote sensing data with the purpose of distributed computing of meteorological data. The experiment was carried out on the partitioning strategy of HDF data stored on HDFS by line segmentation. The result of the experiment is to get the conclusion that the storage efficiency is higher after the data partitionc.