摘要
高光谱影像数据量大、结构复杂,利用集群对海量数据进行分布式存储成为研究热点,但当前研究多将影像像元数据与元数据分离存储,导致系统内聚性差,影响数据存取性能,同时没有充分考虑上层各种遥感算法的数据访问需求,使得系统可用性不高。针对上述问题,基于HDFS(分布式文件系统)模型提出一种分布式影像存储模型,将整个影像数据模型存储在同一个结构中,提升系统内聚性,并且提供灵活的数据分块策略以供上层应用调用,满足多种数据访问需求。通过对比实验表明,该存储模型较现有的方案有更好的存取性能,具有较高的可行性。
Hyperspectral image data relates to the volume and complex structure, so the distributed storage of massive data becomes a research hotspot by cluster, but the current study will image pixel data and metadata in separate storage system which leads to poor cohesion, influences the data access performance, and mostly does not ful is of high availability model based on HDFS, ly consid Aiming in whic er the various remote sensing algorithm for data access needs. The system at the above problems, this paper proposes a distributed image storage h the image data storage model is in the same structure, enhancing the cohesion of the system, and providing a flexible data partition strategy for upper algorithm calls in order to meet a variety of data access requirements. The comparison experiment shows that the storage model has better access performance than the existing scheme, which has higher feasibility.
作者
聂沛
陈广胜
景维鹏
NIE Pei;CHEN Guangsheng;JING Weipeng(College of Information and Computer Engineering,Northeast Forestry University,Harbin 150040,China;Heilongjiang Provincial Engineering Technology Research Center for Forestry Ecological Big Data Storage and High Performance（Cloud）Computing,Harbin 150040,China)
出处
《测绘工程》
CSCD
2018年第11期40-45,共6页
Engineering of Surveying and Mapping
基金
黑龙江省自然科学基金重点项目(ZD201403)
林业公益性行业科研专项(201504307)
