时空数据查询与读取研究现状写作
This commit is contained in:
@@ -0,0 +1,7 @@
|
|||||||
|
---
|
||||||
|
bibliography: references.bib
|
||||||
|
---
|
||||||
|
|
||||||
|
遥感数据的高效时空查询处理已经得到了广泛的研究,早期的工作主要集中在关系数据库系统中的元数据组织和索引级修剪。传统的方法通常扩展基于树的空间索引,如R-tree,四叉树及其时空变体,将图像足迹与时间属性组织在一起,并且通常在关系后端(例如MySQL和PostgreSQL)上实现。这些方法为中等规模的数据集提供了有效的距离过滤,但随着遥感元数据量的快速增长,它们对平衡树结构的依赖往往导致较高的维护开销和有限的可扩展性。随着数据量和数据摄取速度的不断增加,近年来的系统逐渐转向部署在分布式NoSQL存储上的基于网格的时空索引方案。通过使用GeoHash、GeoSOT或空间填充曲线[@li2023]将空间足迹编码为统一的空间网格,并将其与时间标识符结合,这些方法可以实现轻量级索引构建,并在HBase和Elasticsearch等后端具有更好的水平可扩展性。这种基于网格的索引可以通过粗粒度的剪枝有效地减少候选搜索空间,更适合于大规模、持续增长的遥感档案。
|
||||||
|
|
||||||
|
然而,对于遥感工作负载,单靠索引修剪不足以保证端到端的查询效率,因为遥感工作负载中的单个图像通常很大,查询结果需要进一步的像素级处理。为了减少原始I/O的数量,谷歌Earth系统依赖于平铺和多分辨率金字塔,将图像物理地分成小块。而最近的解决方案则利用COG和基于窗口的I/O来实现对整体映像文件的部分读取。OpenDataCube等框架利用这些特性只读取与查询窗口相交的图像区域,从而减少不必要的数据传输。然而,在确定候选图像之后,大多数系统仍然对每个图像执行细粒度的地理空间计算,包括坐标转换和精确的像素窗口推导,当涉及许多图像时,这可能会产生大量开销。
|
||||||
30
国内外研究现状/references.bib
Normal file
30
国内外研究现状/references.bib
Normal file
@@ -0,0 +1,30 @@
|
|||||||
|
|
||||||
|
@article{li2023,
|
||||||
|
title = {TrajMesa: A Distributed NoSQL-Based Trajectory Data Management System},
|
||||||
|
author = {Li, Ruiyuan and He, Huajun and Wang, Rubin and Ruan, Sijie and He, Tianfu and Bao, Jie and Zhang, Junbo and Hong, Liang and Zheng, Yu},
|
||||||
|
year = {2023},
|
||||||
|
month = {01},
|
||||||
|
date = {2023-01},
|
||||||
|
journal = {IEEE Transactions on Knowledge and Data Engineering},
|
||||||
|
pages = {1013--1027},
|
||||||
|
volume = {35},
|
||||||
|
number = {1},
|
||||||
|
doi = {10.1109/TKDE.2021.3079880},
|
||||||
|
url = {https://ieeexplore.ieee.org/abstract/document/9430714},
|
||||||
|
langid = {en}
|
||||||
|
}
|
||||||
|
|
||||||
|
@article{li2023a,
|
||||||
|
title = {TrajMesa: A Distributed NoSQL-Based Trajectory Data Management System},
|
||||||
|
author = {Li, Ruiyuan and He, Huajun and Wang, Rubin and Ruan, Sijie and He, Tianfu and Bao, Jie and Zhang, Junbo and Hong, Liang and Zheng, Yu},
|
||||||
|
year = {2023},
|
||||||
|
month = {01},
|
||||||
|
date = {2023-01},
|
||||||
|
journal = {IEEE Transactions on Knowledge and Data Engineering},
|
||||||
|
pages = {1013--1027},
|
||||||
|
volume = {35},
|
||||||
|
number = {1},
|
||||||
|
doi = {10.1109/TKDE.2021.3079880},
|
||||||
|
url = {https://ieeexplore.ieee.org/abstract/document/9430714},
|
||||||
|
langid = {en}
|
||||||
|
}
|
||||||
Reference in New Issue
Block a user