A Storage Method for Remote Sensing Images Based on Google S2

Xujin Wang,Wen Zhan,Rui Wang,Beibei Yang,Fei Chen,Linyi Li,Lingkui Meng

doi:10.1109/access.2020.2988631

Xujin Wang, Wen Zhan + Show 5 more

Open Access

https://doi.org/10.1109/access.2020.2988631

Copy DOI

Abstract

When using HBase to store tiles of remote sensing images, the spatial position of a tile is often used as the first part of the tile's rowkey so that tiles with high spatial correlations are stored close together to improve query efficiency. We refer to this storage method as the Geo-First model. However, Geo-First models have two problems: the load between nodes is unbalanced, and the accumulation of time-series remote sensing images has a negative impact on storage and query efficiency. Considering these two problems, we proposed a method for storing remote sensing images based on Google S2 and HBase. In our method, two strategies are adopted to eliminate these problems: the balanced placement strategy (BPS) and the periodic storage strategy (PSS). We evaluated our method by focusing on the effectiveness of BPS and PSS. The results show that our method achieves higher tile storage and query efficiency than three Geo-First models based on latitude and longitude, Geohash code, and Google S2 code. BPS effectively balances the load between nodes, while PSS alleviates the negative impact of the accumulation of time-series remote sensing images. Both BPS and PSS greatly improve tile storage and query efficiency.

Highlights

With the development of ground observation technology, remote sensing images have become an important big data category [1] that play an important role in the fields of economic development, environmental protection and national defence construction [2]
RESEARCH CONTRIBUTIONS This paper proposes an efficient method for storing remote sensing images based on Google S2 and HBase
1) We propose a remote sensing image storage method based on Google S2 and HBase

Summary

Introduction

With the development of ground observation technology, remote sensing images have become an important big data category [1] that play an important role in the fields of economic development, environmental protection and national defence construction [2]. The massive, heterogeneous, multiscale data of remote sensing images make storage difficult. The traditional storage methods used to store remote sensing images can be divided into three types: file system storage methods, relational database storage methods, and file-relational database hybrid storage methods. The above methods can effectively store and manage remote sensing images, but all have flaws such as low retrieval efficiency, poor concurrency, and high expansion costs [3], [4]. With the development of Not Only SQL (NoSQL) databases, increasing numbers of researchers have favoured remote sensing image storage based on NoSQL [5]–[14].

Objectives

Methods

Findings

Conclusion