2.1. Current Works in RS Data Processing
To boost the speed of RS data processing, many research have been studied based on different architectures. These research mainly focus on offline RS data processing. Algorithms and processing patterns (Hailiang & LU, 2010; Jun-jun & Gang, 2010; Ke-fei, 2010; LIU et al., 2014) take the advantage of different hardware architectures such as GPU and FPGA. High performance clusters (Wei, Liu, & Wang, 2014; Xue-ping, 2013) enhance the RS data processing performance by using parallel computing. Distributed computing environments (Sekiguchi et al., 2008) can increase the data-storage capacity and expand the diversity of processing services.
There are also works based on cloud computing and RS big data computing. G-Hadoop (Wang et al., 2013) applied MapReduce across distributed data centers. Almeer(2012) employed Hadoop MapReduce framework to implement parallel processing of RS images and built an experimental 112-core high-performance cloud computing system. Wang (2014) estimates the statistical characteristics of RS big data in the wavelet transform domain. Task scheduling strategies (W. Zhang, Wang, Ma, & Liu, 2014) across data centers were also studied. Wang (2015) discusses some research issues about processing distributed internet of things data in clouds.
Apart from these works, Research on RS data streaming processing mainly focuses on data model and data querying. StreamInsight (Kazemitabar, Demiryurek, Ali, Akdogan, & Shahabi, 2010) support RS data querying by adding the Microsoft SQL server spatial library. Deng (2015) discusses the parallel processing of dynamic continuous queries over streaming data flows. Although system design is one of the most important areas in data streaming processing, there is still no mature data streaming processing system designed for RS data processing.