随着全球对地观测技术的快速发展，多源、海量的遥感数据为研究资源、环境、林业、地质、气象、灾害等问题提供了丰富的数据源，然而，海量遥感数据的快速处理需要强大的存储资源和计算资源的支持。因此，基于网络的分布式高性能遥感数据处理技术已逐步成为业界的研究热点。近年来,新兴的云计算技术在诸多应用领域已经逐步进入产业化阶段。因此，本文重点探索研究云计算技术在遥感数据处理服务领域的应用，对探索未来遥感云服务应用具有重要的理论意义和应用价值。

本论文主要研究搭建基于 Hadoop 的云计算实验平台，并结合 WebServices技术开发一套基于 Internet 服务模式遥感数据处理服务云平台系统，通过典型的遥感数据处理算法的并行化实验——在云计算平台上实现 MODIS 数据气溶胶遥感反演、遥感图像分类及土壤水分遥感数据同化等探索研究,发展适合遥感数据处理的 MapReduce 遥感处理并行框架，为其它遥感数据处理算法的快速云并行化提供奠定基础。

本论文的创新点主要包括以下两个方面：

1）提出了基于 Hadoop 平台的遥感云服务系统的设计与实现方法，给出了系统总体设计框架，并选取了 MODIS 数据气溶胶遥感反演、遥感图像分类及土壤水分遥感数据同化 3 个示范性算法进行了云服务系统开发。

2）提出了基于 Hadoop 流机制的遥感数据处理并行框架,并根据特定遥感数据处理服务特点，为每一个遥感处理服务开发设计了相应的控制层程序（Callmapper），桥接客户的服务请求与后台服务端分布式存储及处理，从而实现遥感数据在 Hadoop 平台上的处理。

本论文设计开发的遥感云服务系统探索为用户提供一个简洁、方便、快捷、人性化的服务模式，实现面向用户需求的遥感数据处理系统。该服务系统通过用户与系统界面的交互，用户无需遥感数据处理领域深层次的知识，仅需要了解数据处理需求，根据系统提供的 WEB 界面上的相关参数设置与功能选项，提交处理需求表单，即可获得需求结果，至于数据处理的算法和程序（或软件）、数据的处理过程，完全由系统完成，初步实现了用户无计算资源、无存储资源、无算法资源，只需用户利用访问终端（如未来的智能手机和虚拟化桌面显示器）提交数据及处理需求即可获得用户满足需求的遥感数据处理服务模式，对未来遥感云服务商业化模式具有参考价值。

With the rapid development of the global earth observation technology, more and more sensors provides multi-source and massive remote sensing data for the study of resource, environment, forestry, geology, meteorology, natural disasters and other issues. However, higher surport of many storage resources and computing resources is needed for the fast processing of massive remote sensing data. Thus, the high performance distributed processing technology for remote sensing data has gradually become the research hotspot internationally. Recently, cloud computing technology has been gradually industrialized in many application fields. In this paper the research focuses on applying cloud computing technology into remote sensing data processing services, which is of theoretical and practical significance to explore the future study of remote sensing cloud services.

The research is mainly focused on how to build a cloud computing experiment platform and develop online remote sensing data processing service system based on Hadoop and WebServices technology. Three typical algorithms for remote sensing data processing including aerosol remote sensing retrieval using MODIS data, remote sensing image classification and data assimilation for soil moisture remote sensing were chosen to run on the cloud service system. A new MapReduce framework was developed for parallel processing of remote sensing data, which provides the basis for the development of the other remote sensing data processing algorithms on cloud platform in future.

The two main innovation points are as followed:

1) The method of design and implementation of remote sensing cloud service system based on Hadoop platform was proposed. We proposed a framework of remote 

sensing cloud service system, in which three cloud services based on algorithms of aerosol remote sensing retrieval using MODIS data, remote sensing image classification and data assimilation for soil moisture remote sensing were developed. 

2) A new parallel framework was designed for remote sensing data processing based on Hadoop streaming mechanism, and special remote sensing processing service models (Callmapper) were designed and developed according to the characteristics of each remote sensing data processing algorithm, which bridging the 

service requests from users and the server of distributed storage and processing on Hadoop platform.

Remote sensing cloud service system designed and implemented in this paper can provide users a concise, convenient, fast, personalized service responding to the demands of users. Without much knowledge of remote sensing data processing, the user can submit the related parameters of processing requirements by the web UI of the cloud service system, and the results could be acquired finally. This service mode was realized to meet the processing demands of the users without computing and storage resources, algorithm resources, however, the results could be acquired by accessing terminals (such as smart mobile phone or virtual desktop) to submit remote sensing data. The research provides the reference for business model of remote sensing cloud services in the future.