双态云支持下高分辨率遥感存储与计算一体化研究

发布时间：2018-06-27 17:52

本文选题：高分辨率遥感 + 地学计算　；参考：《浙江大学》2014年博士论文

【摘要】：伴随科学技术的持续发展与应用需求的不断提高,高空间分辨率、高光谱分辨率、高时间分辨率已成为当今卫星遥感发展的主要趋势。当前,高分辨率遥感应用逐渐凸显出数据密集和计算密集的双重特征,对传统的影像数据存储与计算手段提出了巨大挑战。通过对国内外专家学者的相关研究成果进行深入分析,本文围绕高分辨率遥感影像特点与应用需求,通过融合云计算与内存云技术两者的优势,实现了面向高分辨率遥感的双态云存算一体化,解决了目前高分辨率遥感应用中遇到的数据密集与计算密集问题。本文主要研究内容如下： (1)提出了一种面向高分辨率遥感的双态云存储新模式R-D Cloud 首先,对云计算的发展及其优势进行研究,分析了主流云平台的技术架构；然后,以内存硬件的进步为主线,分析了内存存储与计算技术的衍化进程；最后,提出了一种融合云计算与内存云的双态云模式,通过结合硬盘云的稳定与内存云的高效建设一个高性能的高分辨率遥感存储与计算平台。 (2)设计了双态云支持下高分辨率遥感存算一体化策略与技术途径在存算一体化的前提下,研究了双态云环境下的遥感数据组织与存储模式,设计了一种灵活、可扩展的元数据信息管理机制,实现了双态云下多个集群节点的并行计算任务调度,并提出了一种“读-算-写”异步流水线模型,利用GPU众核计算进一步提高了集群节点的高分辨率遥感影像的处理速度。 (3)开展高分辨率遥感存算一体化原型系统设计与实验验证利用当前计算科学领域的多种先进技术,针对高分辨率遥感应用的需求,基于存算一体化策略与关键技术设计并实现了一个原型系统,并利用TB级的多来源、多分辨率的三种国产高分辨率遥感影像实现了对数据存储的可靠性、系统的存储能力和图像处理能力的测试,从实践上论证了双态云下高分辨率遥感存算一体化的性能。研究结果表明,本文提出的面向高分辨率遥感的双态云存算一体化策略,可以有效解决当前高分辨率遥感影像面临的高效存储与快速计算问题。本研究提出高分辨率遥感数据组织与存储机制,同样可以应用于中低分辨率遥感数据。本文研究成果丰富并发展了高性能遥感地学计算领域的理论与方法,对于推动我国高分辨率遥感影像数据的应用具有十分重要的社会价值和现实意义。
[Abstract]:With the continuous development of science and technology and the continuous improvement of application demand, high spatial resolution, high spectral resolution and high time resolution have become the main trend of the development of satellite remote sensing. At present, high resolution remote sensing applications gradually highlight the dual characteristics of data intensive and computation intensive, and the storage and calculation of traditional image data are stored and calculated. Great challenges have been raised by means.
Through the in-depth analysis of the relevant research results of experts and scholars at home and abroad, this paper, around the advantages of high resolution remote sensing image features and application, through the integration of cloud computing and memory cloud technology, has realized the integration of dual state cloud storage for high resolution remote sensing, and solved the number of high resolution remote sensing applications. According to intensive and computationally intensive problems, the main contents of this paper are as follows:
(1) a new dual mode cloud storage model named R-D Cloud for high-resolution remote sensing is proposed.
First, the development and advantages of cloud computing are studied and the technical architecture of the mainstream cloud platform is analyzed. Then, with the progress of memory hardware as the main line, the evolution process of memory storage and computing technology is analyzed. Finally, a two state cloud model, which combines cloud computing and memory cloud, is proposed by combining the stability and memory of the hard disk cloud. Cloud efficient construction of a high-performance high resolution remote sensing storage and computing platform.
(2) design strategy and technical approach for high resolution remote sensing data storage supported by two state cloud.
Under the premise of integration, this paper studies the remote sensing data organization and storage mode under the environment of double state cloud, designs a flexible and extensible metadata information management mechanism, realizes the parallel computing task scheduling of multiple cluster nodes under the dual state cloud, and proposes a "read and write write" asynchronous pipeline model, which uses the GPU crowd nuclear meter. The processing speed of high resolution remote sensing images of cluster nodes is further improved.
(3) design and experimental verification of an integrated prototype system for high resolution remote sensing data storage.
Using many advanced technologies in the field of computing science, a prototype system is designed and implemented based on the integration strategy and key technology for high resolution remote sensing applications. The reliability of the data storage is realized by using the multi resolution TB level multi-source and multi-resolution high resolution remote sensing images, and the system is stored in the system. The test of storage capacity and image processing capability demonstrates the performance of integrated high resolution remote sensing data storage under bistatic cloud.
The research results show that the integration strategy of dual state cloud storage for high resolution remote sensing can effectively solve the problem of high efficient storage and fast calculation for high resolution remote sensing images. This study proposes the mechanism of high resolution remote sensing data organization and storage, and the same sample can be applied to middle and low resolution remote sensing data. The research results enrich and develop the theory and method of the field of high performance remote sensing geoscience. It is of great social and practical significance to promote the application of high resolution remote sensing data in China.
【学位授予单位】：浙江大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TP751;P209

【参考文献】