遥感影像大数据管理系统关键技术研究与系统实现

发布时间：2018-05-02 13:43

本文选题：遥感影像存储 + Hadoop　；参考：《电子科技大学》2016年硕士论文

【摘要】：遥感技术的快速发展使得遥感影像数据的种类成倍增加,同时在数量级上呈现出几何级数的增长。随着遥感影像数据应用与业务需求不断提高,给遥感影像的管理提出了更高的要求,高效存储和快速检索成为当前亟待解决的问题。而在互联网技术不断革新的今天,云存储作为新型存储模型为数据提供无限的存储能力,应用此技术于海量遥感影像的管理和存储,也是当今研究的热点。而对于空间数据的快速检索,离不开高效索引技术的支撑,研究高效的索引机制对于空间数据的检索具有重要价值和意义。除了检索性能方面的要求,遥感影像存储管理的另一个技术指标就是检索准确度,比如从海量遥感影像中精确查找出具有某一地物特征的影像数据。因此,基于语义特征的遥感影像检索也是论文中的研究点。基于以上问题的提出,本论文对遥感影像数据管理和检索做了一系列研究。因此,论文的主要研究工作和成果包括:(1)设计不同解析方法提取所需遥感影像元数据,并将其组成遥感影像数据管理系统的数据表,基于这些元数据对遥感影像实现多种方式的查询和检索。此外,遥感影像数据作为一种空间数据,对其检索性能的优劣直接受索引机制的影响,因此本文还设计了实验验证不同空间索引机制的性能差异,即利用二维空间数据对建立的索引树进行性能测试。该实验基于不同的数据量级与索引项,通过Java程序以可视化的方式实现对不同索引技术的性能测试和验证。(2)设计对遥感影像进行语义特征检索的技术方案,基于该方案完成对遥感影像所包含的地物特征的提取,并将这些特征映射为一组语义ID的字符串,以该字符串作为遥感影像管理系统的元数据存入数据库中。主要思想是:首先,基于分水岭变换和Full Lambda-Schedule合并分块算法,分割原始遥感影像;其次,基于SVM分类算法完成遥感影像特征提取和分类;最后将分类好的特征映射为语义ID。(3)开发遥感影像数据管理系统。通过对系统的需求分析,设计该系统的系统架构,由此架构开发了基于Hadoop的遥感影像数据管理系统。重点研究了Hadoop系统的文件交互流程,并基于HDFS高效地存储管理遥感影像数据。遥感影像数据管理系统是一个B/S架构模型系统,通过设计每层功能模块,最终完成影像的查询检索、快速浏览以及本地下载等。
[Abstract]:The rapid development of remote sensing technology makes the kind of remote sensing image data multiplied, and at the same time it presents the growth of geometric progression. With the application of remote sensing image data and the increasing demand of business, the management of remote sensing image has been raised higher, and the efficient storage and fast retrieval have become the urgent problems to be solved. With the continuous innovation of Internet technology, the cloud storage, as a new storage model, provides unlimited storage capacity for data. The application of this technology to the management and storage of massive remote sensing images is also a hot topic in the present research. The retrieval of spatial data is of great value and significance. In addition to the requirements of the retrieval performance, the other technical index of remote sensing image storage management is the retrieval accuracy, such as the accurate search of the image data with a certain feature from the mass remote sensing images. Therefore, the remote sensing image retrieval based on the semantic features is also in the paper. Based on the above problems, this paper makes a series of research on remote sensing image data management and retrieval. Therefore, the main research work and achievements of this paper include: (1) design different analytical methods to extract the remote sensing image metadata needed for remote sensing image data and make up the data table of remote sensing image data management system, based on these metadata In addition, the remote sensing image data as a kind of spatial data is directly influenced by the index mechanism, so the performance difference of the different spatial index mechanism is also designed in this paper, that is, the performance measurement of the index tree based on the two-dimensional spatial data is used. This experiment is based on the different data magnitude and index terms, through the Java program to visualize the performance test and verification of different index techniques. (2) a technical scheme for semantic feature retrieval of remote sensing images is designed, which is based on the scheme to extract the features of the objects contained in the remote sensing images and reflect these features. As a string of semantic ID, the string is stored as the metadata of remote sensing image management system. The main idea is: first, based on watershed transform and Full Lambda-Schedule merging block algorithm, the original remote sensing image is segmented. Secondly, the feature extraction and classification of remote sensing images are completed based on the SVM classification algorithm; finally, the remote sensing image features are extracted and classified. The classified features are mapped to semantic ID. (3) development of remote sensing image data management system. Through the analysis of the system requirements, the system architecture is designed, and the remote sensing image data management system based on Hadoop is developed. The document interaction process of the Hadoop system is focused on, and the remote sensing image is efficiently stored and managed based on the HDFS. Like data. Remote sensing image data management system is a B/S architecture model system. Through the design of each functional module, the image query and retrieval, rapid browsing and local downloading are completed.

【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP751

【参考文献】