III: Small: Indexing, Querying, and Visualizing Big Spatial and Spatio-temporal Data

该项目进行研究，开发必要的知识，并建立软件基础设施，以支持大空间和时空数据的数据管理。这是对最近由包括智能手机，太空望远镜和医疗设备在内的几种设备产生的空间和时间数据量爆炸的回应。使用这些数据的应用程序以及该项目研究的迫切需要包括研究处理每月时空卫星数据的Terra字节的气候数据，通过将大脑神经元建模为空间数据来了解大脑的结构和功能原理，以及分析数十亿个每月地理标记的社交媒体内容以进行事件检测和分析。该项目将其所有开发的组件打包成一个成熟的免费开源系统，可供广大研究和开发人员社区使用。除了对行业的影响外，该项目还将对社会的多个部分产生更广泛的影响，包括通过使用该项目软件作为研究工具的研究生和本科生教育，通过简单的地图可视化API向K-12学生推广，通过该项目开发软件内的测试实验室开发课程，在国内和国际会议上的指导演讲。虽然有一个迫切需要支持大空间数据，这种需要受到缺乏专门的系统，技术和算法的阻碍。虽然大数据得到了各种通用分布式系统的良好支持，但这些系统都没有为空间或时空数据提供任何特殊支持。在当前系统中支持大空间和时空数据的唯一方法是将其视为非空间数据，或者围绕现有的非空间系统编写代码包装器。然而，这样做并没有利用空间数据的属性，因此导致低于标准的性能。该项目通过为一般当前大数据系统中的空间和时空数据提供原生支持来解决这一研究空白。特别是，该项目利用三个主要的研究课题，即索引，查询和可视化的大空间和时空数据。在索引方面，该项目为Hadoop分布式文件系统（HDFS）构建了新颖，通用和可扩展的空间和时空索引结构，HDFS是当今大多数大数据系统中的事实存储层。在查询方面，该项目为范围查询、最近邻查询和空间连接开发了新的查询处理技术，这些技术利用空间索引HDFS来支持对大空间和时空数据的各种查询操作。在可视化方面，该项目开发了新的可扩展技术，将大空间数据可视化为单级或多级图像。本研究的出版物、技术报告、开放源代码软件和实验数据通过项目网站（http：//www.cs.umn.edu/mokbel/BigSpatial）传播。

This project conducts research, develops requisite knowledge, and builds software infrastructure to support data management for Big Spatial and Spatio-temporal Data. This is a response to the recent explosion in the amounts of spatial and temporal data produced by several devices that include smart phones, space telescopes, and medical devices. Applications using such data and in an urge need for the research of this project include studying climate data that deals with Terra bytes of monthly spatio-temporal satellite data, understanding the brain's architectural and functional principles through modeling brain neurons as spatial data, and analyzing billions of monthly geotagged social media contents for event detection and analysis. The project packages all its developed components into a full-fledged free open-source system, available to the research and developers communities in large. Besides its impact on industry, this project will have significant broader impact across multiple segments of society that include graduate and undergraduate student education by using this project software as a vehicle for their research, outreach to K-12 students through simple map visualization APIs, curriculum development through test labs inside the developed software of this project, and tutorial presentations in domestic and international conferences.While there is an the urge need to support big spatial data, such need is hampered by the lack of specialized systems, techniques, and algorithms. Although big data is well supported with a variety of general purpose distributed systems, none of these systems provide any special support for spatial or spatio-temporal data. The only way to support big spatial and spatio-temporal data in current systems is to either treat it as non-spatial data or to write code wrappers around existing non-spatial systems. However, doing so does not take any advantage of the properties of spatial data, hence resulting in sub-par performance. This project tackles this research gap by providing a native support for spatial and spatio-temporal data inside general current big data systems. In particular, the project exploits three main research topics, namely, indexing, querying, and visualization of big spatial and spatio-temporal data. In terms of indexing, the project builds novel, generic, and scalable spatial and spatio-temporal index structures for Hadoop Distributed File System (HDFS), which is the de facto storage layer in most nowadays big data systems. In terms of querying, the project develops novel query processing techniques for range queries, nearest-neighbor queries, and spatial join, that take advantage of the spatially indexed HDFS to support various query operations on big spatial and spatio-temporal data. In terms of visualization, the project develops new scalable techniques to visualize big spatial data as single- or multi-level images. Publications, technical reports, open-source software, and experimental data from this research are disseminated via the project web site (http://www.cs.umn.edu/~mokbel/BigSpatial).