节点文献

三维运动目标的数据组织与管理

Data Organization and Mangement of 3D Moving Objects

【作者】 叶焕倬

【导师】 龚健雅;

【作者基本信息】 武汉大学 , 摄影测量与遥感, 2004, 博士

【摘要】 无线通讯、空间定位和测量等技术的发展,使得人们对空间目标的信息管理由静态逐渐扩展到动态。特别是活动于现实三维空间的运动目标越来越受到关注。智能交通、电子战场、物流管理、移动电子商务、旅行者服务以及其它各类LBS服务都离不开高效的运动目标信息管理机制。随着三维GIS(地理信息系统)和虚拟现实技术的成熟,运动目标的信息管理也需要由原来的一维或二维扩展到三维。对运动目标的信息处理不仅涉及到空间属性,而且涉及到时间属性,因而相对复杂。然而,传统的数据库管理系统不能完全满足运动目标信息管理的需要。它们是为“静态”数据管理而设计的,数据的任何改变将导致数据库的显式更新,这与目标运动的空间状态持续变化特征不相符,而且,结构化查询语言SQL不是为处理时空数据而设计和优化的,在查询目标运动数据时有明显的不足。另一方面,目前时态数据库主要以变化相对缓慢的以统计为重要目的的数据作为管理和研究对象,它的实时性不强,而且对时间序列的管理较少涉及到空间属性。因此,有必要研究新的数据组织与处理方式来管理与运动目标相关的时空数据。本文研究的重点是三维运动目标的数据组织和管理,包括运动目标的表示和运动数据的索引技术。 对于运动目标信息管理的研究,目前普遍的方法是将运动目标抽象成空间移动的点,忽略目标的形状、大小、外观等因素,集中注意力研究目标空间状态及其变化,通过各种手段采集、传输、存储、索引、查询和处理目标的空间状态数据。在这种情况下,运动目标的表示也就简化成了目标的运动表示。现有的目标运动模型有多种,但可以归纳为点模型、时段模型、折线模型和函数模型四类。这四类模型各有特点,适于不同的应用需要,但它们多基于一维或二维运动目标的管理,要处理三维运动目标,需要做一定的扩展,而且上述所有模型中没有一个模型考虑到目标的姿态,而运动目标的姿态在决定三维目标的空间状态中也是至关重要的。本文在分析了三维运动目标的时间特性、空间特性和运动特性的基础上,改进函数模型,特别是MOST(Moving Object Spatio-Temporal)模型,消除其对运动函数所隐含的显式、增量式和分离式要求,设计了综合运动模型。并提出运动数据的细节层次概念,以不同精度和不同计算复杂度的数据及其相应的计算方法满足不同应用的需要。综合运动模型将运动数据和相关的处理用面向对象的方法进行封装,允许不同类型、不同结构的数据和不同种类的处理方式并存,适用于函数法与样本法的不同运动模型构造方式。针对多数应用系统中数据采集的方式和类型,从目标的运动位置、姿态、运动方向、速度、加速度、旋转速度、旋转加速度等各项指标及其相互关系出发,本文还提出了状态序列模型。该模型用离散的样本加上相应的插值方法,拟合目标运动的过程,对数据和处理过程进行封装,形成简洁的运动表示模型,弥补了现有模型对三维运动目标表达上的缺陷,同时还兼容对一维和二维运动目标的管理。本文还分类讨论了状态序列模型对不同运动目标的表示方式,分析了模型中所表达的信息,包括基本信息和导出信息。 在状态序列模型的基础上。本文提出了以2~n索引树为中心的多向多级运动数据索引机制。运动数据的索引依赖于所采用的运动模型,几乎每一类模型都提出了自己相应的索引方法。通过分析运动数据索引的需求发现,根据运动目标的特点和运动数据访问的需要,所采用的算法应该不仅能够对大量运动目标当前的运动状态建立索引,而且能够管理历史数据,按照运动数据模型所规定的方式提供对将来运动状态的预测。运动数据的追加是大量发生的,所以所建立的索引应该满足这种经常更新的需要,不至于频繁引起大范围或多级结点的大规模操作。运动数据的查询有多种形式,其中有快照式查询,也有连续查询。因此,所建立的索引机制应该兼顾到各种查询的方式。为了满足这些要求,本文采用空间分割的方法,扩展四叉树和八叉树,建立n维空间的2~n索引树。结合运动数据及其处理过程的特点,给出了2~n索引树的数据结构,以及树的建立、状态向量定位、添加、删除、滞后更新和树扩展等相关算法。为满足不同类型查询的需要,特别是连续查询对于相同目标连续一段时间内运动状态的查询要求,在建立运动数据索引时,需要将样本数据与2”索引树相分离,同时按照目标与样本时刻建立索引,从而形成多向多级的索引结构。本文探讨了这种混合索引方式,并给出了状态样本数据的具体存储形式。 在详细分析各种运动数据访问类型的基础上,本文讨论了各种查询方式,给出了基本状态查询和运动过程查询的方法,介绍了正向查询(目标一时刻查询)和反向查询(时间一状态查询)的不同检索程序和步骤,详细讨论了对相容区域和相交区域查询的不同处理方式。从运动数据在n维空间的分布和数据访问过程出发,介绍了运动目标查询的准确性问题,探讨了准确找出所有满足条件的运动目标的不同方法,找出了前期处理法与扩展查询区域法在处理过程和效率上的不同,分析了一般查询与精确查询在查询步骤上的区别。通过分析目标与目标之间的相互关系,提出了目标关系查询的方法,并

【Abstract】 With the development of wireless communication, positioning and photogrametry, people are able to expand the information management of spatial objects from static to dynamic, and the moving objects in 3D real world are drawing more and more attentions. Many applications, such as ITS (Intelligent Transport System), digital battle field, logistics management, mobile e-commerce, traveler service and other LBS (Location Based Service) systems, need efficient MOIM (Moving Objects Information Management). And MOIM is expanding from 1D or 2D to 3D with the application of 3D GIS (Geographical Information System) and VR (Virtual Reality). The information process of moving objects is complicated as it deals with spatial attributes as well as temporal attributes. However, conventional DBMS (Database Management System) can not meet the needs of MOIM. It is designed for static data management, thus any data change will lead to explicit update in database. This is not fit for continuous spatial state change of moving objects. And SQL (Structural Query Language) is not designed and optimized for processing of spatio-temporal data. It is not good enough for moving objects data query. On the other hand, the current study of temporal database is focused on data which change slowly and to which statistic process is one of the important processes. It is not effective enough for real time process and it does not deal with spatial attributes. Therefore, it is necessary to study novel data organization and management algorithms to process the spatio-temporal data of moving objects. This paper is focused on the data organization and management of 3D moving objects, including representation of moving objects and index technique of motion data.The current study of MOIM is to regard the moving objects as the points in space, to pay no attention to the shape, size and appearance of the objects, and to study the objects’ spatial states and their changes, to collect, transfer, store, index, query and process the spatial state data by different means. In this way, the representation of moving objects is simplified as the representation of objects’ moving or objects’ motion. There are various motion models developed. They can be classified into four categories, point model, time interval model, polyline model and function model. Each kind of model has its own advantage and fits for certain application. But most of the models are for ID or 2D moving objects, and need further improvement before they can deal with 3D moving objects. And none of the existing models takes the orientation of objects into consideration, which is very important in spatial state of 3D objects. After analyzing the temporal attributes, spatial attributes and motion attributes, this paper improves function model, especially MOST (Moving Object Spatio-Temporal) model, eliminates the implicit requirement for explicity, increment and separation of elements to motion function, and proposes GMM (General Motion Model). It also introduces LOD (Level of Details) to motion data and satisfies different application requirements by different accuracies and complexities for motion data and their processing. GMM encapsulates motion data and their processing according to object oriented method. It allows different processing methods and data with different types existing in same object, and is suitable for both function method and sampling method of motion model. In compliance with data collecting and types in most application systems, this paper proposes SSM (State Sequence Model) based on the position, orientation, moving direction, moving speed, moving acceleration, rotating speed, rotating acceleration and their relations of the moving objects. SSM restores objects’ motion using discrete sample data with interpolation and extrapolation, and also encapsulates data and their process. It is a concise model for representing moving objects, which isgood for 3D objects as well as ID and 2D objects. This paper also discusses the representation for different kinds of object using S

  • 【网络出版投稿人】 武汉大学
  • 【网络出版年期】2004年 04期
节点文献中: 

本文链接的文献网络图示:

本文的引文网络