Research on 3D Moving Object Databases

本研究探讨了一种运动对象数据库系统的理论和实践，该系统能够回答关于人、汽车等运动对象的各种运动查询。首先，我们提出了一个能够描述运动物体时空特征的运动物体数据模型。它不仅描述了运动物体的质心位置，还描述了运动物体的方向和方位作为运动的数据，还描述了运动物体的轨道以及两个运动物体之间的距离、相互方向和拓扑关系等空间关系。我们提出了一种移动对象查询语言MOQL，通过它我们可以指定各种对象的移动。然而，我们发现，当我们使用WOOL来指定一个动态的移动，比如一个物体像这样移动时，有一个困难。为了解决这个问题，我们引入了一个动作捕捉系统来捕捉一个物体的运动作为一个样本运动，系统可以通过这个样本来检索一组运动与这个样本相似的物体。介绍了两个运动物体的同一性、基本相似度、部分相似度、时间弯曲相似度和时空相似度。结果表明，我们的基本相似度定义可以区分两个运动速度不同但轨道相同的物体。然而，我们也发现我们必须引入一种新的相似性定义来区分运动物体的速度模式。研究结果表明，当一个运动对象被划分为一组组件对象时，基于时间扭曲的相似度定义效果更好。

This research investigated a theory and practice of a moving object database system that is rile to answer to a variety of queries on moves of moving objects such as human and automobiles. First we proposed a moving object data model that can describe the spatio-temporal characteristics of moving objects. It describes not only the centroid position of a moving object but also its direction and orientation as the data of move, it also describes the orbit of a moving object as well as the spatial relationships between two moving objects such as the distance, the mutual direction and the topological relationship. We proposed a moving object query language named MOQL by which we can specify a variety of object moves. However, we found that there was a difficulty when we use WOOL to specify a dynamic move such as an object move moving like this. In order to resolve this problem, we introduce a motion capturing system to capture an object move as a sample move by which the system can retrieve a set of objects whose move are similar to this sample. We have introduced the identity, the basic similarity, the partial similarity, the similarity under time warping, and the spatio-temporal similarity of two moving objects. As a result it is shown that our basic similarity definition can distinguish two object move shaving different moving speed patterns but have the same orbit. However, it is also revealed that we must introduce a new type of a similarity definition that can distinguish the speed patterns of moving objects. It is shown that the definition of the similarity based on time warping works better when an object move is divided into a set of component objects.