Implementation and Evaluation of Multimedia Database for Motion and Sound

本研究的目的是从数据建模和语言方面建立视频图像和音频数据库。我们还开发了几项关键技术来捕捉现实世界的信息。一种是基于一系列立体对帧的“目标检测”算法，该算法基于对目标的“了解”，可以检测出目标的简要轮廓；另一种是“对象追踪器”，它可以通过分析连续的帧来跟踪特定点/对象的空间位置。所有被删除的信息都可以与原始图像一起存储，并构成数据库的一部分。在数据模型方面，我们提出了一种函数数据模型及其图符号。在此基础上，设计了一个“工厂作业”数据库，并将其应用于某载重汽车零部件的装配过程中。例如，在这个数据库中，用户可以在一个单位工作中跟踪一个工人的“右手”，并可以计算手和他的“脸”之间的距离。这些操作被嵌入到功能性查询语言中，因此，计划者可以获得设计和优化制造系统中一系列“工作”的重要信息。

The objective of this research is to establish video image and audio database from data modelling and language aspects. We also developped several key technologies to capture real world's informations. One is the "object detection" algorithm from a series of stereo-pair frames, which can detect brief outline of an object, based on the "knowledge" about the object. Another is the "object chaser" which can trace spatial positions of a specific point/object by analyzing successive frames. All delected informations can be stored accompanied with original images, and form a part of databases. As to data model, we proposed a functional data model and its diagram notation. Using this, we designed a "Factory Work" Database and applied it to an assembling process of truck parts. In this database, the user, for example, can trace "right hand" of a worker in a unit work, and can calculate the distance between the hand and his "face". These operations are embedded in a functional query language, and, as a result, the planner can get important information for designing and optimizing a series of "works" in manufacturing systems.