Databases Reference
In-Depth Information
2
Real World Database
The objective of the RWDB is to provide a total environment of capturing, modeling
and storing physical or logical objects in the real world. Everything in the real
world could be modelled through it and any type of data could be accumulated.
For this purpose, RWDB must involve at least 4 components listed below.
Conceptual architecture of RWDB is shown in Figure.1.
•
Real World Capturer (RWC)
The objective of RWC is to capture the external form of objects in the real
world. There exists various types of 3D or spatial information depending
on capturing devices. The simplest one is a set of video cameras. We can get
a sequence of frames from, for example, a left-eye camera and a right one
simultaneously. Recent technology enables us to get the full-color and high
quality image with no compression of data. However, for the frame
sequences, it is very hard to extract a complete surface model or motion
(kinematics) model of each object. On the other hand, several ”Motion
Capturing Systems” are commercially available but those system can trace
only a small number of marking points and the most part of original
information (surface and texture) are lossed.
Another type of input device is “3D Scanner” by which we can get a perfect
surface (polygon) model for static objects. The practical solution is to get
above two kind of informations from the real world and to combine two
models into one in the database level. This idea is summerized in Figure.2.
•
Real World Modeler (RWM)
RWM is a set of tools each of which analyzes original frame images and
generate a new information. For example, the Outline Chaser
3)
catches the
outline of an object in a certain video frame, and then trace the outline in
preceding and successive frames. On the other hand, the Point Tracer detects
stereo pairs and the range values for (specified) points from a couple of
(Left and Right) frame images, and makes a rough sketch of 3D objects in
the real world. Many algorithms are investigated and evaluated for range
image generation in Image Processing area
4)
. All the results of analysis are
stored into database preserving the correspondences to the original images.
•
Multimedia Database Engine (MMDBE)
MMDBE is a database which treats a variety of data types such as full
texts, graphic drawing, bitmap images and image sequences. The features
of such data are quite different from the conventional DBMS's ones, because
some of them are continuous and might occupy much more space than
traditional data types. As to data model, in order to integrate all types of
data, introduction of simple primitives to describe real world entities and
