lower left corner and the upper right corner of the 
image to fit it on the model. So the database has to 
store these data as well in an image dataset. In this 
coherence a tool to apply a coordinate offset is also 
important. 
Last but not least there has to be a possibility to 
transfer data from the image database to other 
applications especially to AutoCAD and to import data 
to the image database for example from the 
Rolleimetric CDW program. 
3. FUNDAMENTALS OF RELATIONAL 
DATABASES 
3.1 Tables & Datasets 
The structure of the image database follows the 
concept of relational databases. Before we discuss 
the structure of the image database in detail we have 
to define what a relational database is. 
In common a database is an amount of logical bound 
up datasets. A dataset represents a real occurence 
and the special features of it. A relational database 
includes all the logical bounded datasets in a table. 
The rows of the table define the datasets. The 
columns represents the features of the dataset. Every 
dataset has an individual attribute (feature), so that it 
is unique. Thus you can easily find and select a 
dataset from the table. The unique feature is called 
"primary key". The dependencies between rows, 
columns, datasets and attributes are illustrated in 
figure 3. 
Mp] ID — 
  
  
  
Figure 3 
Img ID 
Img No. 
Img Res 
Img file 
0001 
1005 
640x480 
tif 
002 
1006 
640x480 
Gif 
232 
3.2 References & Relations 
In the image database we have to store different 
types of datasets. As discussed before we have to 
represent the hierarchical structure of 
photogrammetric projects. We did this by defining two 
tables one for the so called "Masterproject" the other 
for the "subproject". The Masterproject table and the 
subproject table are very similar. Both tables contains 
datasets which specify the project by a title, a date 
and comment. As seen before a subproject is part of 
a Masterproject. So there is a dependency between 
the two tables. The relationship is a 1:n - relation. This 
means one dataset in the Master-project table may 
have n corresponding datasets in the subproject table. 
The subprojects are build up by the images which 
belongs to one special subproject. So the we defined 
a "Image-table" which has also a 1:n - relation to the 
subproject table. 
Beside the camera standpoint coordinates we have to 
store the image coordinates and further informations 
to the image itself. Information on the images are 
stored in the above explained "Image-table". The 
image coordinates were stored in the "Image data 
table" Each image only has one corresponding 
dataset in the image data table. This is called a 1:1 - 
relation. 
Finally we have the camera data to be stored in a 
table. The camera data depence on the type of 
camera you used to get your photogrammetric 
images. We stored this data in the "camera table". An 
image may have one corresponding camera dataset. 
In addition to the explainations of the dependencies 
we have to discuss how they can be realized. As we 
saw before each dataset has a unique feature. To 
connect datasets from different tables we use this 
unique feature in the corresponding table. This means 
we expand the master dataset with the unique feature 
of the depending dataset. For example we store the 
camera number in the dataset of the image. So we 
can easily access on the camera information for one 
special image. But what happens if there is no 
corresponding dataset? 
To avoid this circumstance we define a "referentiell 
integrity". This means if a dataset needs a 
corresponding dataset it only can be stored when the 
corresponding dataset exists. This technique also 
prevents the database from corruption when deleting 
a dataset but not the corresponding ones. For 
example: Before deleting a Masterproject you have to 
delete all subprojects which belong to this 
Masterproject. The subproject itself only can be 
deleted when all images that correspond to this 
subproject were deleted and so on. Figure 4 shows 
the relations between the tables of the image 
database and their referentiell integrities. 
  
   
   
|] BDB-Explorer 3.01 
] Hife zur Filebox 
] "3m 
   
Figure 4 
3.3 Questic 
A great adv 
ability to get 
To create tl 
allows you 
datasets. ' 
language (S 
basic keyv 
keywords y 
database. / 
could be: 
subproject 
adjustment. 
* FROM Im 
- bundle a 
The introdu 
consistent 
relational d 
language t: 
and select 
language is 
SQL has 
standard. 
The SQL 
definition, c 
administrat 
quantity ori 
relational a 
about relat 
the special 
For examp 
which belo 
SQL stater 
"SELECT 
WHERE S 
The "selec 
creates a rt 
statements 
statements 
operators 
statement 
have to fit | 
the staten 
"Show me