Figure 4: Example for the different modelling of the same
crossroads (after [Walter & Fritsch 1995])
2.3 Databases of Different Scale
==) |
E
c)
Figure 5: a) ALK data, b) extracted roadsides, b) gener-
alized ATKIS road-objects
Besides the data revision based on digital aerial pho-
tographs it is also possible to use an existing digital map.
This dataset can include more information (larger scale)
or information which is captured with a higher accuracy
than the digital map which has to be revised. The official
digital cadastral map (ALK) of Germany contains spatial
information of large scale (1:250 - 1:2500) in several layers,
but is unfortunately not based on objects. Therefore the
usage of modern object oriented techniques requires data
conversion to built up object classes and object hierarchies
for further processing. On the contrary the medium scale
geographic information system ATKIS is based on classes
of several landscape objects and includes very similar infor-
mation contents as the ALK. The large scale dataset may
serve as data source to deduce medium scale information,
e.g. roads and further linear elements.
One aim of our work is to deduce ATKIS road objects
from the ALK because in both DLM’s exists the spatial
object road, although in the ALK roads are captured with
92
a higher accuracy. The effort in time and costs of data
capture for the information system ATKIS could be re-
duced. A two step approach is applied which consists of
object recognition (figure 5b) and object generalization
(figure 5c). The object recognition is carried out using
well-known methods of image processing to extract the
geometry of the roads. Among those especially operators
for the similarity, continuation, unity, symmetry, close-
ness and parallelism are used. The object generalization
is necessary because the ALK describes the geometry of
roads by areas and ATKIS represents the geometry by the
central axes of the lanes. Therefore we will use a vec-
tor geometry based algorithm described in [Olson 1995].
Additionally it is necessary to build correct ATKIS road
objects using the central axes.
3 SYSTEM DESIGN
The aim of our work is to develop a shell onto a given
object oriented spatial database to deduce implicit infor-
mation we are looking for. Our system setup (figure 6)
is based on the object oriented database Objectivity /DB
from Objectivity Inc., because an object oriented data
model has some advantages described in the next chapter.
A general interpretation process needs a type of knowledge
representation to describe the semantic models which will
be used by the interpretation process to control the us-
age of object methods (described in chapter 5) and the
symbolic data processing. In principal the knowledge rep-
resentation can be done by rule base systems, blackboards,
semantic networks or frame based systems, a detailed de-
scription of the different types of knowledge representation
can be found in [Reimer 1992]. We will use a frame based
system because of the direct link between frames and ob-
ject oriented data models. Frames are an object oriented
type of knowledge representation and well suited for the
object oriented data model.
Knowledge Representation
Knowledge System
Based She:
i
y OODBS
Object Models Y
AX
Object Methods
Interface
xX
Figure 6: Scheme of the system design
4 DATA MODEL
Most of the existing spatial information systems are based
on different models of the landscape because they include
only special parts of the landscape (e.g. traffic informa-
tion systems) and have differences in the generalization or
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996
acci
the
bet
tion
forn
an (
toc
moc
repr
tion
in r
Our
the
tial
Arc
leas
in «
Lore
repr
rela
inhe
uses
dat:
clas:
are
The
and
obje
Spat
refe:
The
sic g
Are
