data 
)e re- 
sts of 
ation 
using 
t the 
'ators 
close- 
ation 
ry of 
y the 
| vec- 
995]. 
road 
given 
nfor- 
re 6) 
/DB 
data 
pter. 
ledge 
| will 
e us- 
| the 
rep- 
ards, 
] de- 
ation 
ased 
1 ob- 
nted 
the 
  
  
ased 
lude 
n or 
accuracy of the captured landscape objects (differences in 
the scale of the DLM's). Also, there exists a difference 
between the data representation of the different informa- 
tion systems. The data can be made available in a graphic 
format (e.g. DXF), in a layer format (e.g. ALK) or in 
an object oriented format (e.g. SAIF). Therefore, we have 
to convert the corresponding data models into a uniform 
model. We have chosen an object oriented data model to 
represent spatial objects because of the natural representa- 
tion of real world objects in contrast to their representation 
in relational databases [Riekert 1993, L.Raynal, B.David 
& G.Schorter 1995, Fritsch & Anders 1996]. 
DINER 
  
  
   
  
GeoObjectComplex 
   
Thematic 
Q 
RNC 
    
  
  
  
  
| Geometry | |ReferenceSystem | 
  
Figure 7: Description of the spatial object class 
      
/\ ; Attribut 
Thematic 
Name : String 
  
  
Value : String 
  
Figure 8: Thematic class 
Our object oriented data model used is derived from 
the data model of the Canadian standard SAIF (Spa- 
tial Archive and Interchange Format) described in [Spatial 
Archive and Interchange Format: Formal Definition Re- 
lease 3.2 1995]. In the following figures we use the 
OMT-Notation (Object Modeling Technique) described 
in detail in [Rumbaugh, Blaha, Premerlani, Eddy & 
Lorensen 1991]. In the OMT-Notation the rhomb symbol 
represents the object oriented feature aggregation (has.a 
relation) and the triangle symbol represents the feature 
inheritance (is.a relation). Our developed data model 
uses container classes provided by the used object oriented 
database system. Geographic objects are modelled by the 
class GeoObject (figure 7). Parts of the class GeoObject 
are the classes Thematic and SpatialRelation. The class 
Thematic (figure 8) is implemented as a associative map 
and represents the non-spatial attributes of a geographic 
object. The class SpatialRelation is used to represent the 
spatial attributes and describes the kind of geometry and 
reference system which belongs to the geographic object. 
The class Geometry shown in figure 9 provides the ba- 
sic geometric objects described by the classes Point, Line, 
Area and GeometryComplez. The class GeometryComplez 
93 
is used to build geometries, which are composed of the 
classes Point, Line and Area. Every object from the class 
Line is defined by two objects of the class Point (start 
point and end point of the line) and any number of addi- 
tional points depending on the type of line interpolation 
(e.g. polygon, spline). An object of the class Area is de- 
scribed by at least one border line of the area (areas with 
holes includes two or more border lines), which is modelled 
by the class Line. 
     
Area 
  
    
       
Point (abstract) 
2+ 
size_of_area 
Figure 9: Geometry class (point, line, area) 
5 METHODS 
We can distinguish the following methods which every spa- 
tial object in the database has to provide : 
e relational operators (figure 10) for 
> semantic relations (e.g. typeof, is_a) 
The interpretation process has to be able de- 
termining the kind of object it is looking at. 
For example, with this method we are able 
to determine the semantics of a spatial object 
(e.g. road or building). 
> structural relations (e.g. part_of, has.a) 
In order to get information about the compo- 
sition of an object it provides methods to get 
all objects, which are a part of this spatial ob- 
ject, or all objects to which the spatial object 
belongs. 
> topological relations (e.g. inside, disjoint, 
asically we need information about the topo- 
logical relations of a spatial object because we 
have implemented methods to test all topo- 
logical relations as described in [Egenhofer & 
Herring 1990]. 
> neighbourhood relations (e.g. lies beside, 
near_to) 
The methods for the neighbourhood relations 
are not implemented at the moment. They 
will be used to deduce relations which cannot 
be determined only by topological relations. 
For example, the database query Retrieve all 
buildings which lie beside road A. This can only 
be solved using geometrical relations. Another 
problem is, that there exists no mathematical 
definition of this kind of relation. 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996