ntstages 
and 11. 
then we 
lentified, 
gregated 
sented by 
to more 
ave been 
> Objects 
ons have 
ts at the 
resented 
situation 
mportant 
known in 
ally been 
duce the 
rea could 
cess has 
e graphic 
sults in a 
bjects to 
e smaller 
lification 
robjects, 
n, object 
ition in a 
graphical 
y so. The 
e. alower 
arns and 
e detailed 
ss is very 
ion step 
to obtain 
Je extend 
| agriculture | 
  
natgrsind| | forest | “agriculture | nat 
  
  
  
x E 
DIT 
27) 881) 4,10) (1,6 (6,8,12) 
Aa a ee a = 
T TNR — 
L | agriculture. | 
| nat veg ] 
"nat veg | (agriculture 
Ey L } 
1 
64, 65^» (28 A 
SF 
   
  
grsind | ‘forest |” agriculture | 
27) 86,88) 029 
     
   
  
object generalisation step 2 
object generalisation step 1 
fig. 13: A diagram representing the object generalization steps of figures 4 and 11. 
be defined in the form of database operations for 
databases that are implementations of the formal data 
structure (FDS) as explained in chapter 2. Such databases 
will be called shortly FDS-databases. 
A spatial database may containinformation about different 
aspects of a particular area, as we saw in the example 
of section 3.3. A generalization process may keep one 
aspect invariant, let us call that the primary interpretation 
of the database. The other aspects may be affected so 
that the information is not reliable after the process, we 
will call these the secondary interpretations of the 
database. If we consider generalization operations as a 
type of transformation of a spatial data base, then we 
should make explicit decisions about which aspects of the 
original data bases are to remain invariant, so we should 
decide what is to be considered as the primary interpreta- 
tion of the data base. This choice will be made within 
some users context of the data base, i.e. the user will be 
interested in the correct representation of some spatial 
characteristics, while others may be deformed by the 
transformation. 
A good understanding of database generalization may be 
useful for the design of procedures for spatial data 
acquisition. Information extraction from images is partly 
a reverse process to generalization. Generalization is a 
process with a stepwise data reduction, going from high 
resolution to low resolution. The information of the high 
resolution objects is merged into low resolution objects. 
Image interpretation can often be formulated as a process 
where data are produced stepwise. We can learn from 
generalization processes what information low resolution 
objects carry about their constituting high resolution 
objects. This knowledge may help us in image interpreta- 
tion, where large image segments can be seen as low 
resolution objects. These should then contain thematic 
information in addition to the radiometric and spectral 
information of the image itself, to identify smaller 
segments that may represent high resolution objects. 
REFERENCES 
Brodie, M.L. (1984): On the Development of Data Models. 
in: On Conceptual Modelling (eds. Brodie, Mylopoulus, 
Schmidt). Springer Verlag, New York. 
Brodie, M.L. & D. Ridjanovis (1984): On the Design and 
Specification of Database Transactions. in: On Conceptual 
Modelling (eds. Brodie, Mylopoulos, Schmidt), Springer- Ver- 
lag, New York. 
Egenhofer, M.J. & A.U. Frank (1989): Object-Oriented 
Modelling in GIS: Inheritance and propagation. Auto-Carto 
9. p.588. 
Frank, A.U. and W. Kuhn (1986): Ce// Graphs: A Provable 
Correct Method for the Storage of Geometry. Proceedings 
of the 2nd International Symposium on Spatial Data 
Handling, Seatle. 
Gersting, J. L. (1992): Mathematical structures for 
computer science. Computer Science Press, New York, 
Third edition. 
Hesse, W. and F.J. Leahy (1992): Authorative Topographic- 
Cartographic Information System ATKIS. Landes Vermes- 
sungamt Nordrhein-Wetsfalen, Bonn, 22pp. 
Marx, R.W. (1990): The TIGER system: automating the 
geographic structure of the United States census. In: 
Introductory readings in GIS, (Peuquet, D.J. and D.F. 
Marble eds.), Taylor and Francis, London, pp120-141. 
Martinez Casasnovas, J.A. (1994): Hydrographic Informa- 
tion Abstraction for Erosion Modelling at Regional Level. 
MSc. thesis, Dept of Landsurveying and Remote Sensing, 
Wageningen Agricultural University, Wageningen, 92pp. 
553 
International Archives of Photogrammetry and 
Remote Sensing. Vol. XXXI, Part B3. Vienna 1996