Figure le. Subtraction of both objects from the result 
in figure 1d. Segmentation of the new areas to unique 
objects. 
Figure 1f. Acceptance test for the objects created in the 
previous figure. The left object is rejected and the right 
object is accepted. 
  
Figure 1g. Final result when the accepted object is 
added together with the two original objects into the 
raster data base. 
  
137 
RESULTS 
With this amalgamation method, we have obtained 
two improvements compared with earlier strategies; 
1. With the retrieval of the shortest distance between 
the two objects a good estimator is found for the 
amount of region grow and shrink that should be 
performed to get the areas to join together. 
2. The procedure to segment the newly created 
objects makes it possible to control that concave bays 
of the original objects are not filled in the process 
(figure 2). 
This idea has resulted in a prototype tool which has 
been built within the software GRASS(Geographic 
Resources Analysis Support System)(Westervelt, 
1991). Bourne shell scripts control the user interface 
and the execution of the proper modules in GRASS. 
This tool should be seen as one tool among others in 
a cartographic workbench for performing genera- 
lization. In the hands of a cartographically skilled 
person, it will make it possible to collect rules about 
when amalgamation can be carried out and how this 
can be done. 
DISCUSSION AND CONCLUSIONS 
In this paper an improved method for cartographic 
amalgamation of area objects has been described. In a 
first phase, it is supposed to be used in an interactive 
system for the purpose of gaining experience about 
the generalization process amalgamation. In a later 
phase the tool can be built into a semi-automatic or 
automatic system, based on a rule based approach 
with an expert system to guide the generalization 
process. 
The rules that should be collected are of two kinds; 
1. Onto which cartographic object types the amal- 
gamation should be applied, it means which objects 
to select. In the currently used data set with 
topographic data the amalgamaion process could be 
applied on small wetland, forest and open field areas. 
2. Which rules should be used for acceptance or 
rejection of the newly created areas. 
The second question regarding acceptance criteria is 
that the newly created areas should have both 
original objects as neighbours as mentioned above. 
Another rule that could be resonable is to state that a 
newly created area is not allowed to cover a certain 
cartographic class. E.g., in most cases, we do not like 
the new area to cover water. A rule base that 
specifies these priorities between the original 
cartographic classes in the map could be written here. 
These rules could also be written as explicit rules, 
that handles all occuring cases if the data base does 
not contain to many cartographic classes.