CARTOGRAPHIC AMALGAMATION OF AREA OBJECTS 
Lars Schylberg 
Department of Photogrammetry 
Royal Institute of Technology, Stockholm, Sweden 
email: larss@fmi.kth.se 
ISPRS Commission IV 
ABSTRACT: 
A new method to control the cartographic generalization operator amalgamation is presented. An interactive 
prototype system is developed, which enables an user to select two areas that are subject for amalgamation. The 
areas are processed with a grow and shrink operator. The newly created areas are tested against acceptance 
criteria for acceptance or rejection. The idea is to use this system for knowledge acquisition for building a 
knowledge based system that can automate this task. 
KEY WORDS: Amalgamation, Cartographic Generalization, Image Processing, Grow and Shrink Operation. 
INTRODUCTION 
Amalgamation is the cartographic process where two 
or several individual areal objects are merged into 
one object. This is one kind of simplification 
operator for areas (Shea and McMaster, 1989). 
The reason for this work is a need to generalize 
topographic map data and to automate this process. 
Map data at the scale 1:50000 that should be used in 
the map scale 1:100000 needs to be simplified. At the 
National Land Survey of Sweden map separations 
are scanned from the 1:50000 topographic map series 
and combined into a digital raster map. It is then 
edited manually on screen with an interactive tool to 
simplify some cartographic objects and to remove 
errors introduced by the scanning process. 
Amalgamation is in some cases carried out during 
this step. This interactive editing is however tedious 
and expensive. Therefore, many small objects are 
left, even if they can not be presented on a computer 
screen or on a paper map at a map scale of 1:100000. 
This paper concentrates on a method how and when 
amalgamation can be carried out within a raster 
based GIS. The main idea is to select two carto- 
graphic objects of the same cartographic class and 
process them pair wise. The two objects are merged 
together and we are given the control to check if the 
result is acceptable or if it should be rejected. It is 
inspired by the idea of making tools available in a 
cartographic work bench for performing cartographic 
generalization on digital data to collect rules for later 
use in a knowledge based system( Weibel, 1991). 
135 
BACKGROUND 
Amalgamation of raster data can be carried out with 
a combination of thickening and eroding technique 
(Weber, 1982). The main problem with this method 
is that it is globally applied on all individual carto- 
graphic objects within the cartographic class pro- 
cessed. This gives the desired result in some cases, 
but also causes amalgamation of objects where it is 
not desired. It is difficult to control the process. The 
parameter of how many pixels that are used for the 
thickening and eroding steps is the only variable that 
can be changed. In a recent paper by Person and 
Jungert (1991) it was shown that this same approach, 
which they call the expand and shrink method, does 
not work very well when creating multi-resolution 
maps for path finding. 
Amalgamation can also be accomplished by a gap 
bridging technique (Monmonier, 1983). This tech- 
nique implies that a gap bridging operation is carried 
out in several iterations. It performs row wise and 
column wise gap bridging and gap bridging along the 
the two principal diagonals. This technique suffers 
from the same weakness as the method presented by 
Weber, i.e., it is applied globally over a map for one 
object class at the time. A largest gap measure is here 
also the only control parameter. 
In the following section a procedure is presented 
how the expand and shrink algorithm can be im- 
proved to permit a higher degree of control of the 
outcome of the amalgamation operation.