UPDATING CARTOGRAPHIC DATABASES 
- MERGING OLD AND NEW DATA 
Ole Jacobi 
Professor 
Institute of Surveying and Photogrammetry 
Technical University of Denmark 
DK 2800 Lyngby, Denmark 
ISPRS Commission IV 
  
ABSTRACT: 
The merging of old and new cartographic data requires an adjustment of the control networks giving the two 
sets of data a common orientation in the reference system. The discussion centers around the problems inherent 
in a database consisting of objects of varying accuracy and of different levels of generalization. To provide the 
user of a map with a true picture of the contents of the under-lying database, the graphical outputs whether on 
  
  
Screen or on a map must reflect the variations. 
KEY WORDS : Updating, Generalization, Error propagation. 
1. INTRODUCTION 
Updating digital databases is a problem that arises 
when digital maps have been in use for some time. In 
principle, it is simple to do an updating: outdated ob- 
jects are removed from the database and replaced by 
new objects. The updating can be done from day to 
day, when information about the changes are at 
hand. Maps drawn on the basis of the database will 
automatically include all the changes that have been 
introduced in the database until the moment of ex- 
traction. 
The objects of the database are not independent of 
each other. The relative placement of the objects has 
been determined by making all measurements in a 
common reference system. In order to place the new 
or revised objects in a correct relationship to the ob- 
jects in the existing database, the same reference sy- 
stem must be used, and a least square adjustment of 
the new and old observations must be carried out (Ja- 
cobi,1988). If the coordinates of the existing database 
are kept fixed, and the new observations are fitted in- 
to these coordinates, an error propagation will arise. 
By successive updatings, this error propagation will 
destroy the geometric quality of the database. If the 
updating is carried out by a least square adjustment 
of old and new observations, the geometric quality of 
the database will improve. 
With a least square adjustment it is possible to make 
a correct numerical model consisting of observations 
of varying precision. On the basis of the adjustment, 
the accuracy of every point in the database can be 
computed. 
2. GENERALIZATION 
The geometric quality of the database will vary from 
object to object. But this is not the only variable fac- 
tor, the level of generalization of the objects in the 
map will vary as well. Standards of mapmaking des- 
cribing which objects must be included in the map, 
and by what detail each object must be described will 
576 
change from time to time and from survey to survey. 
In some parts of the database, objects may be repre- 
sented by many points detailing the objects, while in 
other parts of the database, the same type of objects 
may be represented by a few points giving only a co- 
arse outline of the object, thus posing the problem of 
connecting such diverse descriptions. 
In contrast to the adjustment theory for treating ob- 
servations with varying precision, there is no corres- 
ponding theory for integrating a minute survey of ob- 
jects with a coarse survey of objects in the same data- 
base. 
A cartographic description of the entities of nature is 
based on a generalization. As a function of the scale 
of the map, the objects in the map may include vary- 
ing degrees of details. If it were possible to establish 
a mathematical description of the cartographic gene- 
ralization, it would be possible to solve the above- 
mentioned problem. The generalization would be a 
direct function of the scale of the map, and it would 
be possible to compute a lower level of detail from ob- 
jects representing a higher level of detail. 
Methods for automatic generalization are the subject 
of many investigations and much research, and we 
may hope that a theory will turn up for solving these 
problems. 
3. COMBINING OLD AND NEW DATA 
If the update is done photogrammetrically, we have a 
new set of pictures which are to be orientated accor- 
ding to the reference system. This can be done in the 
usual way by an aerotriangulation, perhaps supple- 
mented by geodetic measurements. If the new aero- 
triangulation has only the geodetic reference points 
in common with the aerotriangulation which was the 
basis of the earlier measurements, there is no basis 
for a change of the existing data. Even if we made the 
two aerotriangulation adjustments as one big joint 
adjustment, it would not result in changed coordina- 
tes to the existing points, as no new links have been 
added to the adjustment.