AUTOMATIC DATA ACQUISITION FROM TOPOGRAPHIC MAPS
USING A KNOWLEDGE-BASED IMAGE ANALYSIS SYSTEM*
N. Ebi, B. Lauterbach, Ph. Besslich
Working Group Digital Systems,
University of Bremen, FB1,
P.O.Box 330440, D-2800 Bremen 33, Germany
ABSTRACT:
A system for automatic data acquisition from topographic maps using knowledge-based image analysis methods is
presented. The investigations are part of the interdisciplinary project Environmental Planing System. The goalis to generate
a symbolic description of the map contents that may be imported into the ARC/INFO GIS for supporting the geographical
tasks of the project.
High resolution color scanned topographic maps (scale 1:25 000 and 1:5 000) serve as a data source. Binary color map
layers are produced by a HSI color space guided multi-level segmentation. The processing of each layer includes
vectorization as well as application of methods like neural network-based symbol and object recognition for the extraction
of attributed structure primitives. Subsequent analysis is based on a hierarchical structuring of the map scene with map
objects and their relations. A frame mechanism is utilized for modeling the concepts of all types of map objects. A control
module driven by the data model supervises the creation of instances of the concepts. The map objects located at the lowest
hierarchy level correspond to the attributed structure primitives. The interface to ARC/INFO is represented by the instances
of the upper levels of hierarchy.
KEY WORDS: Topographic Maps, Raster Image Analysis, Attributed Structure Primitives,
Knowledge-Based Interpretation, Data Base, GIS
1. INTRODUCTION created using a color scanner. Symbolic image information
is extracted by splitting the map image into color layers
which are processed using raster object recognition and
vectorization. The extracted symbolic information, called
attributed structure primitives serves as data source for the
knowledge-directed image analysis. The analysis is real-
ized by the control module, the image model, the instance
storage and the image based conflict solving module. The
instances represent the extracted map information and
have finally to be converted to the ARC/INFO data base
format.
Geographic information systems (GIS) are gaining impor-
tance for environmental planning tasks. For an efficient
and flexible use of these systems itis necessary to combine
data acquisition, creation of a valuation scheme and GIS
in an integrated concept. This is the objective of the inter-
disciplinary project Environmental Planing System *. For
an efficient spatial valuation it is necessary to have an
adequate data base. Maps are an important source of infor-
mation for this data base. At present a lot of relevant maps
have still to be digitized manually, which is a time con-
suming and error prone process. To improve the situation
we are developing image analyzing methods for automatic
data aquisition from maps. These methods are described in
this paper as part of the above mentioned project. As
primary data German topographic maps of scale 1:25 000
and 1:5 000 are used.
Map raster image
Manual input e.g. from
terrain mapping or
Soil analysis
2. SYSTEM OVERVIEW
An overview of the proposed system is shown in Fig. 1.
The system kernel contains modules for storage, modifica-
tion, manual digitization, graphic representation and eval-
uation of spatial data. Considering the complexity of the
system kernel we preferred to use a commercial GIS. We
decided to use ARC/INFO from ESRI, because it is a
powerful tool and is in wide-spread use over Germany.
Map interpretation
Data conversion
The acquisition of spatial data is done by means of knowl-
edge-directed map analysis and interactive input of addi-
tional information obtained from soil analysis, remote
sensing techniques, terrain mapping or other sources. The System user
System user will have the opportunity to evaluate the spa-
tial information via an expert system connected to the data
base.
Geographic
expert system
Fig.1: System overview.
The principle of the proposed knowledge-directed image
analysis is shown in Fig 2. A raster o of the map Is 3. SCANNING OF TOPOGRAPHIC MAPS
The topographic maps we use are mainly printed in four
colors printing technique (Schoppmeyer, 1991). The
* Chis work was supported in part by DFG (German Research colors are cyan, magenta, yellow and black. Therefore, a
Association) under contract Stäblein/Besslich: 24-Bit-RGB color scanner is necessary to create the raster
Environmental Planing System (Sta 126/19-1). images. The smallest objects contained in the maps are
655
