KNOWLEDGE-BASED INTERPRETATION OF SCANNED LARGE-SCALE MAPS
USING MULTI-LEVEL MODELLING
H. Mayer, C. Heipke, G. Maderlechner
D621: Competence Center GIS
Siemens Nixdorf Informationssysteme AG, Otto-Hahn-Ring 6, D-8000 Munich 83, Germany
Tel: + 49-89-636 2166; Fax: + 49-89-636 45202; E-mail: hm@ssym1.zfe.siemens.de
Chair for Photogrammetry and Remote Sensing
Technical University Munich, Arcisstr. 21, D-8000 Munich 2, Germany
Tel: + 49-89-2105 2671; Fax: + 49-89-2809573; E-mail: heipke@photo.verm.tu-muenchen.de
Corporate Research and Development
Siemens AG, Otto-Hahn-Ring 6, D-8000 Munich 83, Germany
Tel: +49-89-636 3389; Fax: + 49-89-636 2393; E-mail: gm@bvax4.zfe.siemens.de
Commission III
ABSTRACT:
Present results of map interpretation systems still re-
quire significant manual interaction effort. This is mainly
caused by a lack of high level knowledge in the systems.
In this paper we introduce a multi-level model for a
large-scale map as the key to improved and robust auto-
matic map interpretation. The levels of the model are
called semantic objects, graphics and text, image graph
and image. Objects, operations to be performed on the
objects and relations between the objects are presented
for each level.
Mixed bottom-up and top-down reasoning is performed
by instantiating the concepts of a semantic network built
up from the semantic objects, the graphics and text and
the image graph levels. The interpretation uses relations
on attributed arcs and nodes as well as template mat-
ching on the image.
First results for the interpretation of a Bavarian cada-
stral map of scale 1 : 1000 are presented.
Key Words: automatic map interpretation, model based
interpretation, semantic network, top-down and bot-
tom-up processing
578
1. INTRODUCTION
For all kinds of spatial planning purposes large amounts
of topographic and cadastral data are required. This
information is more and more needed in digital form to
build up geographical information systems (GIS). Com-
pared to photogrammetry, surveying etc existing paper
maps are the most inexpensive source of input data for
the GIS. A GIS whose level of detail corresponds to
large-scale maps (scales from 1 : 100 to 1 : 5000) is
composed of attributed (ie meaningful) high quality
vector data. Maps of this scale are usually plotted as
black and white (ie binary) line drawing.
If a map of this kind is scanned, the result is a raster
image. This is not the type of data needed for the vector
GIS. Today the conversion of paper maps to GIS data is
accomplished through manual digitisation of the maps.
Another possibility is the interactive digitisation of the
scanned image on the screen. This is known as "overlay
technique" and has the advantage that the map and the
digitised data are directly comparable. The drawback of
both approaches is that they are very labour intensive
and that the results depend on the skills of the operator.
Because there are a lot of maps to be digitised, the
automatic conversion of the scanned maps to attributed
vector data is a great challenge. The basic tasks are the
extraction and the interpretation of the lines.
NN. v" wr AM A PN
Un oM em be
C) e+
