International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004
better than the map production systems of the 1990s in finding a
reasonable solution to the challenge of deriving multiple scale data
from a master database (Lee, 2003), and hardware performance and
cost makes them suitable for implementation in a full production
setting (Forghani er al, 2003). For example, ESRI’s current object-
oriented ArcGIS software (version 9.0) provides à spatial framework
to support generalization needs, by introducing geoprocessing
concepts and map generalization tools that have been enhanced and
implemented in the geoprocessing framework (Lee, 2003). The issue
is still the incorporation of cartographer knowledge into the
generalization process, as well as finding situations where high
accuracy and other automatically derived information are both useful
and valuable.
5.0 OVERVIEW OF MAJOR GENERALIZATION SYSTEMS
Despite considerable R&D efforts directed toward automation of
cartographic generalization by academics and the GIS industry,
existing software tools are not able to play a more significant role
than graphic editing and statistical calculation (Meng, 1997). This is
due to inadequate "intelligence" (compared to cartographers), in
determining ‘how’ and ‘when’ to generalize (McKeown er al. 1999;
lwaniak and Paluszynski 2001). However, to remedy this
shortcoming, rule-based systems were introduced to incorporate
topological, geographical and cartographical expert knowledge in
order to build a map generalization expert system. Examples of such
expert systems (eg for generalization of roads) are given in Peschier
(1997) and Skopeliti and Tsoulos (2001). This implies a lack of fully
automated generalization tools. A number of commercial GIS
vendors (e.g. Intergraph, ESRI, and LaserScan) have worked with
various mapping agencies to use these generalization tools for the
production of maps at various scales (e.g. Kilpelainen, 1997; Meng,
1997) while developing tools to automate generalization.
ESRI’s recent ArcGIS (version 9.0) product offers a spatial
framework to support GIS and mapping needs. Geoprocessing,
combining its earlier command operation with a modern user
interface. has become an important part of upcoming software
releases. Developing generalization tools within a geoprocessing
framework has opened opportunities to explore new technology and
data models, and to make enhancements using better techniques (Lee,
2003). In principle, it embedded the Douglas & Peucker algorithm
for line generalization. However research shows that ArcGIS
(versions 8.1-8.3) Generalize does not provide total solutions for
generalization (Limeng and Lixin, 2001; Kazemi, 2003), because
after the point, line and the feature are simplified, manual editing was
still required. The reason is that topological errors are produced when
applying the Generalize tools, such as line crossing and line
overlapping; for polygon coverage, errors such as no label or
multiple labels were introduced. To deal with these problems manual
editing is necessary (Limeng and Lixin, 2001; Kazemi, 2003).
Detailed generalization capabilities of this product are described by
Lee (2002, 2003).
The CHANGE software developed by the Institute for Cartography
of Hanover University is capable of generalizing building and road
objects at a scale ranging from 1:1,000 to 1:25,000. The CHANGE
software generalizes buildings through its sub-program of CHANGE-
Buildings, and for roads CHANGE-roads — (www.ikg.uni-
hannover.de).
The Intergraph Corporation developed the MGE DynaMap
Generalizer as an interactive platform that works under Unix and
Windows NT. It deals with small-scale derivation from large-scale
databases, theoretically without limitation of scale range. À number
of visualization tools in DynaMap Generalizer are also available to
assist the interactive generalization processes (Lee, 1993). Iwaniak
and Paluszynski (2001) combined the expertise of a cartographer
with DynaMap Generalizer in batch mode to perform the actual map
transformations, and a rule-based system for controlling the process.
They noted that this system does not have a mechanism for
controlling topology. Unlike CHANGE, when making essential
decisions in DynaMap Generalizer (such as tuning generalization
sequence), system users must select parameters for each algorithm
and the number of iterations to be applied for each particular task.
1224
DynaMap Generalizer has been tested for different generalization
tasks in several countries, the USA, UK, Germany, Spain, Sweden,
the Netherlands, and China. In Spain, for example, DynaMap
Generalizer is used to derive a topographic map at 1:100,000 from
1:50,000 scale data, and to produce an atlas composed of different
maps at different scales (Baella er a/., 1995).
Since 1990, LaserScan has been developing an Open Systems object-
oriented Application Development Environment (ADE) named
"Gothic". Since 1994, LaserScan has been developing a new
generation Mapping and Charting application using the Gothic ADE.
The new application, named LAMPS2, uses a central database of
map data to generate a range of products. Operations are performed
in two phases: compilation (database creation and maintenance from
a range of sources), and product generation (extraction,
symbolization and generalization).
6.0 REMARKS AND CHALLENGES
A review of the literature demonstrates that future research and
development work on automatic generalization should focus on the
following major streams. This judgment is supported by other researchers
in the field of map generalization (Meng, 1997; Costello ef al., 2001; Le,
2002 and 2003). To build an automatic generalization tool, Lee (2002) and
Kazemi (2003) highlighted a number of major streams:
e A need to evaluate and validate existing generalization
tools as identified by researchers (e.g. Visvalingam, 1999),
as well as improvements in editing tools (e.g. Muller, 1995)
for both area generalization and line generalization
applications. To fulfill the need to evaluate and validate
existing generalization tools, the authors' research will
focus on the development of a detailed generalization
framework to derive multi-scale GEODATA. It focuses on
integration and utilization of generalization operators as
well as cartographer's intuition/skills using the ArcGIS 8.3
Generalize (and possibly DynaMap Generalizer) software
in order to generalize a road network database from
GEODATA TOPO-250K Series 2 to produce smaller scale
maps at 1:500,000 and 1:1000,000.
e Maintaining a single sophisticated database that supports
many applications (rather than multiple simplistic map
layers), as well as a well-designed database, provides a
platform to support data derivation, generalization,
symbolization, and updating (Lee, 2002). The idea is to
associate geographic objects/features to multiple scales and
maintain the cartographic quality of spatial data products.
This requires the development of data models that support
derivative mapping concepts. Many geographic objects
vary in their appearance with scale, so that it is difficult to
encapsulate all possible details for all probable scales
within a single data model. The way forward is to model
data in an object-oriented solution
e. Development of universal guidelines to derive smaller scale
products from a master database. As NMAs (e.g. Land
Information New Zealand, Geoscience Australia, and
Ordnance Survey) migrate their dataset into multi-scale
national seamless coverage, it is essential to develop
guidelines and tools to derive smaller scale products from
their fundamental spatial information (e.g. GA's
“GEODATA TOPO-250K” national coverage) at à
consistent level, as well as providing a basis for
generalizing other data sets at different levels of
generalization. The guidelines should also highlight both
essential and desirable steps for generating smaller scale
maps in line with a production environment focus. These
include topological relations between the object types and
classes, how the objects have to be selected, how [©
generalize, when to smooth, when to delete, when to merge.
how to do reclassification of roads, and so on.
e A set of automatic generalization tools and a set of efficient
post-editing and cartographic editing tools is needed (Lee,
2002). For example, ESRI has begun developing a sel ol
commonly used generalization tools for simplification.
Intei
7.0 At
We w
Profes
Unive
topic c
8.0 RI
Arnolc
behavi
updati
2003, ‘
Baelia.
practic
Map G
Barrau
placem
Autoca
Badler
displac
pp. 1-1
Brassel
framew
Geogra
Brooks.
Compu
Resour
2001, C
Brooks,
framew
Forum 2
Costellc
commu
AUSLIC
the AUF
Hyatt, N
Cecconi
generali:
Proceed
processi
Davis, A
GIS: M
spatial d
pp. 60-6.
Duda, R.
New Yor
Douglas,
reduction
line or its
192.
Forghani
satellite
Mapping
2003, Sey
Forghani.
linear fea
Congress.
Gross, J
press, Loi
