A. Ozgur Dogru is a research assistant of the Cartography
division at Department of Geodesy and Photogrammetry in
Istanbul Technical University. He has done his PhD thesis
on Database and map design for navigational purposes in
the same department. His main interests are cartography,
GIS, car navigation, generalization.
Assoc. Prof. Dr. N. Necla ULUGTEKIN is a senior
lecturer in the Cartography Division of Department of
Geodesy and Photogrammetry in Istanbul Technical
University – ITU. Her main interests recently are
cartography,
visualization,
GIS
and
small
display
cartographic design.
She was involved in a number of individual and team
studies and projects on Electronic Atlas, Mountain Maps,
Map Design for Hand Held Devices and etc. She actively
remains supporting chambers activities and taken position
as a string head or a member in the technical commissions,
like Cartography Commission Secretary from Standing
Scientific and Technique Commission of Chamber of
Surveying, Geographic Information System Commission,
and etc. She is working on cartography in national and
international activities.
Assis. Prof. Dr. Seval ALKOY, Epidemiologist and Public
Health Specialist.
She is senior lecturer of Abant Izzet Baysal University Izzet
Baysal Medical Faculty, Public Health Department. Her
main
interests
are
epidemiology,
biostatistics
and
environmental health. She is studying for generalizing the
use of GIS for epidemiological purposes. In addition to her
academic studies, she is also president of Association of
Doctors for Environmental which is a member of
International Society of Doctors for Environmental-ISDE.
CLASSIFICATION OF ROAD JUNCTIONS BASED ON
MULTIPLE REPRESENTATIONS: ADDING VALUE BY
INTRODUCING ALGORITHMIC AND CARTOGRAPHIC
APPROACHES
A.O. Dogru1, N. Van de Weghe2, N.N. Ulugtekin1, Ph. De Maeyer2
1
Istanbul Technical University, Cartography Division, Istanbul, Turkey
(dogruahm,ulugtek@itu.edu.tr)
2
Ghent University, Department of Geography, Gent, Belgium (nico.vandeweghe,
philippe.demaeyer@ugent.be)
ABSTRACT
In this study, car navigation is considered as the basic activity and the multiple
representation approach for the production of the navigation maps is taken into account to
derive road network data for different representation levels. In this context, road junctions,
which are the most complicated parts of road networks, are classified considering two
different approaches, an approach based on cartographic aspects and an approach based on
algorithmic aspects. In the first approach, road junctions are classified according to their
aspects at different representation levels. In this concept, three representation levels, which
are derived from 1:5000 scaled base level, are used and representations of possible road
junctions are structured according to characteristics of these levels. Finally, they are
classified by considering their geometrical shapes in each level. In the second approach,
road junctions are classified according to algorithmic concerns. Junctions are defined in
matrices based on their specific characteristics and also these matrices are classified
according to their similarity. Finally, common representations are determined for these
classes.
Key
Words:
Multiple
representations,
road
junctions,
classification,
junction
representation.
1. INTRODUCTION
Navigation is one of the fundamental human activities. Thanks to the technological
developments in computer and electronic sciences, navigation industry and the navigation
technology developed rapidly and has become an integral part of everyday life. In addition
to the technical components, maps are considered as the basic component of navigation
systems. Moreover, database models and representation approaches, which form the basic
infrastructure of these maps, are another fundamental component of these systems. These
maps should be considered in two different dimensions: the base maps which involve the
road network data on which network algorithms, such as finding shortest path and optimal
way, are executed, and the maps which are used as user interface (Dogru, 2004; Yuefeng et
al 2005; Dogru and Ulugtekin, 2006). In addition to this development, since the display
media of the navigation systems are small in size, it is impossible to present the base map
as the user interface during the navigation process because base map data is quite complex
with its network data of multi-lanes, one-way streets, cloverleaf junctions etc. Hence,
abstracted representations of road network and related data are used in this manner.
Detailed studies on deriving multiple representations from related databases (Multiple
Representation Databases, MRDB) were executed by Kilpelainen (1997) and Dunkars
(2004).
It is clear that deriving multiple representations of a base data requires generalization
works. Several studies have been succeeded on road network generalization. It is obvious
that, in addition to the studies executed by Mackaness and Beard (1993) and Thomson and
Richardson (1995), one of the most important developments in the road network
generalization was succeeded by Mackanness and Mackechnie (1999). This approach
proposes a method for the detection and simplification of road junctions using a
combination of spatial clustering and graph theory. In this concept, junctions are identified
via cluster analysis and generalized based on graph theoretical methods. As a result, each
junction formed by set of connections is represented by a single vertex representing the
centroid of determined junction clusters. This approach was basically developed to derive
topographic maps for National Mapping Agencies. However, some application-based
requirements should also be taken into account while developing generalization
algorithms.
This study introduces a method for representing more detailed road networks, in particular
road junctions, although they are generalized. The main aim of the study is to derive road
maps, which can be used as different levels of user interfaces for navigational purposes, by
clustering the junctions and determining a common representation for each cluster. In this
concept, road junctions and related works are presented in the second part of the paper.
Additionally, final remarks and benefits of the proposed methods for the use navigation
systems or as a supplementary part of current generalization algorithms are presented in
the third part of the study.
2. ROAD JUNCTIONS
The identification of what is a “junction” is essentially a scale dependent issue. For
example, a collection of roads comprising a town, when viewed at 1: 250,000 could be
viewed as a complex junction required to be reduced to a single point (Mackanness and
Mackechnie, 1999). When the term junction is considered in terms of navigation system it
can also refer to the intersection of two lines or a set of dense intersecting lines (junctions),
which is also called as complex junctions (Zhang, 2004). Mostly, road network
generalization approaches collapse these complex junctions according to the target scale.
However, they should be represented in more detail for navigation systems. Therefore,
more useful organization can be imposed on the maze of junctions. In this study, an
efficient classification of junctions is considered as the first step for better organizations of
junctions.
2.1. Current Classification Approaches
Design and the classification of the junctions differ according to the road engineering
principles and fundamental usage aim of the junctions. In this concept, junctions can be
classified in terms of accident assessment, capacity of the road, turning rights, etc. For
example, junctions are classified as major/minor, signals and roundabouts in terms of
accident assessment (Watkiss et al, 2000). Additionally, each of these classes is subdivided
into standard, small and mini and totally 96 different junction types are obtained in this
manner (DMRB, 1997). As a second example, Geographic Data Files standard, European
Union standard for storing and using road data for navigational purposes, defines two main
types of road junctions called intersection and junction. These junctions combine 9
different road types to each other. According to this classification intersections are divided
in three groups as freeway intersection, roundabout and crossing (GDF3.0, 1995).
However, most common and basic classification of junctions is based on their grade
separation. If a road junction has a grade separation, it is called an intersection. In this type
of junctions, roads cross each other directly. It is called as interchanges if roads pass above
or below one another, preventing a single point of conflict by utilizing grade separation
and slip roads. The terms “motorway junction” and “highway junction” typically refer to
this layout (Dogru, 2004).
2.2. New Approaches
In this study, totally 29 different kinds of junctions (intersections and interchanges) that
can be encountered as a real world road structure were determined to use in practice and
we tried to find a clustering method for them. Potential behaviors of the car driver passing
a junction are considered to obtain the optimum clustering for navigation purposes.
Although there are several classifications including these junction types, new approaches
are tried to be applied to obtain better results of classification. In this concept, these
junctions are classified by using algorithmic methods. First of all junctions were
represented as matrices by considering their complexity and turning rights in all directions.
As it is seen from Figure 1, first row and column of the matrix represent the directions of
the crossing roads as left, right, up, and down (L, R, U, and D). Complexity matrix is
structured by using the topological relations of the line segments while the turning rights
matrix is structured according to the turning rights defined on the junctions of road
network. If it is possible to turn from a direction to another it was represented as 1
otherwise 0. As it is represented in Figure 1, it is not possible to turn in all directions while
driving through the sample junction.. After determining the matrix representations of all
junctions, they were classified according to the similarities of these matrixes. As a result of
this classification, 29 junction structures, including both interchanges and intersections
classified in 7 and 9 sub classes in terms of complexity and turning rights respectively.
Figure 1. Matrix representation of parclo junction
As a result of the discussions on these classification results, it is determined that matrix
representations individually are not sufficient to completely represent the characteristics of
road junctions in terms of navigation purposes. Hence a new methodology was proposed to
succeed the classification task.
According to stated methodology; junctions were represented as tree structures by
considering all potential maneuvers of a driver passing the junction. In this concept, a
junction (J) is considered with its all directions of connecting roads (Left, Right, Up, and
Down) and all maneuvers of a driver, entering to the junction from any direction, at a
connection point of the junction are defined as representing the route selection with a digit
starting from 1; 1 for the first path on the right side (see Figure 2). The junction is
represented as a whole by merging these tree structures into one according to merging
algorithms (Kim 2000; Chen et al 2003).
Figure 2. Tree representation for all routes of a junction
3.
CONCLUDING REMARKS AND FUTURE WORKS
The proposed methodology is a part of ongoing study being executed to implement an
automated junction recognition and representation system in terms of MRDB. This
approach will be used to determine the levels of detail for road networks in representation
levels emphasized by Dogru and Ulugtekin (2006). In this concept clustering and
representing process will be repeated to determine details of the higher levels of
representations. Such an implementation has several benefits in terms of increasing the
quality of road maps used for different purposes. Hence it is obvious that this approach will
also add value to the current navigation systems by enabling them to produce road maps by
considering the behaviors of the driver. Integration of proposed approach, as an automated
system to MRDB will be useful for both navigation systems and automated map
production studies by enabling system to produce different representations for different
applications automatically.
It is obvious that some modifications can also be applied to this tree representation by
using reduction algorithms.
In the following stages of the clustering methodology,
obtained tree structures are planned to be clustered based on the shock graph approach
implemented by Sebastian et al. (2001), Torsello et al. (2003), and Torsello and Hancock
(2004). Finally, graph similarity algorithms will be used to determine a common
representation for each cluster (Blondel et al 2004).
Implementation of proposed approach in an automated way is the future work of the
authors. Moreover, studies on integration of the implemented system with MRDB studies
are considered as another future work. In addition to these works some additional studies
should be taken into account to integrate the real-time map production capability to current
navigation systems.
4. REFERENCES
Blondel, V. D., Gajardo, A., Heymans, M., Senellart, P., and Dooren, P. V. 2004. A Measure
of Similarity between Graph Vertices: Applications to Synonym Extraction and Web
Searching.
SIAM
Rev.
46,
4
(Apr.
2004),
647-666.
DOI=
http://dx.doi.org/10.1137/S0036144502415960
Chen Z., Ding J., Zhang M., Tavanapong W., and Johnny S W. J.S., 2003. Hierarchical
Clustering-Merging for Multidimensional Index Structures, E. M. Bakker et al. (Eds.): CIVR
2003, LNCS 2728, pp. 81–90, 2003. Springer-Verlag Berlin Heidelberg.
DMRB, 1997. Economic Assessment of Road Schemes: COBA Manual, Design Manual for
Roads and Bridges, Volume 13, The Highways Agency.
Dogru, A.O., 2004. Multiple Representations For Junction Modeling In Car Navigation Map
Design, MSc. Thesis, Istanbul Technical University, Institute of Science and Technology,
Turkey (in Turkish).
Dogru, A.O., Ulugtekin, N., 2006. Car Navigation Map Design in terms of Multiple
Representations, First International Conference on Cartography and GIS, 25-28 January,
Borovets, Bulgaria (CD-rom).
Dunkars, M., 2004. Multiple Representation Databeses for Topographic Information, PhD
Thesis, KTH Royal Institute of Technology, Sweden.
GDF 3.0, 1995. Geographic Data Files, European Committee for Standardisation, European
Standard,
CEN
TC
278.
http://www.ertico.com/en/links/links/gdf_-
_geographic_data_files.htm (accessed, 02.03.2007)
Kilpelainen, T., 1997. Multiple Representation and Generalization of Geo-Databases for
Topographic Maps, PhD Thesis, Finnish Geodetic Institute, Finland.
Kim Y., 2000. Simple and Fault-Tolerant Key Agreement for Dynamic Collaborative Groups,
CCS '00, Athens, Greece. 2000 1-58113-203-4/00/0011.
Mackaness, W.A., and Beard, M.K., 1993. Use of Graph Theory to Support Map Generalisation. Cartography and Geographic Information Systems 20: 210 – 221.
Mackaness W., Mackechnie G., 1999., Automating the Detection and Simplification of
Junctions in Road Networks, GeoInformatica 3:2, 185-200.
Sebastian T.B., Klein P.N., and Kimia B.B., 2001. Recognition of Shapes by Editing Shock
Graphs, ICCV 2001, pages 755-762.
Thomson, R.C., and Richardson, D.E., 1995. A Graph Theory Approach to Road Network
Generalisation. In: Proceeding of the 17th International Cartographic Conference, pp 1871–
1880.
Torsello A., Hancock E.R., Petkov N., and Westenberg M.A., 2003. Graph Clustering with
Tree-Unions, Lecture notes in computer science [0302-9743] N. Petkov and M.A.Westenberg
(Eds.): CAIP 2003, LNCS 2756, pp. 451–459, 2003.
Torsello A. and Hancock E.R., 2004. Learning Mixtures of Weighted Tree-Unions by
Minimizing Description Length, Lecture notes in computer science [0302-9743], T. Pajdla
and J. Matas (Eds.): ECCV 2004, LNCS 3023, pp. 13–25, 2004.
Watkiss, P., Brand, C., Hurley, F., Pilkington, A., Mindell J., Joffe, M., and Anderson, R.,
2000. Informing transport health impact assessment in London, Project Report, London.
www.nice.org.uk/media/hiadocs/informing_transport_hia_in_london.pdf (accessed
04.03.2007)
Yuefeng, L., Yiqin, X., Zhiming, G., Jianghua, Z., and Shi, Q., 2005. Feature-based two level
structure road network model for navigation, Geoscience and Remote Sensing Symposium,
IGARSS '05. Proceedings vol. 2 pp. 4.
Zhang, Q., 2004. "Road Network Generalization Based on Connection Analysis". 11th
International Symposium on Special Data Handling.V 343-53.