3. The Centre for Geospatial Science (CGS)
Th C t f G ti l S i
Established November 2005 as a cross-
faculty post-graduate research centre.
Research focus:
– spatial data infrastructures (SDI),
– geospatial intelligence,
– geospatial interoperability
– location-based services.
location based
4. Creating Intelligent Applications
Goal: Make it easier for geospatial researchers to
incorporate proven geospatial techniques into their
workflow.
The idea is to create generic frameworks that are
customized with the problem-specific details.
problem specific
Initial efforts have concentrated on an object-level
object level
optimization framework using state-space search.
5. Bigger picture
Maturity of open
Geospatial Standards source software (for
(for ex. OGC spec.) ex. OSGeo stack)
6. OS Geo Product development statistics 2008
http://wiki.osgeo.org/wiki/Project_Stats
7. CGS Optimization Framework
• Currently implemented:
– Hillclimbing, Simulated Annealing
– (Reactive) Tabu Search
– Simple Genetic Algorithm
• Implementation targets JVM and hence easily
integrated with Geotools52North WPS etc
ith etc.
8. Why – G
h Generalization ?
li i
The process of simplifying the form or shape of map
features, usually carried out when the map is changed from
a large scale to a small scale, is referred to as
generalisation.
Map g
p generalisation is a pprocess of extracting the important
g p
and relevant spatial information from reality.
9. Map Generalization operators
M G li ti t
Simplification
Amalgamation
l
Elimination
Typification
Exaggeration
Displacement
10. Simplification
Si lifi ti
Douglas-Peucker algorithm (1973)
11. Amalgamation
DeLucia and Black (1987) - triangulation-based area amalgamation procedures. These ideas are
taken up and advanced in Jones et al (1995)
Su et al (1997) - A raster-based aggregation method (forms the basis of ESRI's AreaAggregate
function)
Ai and van Oosterom (2002) - displacement vectors p o to a a ga at o
a d a Ooste o ( 00 ) d sp ace e t ecto s prior amalgamation
13. Typification
f
Feature clusturing - (Mackaness 1994, Ormsby and Mackaness 1999, Mackaness
and Mackechnie 1999)
Sester (2003) and Moulin (2003) -Kohonen Self Organizing Maps
Regnauld (1996) - Minimum Spanning Trees
14. Exaggeration
gg
Mackaness (1995) - alpha analysis for classifying urban road
ac a ess ( 995) a p a a a ys s o c ass y g u ba oad
networks hierarchically, providing a means for removing roads at
smaller scale while still conveying essential characteristics of the
network
15. Displacement
Lonergan and Jones (2001) - map quality is measured in terms of minimum distance violations, and polygon displacement
achieved by calculating displacement vectors in an iterative fashion
Li et al (2002) - polygon displacement using a two-level agent-based architecture.
17. Schematic Map - Characteristics
•Topologically consistent.
T l i ll i t t
•Simplified lines (Douglas-Peucker).
•May be desirable to re-orient lines so that they
are horizontal, vertical or diagonal.
•Scale in congested areas expanded at the
expense of scale in areas that are less so
so.
18. Graphic manipulations for producing
a schematic map
Lines are simplified and re-oriented to conform to a regular grid. Congested areas are
increased in scale at the expense of scale in areas of lesser node density
20. Topological
Original network and derived schematic map should be
topologically consistent
Topological – original (Left), topological error (Middle) and
acceptable solution (Right)
21. Orientation
If possible, network edges should lie in horizontal, vertical
or diagonal direction
Orientation – original (L) and schematized (R)
22. Angle
If possible, the angle between a pair of connected
edges should be greater than some minimum
angle
Angle – edges re-oriented but Angle constraint violated (L)
and acceptable solution
23. Rotation
An edge’s orientation should remain as close to its starting
orientation as possible
:Rotation – original (L), acceptable solution (M) and better solution (R)
24. Clearance
If possible, the distance between disjoint features should be
g
greater than some minimum distance
Clearance – constraint violated (L) and resolved (R)
25. Displacement
Vertices should remain as close to their starting positions as possible
possible.
Displacement – original (L), acceptable solution (M) and better solution (R)
27. Core Process
•Evaluate
– For each vertex:
Count topological errors
Measure constraint violations
Heuristic l i th
H i ti value is the sum of the above
f th b
Modify
Displace vertices
30. Conclusions
•Implements a usecase for automatic
production of schematic maps
•Proof-of-concept implemented for WFS, using
schematization as transformation exemplar.
p