A Knowledge Based Genetic Algorithm Approach to Automating Cartographic Generalization

1
A Knowledge Based Genetic Algorithm Approach to
Automating Cartographic Generalization

• Mark Ware, Ian Wilson and Andrew Ware
• School of Computing
• University of Glamorgan

2
Problem Definition

• Automated production of reduced scale products
  from large scale digital map source.
• Automated map generalization.
• Deriving new products from OS Mastermap.

3
(No Transcript)
4
Generalization

• Manual
• Specification
• Experience
• Operators
• delete
• simplify
• displace
• exaggerate
• aggregate

5
(No Transcript)
6
Research to Date

• Individual operators (interactive)
• e.g. line simplification, polygon aggregation
• Features in isolation
• Multiple features, multiple operators ?
• Rule based approach
• Energy minimization
• Multi-Agent

7
Specific Problem

• Graphic conflict caused by road symbolization
• Road symbol thickened
• Roads interfere with buildings

8
(No Transcript)
9
(No Transcript)
10
Graphic Conflict Resolution byBuilding Feature
Displacement

• Road features symbolized
• Assume position of road features remains fixed
• Assume building features can be displaced from
  origin up to some maximum distance

11
(No Transcript)
12
Research

• Agent-based solutions
• Least squares
• Finite elements
• Snakes
• Elastic beams
• Slow, difficult to set parameters, unpredictable,
  just plain hard

13
A Trial Position Approach

• Trial position technique borrowed from the
  related problem of map labeling
• Widely used as a method of labeling point
  features

14
(No Transcript)
15
Apply PFLP trial position techniques to Polygon
Displacement

• - we have map, n objects, containing conflict

- too many to generate and test all (e.g. k8, n
10, gt 1 billion configurations) -need some
strategy for limiting number tested
16
Metaheuristic Procedures

• Simulated Annealing
• Tabu Search
• Genetic Algorithms

17
Genetic Algorithm Approach

• A particular representation is represented by a
  genome (which describes the location of each
  building feature relative to its local origin)
• Starting with initial population (a number of
  genomes, representing a range of realizations),
  solutions are evolved
• Evolution controlled by GA
• initialization
• crossover
• mutation

18
Genome Records x and y offset for each object
19
Initialise pm (randomly) generated realizations
(but including original)
1
2
3




pm
20
Crossover single (random) point
1 0 0 1 1 0 1 0 1 0 1 0 0 1 1 0 1 0 1 0 1 0 1 0 1
0 0 1 0 0 0 1
0 0 1 1 1 0 0 0 1 0 1 0 1 1 1 1 1 0 1 1 1 0 1 0 1
0 1 0 0 1 0 0
21
Mutation (random), based on mutation probability
22
Realization Evaluation Constraints

• Minimum separation between building and road
• Minimum separation between buildings
• Minimum amount of map modification

23
Cost Function

• C1 x Number of building/road minimum separation
  violations
• C2 x Number of building/building minimum
  separation violations
• C3 x Total displacement

24
Genetic Algorithm (steady state)

• initialize population (pm randomly generated
  realizations)
• while !stop
• select individuals for breeding
• breed offspring
• mutate offspring
• update population
• check stopping conditions
• endwhile

25
Experiments

• IGN BDTopo Data
• OS MasterMap

26
IGN BDTopo Data
27
(No Transcript)
28
(No Transcript)
29
(No Transcript)
30
OS Mastermap
31
(No Transcript)
32
(No Transcript)
33
(No Transcript)
34
Additional Operator

• Shrink

b1
. . . .
0 0 1 1 0 0 1 0 1 1 1 0 0 1 0 0 1 0 0 0 1 1 1 0 1

0 0 1 1 0 0 1 0 1
x-offset
y-offset
s-factor
35
(No Transcript)
36
(No Transcript)
37
(No Transcript)
38
(No Transcript)
39
(No Transcript)
40
(No Transcript)
41
Conclusions

• GA trial positions is successful in resolving
  conflict in reasonable time
• Displacement not always sufficient
• Introducing additional shrink operator
  straightforward and improves result

42
Future Work

• Much more experimentation with GA
• More operators
• Additional feature classes
• Compare with SA and TS
• Incorporate into more complete production system
  (in which GA solution becomes just one of a
  number of functions available)