Utilisation of computational intelligence for simplification of - - PowerPoint PPT Presentation

▶
utilisation of computational intelligence for
SMART_READER_LITE
LIVE PREVIEW

Utilisation of computational intelligence for simplification of - - PowerPoint PPT Presentation

Mar 23, 2024 101 likes •278 views

Utilisation of computational intelligence for simplification of linear objects using extended WEA algorithm 1 Anna Fiedukowicz Agata Pillich- Kolipiska Robert Olszewski 1 This work has been supported by the European Union in the framework of

slide-1
SLIDE 1

1 Anna Fiedukowicz

Agata Pillich-Kolipińska Robert Olszewski

Utilisation of computational intelligence for simplification of linear objects using extended WEA algorithm

16th ICA Generalisation Workshop, Dresden, 23-24.08.2013

1This work has been supported by the European Union

in the framework of European Social Fund through the Warsaw University of Technology Development Programme.

slide-2
SLIDE 2

 Widely known and described  Selection of „important”

vertices

 Different algorithms, mostly

  • n
  • ne-parameter ones
  • Douglas-Peucker (1973) -

distance

  • Visvalingam-Whyatt (1993) –

„effective area”

  • Wang (1996) – curve radius
  • many others…

2

slide-3
SLIDE 3

 „Shape-aware” algorithm  S. Zhou, C.B. Jones (2004) – 4 parameters:

  • „effective area” from Visvalingam-Whyatt algorithm
  • flatness (H/W)
  • skewness (H/ML)
  • convexity (↻ – convex, ↺ – concave)

 Defined filters:

  • Wflat
  • Wskew
  • Wconvex

WEA = Wflat * Wskew * Wconvex * EA

3

slide-4
SLIDE 4

Non-deterministic approach: knowledge base Explicit

set of rules defined by an expert

Implicit

derived from examples provided by an expert

computational intelligence Fuzzy Inference Systems Artificial Neural Networks Modified „WEA” (weight of a vertex)

Vertex weight = f(flatness, skewness, convexity, EA)

4

slide-5
SLIDE 5

Vertex weight = f(flatness, skewness, convexity, EA)

rules examples Defined by an expert, eg.

 if (EA > 100) and (flatness

> 1) and (skewness > 0.8 ) then weight = 99; Or more fuzzy definition:

 if EA is big and flatness is

high and skewness is low then vertex is important Important points derived from Topographic DBs and maps 1:10k, 1:50k, 1:250k

 Polish coastline (SABE 30

and surveying data)

 Polish rivers

Vertices weights proposed by an expert

5

slide-6
SLIDE 6

 STATISTICA Neural Networks software TEACHING VALIDATION INPUT OUTPUT

6

slide-7
SLIDE 7

 Three types of ANNs tested Different teaching algorithms tested (backward

errors propagation, quasi-Newton, Delta-Bar-Delta, Levenberg-Marquardt etc.)

7

slide-8
SLIDE 8

 Testing different structures of same ANN type  More neurons ⇒ higher precision ⇒ output

less generalized ⇒ learning more time- consuming ⇒ longer data processing

 Test data – several neurons  Actual data – several thousands neurons?

8

slide-9
SLIDE 9

9

slide-10
SLIDE 10
  • IF building<400m2 THEN y=0,05
  • IF building is small THEN simplify a bit

How you define „small”?

A B

10

slide-11
SLIDE 11

1

SMALL

BIG

0.7 0.3

11

slide-12
SLIDE 12

A B

  • IF x = A THEN y = B
  • IF x = A THEN y = B
  • And what IF x = A’ ? THEN y = B’

12

slide-13
SLIDE 13

IF area =„big” and flatness=„big” then significance =„big”

14

slide-14
SLIDE 14

15

slide-15
SLIDE 15

NEURO FUZZY

 Easy to build  Black box for user  Long computation time

for complex tasks

 Difficult to define  Allows process

understanding

 Builded once

works fast

16

slide-16
SLIDE 16

Up to now Future plans

 Tools independent

  • n GIS

 Preliminary tests

  • n polish coast line

 Combine fuzzy and

neuro tools with GIS sofware

 Check tools on

different type of data

17

slide-17
SLIDE 17

r.olszewski@gik.pw.edu.pl a.pillich@gik.pw.edu.pl a.fiedukowicz@gik.pw.edu.pl

Here it comes, here comes the weekend…