Automated Cell Based Generalization of Virtual 3D City ......Generalization Lenses Navigation...

20
Automated Cell Based Generalization of Virtual 3D City Models with Dynamic Landmark Highlighting Tassilo Glander, Jürgen Döllner Hasso-Plattner-Institut Dep. Computer Graphics Systems Prof. Dr. Jürgen Döllner University of Potsdam www.hpi.uni-potsdam.de/3d www.3dgi.de

Transcript of Automated Cell Based Generalization of Virtual 3D City ......Generalization Lenses Navigation...

Page 1: Automated Cell Based Generalization of Virtual 3D City ......Generalization Lenses Navigation scenarios LOD4 LOD3 LOD2 LOD1 LOD0 LOA0 LOA1 LOA2 LOAn CityGML 21.06.2008 ICA Workshop

Automated Cell Based Generalization of Virtual 3D City Models

with Dynamic Landmark Highlighting

Tassilo Glander, Jürgen Döllner

Hasso-Plattner-InstitutDep. Computer Graphics Systems

Prof. Dr. Jürgen DöllnerUniversity of Potsdam

www.hpi.uni-potsdam.de/3dwww.3dgi.de

Page 2: Automated Cell Based Generalization of Virtual 3D City ......Generalization Lenses Navigation scenarios LOD4 LOD3 LOD2 LOD1 LOD0 LOA0 LOA1 LOA2 LOAn CityGML 21.06.2008 ICA Workshop

21.06.2008 ICA Workshop on Generalization, Tassilo Glander, University of Potsdam 2

Outline

1. Introduction / Generalization in 3D2. Related Work

3. Preprocessing4. Dynamic highlighting5. Demo

6. Conclusion & Outlook

Page 3: Automated Cell Based Generalization of Virtual 3D City ......Generalization Lenses Navigation scenarios LOD4 LOD3 LOD2 LOD1 LOD0 LOA0 LOA1 LOA2 LOAn CityGML 21.06.2008 ICA Workshop

21.06.2008 ICA Workshop on Generalization, Tassilo Glander, University of Potsdam 3

1 Introduction

Current city models reach critical size (100000+ buildings)■ Need for reduction of unnecessary details■ Usage beyond pretty photorealistic visualization?■ Adaptation of cartographic generalization principles

3D generalization has similar problems■ Present appropriate information density on limited space■ Large datasets underlying continuous updates automatic derivation

needed

…and specific problems■ Occlusion due to perspective■ Dynamic (real time) updates of scale■ Continuous scale in one image

Page 4: Automated Cell Based Generalization of Virtual 3D City ......Generalization Lenses Navigation scenarios LOD4 LOD3 LOD2 LOD1 LOD0 LOA0 LOA1 LOA2 LOAn CityGML 21.06.2008 ICA Workshop

21.06.2008 ICA Workshop on Generalization, Tassilo Glander, University of Potsdam 4

2 Related Work

2D – many generalization models / frameworks■ Iterative, step-by-step based (e.g. Agents)■ Global (Least Squares Adjustment, Simulated Annealing, Spring-

based)Matured frameworks for productive use

3D – first steps with single generalization operators■ Single building simplification [Kada2005, Thiemann&Sester2004, Forberg2002, Rau et

al. 2006]

■ Building aggregation & simplification [Sester2004, Anders2005]

Our previous work■ Cell-based generalization [Glander&Döllner 2007]

■ Landmark highlighting [Glander et al. 2007]

Page 5: Automated Cell Based Generalization of Virtual 3D City ......Generalization Lenses Navigation scenarios LOD4 LOD3 LOD2 LOD1 LOD0 LOA0 LOA1 LOA2 LOAn CityGML 21.06.2008 ICA Workshop

21.06.2008 ICA Workshop on Generalization, Tassilo Glander, University of Potsdam 5

3 Preprocessing

Cell-based generalization■ Calculate arrangement (roads cells)■ Map buildings to cells (point-in-polygon tests)■ Per cell: calculate mean height & variance■ Identify initial set of landmarks

(CAD models + outliers)

For smaller scales, remove less important streets and repeat.

Cell-based generalization■ Calculate arrangement (roads cells)■ Map buildings to cells (point-in-polygon tests)■ Per cell: calculate mean height & variance■ Identify initial set of landmarks

(CAD models + outliers)

For smaller scales, remove less important streets and repeat.

Page 6: Automated Cell Based Generalization of Virtual 3D City ......Generalization Lenses Navigation scenarios LOD4 LOD3 LOD2 LOD1 LOD0 LOA0 LOA1 LOA2 LOAn CityGML 21.06.2008 ICA Workshop

21.06.2008 ICA Workshop on Generalization, Tassilo Glander, University of Potsdam 6

3 Preprocessing

Create landmark hierarchy [Winter et al. 2008]

■ Use centroids of initial landmark buildings for a Delaunay triangulation■ Choose set of landmarks for next layer:

□ For each node i, vote for one node with the highest saliencyvalue within neighborhood (including i)

□ In absence of a better measure for saliency, we use the height

■ Repeat, until just one landmark is left

Integration with generalized blocks■ Alignment of landmark hierarchy with levels of abstraction (LOA)

average reduction to 1/3 in each subsequent LOA■ Cut out landmarks from block cells, place in the scene

i

Page 7: Automated Cell Based Generalization of Virtual 3D City ......Generalization Lenses Navigation scenarios LOD4 LOD3 LOD2 LOD1 LOD0 LOA0 LOA1 LOA2 LOAn CityGML 21.06.2008 ICA Workshop

21.06.2008 ICA Workshop on Generalization, Tassilo Glander, University of Potsdam 7

3 Preprocessing

Creating several levels of abstraction■ Exploiting different road weights

Page 8: Automated Cell Based Generalization of Virtual 3D City ......Generalization Lenses Navigation scenarios LOD4 LOD3 LOD2 LOD1 LOD0 LOA0 LOA1 LOA2 LOAn CityGML 21.06.2008 ICA Workshop

21.06.2008 ICA Workshop on Generalization, Tassilo Glander, University of Potsdam 8

3 Preprocessing

Creating several levels of abstraction■ Exploiting different road weights

Page 9: Automated Cell Based Generalization of Virtual 3D City ......Generalization Lenses Navigation scenarios LOD4 LOD3 LOD2 LOD1 LOD0 LOA0 LOA1 LOA2 LOAn CityGML 21.06.2008 ICA Workshop

21.06.2008 ICA Workshop on Generalization, Tassilo Glander, University of Potsdam 9

3 Preprocessing

Creating several levels of abstraction■ Exploiting different road weights

Page 10: Automated Cell Based Generalization of Virtual 3D City ......Generalization Lenses Navigation scenarios LOD4 LOD3 LOD2 LOD1 LOD0 LOA0 LOA1 LOA2 LOAn CityGML 21.06.2008 ICA Workshop

21.06.2008 ICA Workshop on Generalization, Tassilo Glander, University of Potsdam 10

3 Preprocessing

Creating several levels of abstraction■ Exploiting different road weights

Page 11: Automated Cell Based Generalization of Virtual 3D City ......Generalization Lenses Navigation scenarios LOD4 LOD3 LOD2 LOD1 LOD0 LOA0 LOA1 LOA2 LOAn CityGML 21.06.2008 ICA Workshop

21.06.2008 ICA Workshop on Generalization, Tassilo Glander, University of Potsdam 11

3 Preprocessing

Creating several levels of abstraction■ Exploiting different road weights

Page 12: Automated Cell Based Generalization of Virtual 3D City ......Generalization Lenses Navigation scenarios LOD4 LOD3 LOD2 LOD1 LOD0 LOA0 LOA1 LOA2 LOAn CityGML 21.06.2008 ICA Workshop

21.06.2008 ICA Workshop on Generalization, Tassilo Glander, University of Potsdam 12

3 Preprocessing

Creating several levels of abstraction■ Exploiting different road weights

Page 13: Automated Cell Based Generalization of Virtual 3D City ......Generalization Lenses Navigation scenarios LOD4 LOD3 LOD2 LOD1 LOD0 LOA0 LOA1 LOA2 LOAn CityGML 21.06.2008 ICA Workshop

21.06.2008 ICA Workshop on Generalization, Tassilo Glander, University of Potsdam 13

3 Preprocessing

Creating several levels of abstraction■ Exploiting different road weights

Page 14: Automated Cell Based Generalization of Virtual 3D City ......Generalization Lenses Navigation scenarios LOD4 LOD3 LOD2 LOD1 LOD0 LOA0 LOA1 LOA2 LOAn CityGML 21.06.2008 ICA Workshop

21.06.2008 ICA Workshop on Generalization, Tassilo Glander, University of Potsdam 14

3 Preprocessing

Creating several levels of abstraction■ Exploiting different road weights

Page 15: Automated Cell Based Generalization of Virtual 3D City ......Generalization Lenses Navigation scenarios LOD4 LOD3 LOD2 LOD1 LOD0 LOA0 LOA1 LOA2 LOAn CityGML 21.06.2008 ICA Workshop

21.06.2008 ICA Workshop on Generalization, Tassilo Glander, University of Potsdam 15

3 Preprocessing

Creating several levels of abstraction■ Exploiting different road weights

Page 16: Automated Cell Based Generalization of Virtual 3D City ......Generalization Lenses Navigation scenarios LOD4 LOD3 LOD2 LOD1 LOD0 LOA0 LOA1 LOA2 LOAn CityGML 21.06.2008 ICA Workshop

21.06.2008 ICA Workshop on Generalization, Tassilo Glander, University of Potsdam 16

4 Dynamic Highlighting

Dynamic Highlighting■ Emphasize most important landmarks by enlarging them

peak of the landmark hierarchy (e.g., restricted to Top 10)■ Calculate a scaling factor

□ depending on the camera distance□ quadratic scaling function parameterized with a distance interval

[dstart,dend]

■ Use 2x maximum distance to neighbor in hierarchy for dend

□ dstart is constant for all landmarks (e.g. dstart = 2500m)

distance

Camera

startd endd

Page 17: Automated Cell Based Generalization of Virtual 3D City ......Generalization Lenses Navigation scenarios LOD4 LOD3 LOD2 LOD1 LOD0 LOA0 LOA1 LOA2 LOAn CityGML 21.06.2008 ICA Workshop

21.06.2008 ICA Workshop on Generalization, Tassilo Glander, University of Potsdam 17

4 Dynamic Highlighting

Demo

Page 18: Automated Cell Based Generalization of Virtual 3D City ......Generalization Lenses Navigation scenarios LOD4 LOD3 LOD2 LOD1 LOD0 LOA0 LOA1 LOA2 LOAn CityGML 21.06.2008 ICA Workshop

21.06.2008 ICA Workshop on Generalization, Tassilo Glander, University of Potsdam 18

Conclusion / Outlook

LOD vs LOA■ Generalization in city models is only known in

terms of level of detail so far■ Building aggregation is necessary element

Applications■ Dynamic and continuous scale■ Generalization Lenses■ Navigation scenarios

LOD4LOD3LOD2LOD1

LOD0

LOA0

LOA1LOA2

LOAn

CityGML

Page 19: Automated Cell Based Generalization of Virtual 3D City ......Generalization Lenses Navigation scenarios LOD4 LOD3 LOD2 LOD1 LOD0 LOA0 LOA1 LOA2 LOAn CityGML 21.06.2008 ICA Workshop

21.06.2008 ICA Workshop on Generalization, Tassilo Glander, University of Potsdam 19

Conclusion / Outlook

Further work■ Many detail improvements (better aggregation, …, CityGML export)■ Build up on existing platforms & extend towards 3D visualization?

What is gained?■ Potential visualization of 3D generalization■ Usable tourist map

Page 20: Automated Cell Based Generalization of Virtual 3D City ......Generalization Lenses Navigation scenarios LOD4 LOD3 LOD2 LOD1 LOD0 LOA0 LOA1 LOA2 LOAn CityGML 21.06.2008 ICA Workshop

21.06.2008 ICA Workshop on Generalization, Tassilo Glander, University of Potsdam 20

Contact

Tassilo [email protected]

Department for Computergraphic SystemsProf. Dr. Jürgen Döllnerwww.hpi.uni-potsdam.de/3d

Research Group 3D-Geoinformationwww.3dgi.de