Detection of constrictions on closed polyhedral surfaces Franck H - - PowerPoint PPT Presentation

detection of constrictions on closed polyhedral surfaces
SMART_READER_LITE
LIVE PREVIEW

Detection of constrictions on closed polyhedral surfaces Franck H - - PowerPoint PPT Presentation

Detection of constrictions on closed polyhedral surfaces Franck H etroy , Dominique Attali Laboratoire des Images et Signaux, I.N.P. Grenoble, France VisSym2003 Detection of constrictions on closed polyhedral surfaces p.1


slide-1
SLIDE 1

Detection of constrictions

  • n closed polyhedral surfaces

Franck H´ etroy

  • , Dominique Attali

Laboratoire des Images et Signaux, I.N.P. Grenoble, France

VisSym2003 Detection of constrictions on closed polyhedral surfaces – p.1

slide-2
SLIDE 2

Context

Decompose 3D objects in several components connected by narrower parts detect these narrower parts

VisSym2003 Detection of constrictions on closed polyhedral surfaces – p.2

slide-3
SLIDE 3

Objects

Boundary = closed triangulated surface

✂ ✄
  • manifold without boundary

VisSym2003 Detection of constrictions on closed polyhedral surfaces – p.3

slide-4
SLIDE 4

Other works

Other works to decompose polyhedral surfaces:

Decomposition into meaningful patchs: Tal et al. 1995 to 2003, Gregory et al. 1998

Topological decomposition: Erickson and Har-Peled 2002, Colin de Verdière and Lazarus 2002

Dynamical systems approach: Dey et al. 2002

...

VisSym2003 Detection of constrictions on closed polyhedral surfaces – p.4

slide-5
SLIDE 5

Plan

  • 1. Definition and characterization of constrictions
  • 2. Algorithm
  • 3. Results and comments

VisSym2003 Detection of constrictions on closed polyhedral surfaces – p.5

slide-6
SLIDE 6

Definition

A constriction is a locally

  • length-minimizing

simple, closed curve on a constriction is a closed geodesic

  • curve
  • for the Hausdorff distance
  • follows a shortest path between any two sufficiently

close points

VisSym2003 Detection of constrictions on closed polyhedral surfaces – p.6

slide-7
SLIDE 7

What do constrictions look like

✆ ✆✞✝ ✝ ✟ ✟✞✠ ✠ ✡ ✡✞☛ ☛ ☞ ☞✞✌ ✌ ✍ ✍✞✎ ✎ ✏ ✏✞✑ ✑ ✒ ✒✞✓ ✓

pivot vertices = vertices through which a constriction goes

VisSym2003 Detection of constrictions on closed polyhedral surfaces – p.7

slide-8
SLIDE 8

What do constrictions look like

✔ ✔✞✕ ✕ ✖ ✖✞✗ ✗ ✘ ✘✞✙ ✙ ✚ ✚✞✛ ✛ ✜ ✜✞✢ ✢ ✣ ✣✞✤ ✤ ✥ ✥✞✦ ✦

between two successive pivots, a constriction intersects a sequence of faces

VisSym2003 Detection of constrictions on closed polyhedral surfaces – p.8

slide-9
SLIDE 9

What do constrictions look like

✧ ✧✞★ ★ ✩ ✩✞✪ ✪ ✫ ✫✞✬ ✬ ✭ ✭✞✮ ✮ ✯ ✯✞✰ ✰ ✱ ✱✞✲ ✲ ✳ ✳✞✴ ✴ ✵ ✵✞✶ ✶ ✷ ✷✞✸ ✸

the planar unfolding of a constriction in the planar unfolding of the sequence is a straight line segment

VisSym2003 Detection of constrictions on closed polyhedral surfaces – p.9

slide-10
SLIDE 10

What do constrictions look like

✹ ✹✞✺ ✺ ✻ ✻✞✼ ✼ ✽ ✽✞✾ ✾ ✿ ✿✞❀ ❀ ❁ ❁✞❂ ❂ ❃ ❃✞❄ ❄ ❅ ❅✞❆ ❆ ❇ ❇✞❈ ❈ ❉ ❉✞❊ ❊

the angle made by a constriction at a pivot vertex is

VisSym2003 Detection of constrictions on closed polyhedral surfaces – p.10

slide-11
SLIDE 11

What do constrictions look like

constriction simple, closed curve with angle

at each pivot and geodesic curve between any two successive pivots References about geodesics:

Sharir/Schorr 1986

Mitchell/Mount/Papadimitriou 1987

Chen/Han 1990

Mitchell 1998 (survey)

...

VisSym2003 Detection of constrictions on closed polyhedral surfaces – p.11

slide-12
SLIDE 12

Plan

  • 1. Definition and characterization of constrictions
  • 2. Algorithm
  • 3. Results and comments

VisSym2003 Detection of constrictions on closed polyhedral surfaces – p.12

slide-13
SLIDE 13

Algorithm

Idea to detect constrictions: surface will disconnect in their area when simplified

VisSym2003 Detection of constrictions on closed polyhedral surfaces – p.13

slide-14
SLIDE 14

Overview of algorithm

seed curve detection surface simplification curve construction

VisSym2003 Detection of constrictions on closed polyhedral surfaces – p.14

slide-15
SLIDE 15

Surface simplification

Edge collapse operator:

  • ■❍
  • ■❏
  • Several existing algorithms

Used algorithm: Garland and Heckbert’s (SIGGRAPH’97)

VisSym2003 Detection of constrictions on closed polyhedral surfaces – p.15

slide-16
SLIDE 16

Surface simplification

Stop when a seed curve

❑▼▲ ◆ ❖ ◆◗P ❖ P ▲ ❘

is found

the triangle

▲ ◆◗P

is not a face of

❙ ❚

c b a

VisSym2003 Detection of constrictions on closed polyhedral surfaces – p.16

slide-17
SLIDE 17

Seed curve

After one more simplification step, the surface is not a manifold anymore

❱ ❱ ❲ ❱ ❱ ✂

2 different edges between

and

P

VisSym2003 Detection of constrictions on closed polyhedral surfaces – p.17

slide-18
SLIDE 18

Overview of algorithm

seed curve detection surface simplification curve construction

VisSym2003 Detection of constrictions on closed polyhedral surfaces – p.18

slide-19
SLIDE 19

Curve construction

❳ ❨❬❩ ❭ ❪

Case 1: the curve is not modified

VisSym2003 Detection of constrictions on closed polyhedral surfaces – p.19

slide-20
SLIDE 20

Curve construction

❨❬❩ ❭ ❪ ❳ ❫ ❪ ❫❬❴

Case 2: the curve is modified between 2 pivot vertices

VisSym2003 Detection of constrictions on closed polyhedral surfaces – p.20

slide-21
SLIDE 21

Curve construction

Geodesic computation between the two pivot vertices

Pham-Trong et al.’s algorithm (Num. Algo. 2001)

needs an initial sequence of faces

❵ ❛❝❜ ❞ ❡ ❞ ❜ ❵ ❛ ❡ ❢ ❜ ❢ ❡ ❞ ❡ ❞ ❜ ❵ ❛ ❣ ❜ ❜ ❵ ❛❝❜ ❞ ❡ ❞ ❜ ❢ ❜ ❢ ❡ ❵ ❛ ❡ ❵ ❛ ❣ ❜ ❡ ❢ ❤ ❛ ❣ ❜

two geodesics: we keep the shortest one

VisSym2003 Detection of constrictions on closed polyhedral surfaces – p.21

slide-22
SLIDE 22

Curve construction

❨❬❩ ❭ ❪ ❳

Case 3: the curve must have at least one pivot vertex

VisSym2003 Detection of constrictions on closed polyhedral surfaces – p.22

slide-23
SLIDE 23

Curve construction

❩ ❭ ❪ ❪ ❳ ❪ ❩ ❪ ❳ ❴ ❳ ❪ ❩ ❴ ❨✐❩ ❳ ❴ ❩ ❭ ❪ ❴ ❩ ❴ ❩ ❪ ❨✐❩ ❨❥❩ ❭ ❪ ❳ ❳ ❪

and/or

❳ ❴

will be pivot vertices of the new curve

VisSym2003 Detection of constrictions on closed polyhedral surfaces – p.23

slide-24
SLIDE 24

Curve constructions

Conclusion about this stage:

  • each
❨❥❩

is a closed piecewise geodesic

  • the curve is only partially modified at each step

(except one special case)

  • we are not sure the final curve
❨❬❦
  • n

is a constriction: constriction = straight between two successive pivot vertices + angle

at each pivot vertex first condition OK, but what about the second one ??

VisSym2003 Detection of constrictions on closed polyhedral surfaces – p.24

slide-25
SLIDE 25

Plan

  • 1. Definition and characterization of constrictions
  • 2. Algorithm
  • 3. Results and comments

VisSym2003 Detection of constrictions on closed polyhedral surfaces – p.25

slide-26
SLIDE 26

Results on synthetic surfaces

VisSym2003 Detection of constrictions on closed polyhedral surfaces – p.26

slide-27
SLIDE 27

Results on synthetic surfaces

VisSym2003 Detection of constrictions on closed polyhedral surfaces – p.27

slide-28
SLIDE 28

Remarks

Detection of several constrictions:

  • forbid the collapse of the 3 edges of seed curves and

continue surface simplification

  • construct a sequence of closed piecewise geodesic

for each seed curve Constrictions can cut the object or its complementary

VisSym2003 Detection of constrictions on closed polyhedral surfaces – p.28

slide-29
SLIDE 29

Results on scanned data

VisSym2003 Detection of constrictions on closed polyhedral surfaces – p.29

slide-30
SLIDE 30

Results on scanned data

VisSym2003 Detection of constrictions on closed polyhedral surfaces – p.30

slide-31
SLIDE 31

Conclusion

Decomposition of a 3D object in separate components:

  • surface approach (boundary of the object)
  • progressive simplification (almost) until

deconnection of the surface

  • curve construction by iterative geodesic computation
  • final curve is a closed piecewise geodesic but may

not be a constriction

VisSym2003 Detection of constrictions on closed polyhedral surfaces – p.31

slide-32
SLIDE 32

Perspectives

“Loosening” algorithm to slide the curve to a constriction

done, other paper

VisSym2003 Detection of constrictions on closed polyhedral surfaces – p.32

slide-33
SLIDE 33

Perspectives

Not all constrictions are detected:

  • seed curves can degenerate to a single point during

reconstruction

  • some constrictions are not simplified to seed curves

interesting to decompose the object in a few number of components (noisy surfaces) but may not be enough

replace seed curves by another configuration ?

theoretical justification ?

VisSym2003 Detection of constrictions on closed polyhedral surfaces – p.33

slide-34
SLIDE 34

Perspectives

Results highly depend on the simplification algorithm must preserve as much as possible the shape of the

  • bject

theoretical study, comparison between existing algorithms

VisSym2003 Detection of constrictions on closed polyhedral surfaces – p.34

slide-35
SLIDE 35

Thank you

VisSym2003 Detection of constrictions on closed polyhedral surfaces – p.35