Affine Structure from Motion - PowerPoint PPT Presentation

1 / 28
About This Presentation
Title:

Affine Structure from Motion

Description:

Affine Structure from Motion. Frank Dellaert. Slides all my own this time ... Facorization is better algorithm to estimate affine geometry than linear FA ... – PowerPoint PPT presentation

Number of Views:63
Avg rating:3.0/5.0
Slides: 29
Provided by: FrankDe62
Category:

less

Transcript and Presenter's Notes

Title: Affine Structure from Motion


1
Affine Structure from Motion
  • Frank Dellaert
  • Slides all my own this time ?

2
Structure and Motion
3
Book Sequence
4
Book Structure Motion
5
2 Problems !
  • Correspondence
  • Optimization

6
A Correspondence Problem
7
An Optimization Problem
  • Find the most likely structure and motion ?

8
Optimization
Image i
Image i
uik
9
Affine Cameras
  • Easy projection equation

10
Geometric Intuition !!
  • From first lecture on cameras !

(0.6,2.5,1) ?
Z
Y
X
11
Geometric Intuition
  • Projection matrix form


Z
Y
X
12
Translation of image i ?
  • Take derivative wrpt to ti

13
World origin trick
  • Move scene so centroid origin

14
New Problem !
  • Just subtract centroid from all measurements

15
Matrix Form
  • Projection in one image (i)

16
Matrix Form
  • Projection in two images

17
Matrix Form
  • Projection all points in all images

n
16
13
14
M1
1
15
2
3
11
12
4
5
6

24
M2
25
26
21
22
23
2m
34
M3
35
36
31
32
33
44
M4
45
46
41
42
43
18
Geometric Intuition !!
  • Joint Image of point 8-dimensional

Z
Y
X
19
Subspace Constraints
  • Does it really span 8-dim space ??

Z
Y
X
20
Joint Camera Mapping
83
R3
R8
Mapping from 3D to 8D space (for 4 cameras)
21
But Range is only 3D !!!
83
R3
R3
R5
5-dimensional null-space !
22
What does SVD look like ?
AU S VT
3x3
83
3x3
R3
R3P2
R5
Rank 3, economy SVD
23
And measurement matrix ?
Also Rank 3 !!!
24
SVD Approach to SFM
  • Form measurement matrix
  • Do rank 3 economy size SVD
  • Voila

Motion
Structure
25
Problem
  • Where to stick the extra 3x3 ?

Motion
Structure
3x3
3x3
83

26
Euclidean Upgrade
  • Uncalibrated not so easy
  • Calibrated cameras
  • We know they are rotation matrices
  • Impose constraints
  • unit length 2m
  • orthogonal m
  • Solve for Q

3x3
3x3
27
House Sequence
28
Epilogue (no pun intended)
  • Facorization is better algorithm to estimate
    affine geometry than linear FA
  • In RANSAC use factorization
  • Sample set of 4 correspondences
  • Form measurement matrix
  • Do SVD
  • Calculate FA from M, or, better
  • Calculate points with
Write a Comment
User Comments (0)
About PowerShow.com