Geometric Camera Calibration Chapter...

Post on 28-May-2020

3 views 0 download

Transcript of Geometric Camera Calibration Chapter...

Geometric Camera Calibration Chapter 2

Guido GerigCS 6643 Spring 2017

Slides modified from Marc Pollefeys, UNC Chapel Hill, Comp256, Other slides and illustrations from J. Ponce, addendum to course book, and Trevor Darrell, Berkeley, C280 Computer Vision Course.

Equation: World coordinates to image pixels

=

1.........

1

1 3

2

1

zW

yW

xW

T

T

T

ppp

mmm

zvu

pMz

p W 1=

PmPmv

PmPmu

⋅⋅

=

⋅⋅

=

3

2

3

1

pixel coordinatesworld coordinates

Conversion back from homogeneous coordinates leads to:

W. Freeman

Calibration target

http://www.kinetic.bc.ca/CompVision/opti-CAL.html

Find the position, ui and vi, in pixels, of each calibration object feature point.

Camera calibration

0)(0)(

32

31

=⋅−=⋅−

ii

ii

PmvmPmum

So for each feature point, i, we have:

PmPmv

PmPmu

⋅⋅

=

⋅⋅

=

3

2

3

1From before, we had these equations relating image positions,u,v, to points at 3-d positions P (in homogeneous coordinates):

W. Freeman

Camera calibration

0)(0)(

32

31

=⋅−=⋅−

ii

ii

PmvmPmum

Stack all these measurements of i=1…n points

into a big matrix:

=

−−

−−

00

00

00

00

3

2

1111

111

mmm

PvPPuP

PvPPuP

Tnn

Tn

T

Tnn

TTn

TTT

TTT

W. Freeman

=

−−

−−

00

00

00

00

3

2

1111

111

mmm

PvPPuP

PvPPuP

Tnn

Tn

T

Tnn

TTn

TTT

TTT

=

−−−−−−−−

−−−−−−−−

00

00

1000000001

1000000001

34

33

32

31

24

23

22

21

14

13

12

11

1111111111

1111111111

mmmmmmmmmmmm

vPvPvPvPPPuPuPuPuPPP

vPvPvPvPPPuPuPuPuPPP

nnznnynnxnnznynx

nnznnynnxnnznynx

zyxzyx

zyxzyx

Showing all the elements:

In vector form:Camera

calibration

W. Freeman

We want to solve for the unit vector m (the stacked one) that minimizes

2Qm

=

−−−−−−−−

−−−−−−−−

00

00

1000000001

1000000001

34

33

32

31

24

23

22

21

14

13

12

11

1111111111

1111111111

mmmmmmmmmmmm

vPvPvPvPPPuPuPuPuPPP

vPvPvPvPPPuPuPuPuPPP

nnznnynnxnnznynx

nnznnynnxnnznynx

zyxzyx

zyxzyx

Q m = 0

The eigenvector assoc. to the minimum eigenvalue of the matrix QTQ gives us that because it is the unit vector x that minimizes xT QTQ x.

Camera calibration

W. Freeman

Calibration Problem

Analytical Photogrammetry

Non-Linear Least-Squares Methods

• Newton• Gauss-Newton• Levenberg-Marquardt

Iterative, quadratically convergent in favorable situations

Homogeneous Linear Systems

A

A

x

x 0

0=

=

Square system:

• unique solution: 0

• unless Det(A)=0

Rectangular system ??

• 0 is always a solution

Minimize |Ax| under the constraint |x| =12

2

How do you solve overconstrained homogeneous linear equations ??

The solution is e .1

remember: EIG(UTU)=SVD(U), i.e. solution is Vn

Matlab Solution

Example:60 point pairs

Degenerate Point Configurations

Are there other solutions besides M ??

• Coplanar points: (λ,µ,ν )=(Π,0,0) or (0,Π,0) or (0,0,Π )

• Points lying on the intersection curve of two quadricsurfaces = straight line + twisted cubic

Does not happen for 6 or more random points!

Once you have the M matrix, can recover the intrinsic and extrinsic parameters.

Camera calibration

W. Freeman

Estimation of the Intrinsic and Extrinsic Parameters, see pdf slides S.M. Abdallah.

Once M is known, you still got to recover the intrinsic andextrinsic parameters !!!

This is a decomposition problem, not an estimationproblem.

• Intrinsic parameters

• Extrinsic parameters

ρ

Presenter
Presentation Notes
Rho times b3 represents tz, -> sign of rho can be chosen based on plausibility

Orthogonality of r3 and r2

ρ is known -> calculate v0 via dot product of a2 and a3

Slide Samer M Abdallah, Beirut

Slide Samer M Abdallah, Beirut

Typical Result

Camera was roughly 72 cm up from thefloor, and 136 cm back from the pattern.

There is considerable rotation around the y axis (up) but the rotation around x and z was kept to a minimum.

Sensor:

Degenerate Point Configurations

Are there other solutions besides M ??

• Coplanar points: (λ,µ,ν )=(Π,0,0) or (0,Π,0) or (0,0,Π )

• Points lying on the intersection curve of two quadricsurfaces = straight line + twisted cubic

Does not happen for 6 or more random points!

Other Slides following Forsyth&Ponce

Just for additional information on previous slides

Linear Systems

A

A

x

x b

b=

=

Square system:

• unique solution

• Gaussian elimination

Rectangular system ??

• underconstrained:infinity of solutions

Minimize |Ax-b|2

• overconstrained:no solution

How do you solve overconstrained linear equations ??

Homogeneous Linear Systems

A

A

x

x 0

0=

=

Square system:

• unique solution: 0

• unless Det(A)=0

Rectangular system ??

• 0 is always a solution

Minimize |Ax| under the constraint |x| =12

2

How do you solve overconstrained homogeneous linear equations ??

The solution is e .1

remember: EIG(UTU)=SVD(U), i.e. solution is Vn

Linear Camera Calibration

Useful Links

Demo calibration (some links broken):• http://mitpress.mit.edu/e-

journals/Videre/001/articles/Zhang/CalibEnv/CalibEnv.html

Bouget camera calibration SW:• http://www.vision.caltech.edu/bouguetj/

calib_doc/CVonline: Monocular Camera calibration:• http://homepages.inf.ed.ac.uk/cgi/rbf/C

VONLINE/entries.pl?TAG250