代写代考 IJCV 2000

PowerPoint 프레젠테이션

Changjae Oh

Copyright By PowCoder代写 加微信 powcoder

Computer Vision
– Calibration –

Semester 1, 22/23

Objectives

• Understanding the concept of camera calibration

• Understanding the relationship between image coordinate,
camera coordinate, and world coordinate

• Understanding a linear method for camera calibration

Our goal: Recovery of 3D structure

J. Vermeer, Music Lesson, 1662

A. Criminisi, M. Kemp, and A. Zisserman,Bringing Pictorial Space to Life: computer techniques for the analysis of painti

ngs, Proc. Computers and the History of Art, 2002

http://research.microsoft.com/apps/pubs/default.aspx?id=67260

Things aren’t always as they appear…

http://en.wikipedia.org/wiki/Ames_room

http://en.wikipedia.org/wiki/Ames_room

Single-view ambiguity

Single-view ambiguity

Single-view ambiguity

shadow sculptures

Rashad Alakbarov

Our goal: Recovery of 3D structure

• When certain assumptions hold, we ca
n recover structure from a single view

•In general, we need multi-view geometry

Image source

• But first, we need to understand the geometry of a single camera…

https://www.3dflow.net/elementsCV/S4.xhtml

Camera calibration

• Camera calibration:

̶ figuring out transformation from world coordinate system to image coordinate system

• Normalized (camera) coordinate system: camera center is at the origin,
the principal axis is the z-axis;
x and y axes of the image plane are parallel to x and y axes of the world

world coordinate system

)/,/(),,( ZYfZXfZYX 

Review: Pinhole camera model

Principal point

• Principal point (𝒑): point where principal axis intersects the image plane

• Normalized coordinate system: origin of the image is at the principal point

• Image coordinate system: origin is in the corner

Principal point offset

pZYfpZXfZYX ++

We want the principal point to
map to (px, py) instead of (0,0)

Principal point offset

principal point: ),(

Principal point offset

calibration matrix

 0|IKP =

principal point: ),(

projection matrix

Pixel coordinates

• mx pixels per meter in horizontal direction,
my pixels per meter in vertical direction

Pixel size:

pixels/m m pixels

( )C~X~RX~

Camera rotation and translation

• In general, the camera coordin
ate frame will be related to th
e world coordinate frame by a
rotation and a translation

coords. of point
in camera frame

coords. of camera center
in world frame

coords. of a point
in world frame

• Conversion from world to camera coordinate system
(in non-homogeneous coordinates):

camera coordinate
system world coordinate

Camera rotation and translation

( )C~X~RX~

3D transformation
matrix (4 x 4)

camera coordinate system world coordinate system

Camera rotation and translation

( )C~X~RX~

3D transformation
matrix (4 x 4)

camera coordinate system world coordinate system

Camera rotation and translation

3D transformation
matrix (4 x 4)

perspective project
ion matrix (3 x 4)

2D transformatio
n matrix (3 x 3)

camera coordinate system world coordinate system

Camera rotation and translation

camera coordinate system world coordinate system

Camera rotation and translation

camera coordinate system world coordinate system

Camera parameters

• Intrinsic parameters

̶ Principal point coordinates

̶ Focal length

̶ Pixel magnification factors

̶ Skew (non-rectangular pixels), Radial distortion

Camera parameters

• Intrinsic parameters

̶ Principal point coordinates

̶ Focal length

̶ Pixel magnification factors

̶ Skew (non-rectangular pixels), Radial distortion

• Extrinsic parameters

̶ Rotation and translation relative
to world coordinate system

̶ What is the projection of the
camera center?

 CRRKP ~−=

coords. of camera center
in world frame

CRRKPC The camera center is the null space

of the projection matrix!

Camera calibration

 XtRKx =

Camera calibration

• Given n points with known 3D coordinates Xi and known image projections
xi, estimate the camera parameters

Camera calibration: Linear method

• 𝑷 has 11 degrees of freedom

• One 2D/3D correspondence gives us two linearly independent equations

̶ 6 correspondences needed for a minimal solution

Camera calibration: Linear method

• 𝑷 has 11 degrees of freedom

• One 2D/3D correspondence gives us two linearly independent equations

̶ 6 correspondences needed for a minimal solution

Recall: Week1 quiz

Camera calibration: Linear method

• Directly estimate 11 unknowns in the P matrix using known 3D points
(𝑋, 𝑌, 𝑍) and measured (𝑋𝑖 , 𝑌𝑖 , 𝑍𝑖) and measured feature positions (𝑢𝑖 , 𝑣𝑖)

Camera calibration: Linear method

• Directly estimate 11 unknowns in the P matrix using known 3D points
(𝑋, 𝑌, 𝑍) and measured (𝑋𝑖 , 𝑌𝑖 , 𝑍𝑖) and measured feature positions (𝑢𝑖 , 𝑣𝑖)

Camera calibration: Linear method

• Solve for Projection Matrix 𝑷 using least-square techniques

Camera calibration: linear vs. nonlinear

• Linear calibration is easy to formulate and solve, but it doesn’t directly
tell us the camera parameters

• In practice, non-linear methods are preferred

̶ Write down objective function in terms of intrinsic and extrinsic parameters

̶ Define error as sum of squared distances between measured 2D points and estimated
projections of 3D points

̶ Minimize error using Newton’s method or other non-linear optimization

̶ Can model radial distortion and impose constraints such as known focal length and or
thogonality

 XtRKx =vs.

Application?

• Calibration is fundamental task for various computer vision tasks

https://ch.mathworks.com/help/vision/ug/single-camera-calibrator-app.html

Application?

• Calibration is fundamental task for various computer vision tasks

Xiang et al. “PoseCNN: A Convolutional Neural Network for 6D Object Pose Estimation in Cluttered Scenes.” RSS, 2018

Changjae Oh

Computer Vision
– Single-view Modeling –

Semester 1, 22/23

Objectives

• To understand calibration from vanishing points

• To understand measuring height without ruler

Application: Single-view modelling

A. Criminisi, I. Reid, and A. Zisserman, Sin

gle View Metrology, IJCV 2000

http://dhoiem.cs.illinois.edu/courses/vision_spring10/sources/criminisi00.pdf

Camera calibration revisited

• What if world coordinates of reference 3D points are not known?

• We can use scene features such as vanishing points

Camera calibration revisited

• What if world coordinates of reference 3D points are not known?

• We can use scene features such as vanishing points

Vertical vanishing

(at infinity)

Slide from Efros, Photo from Criminisi

Recall: Vanishing points

• All lines having the same direction share the same vanishing point

image plane

line in the scene

vanishing point v

Computing vanishing points

̶ X∞ is a point at infinity, v is its projection: v = PX∞
̶ The vanishing point depends only on line direction

̶ All lines having direction d intersect at X∞

Calibration from vanishing points

• Consider a scene with three orthogonal vanishing directions:

• Note: v1, v2 are finite vanishing points and v3 is an infinite vanishing point

Calibration from vanishing points

• Consider a scene with three orthogonal vanishing directions:

• We can align the world coordinate system with these directions

Calibration from vanishing points

• p1 = P(1,0,0,0)
T – the vanishing point in the x direction

• Similarly, p2 and p3 are the vanishing points in the y and z directions

• p4 = P(0,0,0,1)
T – projection of the origin of the world coordinate system

• Problem: we can only know the four columns up to independent scale factors, addition
al constraints needed to solve for them

Calibration from vanishing points

• Let us align the world coordinate system with three orthogonal vanishing
directions in the scene:

Calibration from vanishing points

• Let us align the world coordinate system with three orthogonal vanishing
directions in the scene:

• Orthogonality constraint:

Calibration from vanishing points

• Let us align the world coordinate system with three orthogonal vanishing
directions in the scene:

• Orthogonality constraint:

• Rotation disappears, each pair of vanishing points gives constraint on focal
length and principal point

Calibration from vanishing points

Can solve for focal length, principal pointCannot recover focal length, principal
point is the third vanishing point

Rotation from vanishing points

• Constraints on vanishing points:

• After solving for the calibration matrix:

• Get λi by using the constraint ||ri||

Calibration from vanishing points: Summary

• Solve for K (focal length, principal point) using three orthogonal vanishing points

• Get rotation directly from vanishing points once calibration matrix is known

• Advantages

̶ No need for calibration chart, 2D-3D correspondences

̶ Could be completely automatic

• Disadvantages

̶ Only applies to certain kinds of scenes

̶ Inaccuracies in computation of vanishing points

̶ Problems due to infinite vanishing points

Making measurements from a single image

http://en.wikipedia.org/wiki/Ames_room

http://en.wikipedia.org/wiki/Ames_room

Recall: Measuring height

Camera height

Measuring height without a ruler

ground plane

Compute Z from image measurements

• Need more than vanishing points to do this

Projective invariant

• We need to use a projective invariant: a quantity that does not change u
nder projective transformations (including perspective projection)

̶ What are some invariants for similarity, affine transformations?

Projective invariant

• We need to use a projective invariant: a quantity that does not change u
nder projective transformations (including perspective projection)

̶ The cross-ratio of four points:

image cross ratio

Measuring height

B (bottom of object)

T (top of object)

R (reference point)

ground plane

scene cross ratio

Measuring height without a ruler

image cross ratio

vanishing line (horizon)

A. Criminisi, I. Reid, and A. Zisserman, Single View Metrology, IJCV 2000

Figure from UPenn CIS580 slides

http://dhoiem.cs.illinois.edu/courses/vision_spring10/sources/criminisi00.pdf
http://cis.upenn.edu/~cis580/Spring2015/Lectures/cis580-04-singleview.pdf

Measurements on planes

Approach: unwarp then measure

What kind of warp is this?

Image rectification

• To unwarp (rectify) an image

̶ solve for homography H given p and p′

̶ how many points are necessary to solve for H?

Image rectification: example

Piero della Francesca, Flagellation, ca. 1455

Application: 3D modeling from a single image

A. Criminisi, M. Kemp, and A. Zisserman,Bringing Pictorial Space to Life: computer techniques for the analysis of paintings,

Proc. Computers and the History of Art, 2002

http://research.microsoft.com/apps/pubs/default.aspx?id=67260

Application: 3D modeling from a single image

J. Vermeer, Music Lesson, 1662

A. Criminisi, M. Kemp, and A. Zisserman,Bringing Pictorial Space to Life: computer techniques for the analysis of paintings,

Proc. Computers and the History of Art, 2002

http://research.microsoft.com/apps/pubs/default.aspx?id=67260

Application: Fully automatic modeling

D. Hoiem, A.A. Efros, and M. Hebert, Automatic Photo Pop-up, SIGGRAPH 2005.

http://dhoiem.cs.illinois.edu/publications/popup.pdf

Application: Object detection

D. Hoiem, A.A. Efros, and M. Hebert, Putting Objects in Perspective, CVPR 2006

https://web.engr.illinois.edu/~dhoiem/publications/hoiem_cvpr06.pdf

Next Topic

• How about using two cameras?

̶ Prerequisite

• Review Part2-3: Calibration (this content!)

• Review Part1-3: Bilateral filtering

程序代写 CS代考 加微信: powcoder QQ: 1823890830 Email: powcoder@163.com