Calibrating a single camera. Odilon Redon, Cyclops, 1914

Similar documents
Geometry of a single camera. Odilon Redon, Cyclops, 1914

Structure from Motion

What are the camera parameters? Where are the light sources? What is the mapping from radiance to pixel color? Want to solve for 3D geometry

Image Alignment CSC 767

Computer Vision. Exercise Session 1. Institute of Visual Computing

LEAST SQUARES. RANSAC. HOUGH TRANSFORM.

Image warping and stitching May 5 th, 2015

Computer Animation and Visualisation. Lecture 4. Rigging / Skinning

Multi-stable Perception. Necker Cube

Geometric Transformations and Multiple Views

Prof. Feng Liu. Spring /24/2017

AIMS Computer vision. AIMS Computer Vision. Outline. Outline.

Range images. Range image registration. Examples of sampling patterns. Range images and range surfaces

An efficient method to build panoramic image mosaics

Radial Basis Functions

Finding Intrinsic and Extrinsic Viewing Parameters from a Single Realist Painting

New dynamic zoom calibration technique for a stereo-vision based multi-view 3D modeling system

New Extensions of the 3-Simplex for Exterior Orientation

Support Vector Machines

Introduction to Geometrical Optics - a 2D ray tracing Excel model for spherical mirrors - Part 2

CS 534: Computer Vision Model Fitting

Recognizing Faces. Outline

Slide 1 SPH3UW: OPTICS I. Slide 2. Slide 3. Introduction to Mirrors. Light incident on an object

Reading. 14. Subdivision curves. Recommended:

Accounting for the Use of Different Length Scale Factors in x, y and z Directions

Model-Based Bundle Adjustment to Face Modeling

RELATIVE ORIENTATION ESTIMATION OF VIDEO STREAMS FROM A SINGLE PAN-TILT-ZOOM CAMERA. Commission I, WG I/5

Introduction to Multiview Rank Conditions and their Applications: A Review.

Graph-based Clustering

2D Raster Graphics. Integer grid Sequential (left-right, top-down) scan. Computer Graphics

LECTURE : MANIFOLD LEARNING

Non-Parametric Structure-Based Calibration of Radially Symmetric Cameras

Angle-Independent 3D Reconstruction. Ji Zhang Mireille Boutin Daniel Aliaga

Feature Reduction and Selection

Lecture 4: Principal components

Fitting & Matching. Lecture 4 Prof. Bregler. Slides from: S. Lazebnik, S. Seitz, M. Pollefeys, A. Effros.

LU Decomposition Method Jamie Trahan, Autar Kaw, Kevin Martin University of South Florida United States of America

Calibration of an Articulated Camera System with Scale Factor Estimation

A Range Image Refinement Technique for Multi-view 3D Model Reconstruction

EVALUATION OF RELATIVE POSE ESTIMATION METHODS FOR MULTI-CAMERA SETUPS

Programming in Fortran 90 : 2017/2018

STRUCTURE and motion problems form a class of geometric

12/2/2009. Announcements. Parametric / Non-parametric. Case-Based Reasoning. Nearest-Neighbor on Images. Nearest-Neighbor Classification

Abstract metric to nd the optimal pose and to measure the distance between the measurements

A 3D Reconstruction System of Indoor Scenes with Rotating Platform

Classification / Regression Support Vector Machines

Line-based Camera Movement Estimation by Using Parallel Lines in Omnidirectional Video

of Multiple Cameras Abstract This paper addresses the problem of calibrating camera parameters using variational methods.

Exact solution, the Direct Linear Transfo. ct solution, the Direct Linear Transform

An Application of the Dulmage-Mendelsohn Decomposition to Sparse Null Space Bases of Full Row Rank Matrices

Kinematics Modeling and Analysis of MOTOMAN-HP20 Robot

4 x 10 3 Residual during bilinear iteration error residual iteration

Scan Conversion & Shading

Towards Direct Recovery of Shape and Motion Parameters from Image Sequences

Color in OpenGL Polygonal Shading Light Source in OpenGL Material Properties Normal Vectors Phong model

A Revisit of Methods for Determining the Fundamental Matrix with Planes

Scan Conversion & Shading

Amnon Shashua Shai Avidan Michael Werman. The Hebrew University, objects.

A Comparison and Evaluation of Three Different Pose Estimation Algorithms In Detecting Low Texture Manufactured Objects

ROBOT KINEMATICS. ME Robotics ME Robotics

Machine Learning. Support Vector Machines. (contains material adapted from talks by Constantin F. Aliferis & Ioannis Tsamardinos, and Martin Law)

Computer Vision I. Xbox Kinnect: Rectification. The Fundamental matrix. Stereo III. CSE252A Lecture 16. Example: forward motion

Kent State University CS 4/ Design and Analysis of Algorithms. Dept. of Math & Computer Science LECT-16. Dynamic Programming

Alignment and Object Instance Recognition

Face Recognition University at Buffalo CSE666 Lecture Slides Resources:

A NEW METHOD FOR STEREO- CAMERAS SELF-CALIBRATION IN SCHEIMPFLUG CONDITION

Any Pair of 2D Curves Is Consistent with a 3D Symmetric Interpretation

Fitting and Alignment

Kinematics of pantograph masts

Geometric Primitive Refinement for Structured Light Cameras

Self-Calibration from Image Triplets. 1 Robotics Research Group, Department of Engineering Science, Oxford University, England

Region Segmentation Readings: Chapter 10: 10.1 Additional Materials Provided

Two Dimensional Projective Point Matching

ESTIMATION OF INTERIOR ORIENTATION AND ECCENTRICITY PARAMETERS OF A HYBRID IMAGING AND LASER SCANNING SENSOR

Research on laser tracker measurement accuracy and data processing. Liang Jing IHEP,CHINA

A Robust Method for Estimating the Fundamental Matrix

Robust Computation and Parametrization of Multiple View. Relations. Oxford University, OX1 3PJ. Gaussian).

Image Fusion With a Dental Panoramic X-ray Image and Face Image Acquired With a KINECT

USING GRAPHING SKILLS

Exterior Orientation using Coplanar Parallel Lines

Feature Extraction and Registration An Overview

R s s f. m y s. SPH3UW Unit 7.3 Spherical Concave Mirrors Page 1 of 12. Notes

1. Answer the following. a. A beam of vertically polarized light of intensity W/m2 encounters two polarizing filters as shown below.

INF Repetition Anne Solberg INF

CS 231A Computer Vision Midterm

Announcements. Supervised Learning

Ecient Computation of the Most Probable Motion from Fuzzy. Moshe Ben-Ezra Shmuel Peleg Michael Werman. The Hebrew University of Jerusalem

Shape Representation Robust to the Sketching Order Using Distance Map and Direction Histogram

arxiv: v1 [cs.ro] 8 Jul 2016

PROJECTIVE RECONSTRUCTION OF BUILDING SHAPE FROM SILHOUETTE IMAGES ACQUIRED FROM UNCALIBRATED CAMERAS

3D Modeling Using Multi-View Images. Jinjin Li. A Thesis Presented in Partial Fulfillment of the Requirements for the Degree Master of Science

6.1 2D and 3D feature-based alignment 275. similarity. Euclidean

A Scalable Projective Bundle Adjustment Algorithm using the L Norm

y and the total sum of

Wavefront Reconstructor

Hand Eye Calibration Applied to Viewpoint Selection for Robotic Vision Yuichi Motai, Member, IEEE, and Akio Kosaka, Member, IEEE

Machine Learning 9. week

Application of Visual Tracking for Robot-Assisted Laparoscopic Surgery

The Research of Ellipse Parameter Fitting Algorithm of Ultrasonic Imaging Logging in the Casing Hole

Planar Catadioptric Stereo: Multiple-View Geometry and Image-Based Camera Localization

Transcription:

Calbratng a sngle camera Odlon Redon, Cclops, 94

Our goal: Recover o 3D structure Recover o structure rom one mage s nherentl ambguous???

Sngle-vew ambgut

Sngle-vew ambgut Rashad Alakbarov shadow sculptures

Sngle-vew ambgut Ames room

Our goal: Recover o 3D structure We wll need mult-vew geometr

Revew: Pnhole camera model world coordnate sstem Normalzed (camera) coordnate sstem: camera center s at the orgn, the prncpal as s the z-as, and aes o the mage plane are parallel to and aes o the world Goal o camera calbraton: go rom world coordnate sstem to mage coordnate sstem

) /, / ( ),, ( Z Y Z Z Y! Z Y Z Y Z Y! Revew: Pnhole camera model P

Prncpal pont p p Prncpal pont (p): pont where prncpal as ntersects the mage plane Normalzed coordnate sstem: orgn o the mage s at the prncpal pont Image coordnate sstem: orgn s n the corner

) /, / ( ),, ( p Z Y p Z Z Y + +! + + Z Z p Y Z p Z Y! Prncpal pont oset We want the prncpal pont to map to (p, p ) nstead o (,) p p Z Y p p

+ + Z Y p p Z Zp Y Zp Prncpal pont oset p p K calbraton matr [ ] P K I prncpal pont: ), ( p p p p

p p m m K β α β α Pel coordnates m pels per meter n horzontal drecton, m pels per meter n vertcal drecton Pel sze: m m pels/m m pels

Camera rotaton and translaton In general, the camera coordnate rame wll be related to the world coordnate rame b a rotaton and a translaton Converson rom world to camera coordnate sstem (n non-homogeneous coordnates): ~ cam ( ) R ~ C ~ coords. o pont n camera rame coords. o a pont n world rame coords. o camera center n world rame

Camera rotaton and translaton ~ cam ( ) R ~ C ~ cam ~ cam R RC ~ ~ R RC ~ [ ] K[ R ] K I RC ~ cam P K[ R t], t RC ~

Camera parameters Intrnsc parameters Prncpal pont coordnates Focal length Pel magncaton actors Skew (non-rectangular pels) Radal dstorton p p m m β α β α K [ ] t P K R

Camera parameters P K[ R t] Intrnsc parameters Prncpal pont coordnates Focal length Pel magncaton actors Skew (non-rectangular pels) Radal dstorton Etrnsc parameters Rotaton and translaton relatve to world coordnate sstem P [ ~ ] K R RC What s the projecton o the camera center? PC ~ ~ C [ RC] K R coords. o camera center n world rame he camera center s the null space o the projecton matr!

Camera calbraton Z Y λ λ λ [ ] t K R Source: D. Hoem

Camera calbraton Gven n ponts wth known 3D coordnates and known mage projectons, estmate the camera parameters P?

P λ Camera calbraton: Lnear method P 3 P P P 3 P P P wo lnearl ndependent equatons

Camera calbraton: Lnear method P has degrees o reedom One D/3D correspondence gves us two lnearl ndependent equatons 6 correspondences needed or a mnmal soluton Homogeneous least squares: nd p mnmzng Ap Soluton gven b egenvector o A A wth smallest egenvalue p A 3 P P P n n n n n n!!!

Camera calbraton: Lnear method Note: or coplanar ponts that sats Π, we wll get degenerate solutons (Π,,), (,Π,), or (,,Π) Ap 3 P P P n n n n n n!!!

Camera calbraton: Lnear method he lnear method onl estmates the entres o the projecton matr: What we ultmatel want s a decomposton o ths matr nto the ntrnsc and etrnsc parameters: State-o-the-art methods use nonlnear optmzaton to solve or the parameter values drectl [ ] t K R Z Y λ λ λ

Camera calbraton: Lnear method Advantages: eas to ormulate and solve Dsadvantages Doesn t drectl tell ou camera parameters Doesn t model radal dstorton Can t mpose constrants, such as known ocal length and orthogonalt Non-lnear methods are preerred Dene error as sum o squared dstances between measured D ponts and estmated projectons o 3D ponts Mnmze error usng Newton s method or other non-lnear optmzaton Source: D. Hoem

A taste o mult-vew geometr: rangulaton Gven projectons o a 3D pont n two or more mages (wth known camera matrces), nd the coordnates o the pont

rangulaton Gven projectons o a 3D pont n two or more mages (wth known camera matrces), nd the coordnates o the pont? O O

rangulaton We want to ntersect the two vsual ras correspondng to and, but because o nose and numercal errors, the don t meet eactl? O O

rangulaton: Geometrc approach Fnd shortest segment connectng the two vewng ras and let be the mdpont o that segment O O

rangulaton: Nonlnear approach Fnd that mnmzes d (,P ) + d (,P )? P P O O

rangulaton: Lnear approach b a b a ] [ z z z b b b a a a a a a P P λ λ P P ]P [ ]P [ Cross product as matr multplcaton:

rangulaton: Lnear approach λ λ P P P P [ [ ]P ]P wo ndependent equatons each n terms o three unknown entres o