Introduction To Digital Image Correlation - ResearchGate

SEM 2009 Workshop
Digital Image Correlation:

Overview of
Principles and Software
University of South Carolina

Correlated Solutions, Inc.
2D Image Correlation
Fundamentals
Outline
Introduction
Purpose
Basic Idea
Principle
Measuring Displacem.
Introduction Strain Computation

Matching by DIC
Principle Purpose Method
Photometric Mapping
Displacement Mapping
Basic Idea
Advanced Topics
Conclusions
Principle
Strain Computation
Limitations, Benefits
Method
Acknowledgments
Conclusions
Acknowledgments
Matching by DIC
Principle
Photometric Mapping
Advanced Topics
Introduction
Introduction
Purpose
Basic Idea
Principle

Matching by DIC
Photometric Mapping
Basic Idea
Advanced Topics
Conclusions
Principle
Strain Computation
Method
Acknowledgments
Conclusions
Acknowledgments
Matching by DIC
Principle
Photometric Mapping
Advanced Topics
Purpose
Introduction
Purpose
Basic Idea
Principle
Measuring Displacem. Deformation
Matching by DIC
Principle
measurement
Photometric Mapping
Advanced Topics
▫ Full-field
Strain Computation measurement
Method
▫ Non-intrusive
Conclusions
▫ Planar specimen only
Acknowledgments
▫ No out of plane motion
Purpose
Introduction
Purpose
Basic Idea
Principle
Matching by DIC
▫ Large range of size scales

Principle
Photometric Mapping
Advanced Topics (10-9 to 102 m.)
Strain Computation
Method
Conclusions
▫ Large range of time scales
Acknowledgments (static to 200MHz)
Principle
Introduction
Purpose
Basic Idea
Principle
Matching by DIC ▫ Monocular (cyclopean)

Principle
Photometric Mapping
vision cannot determine
Displacement Mapping the size of objects
Advanced Topics
▫ Consequence: a 200%
Strain Computation
isotropic deformation of
Method
an object produces the
Conclusions same image as if the
Acknowledgments
object was moved to one-
half its original distance
from the visual sensor
▫ We must assume the
object is planar, parallel
to and at a constant
distance from the visual
sensor during the entire
experiment
Image Correlation Technique
Introduction
Purpose
Basic Idea
▫ How? Given a point and its signature in the
Principle
undeformed image, search/track in deformed
Matching by DIC
image for the point which has a signature which
Principle
Photometric Mapping
maximizes a similarity function
Advanced Topics
Strain Computation
Method
Conclusions
Acknowledgments
tracking
tracking
Time t Time t’ Time t”

Introduction
Purpose
Basic Idea
▫ In practice, a single value is not a unique
Principle
signature of a point, hence neighboring pixels
Matching by DIC
are used
Principle
Photometric Mapping ▫ Such a collection of pixel values is called a
Advanced Topics subset or window
Strain Computation
Method
Conclusions
Acknowledgments
Time t Time t’
Introduction
Purpose
Basic Idea
Principle
Repetitive
Matching by DIC ▫ The uniqueness of Anisotropic
each signature is
Principle
Photometric Mapping
High-contrast
only guaranteed if
Advanced Topics
Strain Computation the surface has a Non-repetitive

Anisotropic
non repetitive,
Method
High-contrast
Conclusions
Limitations, Benefits isotropic, high
contrast pattern
Acknowledgments
Non-repetitive
▫ Random textures Isotropic
fulfill this constraint Low-contrast
(speckle pattern)
Non-repetitive
Isotropic
High-contrast
Matching by DIC
Introduction
Purpose
Basic Idea
Principle

Matching by DIC
Photometric Mapping
Basic Idea
Advanced Topics
Conclusions
Principle
Strain Computation
Method
Acknowledgments
Conclusions
Acknowledgments
Matching by DIC
Principle
Photometric Mapping
Advanced Topics
Principle
Introduction
Purpose ▫ Example
Basic Idea
Principle ▫ The camera acquires 9x9 pixel images
▫ The specimen is marked with a cross-like pattern
Matching by DIC
Principle
Photometric Mapping
Advanced Topics
Strain Computation Image, on screen

Method
Conclusions
Acknowledgments
Pixels
Principle
▫ Example
Introduction
Purpose
Basic Idea
Principle
▫ White pixels are gray level 100
Measuring Displacem. ▫ Black pixels are gray level 0
Matching by DIC ▫ An image is a matrix of natural integers
Principle
Photometric Mapping
Advanced Topics
Strain Computation Image, in memory Image, on screen

Method
100 100 100 0 0 0 100 100 100

Conclusions
Limitations, Benefits 100 100 100 0 0 0 100 100 100
Acknowledgments
100 100 100 0 0 0 100 100 100
0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0
100 100 100 0 0 0 100 100 100
100 100 100 0 0 0 100 100 100
100 100 100 0 0 0 100 100 100

Principle
Introduction
Purpose ▫ Example
Basic Idea
Principle ▫ The specimen moves such that its image moves 1 pixel up
and right
Matching by DIC
Principle
Photometric Mapping
Advanced Topics
Strain Computation Image after motion, in memory Image after motion, on screen
Method
100 100 100 100 0 0 0 100 100

Conclusions
Acknowledgments
0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0
100 100 100 100 0 0 0 100 100
100 100 100 100 0 0 0 100 100
100 100 100 100 0 0 0 100 100
100 100 100 100 0 0 0 100 100

Principle
Introduction
Purpose ▫ Example: we define a 5x5 subset in the reference image
Basic Idea
Principle
(before motion)
▫ Problem: find where the subset moved (matching)
Matching by DIC
Principle
Photometric Mapping
Advanced Topics
Strain Computation Image before motion Image after motion

Method
100 100 100 0 0 0 100 100 100 100 100 100 100 0 0 0 100 100
Conclusions
Limitations, Benefits 100 100 100 0 0 0 100 100 100 100 100 100 100 0 0 0 100 100
Acknowledgments
100 100 100 0 0 0 100 100 100 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
?
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 100 100 100 100 0 0 0 100 100
100 100 100 0 0 0 100 100 100 100 100 100 100 0 0 0 100 100
Subset
100 100 100 0 0 0 100 100 100 100 100 100 100 0 0 0 100 100
100 100 100 0 0 0 100 100 100 100 100 100 100 0 0 0 100 100
Principle
▫ Solution: check possible matches at several locations and
Introduction
Purpose
Basic Idea
Principle
use a similarity score (correlation function) to grade them
Measuring Displacem. ▫ Classic correlation function: sum of squared differences
Matching by DIC
(SSD) of the pixel values (smaller values = better similarity)
Principle
Photometric Mapping
Advanced Topics
Strain Computation
Method
Pixel coord., reference image Pixel value at (x+i; y+j) Pixel value at (x+u+i; y+v+j)
Conclusions
n/2

Acknowledgments
C ( x, y, u, v)  ( I ( x  i , y  j )  I *
(x  u  i , y  v  j )) 2
i , j  n / 2
Image before motion Image after motion
n: subset size (9 in our example)
Displacement (disparity)
Correlation function
Principle
Introduction
Purpose
▫ Example: subset at (x;y)=(5;5), displacement candidate (u;v)=(-2;-2)
Basic Idea
Principle
2
Matching by DIC
C (5,5,2,2)   (I (5  i,6  j)  I (5  2  i,5  2  j))
i , j  2
* 2
Principle (100  0) 2  (0  0) 2  (0  0) 2  (0  0) 2  (100  0) 2 

Photometric Mapping
(0  100) 2  (0  100) 2  (0  100) 2  (0  100) 2  (0  0) 2 
(0  100) 2  (0  100) 2  (0  100) 2  (0  100) 2  (0  0) 2 
Advanced Topics
(0  100) 2  (0  100) 2  (0  100) 2  (0  100) 2  (0  0) 2 
Strain Computation (100  100) 2  (0  100) 2  (0  100) 2  (0  100) 2  (100  0) 2  18,000
Method
100 100 100 0 0 0 100 100 100 100 100 100 100 0 0 0 100 100
Conclusions
(x;y)=(5;5)
100 100 100 0 0 0 100 100 100 100 100 100 100 0 0 0 100 100
Acknowledgments
Image before motion

100 100 100 0 0 0 100 100 100 0 0 0 0 0 0 0 0 0
Image after motion

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 100 100 100 100 0 0 0 100 100
100 100 100 0 0 0 100 100 100 100 100 100 100 0 0 0 100 100
100 100 100 0 0 0 100 100 100 100 100 100 100 0 0 0 100 100
(u;v)=(-2;-2)
100 100 100 0 0 0 100 100 100 100 100 100 100 0 0 0 100 100
Principle
Introduction
▫ Example: subset at (x;y)=(5;5), displacement candidate
Purpose
(u;v)=(1;1)
C (5,5,1,1)  0
Basic Idea
Principle
Matching by DIC
▫ Better correlation score than candidate (u;v)=(-2;-2) [18,000]
Principle
Photometric Mapping
Indeed it is the smallest score achieveable (perfect match)
Advanced Topics
Strain Computation
Method
100 100 100 0 0 0 100 100 100 100 100 100 100 0 0 0 100 100
Conclusions
(x;y)=(5;5)
100 100 100 0 0 0 100 100 100 100 100 100 100 0 0 0 100 100
Acknowledgments
Image before motion

100 100 100 0 0 0 100 100 100 0 0 0 0 0 0 0 0 0
Image after motion

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 100 100 100 100 0 0 0 100 100
100 100 100 0 0 0 100 100 100 100 100 100 100 0 0 0 100 100
100 100 100 0 0 0 100 100 100 100 100 100 100 0 0 0 100 100
(u;v)=(1;1)
100 100 100 0 0 0 100 100 100 100 100 100 100 0 0 0 100 100
Principle
Introduction
Purpose ▫ In reality, images are corrupted by some noise
Basic Idea
Principle ▫ The SSD function will likely never be 0 for a perfect match
Matching by DIC
Principle
Photometric Mapping
Advanced Topics
Strain Computation Image before motion Image after motion

Method
103 101 99 2 0 1 105 100 96 99 100 101 102 3 0 2 100 102

Conclusions
Limitations, Benefits 101 104 98 1 4 3 101 98 100 101 97 98 101 1 2 0 96 102
Acknowledgments
103 96 99 0 2 2 102 103 98 0 1 3 3 2 0 1 2 0
2 3 0 1 1 2 3 0 1 1 0 3 0 2 1 1 0 3
1 3 3 0 2 1 0 3 0 1 3 2 0 1 1 2 2 0
0 0 2 0 3 0 2 0 0 101 100 100 103 0 2 1 102 101
98 101 102 0 1 0 96 97 102 97 99 100 101 3 2 0 97 101
97 98 103 0 2 0 103 98 100 101 103 98 101 0 1 1 99 96
102 99 101 2 0 0 104 102 101 102 99 96 103 2 3 3 102 100

Principle
Introduction
Purpose
Basic Idea
▫ Example
Principle
▫ The specimen moves such that its image moves 0.5
pixel to the right
Matching by DIC
Principle ▫ Need to interpolate the image at non-integer locations
Photometric Mapping
Advanced Topics
Strain Computation
Method
103 101 99 52 0 1 55 100 96

Conclusions
Acknowledgments
103 96 99 49 2 2 52 103 98
2 3 0 1 1 2 3 0 1
1 3 3 0 2 1 0 3 0
0 0 2 0 3 0 2 0 0
98 101 102 52 1 0 46 97 102

Perfect match
97 98 103 51 2 0 53 98 100
102 99 101 48 0 0 54 102 101

Photometric Mapping
Introduction
Purpose
Basic Idea
Principle
Matching by DIC
Principle
Photometric Mapping
▫ During image acquisition:
▫ Lighting conditions may change
Advanced Topics
Strain Computation
Method
▫ Sensor integration time adjusted
Conclusions
▫ Pattern may become lighter/darker when
Acknowledgments
expanded/compressed
Photometric Mapping
Introduction
Purpose
Basic Idea
Principle
Matching by DIC
Principle
Photometric Mapping
▫ The photometric mapping is not
Advanced Topics guaranteed to be an identity, hence the
Strain Computation
Method
DIC algorithm may have false matches
Conclusions ▫ Solution: model the photometric
transformation and use it to design a
Acknowledgments
robust correlation function

Introduction
Purpose
Basic Idea
Principle
▫ General circumstances:
▫ The subset in the deformed image has changed
Matching by DIC
shape, e.g. a square initial subset is likely to be non-
Principle
Photometric Mapping square
Advanced Topics
Strain Computation
Method
Conclusions
Acknowledgments
Specimen deformation
Time t Time t’
Introduction
Purpose
Basic Idea
Principle
▫ Solution: model this displacement
Matching by DIC
Principle
transformation (called subset shape function)
Photometric Mapping
Displacement Mapping and use it to define the deformed subset
Advanced Topics
Strain Computation
Method
Conclusions
Acknowledgments
Image Interpolation
Introduction
Purpose
▫ Optimization algorithms require the criterion

Basic Idea
Principle
to be continuous over its parameter space

Matching by DIC
Principle
Photometric Mapping
▫ Images are discrete, hence we need to
Advanced Topics
reconstruct the continuous information by
Strain Computation
means of interpolation
▫ B-splines are a good choice for that purpose
Method
Conclusions
Acknowledgments
Raw image Bi-linear B-spline interp. Bi-cubic B-spline interp.

Conclusions
Introduction
Purpose
Basic Idea
Principle

Matching by DIC
Photometric Mapping
Basic Idea
Advanced Topics
Conclusions
Principle
Strain Computation
Summary
Method
Acknowledgments
Conclusions
Acknowledgments
Matching by DIC
Principle
Photometric Mapping
Advanced Topics
Summary
Introduction
Purpose
Basic Idea
Principle tracking
tracking
Matching by DIC
Principle
Photometric Mapping
Advanced Topics
Time t Time t’ Time t”
Strain Computation
Method
Conclusions ▫ Matching process

Acknowledgments
▫ Photometric mapping
▫ Shape function
▫ Implications
Acknowledgements
Introduction
Purpose
Basic Idea
Principle
▫ We thank Prof. Jean-José Orteu from

Matching by DIC
Principle
École des Mines d’Albi, France, for
Photometric Mapping providing us with multimedia content
Advanced Topics
Strain Computation
Method
Conclusions
Acknowledgments
▫ We gratefully acknowledge the

University of South Carolina,
Mechanical Engineering Dept.,
support for providing us with
multimedia content
Three-Dimensional
Digital Image Correlation
Overview
Outline
Introduction
Purpose
Basic Idea
Camera Convention
Principle
Introduction Matching by DIC

Calibration
Model Purpose Stereo Matching
Estimation
Basic Idea Tracking

Matching by DIC
Stereo Matching Camera Convention Stereo Tracking
Tracking
Stereo Tracking
Lens Distortion
Principle Lens Distortion
Strain Computation
Principle
Calibration Strain Computation
Model Principle
Conclusions
Estimation
Acknowledgments
Conclusions
Acknowledgments
Introduction
Introduction
Purpose
Basic Idea
Camera Convention
Principle

Calibration
Estimation
Basic Idea Tracking

Matching by DIC
Tracking
Stereo Tracking
Lens Distortion
Strain Computation
Principle
Model Principle
Conclusions
Estimation
Acknowledgments
Conclusions
Acknowledgments
Purpose
Introduction
Purpose
Basic Idea
▫ 3-D shape
Camera Convention
Principle
▫ Strain map
Calibration ▫ Full-field measurement
Model
Estimation
▫ Non-intrusive
▫ Any specimen shape
Matching by DIC
Stereo Matching
Tracking
Stereo Tracking
Lens Distortion
▫ Any motion
Strain Computation
Principle
Conclusions
Acknowledgments
Purpose
Introduction
Purpose
Basic Idea
▫ Large range of size
Camera Convention
Principle
scales
Calibration
(10-9 to 102 m.)
Model
1X 50X 200X
Estimation
Matching by DIC
Stereo Matching
Tracking
Stereo Tracking
Lens Distortion
Strain Computation
Principle
Conclusions
Acknowledgments
3-D shape of letter „R‟ from “E

PLURIBUS UNUM” on U.S. Penny
coin using a Scanning Electron
Microscope
Basic Idea
Introduction
Purpose
Basic Idea
▫ Much like human
Camera Convention
Principle
vision, two imaging
Calibration
sensors provide
Model
Estimation
enough information to
Matching by DIC
perceive the
Stereo Matching
Tracking
environment in three-
Stereo Tracking
Lens Distortion
dimensions
Strain Computation
Principle
Conclusions
Acknowledgments
Basic Idea
Introduction
Purpose ▫ DIC for stereo-
Basic Idea
Camera Convention
matching: two
Principle
3-D shapes
Calibration
Model
▫ DIC for
Estimation tracking:
Matching by DIC
relates the two
Stereo Matching
Tracking
3-D shapes
Stereo Tracking through time
Lens Distortion
Time t
3-D
Strain Computation
Principle Matching by
stereo-correlation
Conclusions
Limitations, Benefits Temporal matching
Acknowledgments (tracking)
Matching
by stereo-correlation
3-D
Time t’
Principle
Introduction
Purpose
Basic Idea
▫ Monocular (cyclopean)
Camera Convention
Principle
vision cannot resolve
Calibration
for the scale of
Model
Estimation
objects
Matching by DIC
Stereo Matching
Tracking
Stereo Tracking
Lens Distortion
Strain Computation
Principle
Conclusions
Acknowledgments
Principle
Introduction
Purpose
Basic Idea
▫ Recovering the three-
Camera Convention
Principle
dimensional structure
Calibration
of the environment
Model
Estimation
using two imaging
Matching by DIC
sensors is called Object coordinate
Stereo Matching
Tracking
stereo-triangulation system
Stereo Tracking
Lens Distortion
Strain Computation
Principle
Conclusions
Acknowledgments
Principle
Introduction
Purpose
Basic Idea
Camera Convention
Principle
▫ Stereo-
Calibration
triangulation
Model
Estimation
requires computing
Matching by DIC
the intersection of ? ?
Stereo Matching
Tracking
two optical rays
Stereo Tracking
Lens Distortion ▫ Feasible only if
Strain Computation these rays are
Principle
formulated in a
Conclusions
common coordinate
Acknowledgments
system
▫ We need to model
and calibrate the
stereo-rig ?
Calibration
Introduction
Purpose
Basic Idea
Camera Convention
Principle

Calibration
Estimation
Basic Idea Tracking

Matching by DIC
Tracking
Stereo Tracking
Lens Distortion
Strain Computation
Principle
Model Principle
Conclusions
Estimation
Acknowledgments
Conclusions
Acknowledgments
Model
Introduction
Purpose
Basic Idea
Camera Convention
Principle
Calibration
Model
Estimation
Matching by DIC
Stereo Matching
Tracking
Stereo Tracking Perspective projection
Lens Distortion
Strain Computation
Principle
Conclusions
Rigid transform Perspective projection
Acknowledgments
▫ Model: one extrinsic rigid transformation,

two intrinsic perspective projections
Data Acquisition
Introduction
Purpose
Basic Idea
▫ Acquire pairs of
Camera Convention
Principle
images of a special
Calibration
target undergoing
Model
Estimation
arbitrary motions
Matching by DIC
▫ Calibration is shape
Stereo Matching
Tracking
measurement process
Stereo Tracking
Lens Distortion (bundle-adjustment)
Strain Computation
Principle
Conclusions
Acknowledgments
Parameter Estimation
Introduction
Purpose
Basic Idea
▫ Dedicated pattern
Camera Convention
Principle
recognition algorithm
Calibration
Model
Estimation
Matching by DIC
Stereo Matching
Tracking
Stereo Tracking
Lens Distortion
Strain Computation
Principle
▫ Identification of the
Conclusions
transfer function
Acknowledgments
Calibration target examples

Left: raw image, right: automatic pattern
recognition
Target Pairs of
3-D points Stereo Rig 2-D points
(approx. (auto-recognized)
known) Transfer function
Parameter Estimation Techniques
Introduction
Purpose
Basic Idea
▫ Bundle Adjustment
▫ No knowledge of calibration target required
Camera Convention
Principle
Calibration aside from a scale.

Model
Estimation ▫ Shape of the target is measured during calibration.
Matching by DIC ▫ Typically, a relatively flat target is used for
Stereo Matching
Tracking initialization through a linear method.
Stereo Tracking
Lens Distortion ▫ Generally regarded as the optimal method to
Strain Computation calibrate cameras.
Principle
Conclusions
▫ Complex to implement due to large equation
Acknowledgments
systems.
▫ Require block matrix solvers to efficiently invert
large equation systems.
▫ Can provide confidence margins on all
parameters estimated.
Parameter Estimation Techniques
Introduction
Purpose
Basic Idea
▫ There are only two constraints required
Camera Convention
Principle to accomplish calibration.
Calibration
Model
Estimation
Matching by DIC
1. The calibration target must not deform
Stereo Matching
Tracking between images.
Stereo Tracking
Lens Distortion
2. A distance between two points on the
target must be known.
Strain Computation
Principle
Conclusions
Acknowledgments
Matching by DIC
Introduction
Purpose
Basic Idea
Camera Convention
Principle

Calibration
Estimation
Basic Idea Tracking

Matching by DIC
Tracking
Stereo Tracking
Lens Distortion
Strain Computation
Principle
Model Principle
Conclusions
Estimation
Acknowledgments
Conclusions
Acknowledgments
3-D Reconstruction
Introduction
Purpose
Basic Idea
▫ 3-D Shape Measurement given a pair of images
Camera Convention
Principle
Calibration
Model
Estimation
Matching by DIC
Stereo Matching
Tracking
Stereo Tracking
Lens Distortion
Time t
3-D
Strain Computation
stereo-correlation
Conclusions
Matching
3-D
Time t’
3-D Reconstruction
Two steps:
Introduction
Purpose
Basic Idea
Camera Convention
Principle
1. Stereo-correlation
Calibration
2. Stereo-triangulation
Model
Estimation
Matching by DIC
Stereo Matching
Tracking
Stereo Tracking
Lens Distortion
Strain Computation
Principle
Conclusions
Acknowledgments
Tracking: Similar to 2-D DIC
Introduction
Purpose
Basic Idea
▫ Allow to
Camera Convention
Principle
relate 3-D
Calibration
shapes
Model
Estimation
through
Matching by DIC
time
Stereo Matching
Tracking
▫ Algorithms
Stereo Tracking
Lens Distortion similar to
Strain Computation
2-D DIC Time t
3-D
stereo-correlation
Conclusions
Matching
3-D
Time t’
Combining everything
Introduction
Purpose
Basic Idea
▫ The technology
Camera Convention
Principle
behind 3-D DIC
Calibration
▫ Stereo-correlation
Model
Estimation
▫ Tracking by
correlation
Matching by DIC
Stereo Matching ▫ Stereo-
Tracking
Stereo Tracking triangulation
Lens Distortion
(needs Time t
Strain Computation calibration) 3-D
stereo-correlation
Conclusions
Matching
3-D
Time t’
Combining everything
Introduction
Purpose
Basic Idea
▫ Example
Camera Convention
Principle
Calibration
Model
Estimation
Matching by DIC
Stereo Matching
Tracking
Stereo Tracking
Lens Distortion
Strain Computation
Principle
Conclusions
Acknowledgments
Strain Computation
Introduction
Purpose
Basic Idea
Camera Convention
Principle

Calibration
Estimation
Basic Idea Tracking

Matching by DIC
Tracking
Stereo Tracking
Lens Distortion
Strain Computation
Principle
Model Principle
Conclusions
Estimation
Acknowledgments
Conclusions
Acknowledgments
Principle
Introduction
Purpose
Basic Idea
▫ Strains are computed from the measured
Camera Convention
Principle
displacements of object points
Calibration
(strains   displacements)
Model
Estimation
Matching by DIC
Stereo Matching
Tracking
Stereo Tracking
Lens Distortion
Strain Computation
Principle
Conclusions 3-D displacement

Acknowledgments
3-D shape at time t’
3-D shape at time t

Principal
Introduction
Purpose
Basic Idea
▫ Strains are only defined in the tangential
Camera Convention
Principle plane of the surface.
Calibration
Model
▫ The DIC data can be converted to a
triangular mesh.
Estimation
Matching by DIC
Stereo Matching
Tracking ▫ Since triangles are planar by definition, it
is simple to compute the strain on each
Stereo Tracking
Lens Distortion
Strain Computation
Principle
triangle.
Conclusions
▫ Same equations as found
Acknowledgments
in FEM.
Smoothing
Introduction
Purpose
Basic Idea
▫ The strains computed on each triangle are
Camera Convention
Principle noisy, unless the triangles are fairly large.
Calibration
Model
▫ The strains are normally smoothed using
low-pass filters.
Estimation
Matching by DIC
Stereo Matching
Tracking ▫ Low-pass filtering decreases spatial
resolution.
Stereo Tracking
Lens Distortion
Strain Computation
Principle
Conclusions
Acknowledgments
Conclusions
Introduction
Purpose
Basic Idea
Camera Convention
Principle

Calibration
Estimation
Basic Idea Tracking

Matching by DIC
Tracking
Stereo Tracking
Lens Distortion
Strain Computation
Principle
Model Principle
Conclusions
Estimation
Acknowledgments
Conclusions
Summary
Acknowledgments
Summary
Introduction
Purpose
Basic Idea
▫ Matching process – identical to 2D
Camera Convention
Principle
▫ Combining matching with triangulation
▫ Combining triangulation with tracking
Calibration
Model
Estimation
Matching by DIC
Stereo Matching
Tracking
Stereo Tracking
Lens Distortion
Strain Computation
Principle
Conclusions
Acknowledgments
Acknowledgements
Introduction
Purpose
Basic Idea
Camera Convention
Principle
Calibration
▫ We thank Prof. Jean-José Orteu
Model from École des Mines d’Albi,
Estimation
France, for providing us with
Matching by DIC
Stereo Matching
multimedia content
Tracking
Stereo Tracking
Lens Distortion
Strain Computation
Principle
Conclusions
Acknowledgments
▫ We gratefully acknowledge the
University of South Carolina,
Mechanical Engineering Dept.,
support for providing us with
multimedia content
Questions?
Introduction
Purpose
Basic Idea
Camera Convention
Principle
Calibration
Model
Estimation
Matching by DIC
Stereo Matching
Tracking
Stereo Tracking
Lens Distortion
Strain Computation
Principle
Conclusions
Acknowledgments
3D Image Correlation
Procedures and
Practicalities
Outline
Camera Setup
Camera Selection
Setup
Synchronization
Focusing
Camera Setup System Calibration

Specimen Preparation
Preparation Methods Camera Selection Why?
Basic Requirements
What Not To Do
Guidelines
Setup Calibration as Shape
Measurement
System Calibration
Synchronization
Suitable Targets
Why
Calib: Shape Measurem.
Focusing
Suitable Targets Procedures
Procedures
Specimen Prep.
Conclusions
Preparation Methods Conclusions
Summary
Acknowledgement Summary
Basic Requirements
Acknowledgement
What Not To Do
Guidelines
Outline
Camera Setup
Camera Selection
Setup
Synchronization
Focusing

Basic Requirements
What Not To Do
Guidelines
Measurement
System Calibration
Synchronization
Suitable Targets
Why
Focusing
Procedures
Specimen Prep.
Conclusions
Summary
Basic Requirements
Acknowledgments
What Not To Do
Guidelines
Camera Selection
Camera Setup
Camera Selection
Setup
▫ Cameras available for different time
Synchronization
Focusing scales
Preparation Methods
▫ Quasi-static (10-30 fps)
Basic Requirements
What Not To Do ▫ Medium speed (500 fps)
Guidelines
System Calibration
▫ High-speed (50,000 fps)
Why
Calib: Shape Measurem. ▫ Ultra-high speed (1,000,000 fps)
Suitable Targets
Procedures
▫ Choose sensitive, monochrome cameras
Conclusions
Summary
Acknowledgement
▫ Pixel size
▫ CCD vs. CMOS
Camera Setup
▫ Cameras must not move relative to each

Camera Setup
Camera Selection
Setup
Synchronization
Focusing
other
Specimen Preparation ▫ Tolerable camera motion depends on
magnification
Preparation Methods
Basic Requirements
What Not To Do
Guidelines
▫ Example:
System Calibration
Why ▫ 1:1 magnification
Suitable Targets
Procedures
▫ 5 micron pixel size
Conclusions
▫ Detectable motion 0.01 pixel
▫ 50 nanometer relative camera motion is
Summary
Acknowledgement
detectable!!!
▫ For high magnifications, a rigid camera
setup is paramount
Camera Setup
Camera Setup
Camera Selection
Setup
Synchronization
Focusing
▫ Mount cameras on
Preparation Methods
Basic Requirements
rigid support
What Not To Do
Guidelines
▫ Tie down camera
System Calibration
Why cables
Suitable Targets
Procedures
Conclusions
▫ Use vibration isolation for

Summary
Acknowledgement
high-magnification work
Minimizing Bias & Noise
Camera Setup
Camera Selection
Setup ▫ Bias: systematic deviations from the
correct result
Synchronization
Focusing
▫ Noise: random, zero-mean deviations

Preparation Methods
Basic Requirements
What Not To Do
Guidelines
from the correct result
System Calibration
Why
▫ Noise and bias are present for location
Suitable Targets (in-plane, out-of-plane), displacement,
and strain
Procedures
Conclusions
▫ Bias can be reduced or eliminated with

Summary
Acknowledgement
proper setup and parameters

▫ Noise is unavoidable, but can be
minimized with careful setup
Camera Setup
Camera Selection
Setup ▫ Noise in displacement and strain is
strongly dependent on stereo angle
Synchronization
Focusing
Preparation Methods
Basic Requirements
What Not To Do
Guidelines
System Calibration
Why
Suitable Targets
Procedures
Conclusions
Summary
Acknowledgement
Camera Setup
Camera Selection
Setup ▫ Noise is lowest near the optical axis
Synchronization
Focusing
Preparation Methods
Basic Requirements
What Not To Do
Guidelines
System Calibration
Why
Suitable Targets
Procedures
Conclusions
Summary
Acknowledgement
Camera Setup
Camera Selection
Setup ▫ For short lenses (8mm, 12mm), use a
stereo angle of at least 25º
Synchronization
Focusing
▫ With longer lenses (35mm, 70mm), use

Preparation Methods
Basic Requirements
What Not To Do
Guidelines
at least 10-15º
System Calibration
Why
▫ If you must use small stereo angles,
Suitable Targets keep the AOI to the center of the
images
Procedures
Conclusions
Summary
Acknowledgement
Camera Synchronization
Camera Setup
Camera Selection
Setup
▫ Stereo cameras must be synchronized
Synchronization
Focusing
▫ Required accuracy depends on speed of
Preparation Methods
event
Basic Requirements
What Not To Do
Guidelines
▫ Synchronization accuracy should be small
System Calibration fraction of exposure time
Why
Suitable Targets
Procedures
Conclusions
Summary
Acknowledgement
Focusing cameras
Camera Setup
Camera Selection
Setup
▫ The entire sample should be focused
Synchronization
Focusing during the entire test
Preparation Methods
▫ It can be difficult to focus a camera
correctly when the depth-of-field is large
Basic Requirements
What Not To Do
Guidelines
System Calibration
Why
Small aperture (high f-number):
Suitable Targets
Procedures
Conclusions
Summary
Acknowledgement
Specimen
Camera Focal plane
Depth-of-field
Focusing cameras
Camera Setup
Camera Selection
Setup
▫ The entire sample should be focused
Synchronization
Focusing during the entire test
Preparation Methods
▫ It can be difficult to focus a camera
correctly when the depth-of-field is large
Basic Requirements
What Not To Do
Guidelines
System Calibration
Why
Large aperture (small f-number):
Suitable Targets
Procedures
Conclusions
Summary
Acknowledgement
Specimen
Camera Focal plane
Depth-of-field
Focusing cameras
Camera Setup
Camera Selection
Setup
▫ Focus with open aperture
Synchronization
Focusing
▫ Close aperture after focusing
Preparation Methods
Basic Requirements
What Not To Do
Guidelines
System Calibration
Why
After closing aperture:
Suitable Targets
Procedures
Conclusions
Summary
Acknowledgement
Specimen
Camera Focal plane
Depth-of-field
Aperture, Exposure, Lighting
Camera Setup
Camera Selection
Setup
▫ Use a short enough exposure to freeze
▫ Calibration is frequently the bottleneck
Synchronization
Focusing
Preparation Methods ▫ Apertures in the middle are best
Basic Requirements
What Not To Do
Guidelines
▫ DOF concerns
System Calibration ▫ Diffraction limit & pixel size
Why
Suitable Targets
▫ Difficult cases
Procedures
Conclusions
▫ Wet/shiny specimen
Summary
Acknowledgement
▫ Metal & glare
▫ Transparent/translucent
▫ 3D and textures
Checking for Contaminations
Camera Setup
Camera Selection
Setup
▫ Contaminations on sensor, IR filter or lens
Synchronization
Focusing can lead to small local bias in data
Preparation Methods
▫ The image needs to be checked if
contaminations are present
Basic Requirements
What Not To Do
Guidelines
System Calibration
Why
Suitable Targets
Procedures
Conclusions Contaminations
cannot be spotted
Summary
Acknowledgement
in speckle image
Checking for contaminations
▫ Contaminations can be seen by

Camera Setup
Camera Selection
Setup
Synchronization
Focusing
▫ Completely defocusing camera
Specimen Preparation ▫ Using a uniform gray background
▫ Move camera
Preparation Methods
Basic Requirements
What Not To Do
Guidelines
▫ If spot stays fixed, contamination present
▫ If contaminations are present, turn lens
System Calibration
Why
Suitable Targets
Procedures
▫ If spot moves with lens rotation,
Conclusions
contamination is on lens
Summary
Acknowledgement
▫ If spot stays in place, spot is on sensor or IR
filter
▫ Use compressed air to blow dust particles
away
▫ Always use lens caps and camera covers to
avoid contaminations in the first place
Summary
Camera Setup
Camera Selection
Setup
▫ Select appropriate camera for speed
Synchronization
Focusing requirement and resolution requirement
Preparation Methods
▫ Use sensitive, low-noise cameras
Basic Requirements
What Not To Do
Guidelines
▫ Set cameras up on rigid supports and tie
System Calibration all cables down
Why
Suitable Targets ▫ Use vibration isolation for high
Procedures
Conclusions
magnifications
▫ Focus at open aperture and close for test
Summary
Acknowledgement
▫ Always check for contaminations against

uniform background or by blurring image
Outline
Camera Setup
Camera Selection
Setup
Synchronization
Focusing

Basic Requirements
What Not To Do
Guidelines
Measurement
System Calibration
Synchronization
Suitable Targets
Why
Focusing
Procedures
Specimen Prep.
Conclusions
Summary
Basic Requirements
Acknowledgments
What Not To Do
Guidelines
Preparation Methods
Camera Setup
Camera Selection
Setup
▫ Uniform random pattern must be applied
Synchronization
Focusing
▫ Large range of application methods:
▫ Lithography and vapor deposition
Preparation Methods
Basic Requirements
What Not To Do
Guidelines
▫ Toner powder on paint
System Calibration ▫ Spray painting
Why
Suitable Targets
▫ Toothbrush technique
Procedures
▫ Electrochemical etching
Conclusions
Summary
Acknowledgement
▫ Screen printing
▫ Adhesive-backed vinyl (e.g., 3M Controltac)
▫ Stencils
▫ …
Basic Requirements
Camera Setup
Camera Selection
Setup
▫ Pattern must deform with sample
▫ No slip for stick-on patterns
Synchronization
Focusing
Preparation Methods
▫ Dye sample for large deformations (100%-
Basic Requirements
What Not To Do
800%)
▫ Must hold up to testing conditions
Guidelines
(temperature, moisture, acceleration etc.)

System Calibration
Why
Suitable Targets
Procedures ▫ Pattern must not reinforce sample
Conclusions
Summary
▫ Use dye penetrant developer for
Acknowledgement
measurements on very thin metal foils
Basic Requirements
Camera Setup
Camera Selection
Setup
▫ It is absolutely critical to avoid specular
Synchronization
Focusing reflections
Preparation Methods
▫ If sensor is saturated, the signal is chopped off
Basic Requirements
What Not To Do ▫ Accurate matching no longer possible
Guidelines
System Calibration
▫ Large artificial spikes in strain
▫ Use matte paints
Why
Suitable Targets
▫ Whenever possible, use back lighting for

Procedures
Conclusions
Summary
Acknowledgement
transparent or semi-transparent samples
▫ Avoid thick paint drops
▫ Use diffuse lighting Paint drop
Aliasing
Camera Setup
Camera Selection
Setup
Synchronization
Focusing
Preparation Methods
Basic Requirements
What Not To Do
Guidelines
System Calibration
Why
Suitable Targets
Procedures
Conclusions
Summary
Acknowledgement
▫ Pattern elements are too fine, causing…

Aliasing
Camera Setup
Camera Selection
Setup
Synchronization
Focusing
Preparation Methods
Basic Requirements
What Not To Do
Guidelines
System Calibration
Why
Suitable Targets
Procedures
Conclusions
Summary
Acknowledgement
▫ Same patch!
What makes a good pattern?
Camera Setup
Camera Selection
Setup
▫ Small speckles approximately 3-5 pixels in
Synchronization
Focusing size
Preparation Methods
▫ Permits analysis with small subset
Basic Requirements
What Not To Do ▫ High spatial resolution
Guidelines
System Calibration
▫ No preferred orientation
Why
Suitable Targets
▫ Uniform (check histogram)
Procedures
Conclusions
▫ Matte
▫ No big paint blobs or uncovered areas
Summary
Acknowledgement
Specimen Preparation Summary
Camera Setup
Camera Selection
Setup
▫ Pattern must deform with sample and not
Synchronization
Focusing reinforce specimen
Preparation Methods
▫ Pattern must not degrade (too much)
during test
Basic Requirements
What Not To Do
Guidelines
System Calibration ▫ A wide variety of application methods

Why
Suitable Targets
exist
Procedures
Conclusions
▫ Pattern should be uniform
▫ Specular reflection must be avoided
Summary
Acknowledgement
▫ Dense speckle pattern is desirable to

achieve maximum spatial resolution
Outline
Camera Setup
Camera Selection
Setup
Synchronization
Focusing

Basic Requirements
What Not To Do
Guidelines
Measurement
System Calibration
Synchronization
Suitable Targets
Why
Focusing
Procedures
Specimen Prep.
Conclusions
Summary
Basic Requirements
Acknowledgments
What Not To Do
Guidelines
Camera and System Calibration
Camera Setup
Camera Selection
Setup
▫ The internal camera parameters have to
Synchronization
Focusing be calibrated (focal length, image center,
Preparation Methods
distortions, skew)
Basic Requirements
What Not To Do
Guidelines
System Calibration ▫ The relative orientation of the stereo

Why
Suitable Targets
cameras has to be known for triangulation
Procedures
Conclusions
▫ Typically, all parameters are calibrated at

Summary
Acknowledgement
the same time using a calibration target

Calibration as Shape Measurement
Camera Setup
Camera Selection
Setup
Synchronization
Focusing
Preparation Methods
Camera calibration is a shape

Basic Requirements
What Not To Do
Guidelines
System Calibration
Why measurement of the calibration
target!!!
Suitable Targets
Procedures
Conclusions
Summary
Acknowledgement
Calibration Target
Camera Setup
Camera Selection
Setup Does one need a precision calibration
target to calibrate a stereo system?
Synchronization
Focusing
▫ Internal camera parameters can be

Preparation Methods
Basic Requirements
What Not To Do
Guidelines
calibrated from a completely arbitrary
calibration target (only constraint: target
System Calibration
Why
Suitable Targets
Procedures
must be rigid)
Conclusions
Summary
▫ Rotation and direction vector between
Acknowledgement
cameras can also be calibrated from
arbitrary target
▫ Scale of translation vector can only be
recovered from known distance on target
Calibration Target
Camera Setup
Camera Selection
Setup
▫ Target does not need to be precisely flat
Synchronization
Focusing or have a known shape to accurately
Preparation Methods
calibrate a stereo system
▫ A distance between two fiducial points has
Basic Requirements
What Not To Do
Guidelines
System Calibration
to be known accurately to recover the
Why
Calib: Shape Measurem. absolute scale
Suitable Targets
Procedures
Conclusions
Summary
Acknowledgement
Calibration Procedure
Camera Setup
Camera Selection
Setup
▫ Typically, a fairly flat calibration target is
Synchronization
Focusing used (simple linear solutions to the
Preparation Methods
calibration problem exist for planar
Basic Requirements
What Not To Do
targets)
Guidelines
System Calibration
▫ Multiple views of the target (15) are
Why
Calib: Shape Measurem. acquired
Suitable Targets
Procedures
▫ A bundle adjustment procedure
determines shape of target and calibration
Conclusions
Summary
Acknowledgement
parameters
What orientations of the target
make a good calibration
sequence?
Importance of Rotations
Camera Setup
Camera Selection
Setup
▫ Calibration equivalent
Synchronization
Focusing to shape
Specimen Preparation measurement Stereo-base
Preparation Methods
▫ Reliable shape
Basic Requirements
What Not To Do
Guidelines
System Calibration
measurement only
Why
Calib: Shape Measurem. possible for large
Suitable Targets
Procedures
stereo-baseline
▫ Rotations serve to
Conclusions
Summary
Acknowledgement
increase baseline
Calibration Procedure
Camera Setup
Camera Selection
Setup
▫ Coded targets are used
Synchronization
Focusing for automatic
Preparation Methods
extraction
Basic Requirements
What Not To Do ▫ Coded markers must be
visible in all views
Guidelines
System Calibration
Why
▫ Avoid glare
Suitable Targets
Procedures
▫ Use short exposure
Conclusions
Summary times to freeze target
motion (rule of thumb:
Acknowledgement
1/focal length seconds)

Calibrating Image Distortions
Camera Setup
Camera Selection
Setup
▫ Distortions are described by high-order
Synchronization
Focusing
polynomials in the distance from the image
center
Preparation Methods
Basic Requirements
▫ Distortions are normally small in central image area
What Not To Do
Guidelines
▫ Distortions rapidly increase at image boundaries
System Calibration ▫ It is critical that the calibration grid covers the
entire image, particularly close to the
Why
boundaries, to accurately calibrate lens

Suitable Targets
Procedures
Conclusions distortions
Summary
Acknowledgement
▫ Short focal-length (and inexpensive) lenses
require higher-order polynomials and can only be
calibrated with a large number of images
High-Magnification Calibration
Camera Setup
Camera Selection
Setup
▫ For high magnifications, glass grids are
Synchronization
Focusing used
Preparation Methods
▫ Background lighting
Basic Requirements
What Not To Do ▫ Positioning
Guidelines
System Calibration
Why
Suitable Targets
Procedures
▫ Strong tilt becomes very difficult at
smaller FOV’s
Conclusions
Summary
Acknowledgement
▫ “High-magnification” options
▫ DOF concerns
▫ Calibrating in front may be impossible
High-Speed Calibration
Camera Setup
Camera Selection
Setup
▫ Full resolution calibration
▫ Crop adjustment
Synchronization
Focusing
Preparation Methods ▫ DOF concerns
Basic Requirements
What Not To Do
Guidelines
▫ Calibrating in front may be impossible
System Calibration
Why
Suitable Targets
Procedures
Conclusions
Summary
Acknowledgement
Calibration Summary
▫ Calibration for camera parameters and

Camera Setup
Camera Selection
Setup
Synchronization
Focusing
relative orientation is required
▫ Calibration is a shape measurement of the target
▫ Target can be arbitrarily shaped and unknown
Preparation Methods
Basic Requirements
What Not To Do
Guidelines ▫ A length scale on the target has to be known
System Calibration
Why
▫ Large rotations of the target are required for
Suitable Targets
accurate calibration
Procedures
▫ Distortions require that target fills entire image
▫ Short or inexpensive lenses require a large
Conclusions
Summary
number of images to calibrate distortions

Acknowledgement
▫ High-magnification
▫ High-speed
Conclusions
Camera Setup
Camera Selection
Setup
Synchronization
Focusing

Basic Requirements
What Not To Do
Guidelines
Measurement
System Calibration
Synchronization
Suitable Targets
Why
Focusing
Procedures
Specimen Prep.
Conclusions
Summary
Basic Requirements
Acknowledgement
What Not To Do
Guidelines
Conclusion
Camera Setup
Camera Selection
Setup
▫ B/W cameras with sufficient frame rate to
Synchronization
Focusing capture the event
Preparation Methods
▫ Rigid setup, well focused and clean
optics/sensor
Basic Requirements
What Not To Do
Guidelines
System Calibration ▫ Synchronized cameras

Why
Suitable Targets ▫ Uniform, dense speckle patterns with no
Procedures
Conclusions
specular reflections
▫ Sufficient number of calibration images in
Summary
Acknowledgement
suitable orientations
▫ Good experimental techniques!
Camera Setup
Camera Selection
Setup
Synchronization
Focusing
Preparation Methods
Basic Requirements
What Not To Do
Questions?
Guidelines
System Calibration
Why
Suitable Targets
Procedures
Conclusions
Summary
Acknowledgement

Introduction To Digital Image Correlation - ResearchGate

Transféré par

Informations du document

Titre original

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

Introduction To Digital Image Correlation - ResearchGate

Transféré par

Droits d'auteur :

Formats disponibles

SEM 2009 Workshop

Digital Image Correlation:

University of South Carolina

Introduction Strain Computation

Introduction Strain Computation

▫ Large range of size scales

Matching by DIC ▫ Monocular (cyclopean)

Time t Time t’ Time t”

Strain Computation the surface has a Non-repetitive

Introduction Strain Computation

Strain Computation Image, on screen

Strain Computation Image, in memory Image, on screen

100 100 100 0 0 0 100 100 100

100 100 100 0 0 0 100 100 100

100 100 100 0 0 0 100 100 100

100 100 100 0 0 0 100 100 100

100 100 100 0 0 0 100 100 100

100 100 100 100 0 0 0 100 100

100 100 100 100 0 0 0 100 100

100 100 100 100 0 0 0 100 100

100 100 100 100 0 0 0 100 100

100 100 100 100 0 0 0 100 100

Strain Computation Image before motion Image after motion

100 100 100 0 0 0 100 100 100 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 100 100 100 100 0 0 0 100 100

Image before motion Image after motion

n: subset size (9 in our example)

Principle (100  0) 2  (0  0) 2  (0  0) 2  (0  0) 2  (100  0) 2 

Image before motion

Image after motion

0 0 0 0 0 0 0 0 0 100 100 100 100 0 0 0 100 100

Image before motion

Image after motion

0 0 0 0 0 0 0 0 0 100 100 100 100 0 0 0 100 100

Strain Computation Image before motion Image after motion

103 101 99 2 0 1 105 100 96 99 100 101 102 3 0 2 100 102

103 96 99 0 2 2 102 103 98 0 1 3 3 2 0 1 2 0

0 0 2 0 3 0 2 0 0 101 100 100 103 0 2 1 102 101

98 101 102 0 1 0 96 97 102 97 99 100 101 3 2 0 97 101

97 98 103 0 2 0 103 98 100 101 103 98 101 0 1 1 99 96

102 99 101 2 0 0 104 102 101 102 99 96 103 2 3 3 102 100

103 101 99 52 0 1 55 100 96

98 101 102 52 1 0 46 97 102

102 99 101 48 0 0 54 102 101

robust correlation function

▫ Optimization algorithms require the criterion

to be continuous over its parameter space

Raw image Bi-linear B-spline interp. Bi-cubic B-spline interp.

Introduction Strain Computation

Conclusions ▫ Matching process

▫ We thank Prof. Jean-José Orteu from

▫ We gratefully acknowledge the

Introduction Matching by DIC

Basic Idea Tracking

Introduction Matching by DIC

Basic Idea Tracking

3-D shape of letter „R‟ from “E

Introduction Matching by DIC

Basic Idea Tracking

▫ Model: one extrinsic rigid transformation,

Calibration target examples

Calibration aside from a scale.

Introduction Matching by DIC