Error Estimation in Digital Image Correlation Caused by Rigid Particles

1
Error Estimation in Digital Image Correlation
Caused by Rigid Particles

• By Xiaodan (Danna) Ke

2
Content

• Introduction
• Simulation Process
• Particle Extraction Algorithm
• Particle Translation Algorithm
• Preliminary Results
• Future Work

3
Introduction to Digital Image Correlation (DIC) 1
Speckle patterns
DIC is a non-contacting measurement method based
on tracking speckle patterns on material surface
before and after deformation to determine
displacement and strain fields.
Force
Reference state
Deformed state
4
Introduction to Digital Image Correlation 2
Speckle patterns

• Working Principles
• Calculate cross-correlation
• factor of subsets in both states
• Center locations of subsets at deformed state
  will be correlated with those at reference state
• Accuracy 0.02 pixels

Subset
F
Reference state
(u,v)
Deformed state
5
Introduction to Digital Image Correlation 3

• DIC assumes no difference bt. background and
  particles in pattern
• ?Background and particles deform the same way
• Problem caused by rigid particles
• ?Error estimation is needed

!
Stretch and translate
Only translate
Flexible particles
Rigid particles
Reference state
Deformed state
Deformed state
6
Simulation Process 1

• Rigid particles are introduced when applying DIC
  to biological materials
• Simulation method is used to evaluate errors
  induced by rigid particles

Mouse carotid vessel
0.5 mm
7
Simulation Process 2
Challenge part!

• Extract individual particles from reference image
• Translate extracted particles according to
  pre-assigned displacement
• Generate deformed image
• Use Vic-2D (a given DIC package) to calculate
  displacement and strain
• Compare result from DIC with pre-assigned
  displacement

8
Particle Extraction Algorithm 2

• Why to extract individual particles
• ?Calculate centroid location and obtain accurate
  displacement assignment

20
29
41
30
105
34
51
40
Centroid
85
16
50
120
10
38
50
99
u
u1
x
9
Particle Extraction Algorithm 3

• Local neighborhood searching algorithm
• Intensity threshold (60)
• Extract continuous pixels for individual particles

Start
24 32 32 33 41 39 36 29 23 30
38 41 58 74 68 52 34 27 34
44 66 100 124 116 75 38 29 39 44
75 130 158 141 89 46 31 38 40 64
96 124 118 70 38 29 31 39 41
55 62 65 45 32 25
10
Particle Extraction Algorithm 3

• Results

Original image
Extracted particles
11
Particle Translation Algorithm 1

• Calculate centroid in subpixels
• Assign displacement to particles
• Translate particles
• Integration
• Deform background pixels
• Fill holes based on interpolation
• Generate deformed image

12
Particle Translation Algorithm 2

• Results

Original image
Pure rigid motion 10 pixel in width direction
Pure 50 stretch in width direction
13
Particle Extraction Algorithm 3

• Results

Rigid particles
Flexible particles
Pure 50 stretch in width direction
14
Preliminary Results of Error Estimation
Deformation Ground truth Digital Image Correl. prediction Digital Image Correl. prediction
Deformation Ground truth Flexible particles Rigid particles
Pure translation 10.0 pixels, 0.0 11.0 pixels, 0.0046 10.8 pixels, 0.07
Pure stretch 50.0 62.57 61.89
Trans.stretch 10.0 pixels, 50.0 62.54 61.89
?
?
?
?
15
Future Work

• Debug codes
• Separate clogged particles
• Estimate errors for complicated deformation

16
Thank you!