Top Banner
3D Digitization Project: Shape from Distortion BY: Agam A. Nugroho Vanya V.Valindria Eng Wei Yong VIBOT 4 2011
31

Shape from Distortion - 3D Digitization

Jun 20, 2015

Download

Technology

Vanya Vabrina

3D digitization, reconstruction of an object by shape from distortion using Tarini Method and Simple Deflectometry
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Page 1: Shape from Distortion - 3D Digitization

3D Digitization Project:

Shape from DistortionBY:

Agam A. NugrohoVanya V.Valindria

Eng Wei Yong

VIBOT 42011

Page 2: Shape from Distortion - 3D Digitization

Outline

Introduction Methods Procedure Tarini Method Simple Deflectometry

Result Conclusion

Page 3: Shape from Distortion - 3D Digitization

Introduction

3D image reconstruction main issues in computer vision.

Many technique: Shading Texture Stereoscopy Structured Light Contour

?

?

?

?

Shape from distortion reconstruct the 3D shape of the mirror from its reflected images

Page 4: Shape from Distortion - 3D Digitization

Aims

Obtain the relationship that is useful in 3D surface reconstruction for specular objects.

depth map

normal map

3D surface

Page 5: Shape from Distortion - 3D Digitization

Tarini Method

Deduce 3D shape of the target object by looking at the way it distorts patterns from a monitor.

Page 6: Shape from Distortion - 3D Digitization

Deflectometry

This method works by measuring a surface slope of an optical beam which is deflected by the surface.

Page 7: Shape from Distortion - 3D Digitization

Experiment

Devices: Monitor: DELL 14” ,

flat monitor Camera: UI-1225

LE-C. CMOS 1/3”. 752 x 480

Object : Small specular object

Page 8: Shape from Distortion - 3D Digitization

Camera Calibration

Bouget toolbox using checkerboardParameter Value

Focal Length fc = [ 4017.658 3145.87 ] ± [ 267.5 380.1]

Principal Point cc = [ 375.5 239.5 ] ± [ 0 0 ]

Skew alpha_c = 0 => angle of pixel axes = 90 degrees

Distortion kc = [ -0.88 -30.02 0.01 0 0 ] ± [ 1.57 134 0.03 0.01 0]

Pixel Error err = [ 2.3 1.9 ]

Page 9: Shape from Distortion - 3D Digitization

Generate Matte Pattern

20 40 60 80 100120

50

100

150

200

250

300

350

400

450

0 20 40 60 80 100 120 1400

50

100

150

200

250

300

1 stripe pattern

Duplicate in row and column

Page 10: Shape from Distortion - 3D Digitization

Generate Matte Pattern

20 40 60 80 100120

50

100

150

200

250

300

350

400

450

Vertical pattern Horizontal pattern

Diagonal 45 pattern

Diagonal 135 pattern

Page 11: Shape from Distortion - 3D Digitization

Project the Matte Patterns Using Specular Object

For normali-zation

Other orientation and positionsX

Page 12: Shape from Distortion - 3D Digitization

Matte Extraction

Using perfect mirror object

Normalization

Page 13: Shape from Distortion - 3D Digitization

Matte Extraction

Which stripes( y)??!

Diophantine Equations:

Page 14: Shape from Distortion - 3D Digitization

Change Method!!

Simple Deflectometry

Page 15: Shape from Distortion - 3D Digitization

Experimental Setup - Set Exposure Time

Curve from 1 line

Clipped Peak Saturation

Page 16: Shape from Distortion - 3D Digitization

Set Exposure Time

Exposure_time = 10.5 Pixel_clock = 30; Frame rate =

max

0 20 40 60 80 100 120 140 160 180 2000

50

100

150

200

250

Curve from 1 line

Page 17: Shape from Distortion - 3D Digitization

Color Calibration

Color projected by screen ≠ perceived by camera Generate Ramp Images in RGB

Red Ramp

Green Ramp Blue Ramp

Page 18: Shape from Distortion - 3D Digitization

Color Calibration

Project ramp images to mirror object

Red Ramp

Green Ramp Blue Ramp

Page 19: Shape from Distortion - 3D Digitization

Color Calibration

Response Curve for each ramp image

0 20 40 60 80 100 120 140 160 1800

20

40

60

80

100

120

140

160

0 20 40 60 80 100 1200

50

100

150

200

250

0 20 40 60 80 100 120 140 160 1800

20

40

60

80

100

120

140

160

RED Ramp Image

RED response curve : linear

Noisy response from other channels

Page 20: Shape from Distortion - 3D Digitization

Color Calibration

Find minimum and maximum values of perceived color values in each channel

MIN MAX

Red 17 132

Green 26 182

Blue 21 247

Page 21: Shape from Distortion - 3D Digitization

Color Calibration

Normalization of the linear Response Curve

0 50 100 150 200 2500

20

40

60

80

100

120

140Response Curve from Red Channel

0 50 100 150 200 2500

50

100

150

200Response Curve from Green Channel

0 50 100 150 2000

50

100

150

200

250Response Curve from Blue Channel

0 50 100 150 200 2500

50

100

150

200

250

300Normalized Response Curve - RED

0 50 100 150 2000

50

100

150

200

250

300Normalized Response Curve - GREEN

0 50 100 150 200 2500

50

100

150

200

250

300Normalized Response Cruve - BLUE

Page 22: Shape from Distortion - 3D Digitization

Geometrical Diagram

Page 23: Shape from Distortion - 3D Digitization

Mirror with various shift

• Vertical pattern with 1˚ angle increment• Obvious pattern shift observed

Θ = 0 Θ = 1

Θ = 3

Θ = 2

Θ = 4 Θ = 5

Page 24: Shape from Distortion - 3D Digitization

Shifted Color Curve

0 200 4000

50

100

150

200

250Theta = 0

0 200 4000

50

100

150

200

250Theta = 1

0 200 4000

50

100

150

200

250Theta = 2

0 200 4000

50

100

150

200

250Theta = 3

0 200 4000

100

200

300Theta = 4

0 200 4000

100

200

300Theta = 5

0 200 4000

100

200

300Theta = 6

0 200 4000

100

200

300Theta = 7

0 200 4000

100

200

300Theta = 8

0 200 4000

100

200

300Theta = 9

0 200 4000

100

200

300Theta = 10

0 200 4000

100

200

300Theta = 11

0 200 4000

50

100

150

200

250Theta = 12

0 200 4000

50

100

150

200

250Theta = 13

Page 25: Shape from Distortion - 3D Digitization

Theta Vs Shift

Page 26: Shape from Distortion - 3D Digitization

Shifted Color Curve

0 200 4000

50

100

150

200

250Theta = 0

0 200 4000

50

100

150

200

250Theta = 1

0 200 4000

50

100

150

200

250Theta = 2

0 200 4000

50

100

150

200

250Theta = 3

0 200 4000

100

200

300Theta = 4

0 200 4000

100

200

300Theta = 5

0 200 4000

100

200

300Theta = 6

0 200 4000

100

200

300Theta = 7

0 200 4000

100

200

300Theta = 8

0 200 4000

100

200

300Theta = 9

0 200 4000

100

200

300Theta = 10

0 200 4000

100

200

300Theta = 11

0 200 4000

50

100

150

200

250Theta = 12

0 200 4000

50

100

150

200

250Theta = 13

Page 27: Shape from Distortion - 3D Digitization

Θ Versus α

Angle

α

1 1.692 1.843 1.784 1.875 1.546 1.557 1.618 1.819 1.8410 1.6511 1.5312 1.6913 1.84

Page 28: Shape from Distortion - 3D Digitization

Angle – Shift Relationship

θ= 1o

α = 1.7 rad

δ= 97 pixels

Page 29: Shape from Distortion - 3D Digitization

Conclusion

3D reconstruction of specular surface can be performed using shape from distortion method

In this project, we succeed to obtain the orientation-shift relationship using the flat surface

This result will be useful for extracting the depth from the specular surface

In the future it can be extend to a more complex shiny surface 3D reconstruction.

Page 30: Shape from Distortion - 3D Digitization

THANK YOU….

Page 31: Shape from Distortion - 3D Digitization

REFERENCES

  M.Tarini, et,al, 3D acquisition of mirroring objects using

striped patterns, Graphical Models 67 233–259.2005.

Hui-Liang Shen, et.al. Estimation of Optoelectronic Conversion Functions of Imaging Devices Without Using Gray Samples, Wiley Periodical. Volume 33, Number 2, April 2008.

V. Hanta. SOLUTION OF SIMPLE DIOPHANTINE EQUATIONS BY MEANS OF MATLAB. Institute of Chemical Technology, Prague.

Y. Francken. Metostructure Acquisition with Planar Illuminants.PhD Dissertation. University of Maastrich