-1- barcelona, june 27 th, 20132 nd itfg workshop thz signal & image processing
TRANSCRIPT
-1-Barcelona, June 27th, 20132nd ITFG Workshop
THz Signal & Image Processing
-2-Barcelona, June 27th, 20132nd ITFG Workshop
Outline
• SECTION 1: Alternative configurations
– Wide beam configuration (synthetic aperture processing)
– Tomographic configuration (tomosynthesis processing)
• SECTION 2: Challenge of data size
– Enhanced visualisation
– (Semi-) Automatic defect detection
-3-Barcelona, June 27th, 20132nd ITFG Workshop
Outline
• SECTION 1: Alternative configurations
– Wide beam configuration (synthetic aperture processing)
– Tomographic configuration (tomosynthesis processing)
• SECTION 2: Challenge of data amount
– Enhanced visualisation
– (Semi-) Automatic defect detection
-4-Barcelona, June 27th, 20132nd ITFG Workshop
Wide beam configuration
THz
THz
dept
h
azimuth
lenses
material under test
focusedbeam
unfocusedbeam
Top / overhead view
THz
THz
Beamwaist
• Small focused beam• Diffraction-limited• Data already focused• Lateral resolution: 1-3 mm
THz
dept
hazimuth
material under test
Top / overhead view
unfocusedbeamT
Hz
THz
THz
• No lenses needed• Wide unfocused beam (16º)• Data is focused in software• Lateral resolution: 4-7 mm
in-depthdefects
Focused FMCW Synthetic Aperture (SA) FMCW
-5-Barcelona, June 27th, 20132nd ITFG Workshop
3-D synthetic aperture imagery
Inserts - A-sandwich
100 GHz – Side A
Near surface
Rohacell core Honeycomb core
-6-Barcelona, June 27th, 20132nd ITFG Workshop
Debonds - A-sandwich - honeycomb core
Near surface
In depth
100 GHz – Side A
3-D synthetic aperture imagery
-7-Barcelona, June 27th, 20132nd ITFG Workshop
100 GHz – Side B
150 GHz – Side B
Near surface
Near surface
Debonds - A-sandwich - Rohacell core
3-D synthetic aperture imagery
-8-Barcelona, June 27th, 20132nd ITFG Workshop
Near surface
Debonds - C-sandwich - honeycomb core
100 GHz – Side B
150 GHz – Side B
3-D synthetic aperture imagery
-9-Barcelona, June 27th, 20132nd ITFG Workshop
100 GHz – Side A
Near surface
Impact (30 J) - Solid laminate - fibre glass
x-axis in mm
y-ax
is in
mm
20 40 60 80 100
10
20
30
40
50
60
70
80
90 0.55
0.6
0.65
0.7
0.75
0.8
0.85
0.9
0.95
1
x-axis in mm
y-ax
is in
mm
20 40 60 80
10
20
30
40
50
60
70
80
90 0.65
0.7
0.75
0.8
0.85
0.9
0.95
1
100 GHz – Side B
3-D synthetic aperture imagery
-10-Barcelona, June 27th, 20132nd ITFG Workshop
Tomographic configuration
Data procedure for 3-D imaging
• Technique for structural reproduction of object
• In THz the beam is close to a Gaussian distribution
• Information from reflected or transmitted beam
Process of acquisition/reconstruction
• Acquisition composed of a set of projections
• Modeled by the Radon transform
Tomosynthesis in transmission mode
-11-Barcelona, June 27th, 20132nd ITFG Workshop
Tomographic configuration
Tomosynthesis in reflection mode
View A
View B
View C• Reconstruction of internal
structure with more accuracy than unique reflection image
• Allow to converge to diffraction limit for BOTH lateral and longitudinal resolution
-12-Barcelona, June 27th, 20132nd ITFG Workshop
Tomosynthesis
Inserts – Solid laminate – Glass fibre
100 GHz – Side A
Images at different depths
-13-Barcelona, June 27th, 20132nd ITFG Workshop
Outline
• SECTION 1: Alternative configurations
– Wide beam configuration (synthetic aperture processing)
– Tomographic configuration (tomosynthesis processing)
• SECTION 2: Challenge of data size
– Enhanced visualisation
– (Semi-) Automatic defect detection
-14-Barcelona, June 27th, 20132nd ITFG Workshop
Challenge of data size
FMCW system
3-D data set – spatial information
TDS
3-D data set – spatial information 3-D data set – frequency information
CHALLENGE FOR THE OPERATOR
DOTNAC
• Enhanced visualisation for flat and curved samples• (Semi-) Automatic defect detection
-15-Barcelona, June 27th, 20132nd ITFG Workshop
Defect visualisation
Raw image representation Processed image representation
Water inclusion- A-sandwich - Honeycomb core
-16-Barcelona, June 27th, 20132nd ITFG Workshop
Defect visualisation
Impact (30 J) - Solid laminate - fibre glass
-17-Barcelona, June 27th, 20132nd ITFG Workshop
Automatic defect detection
Inserts - Solid laminate - fibre glass
Global Background
Input Denoised Image
Objects Boundaries
Mask Area of Interest
Foreground/Object Target
Classification
Segmentation
-18-Barcelona, June 27th, 20132nd ITFG Workshop
Automatic defect detection
Inserts - Solid laminate - fibre glass
-19-Barcelona, June 27th, 20132nd ITFG Workshop
3-D Visualisation
In-situ test on radome
3-D In-depth viewing 1 screenshot
Imaging with FMCW system
-20-Barcelona, June 27th, 20132nd ITFG Workshop
Questions?
-21-Barcelona, June 27th, 20132nd ITFG Workshop 21
Questions?
-22-Barcelona, June 27th, 20132nd ITFG Workshop
2nd ITFG Workshop
BFP: Backprojection of Filtered Projections
W0 corresponds to the projection P filtered by a ramp filter to increase details
A is the weight coefficient between the projection (θ;ρ) and the pixel (i; j)
We sum the filtered projections crossing each point
March 6th 2012
Reconstruction methods: BFP
-23-Barcelona, June 27th, 20132nd ITFG Workshop
2nd ITFG Workshop
SART: Simultaneous Algebraic Reconstruction Technique
OSEM (Ordered Subsets Expectation Method) Make initiak guess Check how well it corresponds
to the measured data Calculate the difference
between the result and real measurement
Correct the values Repeat until results satisfying
Iterative methods