Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
Jakobshavnisbrae using highresolution digital
camera
Rignot, Friedt,Moreau
Data acquisitionand processing
Greenland
French alps
Conclusion
About the tide oscillations and icebergs motion recorded at Jakobshavnisbrae during summer 2007 using high resolution digital camera, Icefjord,
West Greenland.
Discussion about acceleration of glacier flow in Greenland, lubrication atthe ice-rock interface, vertical circulation of melt water (moulins) and
lubrication recorded in situ under Argentire glacier (Mt-blanc).
É. Rignot, J.-M Friedt, L. Moreau
5 mars 2008
1 / 15
Jakobshavnisbrae using highresolution digital
camera
Rignot, Friedt,Moreau
Data acquisitionand processing
Greenland
French alps
Conclusion
Data acquisition
Automated digital image capture using a custom circuit for triggering acommercial digital camera
• low power consumption of control circuitry (< 200 µA) for anautonomy of several years
• camera consumption limits the global autonomy to theoretically'1000 frames
• date & time stored in EXIF header for quantitative processing(problem of powering the camera’s internal clock)
Two testbeds : a 1 month long continuous sequence shot in Greenland, a6 month long webcam archive from Chamonix, France.
2 / 15
Jakobshavnisbrae using highresolution digital
camera
Rignot, Friedt,Moreau
Data acquisitionand processing
Greenland
French alps
Conclusion
Selected analysis area
blue red green magenta black
3 / 15
Jakobshavnisbrae using highresolution digital
camera
Rignot, Friedt,Moreau
Data acquisitionand processing
Greenland
French alps
Conclusion
Measurement horizon (theory)• Fixed window : correlation technique only works as long as reference
and anlyzed frame look similar.
• Short term difference of successive frames provides poor accuracy(motion > 1 pixel between two frames)
• Shift reference window and connect drift curves
Reference frame (fixed)
+ illumination artifact : apparent “Y motion” with 24 hour period dueto sun motion
4 / 15
Jakobshavnisbrae using highresolution digital
camera
Rignot, Friedt,Moreau
Data acquisitionand processing
Greenland
French alps
Conclusion
Measurement horizon (theory)• Fixed window : correlation technique only works as long as reference
and anlyzed frame look similar.
• Short term difference of successive frames provides poor accuracy(motion > 1 pixel between two frames)
• Shift reference window and connect drift curves
Reference frame (fixed)
Frame 1
+ illumination artifact : apparent “Y motion” with 24 hour period dueto sun motion
5 / 15
Jakobshavnisbrae using highresolution digital
camera
Rignot, Friedt,Moreau
Data acquisitionand processing
Greenland
French alps
Conclusion
Measurement horizon (theory)• Fixed window : correlation technique only works as long as reference
and anlyzed frame look similar.
• Short term difference of successive frames provides poor accuracy(motion > 1 pixel between two frames)
• Shift reference window and connect drift curves
Frame 2
Reference frame (fixed)
Frame 1
+ illumination artifact : apparent “Y motion” with 24 hour period dueto sun motion
6 / 15
Jakobshavnisbrae using highresolution digital
camera
Rignot, Friedt,Moreau
Data acquisitionand processing
Greenland
French alps
Conclusion
Measurement horizon (theory)
• Fixed window : correlation technique only works as long as referenceand anlyzed frame look similar.
• Short term difference of successive frames provides poor accuracy(motion > 1 pixel between two frames)
• Shift reference window and connect drift curves
Frame 2
Reference frame (fixed)
Frame 1Frame 3
+ illumination artifact : apparent “Y motion” with 24 hour period dueto sun motion
7 / 15
Jakobshavnisbrae using highresolution digital
camera
Rignot, Friedt,Moreau
Data acquisitionand processing
Greenland
French alps
Conclusion
Motion detection (application)• Basic but robust technique : intercorrelation looks for translation
vector for best match of a reference picture and a measurementpicture (xcorr2() with Matlab)
• We will use an Eulerian descritpion of fluid motion (fixed window,monitor the mass entering and leaving this frame)
image(t)
50 100 150 200
20
40
60
80
100
120
140
image(t+24)
50 100 150 200
20
40
60
80
100
120
140
image(t+24)
50 100 150 200
20
40
60
80
100
120
140
xcorr2
100 200 300 400 475
50
100
150
200
250
295
8 / 15
Jakobshavnisbrae using highresolution digital
camera
Rignot, Friedt,Moreau
Data acquisitionand processing
Greenland
French alps
Conclusion
Measurement horizon (results)
4000 6000 8000 10000 12000 14000−50
0
50
100
150
time (min)
disp
lace
men
t (pi
xel)
4000 6000 8000 10000 12000 14000−20
−10
0
10
20
30
time (min)
disp
lace
men
t (pi
xel)
0.0326 pixel/min
0.0215 pixel/min
24 h
referenceframe 9
4000 6000 8000 10000 12000 14000
−100
−50
0
50
100
time (min)
disp
lace
men
t (pi
xel)
4000 6000 8000 10000 12000 14000
−20
0
20
40
time (min)
disp
lace
men
t (pi
xel)
0.0323 pixel/min
0.0193 pixel/min
referenceframe 40
24 h
X
Y 24 h
4000 6000 8000 10000 12000 14000−100
−50
0
50
100
time (min)
disp
lace
men
t (pi
xel)
4000 6000 8000 10000 12000 14000−10
0
10
20
30
40
time (min)
disp
lace
men
t (pi
xel)
referenceframe 70
0.0321 pixel/min
0.023pixel/min
9 / 15
Jakobshavnisbrae using highresolution digital
camera
Rignot, Friedt,Moreau
Data acquisitionand processing
Greenland
French alps
Conclusion
Influence of tideVisually, obvious influence of tide on “vertical” motion of ice
1.4 1.6 1.8 2 2.2 2.4 2.6
x 104
−30
−20
−10
0
10
20
30
40
time (min)
disp
lace
men
t (pi
xel)
Y motionX motion
referenceframe (130)
low tide
high tide
0.0033pixel/min
3 3.2 3.4 3.6 3.8 4
x 104
−20
−10
0
10
20
30
40
time (min)
disp
lace
men
t (pi
xel)
Y motionX motion
referenceframe 230
0.0055pixels/min
3.4 3.5 3.6 3.7 3.8 3.9 4 4.1 4.2
x 105
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
date (minutes since 01/01/2007)
tide
ampl
itude
(m
)
22/08frame 330
29/07frame 100
01/08frame 130
21/07frame 10
27/07frame 70
13/08frame 260
4.2 4.4 4.6 4.8 5 5.2
x 104
−15
−10
−5
0
5
10
15
time (min)
disp
lace
men
t (pi
xel)
Y motionX motion
Y motionX motion
referenceframe (330)
0.0024 pixel/min
high tide
low tide
• Two tide amplitude maxima over 1 month record : visible signalsynchronous with predicted tide max/min.
• No obvious signal synchronous with tide during amplitude minimum(bot. right)
10 / 15
Jakobshavnisbrae using highresolution digital
camera
Rignot, Friedt,Moreau
Data acquisitionand processing
Greenland
French alps
Conclusion
Influence of windNo visible influence of wind1 : the local influence on a given iceberg isnot tracked : would require a Lagrangian description of fluid motion(object tracking)
0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
x 104
−800
−700
−600
−500
−400
−300
−200
−100
0
time (min.)
disp
lace
men
t (pi
xels
)
green
magenta
blue
red
yellow
wind
1 m
onth
0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
x 104
−120
−100
−80
−60
−40
−20
0
20
40
time (min)di
spla
cem
ent (
pixe
l)
tide amplitudex50, +30
high tide x10, −100
low tide x10, −100
blue
red (ref.)
green
yellow
magenta
Long term “vertical motions” appears related to tide2 amplitude (right)
1Hourly METAR data collected fromhttp://english.wunderground.com/history/airport/BGJN/
2Web interface to xtide athttp://tbone.biol.sc.edu/tide/sites othernorth.html 11 / 15
http://english.wunderground.com/history/airport/BGJN/http://tbone.biol.sc.edu/tide/sites_othernorth.html
Jakobshavnisbrae using highresolution digital
camera
Rignot, Friedt,Moreau
Data acquisitionand processing
Greenland
French alps
Conclusion
Influence of picture qualityWe have demonstrated the use of high resolution (10 Mpixels), lowcompression images on a fast moving object.What about poor quality images of a slow glacier ?
100 200 300 400 500 600 700 800 900
0
20
40
X motion (pixel)
red
100 200 300 400 500 600 700 800 900
0
20
40
blue
100 200 300 400 500 600 700 800 900
0
20
40
gree
n
100 200 300 400 500 600 700 800 900
0
20
40
mag
enta
100 200 300 400 500 600 700 800 900
0
20
40
time (frame number)
blac
k
Requires
1 interpolation of the original image to smooth JPEG compressionboundaries and improve resolution
2 histogram equalization12 / 15
Jakobshavnisbrae using highresolution digital
camera
Rignot, Friedt,Moreau
Data acquisitionand processing
Greenland
French alps
Conclusion
Influence of picture quality
Strong JPEG compression ⇒ artifacts
Matlab’s imresize() bilinear interpolation (weighted average of pixelsin the nearest 2-by-2 neighborhood)
13 / 15
Jakobshavnisbrae using highresolution digital
camera
Rignot, Friedt,Moreau
Data acquisitionand processing
Greenland
French alps
Conclusion
Results and improvements
• Some frames3 with poor weather induce high noise level : wouldrequire pre-processing to avoid noise
• Strong effect of shadow → histogram equalization
⇒ reduce influence of noise when connecting successive windows byusing linear fit of glacier motion
3http://www.compagniedumontblanc.fr/webcam/CMM1MERDEGLACE.jpg14 / 15
http://www.compagniedumontblanc.fr/webcam/CMM1MERDEGLACE.jpg
Jakobshavnisbrae using highresolution digital
camera
Rignot, Friedt,Moreau
Data acquisitionand processing
Greenland
French alps
Conclusion
Results and improvements
• The motion is an average over a given part of the picture :the larger the picture, the longer the horizon, but the worse theaverage
• Connection of analysis periods is not always accurate• Conversion from pixel to meters requires terrain model + camera
characteristics
• Validation using calibrated instruments (GPS)
Yet, digital image processing provides a means of continuous monitoringof ice flow in area where instuments cannot be positioned (icebergs,strong slopes)
15 / 15
Data acquisition and processingGreenlandFrench alpsConclusion