of the method was tested by using time-lapse imagery ob-
tained at the terminus of the Viedma glacier,
SPI
, Argentina,
by using a 16 Mpixel CANON EOS Mark II
DSLR
camera.
Images were acquired once a day for nearly two years, and
preprocessed to correct for sensor modeling errors and then
reduced in size to the glacier area. In addition field surveys
were conducted acquire ground control (
GCP
).
The large displacement optical flow (
LDOF
) method was
selected for the investigation. Any optical flow calculation
is based on the assumption that change in image intensities
are due to changes in object shape, such as deformation or
motion; a hard to achieve condition in real life. Therefore, the
time-lapse image sequence was filtered to remove images with
very different intensity characteristics; mainly, due to varying
cloud cover, snow, and weather that may induce rapid melt-
ing. The
CA
method provided a good solution to measure dif-
ferences between images, and based on our settings, resulted
in 38 percent of the images passed the test and were used for
subsequent processing. The
LDOF
algorithm provided optical
flow in the image domain by providing a motion vector to all
the pixels. To convert the estimated motion to actual velocity
values, a sparse
DEM
was created and triangulation was used
for interpolation. Then, based on the
DEM
, the scaling was
estimated and used to convert the results from image domain
to object domain. Thus, glaciological analysis and interpreta-
tions through the velocity field quantification were obtained.
The resulting glacier surface velocities have reached a
maximum value of 3.5 m/d in the central part that is consistent
with the expected ice flow of a calving glacier with high ve-
locities near to the terminus. The estimation error was evaluat-
ed qualitatively and quantitatively, and the directly computed
mean
TRE
was 0.36 m/d. Factoring in the glacier dynamics, the
changes in velocities are slow, and thus, it is fair to say that
the mean error is rather conservatively estimated. Larger errors
were confined to the farther part of the glacier, where the larger
range limited the estimation accuracy. Also, object occlusion is
more frequent in these areas, also contributing to larger errors.
In summary, the proposed method based on computer vi-
sion and photogrammetric techniques, combining the
CA
and
LDOF
algorithms, has shown very good performance at detect-
ing ice flow movement, and ultimately estimating the surface
velocity of the Viedma glacier. In addition, the time-lapse se-
quence provided a good temporal resolution, helping improve
the results. Although, no reference data was available for an in-
dependent comparison, the limited data from earlier investiga-
tions as well as the known glacier behavior are consistent with
the obtained velocity estimation results. The data acquisition
system is simple and affordable, the selected algorithms pro-
vide robust computation, so the overall performance is primar-
ily dependent on the environmental conditions. In the selected
test areas, Viedma glacier, about in 40 percent of the time, good
quality imagery could be acquired in a two year period, result-
ing an accurate velocity estimation of the glacier surface.
Acknowledgments
Fieldwork was funded by Grants PICT 2921-2012 and PICT
1995-2013, Agencia Nacional de Ciencia y Tecnología Argen-
tina (ANCyT). The authors would like to thank Adalberto Fer-
lito, Marcelo Durand, Andrés Lo Vecchio, and Robert Bruce
for field assistance. Support from the Los Glaciares National
Park is greatly appreciated.
References
Ahn, Y., and J.E. Box, 2010. Instruments and Methods. Glacier
velocities from time-lapse photos: Technique development and
first results from the Extreme Ice Survey Greenland,
Journal of
Glaciology
, 56:198.
Aniya, M., H. Sato, R. Naruse, P. Skvarca, and G. Casassa, 1996. The
use of satellite and airborne imagery to inventory outlet glaciers of
the Southern Patagonia Icefield, South America,
Photogrammetric
Engineering & Remote Sending
, 62(12):1361–1369.
Aniya, M., 2013. Holocene glaciations of Hielo Patagónico (Patagonia
Icefield), South America: A brief review,
Geochemical Journal
,
47(2):97–105.
Bamber, J., and A. Rivera, 2007. A review of remote sensing methods
for glacier mass balance determination,
Global and Planetary
Change
, 59:138–148.
Bartholomé, E., and A.S. Belward, 2005. GLC2000: A new approach
to global land cover mapping from Earth observation data,
International Journal of Remote Sensing
, 26(9): 1959–1977.
Beauchemin, S.S., and J.L. Barron, 1995. The computation of optical
flow,
ACM Computing Surveys
, 27:433–466.
Benn, D., and D. Evans, 2010.
Glaciers & Glaciation
, Routledge
Publishing. New York, 802p.
Bolch, T., A. Kulkarni, A. Kääb, C. Huggel, F. Paul, J.G. Cogley, and
S. Bajracharya, 2012. The state and fate of Himalayan glaciers,
Science
, 336(6079):310–314.
Bouguet, J.Y., 2001.
Pyramidal Implementation of the Affine Lucas
Kanade Feature Tracker Description of the Algorithm
, Intel
Corporation, 5(1–10):4.
Bown, F., 2015.
Dinámica Frontal de un Glaciar de Descarga
Oceánica: Glaciar Jorge Montt, Campo de Hielo Patagónico Sur
,
Ms. dissertation, Universidad de Concepción, Chile, 104 p.
Brox, T., A. Bruhn, N. Papenberg, and J. Weickert, 2004. High
Accuracy optical flow estimation based on a theory for warping,
Proceedings of the 8
th
European Conference on Computer Vision
,
Springer LNCS 3024 ( T. Pajdla and J. Matas, editors), Prague,
Czech Republic, 4:25–36.
Brox, T., C. Bregler, and J. Malik, 2009. Large displacement optical
flow,
Proceedings of the IEEE International Conference on
Computer Vision and Pattern Recognition (CVPR)
.
Brox, T., and J. Malik, 2011. Large displacement optical flow:
Descriptor matching in variational motion estimation,
IEEE
Transactions on Pattern Analysis and Machine Intelligence
,
33(3):500–513.
Chang, C., and S. Chatterjee, 1992. Quantization error analysis in stereo
vision,
Signals, Systems and Computers
, October, 1992 Conference
Record of The Twenty-Sixth Asilomar, IEEE, pp. 1037–1041.
Clarke, T.A., and J.G.Fryer, 1998. The development of camera
calibration methods and models,
The Photogrammetric Record
,
16(91):51–66.
Danielson, B. and M. Sharp, 2013. Development and application of
a time-lapse photograph analysis method to investigate the link
between tidewater glacier flow variations and supraglacial lake
drainage events,
Journal of Glaciology
, 59:287–302.
De la Nuez, A.J.S., 2010.
Métodos Variacionales para la Estimación
del Flujo
Óptico
y Mapas de Disparidad
, Universidad de Las
Palmas de Gran Canaria, pp. 17–24.
Gao, J., and Y. Liu, 2001. Applications of remote sensing, GIS and
GPS in glaciology: A review,
Progress in Physical Geography
,
25(4):520–540.
Garcia Tomaselli, A.M., M.L. Lopes Reiss, 2005. A photogrammetric
method for single image orientation and measurement.
Photogrammetric Engineering & Remote Sensing, 71(6):727–732.
Gleitsmann, L., and M. Kappas, 2006. Glacier monitoring survey
flights below clouds in Alaska: Oblique aerial photography
utilizing digital multiple image photogrammetry to cope with
adverse weather,
EARSel eProc
, 5(1):42–50.
Granshaw, S.I., and C.S. Fraser, 2015. Computer vision and
photogrammetry: Interaction or introspection?,
The
Photogrammetric Record
, 30(149):3–7.
Hashimoto, T., M. Kaneko, A. Rövid, S. Isono, T. Sone, K. Baba,
A. Fukuda, M. Aniya, N., Naito, H. Enomoto, and P. Skvarca,
2009. An introduction of high-precise 3D measurement system
and its applications,
Journal of Automation, Mobile Robotics &
Intelligent Systems
, 3(4).
Harrison, W.D., K.A. Echelmeyer, D.M. Cosgrove, and C.F.
Raymond, 1992. The determination of glacier speed by time
lapse photography under unfavorable conditions,
Journal of
Glaciology
, 38(129):257–265.
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
January 2018
41
