accuracies for fish and rubble. Still, even with such low
producer/user’s accuracies, the overall accuracies obtained
were still high, unlike in the pixel-based classification.
Advantages of OBIA approach: The developed rule set is
transferrable and applicable to similar aims of obtaining benthic
cover maps. With a little tweaking of the rule sets, similar
outputs may be obtained quickly.
Potential improvements: Aside from improving the rule sets,
use of more accurate bathymetric data and image ratioing
(blue/red and green/blue rations) will further increase the
accuracy of OBIA classification.
4. CONCLUSIONS & RECOMENDATIONS
Georeferencing and mosaicking: The generated uncontrolled
mosaic was able to produce an acceptable replica of the actual
transected area and registration to the bathymetric data was able
to yield good results. Therefore, in such cases where limitations
of the available data hinder the production of a useful photo-
mosaic, Adobe Photoshop is a useful and capable alternative.
Performance of OBIA against pixel-based image
classification: Pixel-based classification methods are capable
of classifying underwater photos to benthic cover maps with
high accuracy as long as only a general classification is
required of off them, such as living (coral) or non-living (sand)
classes. Inclusion of other classes, such as rubble and fish,
confuses the classification methods such as supervised and
unsupervised classifications. Hybrid classification was the only
pixel-based algorithm that was able to perform well even with
orders of classifying more than 3 classes. As for the object-
based classification, which was capable of automatically
identifying 4 classes, it was able to perform significantly better
than the individual classifications, which were supervised (p-
value = 0.0152, a = 0.05), unsupervised (p-value = 0.0005, a =
0.05) and hybrid classification (p-value = 0.0377, a = 0.05).
Against all pixel-based classifications, OBIA was significantly
able to generate more accurate results (p-value = 0.0001, a =
0.05). Therefore, OBIA is a method capable of automatically
and accurately classifying benthic cover other than living
(coral) and non-living (sand).
Recommendations: Geotagging by GPS coordinates collected
by a snorkeler swimming directly on top of the diver doing the
video capture can provide approximate coordinates to the
video, which may be enough for some applications and will
also facilitate the matching of photos with the MBES data. But
more accurate georeferencing and mosaicking (or mosaicking
and georeferencing) is needed for accurate monitoring of reef
conditions. In order to obtain the full potential of OBIA,
accurate bathymetric data and underwater videos would be
required. To minimize errors in the recorded underwater videos
and to be able to record at any depth, an AUV with a motion
sensor may be a better alternative in recording transects. If
motion sensors are not available for the underwater camera, a
bubble level may be used instead to maintain the vertical
orientation of the video to the seafloor. If this level is still not
available, recording should just be done in such a way that the
camera is always perpendicular to the seafloor so as to obtain
images that are directly beneath the camera’s lens. This way,
georeferencing and image to image registration will have
minimal distortions and occlusions. Controlled lighting during
image acquisition is also recommended to minimize variations
in the spectral characteristics of the images. Stereo pairs, as
done in photogrammetry, may also be attempted in order to
fully model and recreate image acquisition geometry and to
obtain relief information of the reef, which can also be used to
identify elevation of benthic cover. With such information,
height may be used in the rule sets in order to identify elevated
and non-elevated features. For this case, however, ground
controls are needed for absolute orientation but can be very
difficult to provide.
5. REFERENCES
Addink, E. & Coillie, V., 2010. “Object-based Image Analysis,” GIM
International:
http://www.gim-international.com/issues/articles/id 1469-
Objectbased Image Analysis.html (28 Jan. 2010).
Baatz, M. & Schäpe, A., 2000. Multiresolution Segmentation: an
optimization approach for high quality multi-scale image segmentation.
In... Srmobl, J, Blaschke, T, Griesebner, G. (Eds.),
AngewandteGeographische ^ Informationsverarbeitung, vol. XI.
Wichmann, Heidelberg, p. 3.
Cohen, J., 1960. A coefficient of agreement of nominal scales. In:
Educational and Psychological Measurement, pp. 3-5
Congalton, R.G., 1991. A review of assessing the accuracy of classification
of remotely sensed data. In: Remote Sensing Environment, p. 40.
ENVI User's Guide
http://geol.hu/data/online help/DefiningRegionsOflnterest.html
Ernstsen, V., Noormets, R., Hebbeln, D., Bartholoma, A. & Flemming, B.,
2006. Precision of high-resolution multibeam echo sounding coupled
with high-accuracy positioning in a shallow water coastal environment.
In: Geo-Marine Letters, Vol. 26, Issue 3, pp. 141-149.
ES3PT including ES3 Operating Software User Manual v.14
http://www.odomhydrographic.com
Hill, J. & Wilkinson, C., 2004. Methods for Ecological Monitoring of Coral
Reefs. Australian Institute of Marine Science, pp. 1-100
Kaeli, J., Singh, H. & Armstrong, R., 2005. Morphological Image
Recognition of Deep Water Reef Corals.
Levick, S.R. & Rogers, K.H., 2006. LiDAR and Object-based Image
Analysis as Tools for Monitoring the Structural Diversity of Savanna
Vegetation. In: Int'l Archives for the Photogrammetry, Remote Sensing
and Spatial Information Sciences, Vol. 34, Part XXX, 6 pp.
Marcos, M.S.A.C., David, L.T. & Soriano, M.N., 2008. Area-Calibrated
Automation of Coral Classification for Near and Subsurface Reef Videos.
In: Proceedings of the 11" International Coral Reef Symposium, Ft.
Lauderdale, Florida, Session 16, pp. 1-5.
Marcos, M.S.A.C., Soriano, M. & Saloma, C., 2005. Classification of coral
reef images from underwater video using neural networks. In: Opt.
Express, Vol. 13, Issue 22, 13:8766-8771
Mumby, P.J., Skirving, W., Strong, A.E., Hardy, J.T., LeDrew, E.F.,
Hochberg, E.J., Stumpf, R.P., & David, L.T., 2004. Remote sensing of
coral reefs and their physical environment. In: Marine Pollution Bulletin,
Vol. 48, pp. 219-228.
Richards, J.A., & Jia, X., 2006. Remote Sensing Digital Image Analysis: An
Introduction 4th Edition, Springer-Verlag Berlin Heidelberg, Germany,
pp. 193-263 & 295-302
Roelfsema, C.M., Phinn, S.R. & Joyce, K.E., 2006. Evaluating Benthic
Survey Techniques for Validating Maps of Coral Reefs Derived from
Remotely Sensed Images. In: 11th Int’l Coral Reef Symposium, Okinawa,
Japan, Int’l Coral Reef Society, 10 pp.
Rossiter, D.G., 2004. Technical Note: Statistical methods for accuracy
assessment of classified thematic map. In: Technical Report ITC,
Enscheda, NL, April 2004, 46 pp, also available online at
http://www itc.nl/personal/rossiter/teach/R/R_ac.pdf
Scopélitis J., Andrefouét S., Phinn S., Arroyo L., Dalleau, M., Cros, A.,
Chabanet, P., 2010. The next step in shallow coral reef monitoring:
combining remote sensing and in situ approaches. In: Marine Pollution
Bulletin, 2010, 60 (11), p. 1956-1968. ISSN 0025-326X
Trimble SPS 461
http://gps-is.com/main/page marine gps sps series sps46l.html
Wilkinson, C., 2008. Executive summary of status of coral reefs of the
world: 2008. Australian Government, Australian Institute of Marine
Science (AIMS), Townsville, p. 10.
Zhang, Y., & Maxwell, T., 2006. A fuzzy logic approach to supervised
segmentation for object-oriented classification. In: ASPRS 2006 Annual
Conference, Reno, Nevada, May 1-5, 2006
6. ACKNOWLEDGEMENTS
This research was supported by CECAM project, funded by
ERDT-DOST and had full guidance from Dr. Ariel C. Blanco.