Final Program - Oceans 2007 MTS/IEEE Vancouver
Final Program - Oceans 2007 MTS/IEEE Vancouver
Final Program - Oceans 2007 MTS/IEEE Vancouver
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
Tutorials (cont’d.)<br />
fi rst, since they relate to the principles of image<br />
classifi cation. Near nadir, the amplitudes and<br />
shapes of sounder echoes are rich in sediment<br />
information. Away from vertical incidence, echoes<br />
carry sediment information in their amplitudes and<br />
their noise characteristics, but not in their shapes.<br />
Echoes from imaging sonars, with their wide<br />
horizontal beamwidths, become rasters in sonar<br />
images, so noise in these echoes becomes image<br />
texture. Macro-roughness such as sand waves and<br />
changes in sediment also contribute to texture.<br />
Image amplitude and texture are both heavily<br />
infl uenced by sediment type and are exploited for<br />
segmentation.<br />
Sonar calibration is not necessary for image-based<br />
acoustic classifi cation. Image amplitudes are<br />
made consistent throughout a survey, but remain<br />
in relative, not absolute, units. Since calibrating<br />
imaging sonars is challenging, the ability to use<br />
systems that need only be consistent offers costeffective<br />
practical classifi cation for military and<br />
civil purposes.<br />
Topics in this tutorial include:<br />
Quality control, suppressing system artifacts.<br />
Compensating images for beam patterns and<br />
grazing angle effects.<br />
Features that capture amplitude and texture<br />
characteristics.<br />
Classifi cation with amplitude: backscatter, backscatter<br />
vs. grazing angle.<br />
Classifi cation with texture: Pace, Haralick,<br />
fractal, wavelet.<br />
Differences between classifying multibeam and<br />
sidescan images: resolution, using bathymetric<br />
data for compensation, benefi ts of images<br />
stitched together from backscatter in beams.<br />
Supervised classifi cation, training sets.<br />
Unsupervised classifi cation, PCA, manual and<br />
automated clustering.<br />
Using non-acoustic data to relate acoustic<br />
classes to sediment geoacoustic properties.<br />
Categorical interpolation.<br />
Maps with acoustic classes in similarity colours.<br />
The techniques presented in this tutorial are wide<br />
ranging, and do not concentrate on a selected<br />
technical approach. Participants in this tutorial<br />
can expect to gain a thorough understanding of<br />
the principles and practice of image-based sediment<br />
classifi cation.<br />
Meeting Room 18<br />
T3 – Signal Processing Methods for Underwater<br />
Acoustic Communications<br />
By Dr. Milica Stojanovic, MS, PhD – Principal<br />
Scientist at the Massacheusetts Institute of<br />
Technology & Guest Investigator at the Woods<br />
Hole Oceanographic Institution Lee Freitag,<br />
BS, MS – Senior Engineer at Woods Hole<br />
Oceanographic Institution<br />
Wireless information transmission through the<br />
ocean is one of the enabling technologies for<br />
the development of future ocean-observation<br />
systems, whose applications include gathering<br />
of scientifi c data, pollution control, climate<br />
recording, detection of objects on the ocean<br />
fl oor, and transmission of images from remote<br />
sites. Implicitly, wireless signal transmission is crucial<br />
for control of autonomous underwater vehicles<br />
(AUVs) which will serve as mobile nodes in the<br />
future information networks of distributed underwater<br />
sensors. Wireless communication provides<br />
advantages of collecting data without the need<br />
to retrieve instruments, and maneuvering underwater<br />
vehicles and robots without the burden<br />
of cables.<br />
24 25