SEA BOTTOM CLASSIFICATION FROM ECHO SOUNDING DATA

PDF

Authors
  1. Poeckert, R.H.
  2. Caughey, D.
  3. Prager, B.
  4. Klymak, J.
Corporate Authors
Defence Research Establishment Pacific, Victoria BC (CAN);Quester Tangent Corp, Sidney BC (CAN)
Abstract
Sea bottom classification was attempted using echo sounder profile data. Statistical, spectral and image processing approaches were investigated. Image processing approaches were found to be of limited value due to the limited areal extent of the echo sounder data sets. Statistical approaches involved omputing second and higher order moments of the echo that might be thought of as measures of the "shape" of the echo, and then relating these to ground truth. this approach has been found to be separate bottom types such as rock and mud, but was limited use in separating classes such as gravel and sand. The study focussed on spectral signal processing involving wavelet transforms. Using wavelets, it was possible to identify separate classes of signatures in an unsupervised classification which corresponds well with ground truth. Classification accuracies of about 80% were achieved. The signatures were found to be rather independent of water depth. Class boundaries identified with 200 kHz sounder data agree with those collected with a 40 kHz sounder, but additional classes were indentified with the lower frequency data.
Keywords
Wavelet transforms;Seabed classification
Report Number
DREP-CR-94-56;SC93-019-FR-001 — Contractor Report
Date of publication
01 Mar 1994
Number of Pages
41
DSTKIM No
94-02487
CANDIS No
141435
Format(s):
Document Image stored on Optical Disk;Originator's fiche received by DSIS

Permanent link

Document 1 of 1

Date modified: