Triton Perspective - Manual (Page 5)

iv | P a g e F e b r u a r y 2 0 1 1
SeaClass™ is an advanced seabed
segmentation/classification module that
automatically characterizes bottom types based
on statistical properties of sidescan mosaics or
multibeam backscatter imagery. SeaClass is
based on a multi-layer perceptron supervised
neural network. The classification procedure
consists of two stages: a learning stage and a
classification stage. Training is accomplished
by the operator selecting areas in the mosaic of differing bottom type (e.g. sand, rock,
mud, etc.). Training of the classifier neural network then proceeds thru an automated
statistical analysis of the selected samples and a characterization of each type. The
following classification stage is a completely automated process where the entire mosaic
image is segmented into the different classes. Results of the classification are displayed
overlain on the mosaic image with each area colored or optionally with vector/line
boundaries around each area. Output from SeaClass can be the DXF boundary vectors or
a GeoTIFF mosaic image with the displayed color-coded areas.