Chesapeake Times, Vol 12 | Jan 2023
Next month, I am presenting a talk entitled “Automated ways to distinguish areas of viable targets against a uniform seabed”. It is still a work in progress, but I wanted to share a few of the techniques and results. Note that all the tools used for this project are in SonarWiz, and we’re just using them in a slightly different way.
The idea of the project came from an AUV mission but the approach here can be applied to any survey operation. From a typical sidescan sonar, the data creates a mosaic of the seafloor. Afterwards, features and areas of interest are highlighted and then further investigated. The challenge thus is to focus the secondary search only on targets of interest, specifically excluding the non-feature uniform seabed.
Two approaches were used and combining them would produce the desired result.
- Step one: Seabed classification (use minimal classes)
The initial search would highlight, within boundaries, those areas that are similar content, and more importantly, create large areas of uniform seabed as a single class. - Step two: gridded data with amplitude values (color coded)
The high amplitude values should designate a feature, as items on the seafloor would show a higher return, and the higher intensity could show either a larger feature or more pronounced features. - Step three: overlay the two, looking at the differences between the two layers.
Using a survey done off the coast of Florida, a man-made fish haven is used to show how these steps are able to detect the area of interest.
Single line mosaic. Fish haven is in center (clearly noted), but a feature 200m northwest is a possible contact. Let’s see how the methodology works on this:
Step 1 – Seabed classification
The fish haven is defined within two classes, which was expected. But note the two small areas NW and
NE of the fish have that are also defined as a class, and areas in the nadir section. These are all potential
targets of interest.
Step 2 – Gridding data, highlighting color values for intensity. Grid cell size was 1 meter, color chart set
to Heat
From the gridded data, the feature NW of the fish haven is highlighted and clearly seen as an anomoly.
The other sections indicated by the classification mode (NE area, nadir area) did not present a high
enough threshold to be seen in the color chart of the amplitude data
Step 3 – Combining the two data sets
The idea now is to use the two methodologies and determined where the overlap lies. The uniform
seabed would show the absence of either of the two outcomes.
The NW corner: the classification showing the nadir area and feature. The amplitude color chart only
showing agreement on the feature. The SE Corner: The amplitude color chart is not aligning with the
classification items.
One can imagine a huge survey area with just a few features that was of interest for investigation. This
idea of using the seabed classification and gridding amplitude is an approach to help minimize the
search area, and focus just on the areas of interest. These tools are not new in SonarWiz, although we
have worked on the gridder lately, the use of the two present a new opportunity for data analysis.
I will post my presentation on the website at the end of February with a full report and more in-depth
analysis.