'Using Machine Learning with Scanning Sonar Data and Artificial Targets for Shrimp Biomass Estimation' - Scanning Sonar Code Repository

This repository contains some of the code described in research that is provisionally titled 'Using Machine Learning with Scanning Sonar Data and Artificial Targets for Shrimp Biomass Estimation'. Code was created for OTAQ Ltd's scanning sonar as part of UK government-funded industry-collaborated research.

The research was aimed at investigating if cost-effective low-resolution sonar was able to produce the acoustic data needed for the training and deployment of effective Machine Learning models. An acoustic dataset was created by changing the number of targets in a water tank as well as their spatial distribution. The developed machine learning models were used to interpret echograms to output the number of underwater targets submerged in a water tank based on the acoustic data captured by the scanning sonar. The models were tested and found to have a 0.98 r-square value.

The data was processed using MATLAB so the code files exist as MATLAB files. The Codes folder contains all the codes developed to handle the raw data for processing and visualisation. Machine Learning models created in this research as well as the datasets used are intellectual property that belongs to OTAQ under contractual agreement of the project funding.

For enquiries about the sonar, data and the ML models, please contact OTAQ through their website: https://otaq.com/