Active Learning for Fish School Recognition in Echograms in the Bay of Fundy

OERA use hydroacoustic echosounder surveys to evaluate the impact on marine life of tidal turbines in the Bay of Fundy. OERA use Echoview software to read in the raw sensor data (e.g. voltages) and convert it to a visual representation. Echoview contains some algorithms to detect the bottom of the ocean. However, the Fundy data is very noisy from several sources including air bubbles, “entrained air” pushed below the surface of the water, and irregular surfaces on the bottom of the ocean. In order to analyze the survey data, manual pre-processing is currently required to annotate the data. This manual process delays the turnaround, potential consistency and provides opportunity for inconsistency.
This project will train a machine learning model to detect and filter noise in the hydroacoustic sensor data, allowing OERA to improve the accuracy, consistency and manual effort required to pre-process its data. By identifying “bad regions” in hydroacoustic data collected near underwater turbines in the Bay of Fundy with a model to automatically extract these regions from future survey data, OERA will be able to apply greater speed, consistency and accuracy to data processing to prepare for rigorous analysis.

Faculty Supervisor:

Sageev Oore;Evangelos Milios

Student:

Scott Lowe

Partner:

Offshore Energy Research Association of Nova Scotia

Discipline:

Computer science

Sector:

Mining and quarrying

University:

Dalhousie University