Automatic detection and tracking of fish shoals over large areas using Ocean Acoustic Waveguide Remote Sensing (OAWRS)

Abstract

Accurately tracking multiple fish shoals can help understand fish migration patterns, better estimate fish populations, and lead to better decisions regarding fishery routines and preservation of the related marine ecosystems. Previously, fish shoals and their characteristics were identified manually for each shoal. Here we present techniques to automatically detect, track, and predict small fish shoal characteristics using Ocean Acoustic Waveguide Remote Sensing (OAWRS) over large areas and extended time periods. OAWRS system allows us to instantaneously map, image, and monitor fish populations over continental shelf-scale areas of thousands of square kilometers. Conventional fishery sonars operate at much higher frequencies and so have detection ranges limited the immediate vicinity of research vessels. The methods presented here can provide near real-time analysis during experiments or later analysis. In this work, we continuously tracked the migration of multiple fish schools during the 2014 Norwegian Sea experiment and 2003 Atlantic US coast experiment. Each shoal goes through image processing, tracking, feature extraction, and track management analysis. We take into account special cases such as splitting or merging of shoals. The results of this work can provide reliable tracking of small fish shoals, marine mammals, and underwater vehicles.