Biometric recognition and individual tracking of salmon in large-scale sea cages

Abstract

In this project a method for tracking and biometric identification of salmon in large scale sea cages using machine vision cameras and techniques has been developed. This has been done by finding unique point patterns in the dorsal head region of the fish itself, without the need of physically handling the fish or any form of external or internal tagging.

Automatic recognition of individuals will be an important step forward for the aquaculture industry, as it will facilitate more precisely data driven practices in farming. Combining individual recognition and tracking with monitoring treats such as growth, health parameters, lice counting and social behavior over time would give the farmers the data to manage their stock at a more accurate level.

Salmon were photographed by a under water machine vision camera made by SEALAB AS and tracked as they moved across the field of view. The pattern was found and learned by the algorithm from 333 individuals salmons by using one image of each individual. From a total of 14,652 images test images, 5,922 point patters were recognized as one of the patterns from the database. Among the recognized point patters, 5,902 images were classified as the correct individual, yielding a very promising accuracy of 99,7\% on the recognized point patterns.