Salmon have a unique "fingerprint" in the form of dots on their skin and head. This has been proven to be unique by SINTEF in their research:
The ultimate goal of the project is to create a biometric identification pipeline that is able to identify salmon swimming underwater in a fish farm.
A dataset consisting of multiple labelled images of salmon on land (SalmonLandDB) will be provided by a local company (SalmonID).
Feature extraction will take place, in the first instance by segmenting the dots on the salmon. Other features, like the shape of the head and body, may also be used.
A biometric pipeline for testing this on clearly labeled fish on land (SalmonLandDB) is to be created first.
Then the problems related to live fish swimming past an underwater camera will be analysed and attempted to be solved.
By August 15, 2019, it is envisaged that SalmonLandDB will be made available.
In the Fall (Specialisation) project, it is envisaged that the biometric pipeline will be planned and most components tested on SalmonLandDB.
In the Spring (Master's) project the focus will be on implementing the biometric pipeline for identifying the fish
Ideally both TDT4195 and TDT4230
Supervisors: Prof. T. Theoharis (IDI, NTNU)