Abstract:
Blue whales (Balaenoptera musculus) are the largest mammalian ever living on the planet. They
are long-range migration species because of their breeding and feedings patterns. Therefore, the
population is varied from time to time and location to location. Considering long-range migration;
marine biologists are trying to track this endangered species to ensure their status, health
conditions, and survival rate. Individual identification is a key step to the pursuit of information
regards to population tracking of organisms. Unfortunately, marine species identification is
intricate to accomplish due to the lack of records of living organisms. Blue whale identification
based on unique pigmentation patterns using captured images. Comprehensive images captured
from the right-hand side of the blue whales. Classical identification is performed by observing
pigmentation patterns on a large portion of the dorsal fin, fluke, etc. We have proposed an
automatic method to identify individual blue whales by calculating the ratio of their dorsal fin by
matching its shape to track the blue whale population. The shape of the dorsal fin and the ratio
value combination bring major key points to identify blue whales individually. The extraction of dorsal fin form background image has been done using otsus threshholding and morphological
transformation mechanisms. The starting points of the dorsal fin were tracked and the ratio was
recorded. The key features of the shape of the dorsal fin were extracted using Scale Invariant
Feature Transform (SIFT) and matching pairs count by running test-cases. Finally, using ratio
values and the matched pairs, the individual blue whale is tracked and the population was
calculated. The proposed automated method to identify individual blue whales in a large data set
gives 80% accuracy. Further, the model can be improved in accuracy by applying machine learning
for feature matching purposes.
