BioEd Online

The principles of face recognition software are being used to identify individual whales or dolphins from photos of their fins or flippers. The method promises to make it easier for researchers and conservationists to track marine mammals, without the need for physical brands or tags.

The ability to recognize individual animals within a population is key for researchers who want to understand the behaviour of marine mammals, and for conservationists trying to evaluate the status of different species in the wild. Knowing which whale travels where, for example, makes it possible to understand general migration patterns.

"It's very important to be able to identify marine mammals because this allows you to count how many animals there actually are and produce a more robust population analysis," says Luke Rendell, a research fellow at the University of St. Andrews who has studied social patterns in whales.

In the past, biologists have tagged animals such as whales using branding with heat or liquid nitrogen. But as well as being rather impractical, it is not clear whether such approaches harm the animals or affect their behaviour.

Caught on camera

Instead of relying on physical tags, Chandan Gope at the University of Texas, Dallas, and his colleagues have been studying photographs. They have developed software that analyses the pattern of curves at the edges of dolphin dorsal fins, whale flukes or sea-lion flippers.

The shapes of these body parts change little over time, so they provide a reliable 'fingerprint' that can be spotted in different photos of the same animal, say the researchers.

The computer program works by picking out the distinguishing points along one of these edges and then digitally recreating the contour. It then compares the curve with others from a database and highlights the best potential matches.

Fast matches

Previous studies have tried to use these curves to recognize specific individuals, but they have had problems comparing pictures taken from different angles. The software overcomes this challenge by using a trick known as an "affine transformation". This mathematical method allows you to rotate a line while preserving the relationships between its points, including the ratio of distances between them.

This improved method of matching photos will enable marine biologists to identify individual animals faster, the researchers say. They have already tested the method on a sea-lion database containing 97 images.

The program narrowed down the number of options, producing a shortlist from which the user could make a final match by eye. Half of the time the user found a match within the first seven images suggested by the computer, the researchers report online in Pattern Recognition¹.

According to Rendell, the process of identifying marine mammals has already come a long way since the days when researchers had to search through files containing hard-copy versions of photos. But he adds that there is still room for improvement over current computer programs. "It's laborious," he says. "Anything that saves time can be a real bonus."