Over the past three years, Whisker Patrol key investigator Sylvia Osterreider has been in the field collecting photographs of wild Australian sea lions. Visiting 15 breeding and haul-out islands across 127 different field trips, Sylvia has taken thousands of photographs to be used in the development of computer software to identify individual animals. Now her hard work has paid-off, with the first publication of the research in a scientific journal.
Published in the Journal of Mammalogy, this paper describes the use of whisker spot patterns as a non-invasive method for individually identifying Australian sea lions. Several co-authors also contributed to the article, including Whisker Patrol key investigators Chandra Salgado Kent and Iain Parnum and project collaborators Carlos Anderson and Randall Robinson.
The study aimed to establish whether the variability of whisker spot patterns were large enough to reliably use them for individual identification of Australian sea lions and, if so, to develop and test the accuracy of computer-aided pattern recognition technique.
To first test this technique, photographs were used from known individual Australian sea lions in captivity. This allowed the technique to be “ground-truthed” using 3,036 photos of 16 individuals kept in various zoos and aquaria across Australia. These were then combined with Sylvia’s photos, numbering 5,766 from her field trips and collected using similar methods to that described in the Whisker Patrol Photo Masterclass page.
From this grand total of 8,802 images, only photos of the highest quality which met stringent selection criteria were chosen for further analysis. Individual whisker spots were manually selected within a semi-automated pattern recognition software (which was originally developed for identifying polar bears). The program then standardised the location of each spot point using three reference marks on the animal’s face (corner of eye, nose, and mouth). This allowed researchers to compare the pattern of whisker spots between different animals.
They found that there is sufficient variability in Australian sea lions for reliable matching using whisker spot patterns. However, this is for a relatively small population – the larger the population, the greater the chance of pattern duplication and mis-identification of individual animals.
After a lot of program modifications and trials, the research team were able to create a library of the 53 individual sea lions used in this study. They also created several catalogues, where photos were taken under different scenarios (e.g. varying distances or angles). By comparing the accuracy of the software using photos from different catalogues, the team were able to identify under which conditions photographs should be taken for maximum chance of successful matching, and which software settings gave the most precise results.
Photographs taken at 90° resulted in good matching results, with 90% correct matches. This matching process revealed that the angle at which the photograph is taken is the main variable which affects reliability of results. If the angle at which the photograph is taken shifts, this has a knock-on effect on the accuracy of correct identification. Photographs taken at greater distances did not seem to alter matching success, as long as the images remained of high-quality and were taken from the same angle.
Further work is required since the angles of photos taken of wild sea lions may shift from one field day to the next. It is hoped that further development of this new technique for identifying individual Australian sea lions can eventually render it available for application in assessment of habitat-use and residency in localised areas. Determining the population or resident community size are other potential uses, which could then feed into species management and conservation.
The full article is available to read online. Congratulations to the Whisker Patrol team on their first project publication!