By Randall W. Davis
Recognition of individual animals enables detailed studies of movement patterns, foraging, life histories and survival. It is also important for understanding the ecology and behavior of species. Artificial marks, such as tattoos, dyes, brands, colored or numbered tags, and radio and satellite telemeters have been the primary ways of identifying individual animals in research. However, these systems require that the animal be captured, which may cause stress or injury to the animal and/or the researcher and may modify the animal’s behavior. Increasingly, researchers are using natural color patterns, scars and other features to identify animals in a wide range of taxa for which capture and marking is not desirable or logistically feasible.
Sea otters have naturally occurring nose scars from copulation and fighting, which can be used to identify individuals based on the size, shape and location of the scars. However, matching the scars in digital images of the individuals can require many hours of effort, depending on the size of the catalog. This study, funded by a Christine Stevens Wildlife Award from the Animal Welfare Institute, tested the performance of a new program, Sea Otter Nose Matching Program (SONMaP). SONMaP, which was developed about three years ago by Gilbert Hillman, Ph.D., used blotch-pattern recognition algorithms to match the shape and location of lightly colored scar tissue in relation to normal black pigmentation of sea otter noses.
Our study of the device was conducted in Simpson Bay, located in northeastern Prince William Sound, Alaska. Digital images of sea otters were taken from a six-meter-long skiff with a Nikon D1H digital camera with an 80 to 400 millimeter image-stabilized telephoto lens. When an otter was sighted, the skiff driver approached the animal slowly while the photographer attempted to obtain a frontal image of the animal’s face, usually at a distance of about 30 meters.
A catalog was then created for the 1,638 images of otters. One to four of the best images (based on proximity, sharpness and head orientation) of each individual were cropped to isolate the face from the rest of the image. Next, two researchers independently matched sea otters in these images by visually comparing them with all other images in the catalog, a process that took many hours. Images of the 186 previously matched otters were to test the performance of SONMaP. The nose in each image was first isolated using Adobe Photoshop 7.0 and ranked based on quality (Q1 to Q4) and distinctiveness (D1 to D5) of the scars. After running the images through SONMaP, they were classified as "Best", "Average" or "Worst," based on whether the correct match was within the first 10 percent, 11 to 50 percent, or 51 to 100 percent of images in the catalog, respectively. In 49 percent of the previously visually-matched images, the program accurately selected the correct image in the first 10 percent of the catalog, which compares favorably with other computer-assisted photo identification studies of marine mammals. We concluded that SONMaP performed well enough to provide significant assistance in the process of photo-identification by reducing the time needed to match sea otters in a catalog by 67 percent, and can be used in the field for identification of individual animals under study.