A computer-aided framework for subsurface identification of white shark pigment patterns

Abstract

Subsurface video footage can be used as a successful identification tool for various marine organisms; however, processing of such information has proven challenging. This study tests the use of automated software to assist with photo-identification of the great white shark Carcharodon carcharias in the region of Gansbaai, on the south coast of South Africa. A subsurface photo catalogue was created from underwater video footage. Single individuals were identified by using pigmentation patterns. From this catalogue, two images of the head for each individual were inserted into automated contour-recognition software (Interactive Individual Identification System Beta Contour 3.0). One image was used to search the database, the other served as a reference image. Identification was made by means of a contour, assigned using the software to the irregular border of grey and white on the shark’s head. In total, 90 different contours were processed. The output provided ranks, where the first match would be a direct identification of the individual. The method proved to be accurate, in particular for high-quality images where 88.24% and 94.12%, respectively, were identified by two independent analysts as first match, and with all individuals identified within the top 10 matches. The inclusion of metadata improved accuracy and precision, allowing identification of even low-quality images.

Keywords: Carcharodon carcharias, computer-aided identification, contour recognition, mark-recapture, photo-identification, underwater