Automated marine turtle photograph identification using artificial neural networks, with application to green turtles

Marine turtle population studies to date have relied on flipper tags or other physical markers to identify individuals previously caught and released. This approach is not entirely successful, motivating us to develop a method for producing an automated turtle photograph identification (photo ID) system. This advancement uses artificial neural networks to compare a digital photo of an individual turtle with a database of turtle photos. Unlike many animals, marine turtles have distinctive facial characteristics, making them ideal candidates for automated photo ID systems. It is easy to gather the large number of good photos of tagged turtles needed to train and test the system; the pattern of interest can be distinguished in a relatively small number of pixels; and it is possible to take suitable photos of both nesting and free swimming turtles. We have used this method to develop a photo ID system, MYDAS, for green turtles (Chelonia mydas), with individual animals identified by their distinctive post-ocular scute patterns. MYDAS has a success rate better than 95% in correctly determining whether a new photo matches a photo in a database, and is now being applied to the green turtle population of Lady Elliot Island in the southern Great Barrier Reef.

[1]  Peter Meer,et al.  Are we making real progress in computer vision today? , 2012, Image Vis. Comput..

[2]  I. Douglas‐Hamilton,et al.  On the ecology and behaviour of the Lake Manyara elephants , 1973 .

[3]  M. Chaloupka,et al.  Monitoring green turtles (Chelonia mydas) at a coastal foraging area in Baja California, Mexico: multiple indices to describe population status , 2003, Journal of the Marine Biological Association of the United Kingdom.

[4]  Shaun Mcconkey.  Photographic identification of the New Zealand sea lion: A new technique , 1999 .

[5]  Scarla J. Weeks,et al.  Distribution, site affinity and regional movements of the manta ray, Manta alfredi (Krefft, 1868), along the east coast of Australia , 2011 .

[6]  Martin T. Hagan,et al.  Neural network design , 1995 .

[7]  C. Limpus,et al.  Geographic structure of mitochondrial and nuclear gene polymorphisms in Australian green turtle populations and male-biased gene flow. , 1997, Genetics.

[8]  K. Andrews,et al.  PIT Tagging: Simple Technology at Its Best , 2004 .

[9]  Mohammad Bagher Menhaj,et al.  Training feedforward networks with the Marquardt algorithm , 1994, IEEE Trans. Neural Networks.

[10]  J. B. Foster.  THE GIRAFFE OF NAIROBI NATIONAL PARK: HOME RANGE, SEX RATIOS, THE HERD, AND FOOD , 1966 .

[11]  Peter H. Dutton,et al.  Increase of a Caribbean leatherback turtle Dermochelys coriacea nesting population linked to long-term nest protection , 2005 .

[12]  Marcella J. Kelly,et al.  COMPUTER-AIDED PHOTOGRAPH MATCHING IN STUDIES USING INDIVIDUAL IDENTIFICATION: AN EXAMPLE FROM SERENGETI CHEETAHS , 2001 .

[13]  V. Gelder,et al.  The Serengeti Lion , 1973 .

[14]  J. C. Briand Petersen,et al.  An identification system for zebra (Equus burchelli, Gray) , 1972 .

[15]  R. A. Reijns,et al.  A computer‐aided program for pattern‐matching of natural marks on the spotted raggedtooth shark Carcharias taurus , 2007 .

[16]  Panayotis Dimopoulos,et al.  Investigating the viability of photo-identification as an objective tool to study endangered sea turtle populations , 2008 .

[17]  J. Hobbs,et al.  Photographic identification based on unique, polymorphic colour patterns: A novel method for tracking a marine crustacean , 2007 .

[18]  K. Ullas Karanth,et al.  Application of photographic capture–recapture modelling to estimate demographic parameters for male Asian elephants , 2007 .

[19]  Lawrence D. Jackel,et al.  Backpropagation Applied to Handwritten Zip Code Recognition , 1989, Neural Computation.