Automatic Individual Animal Identification from Photographs

http://compbio.cs.uic.edu/projects.html
http://code.google.com/p/stripespotter/wiki/Walkthrough

Welcome to Computational Population Biology at UIC

Professor Tanya Y. Berger-Wolf


In wild animal populations, collecting behavioral data about a species often entails identifying individual animals between sightings taken at different places and times. This is a primitive operation in ecological analysis that underlies broader aspects of animal behavior research. Electronic tracking devices embedded in animals are one ap proach to identifying individual animals, but can be prohibitively expensive and difficult to design for field conditions, and involve considerable cost and risk for larger an imals. Researchers are therefore left with no alternative other than to manually record data in the field us ing, for example, genetic markers in excrement, capture recapture techniques, or manual identification from photographs. Advances in hardware and the correspond ing drop in prices of digital cameras have resulted in an increase in the availability of digital photographs of wild animal sightings at high resolutions and qualities, making it an attractive candidate for fully-automatic or computer-assisted animal identification.

In collaboration with the Princeton Equid research group, we have developed StripeSpotter, a program to perform automatic individual animal identification of zebras and other striped animals. It is an ongoing project, and has currently been deployed at various nature conservancies in Kenya.