Exploration of Methods for Automatic Whale Identification

This paper describes research into techniques for automating the classification and identification of humpback whales from tail fin images. The need for a system to aid in the visual identification process is becoming greater as the number of whales encountered during wildlife studies increases. Currently manual handling of large numbers of photographs, including matching the whales in each new image of whales photographed with those already cataloged is required and is time consuming and expensive. The location of the fluke in the image is a critical part of an automatic identification system. Although, the overall goal is to create a complete system useful to whale researchers, this report focuses on investigation of techniques to locate the whale fluke as the first step towards this goal. Fluke location makes use of knowledge of whale anatomy and image processing methods to build an active shape model to locate the fluke in the image. A prototype fluke location system was implemented and results are presented.