| Summary: | Many types of ecological studies require identification of individual animals. I developed and evaluated an automated identification system for polar bears (Ursus maritimus) based on their whisker spot patterns. First, I measured the reliability of using whisker spot patterns for identification from polar bear photographs taken in western Hudson Bay. This analysis involved estimating the complexity of each whisker spot pattern in terms of its information content. I found that 98% of patterns contained enough information to be reliable, and this result varied little among three different observers. Based on these results, I implemented a computer-aided identification system for polar bears based on whisker spot pattern recognition. I used standard computer vision techniques to pre-process images and the Chamfer distance transform to compute similary scores between images. In addition, I evaluated the system by testing the effects of photographic quality and angle on system accuracy. I found that excellent and moderate quality/angle provided best results, with system accuracy of 90-95%. These findings suggest that individual identification of polar bears in the field based on whisker spot pattern variation is possible. Researchers studying polar bear behavior or estimating population parameters should benefit from this noninvasive technique. 2007-05-01 M.S. Sciences, Department of Biology Masters This record was generated from author submitted information.
|
|---|