Feature extraction as ellipse of wild-life images

Abstract

In this work we propose a new feature for automatic identification of individuals that belong to the species with spot patterns in their fur such as leopards, giraffes, cheetahs, and jaguars. We specifically worked with ocelots, kind of leopards found in America, using a collection of photographs taken with trap-cameras installed in the forest in Michoacan, Mexico. Our proposed feature extraction technique obtains for each spot automatically found in the fur the smallest ellipse that encloses the spot. Once the spots are replaced by ellipses, we only have to store for each ellipse, its center and a vector in the direction of its major axis. As a result, we end up with a collection of points and vectors which is exactly the same kind of features that are extracted in fingerprints after locating the minutiae which was precisely what motivated our work. We use two methods to determine the ellipses, one is based on geometric construction and the other one is based on the image moments. The preliminary experiments show that a better representation of the elliptical pattern is obtained when the method of moments is used.