Spot detection and pattern matching for individual whale shark identification
Loading...
Authors
Brennan, Nicholas
Issue Date
2017-12
Type
Electronic thesis
Thesis
Thesis
Language
ENG
Keywords
Computer science
Alternative Title
Abstract
Photographic identification of individual animals is a vital tool for conservation efforts. We produce an algorithm to match spot patterns in whale sharks by segmenting a portion of the animal, detecting the locations of its spots, and matching them between images. We make use of convolutional neural networks to perform segmentation and spot detection, and a modification of the RANSAC algorithm for point matching. Our results show that we are able to accurately extract spot locations from the segmented images. However, segmentation is not reliable enough to replace the need for human interaction, and our matching algorithm did not produce effective results.
Description
December 2017
School of Science
School of Science
Full Citation
Publisher
Rensselaer Polytechnic Institute, Troy, NY