Publication Date

2012-05-29

Availability

Open access

Embargo Period

2012-05-29

Degree Name

Master of Science (MS)

Department

Computer Science (Arts and Sciences)

Date of Defense

2012-04-13

First Committee Member

Shahriar Negahdaripour

Second Committee Member

Mitsunori Ogihara

Third Committee Member

Victor Milenkovic

Abstract

The research of underwater creatures is a fascinating work to marine biologists and environmental experts. However, finding and segmenting underwater creatures in underwater images and videos is time consuming. Automated object detection and segmentation can be applied to such problems to accelerate the tedious process. This work describes an approach to detect fish from images or videos of benthic habitats, recorded in Virgin Islands, during the yellow-tail grouper spawning activities. The method involves first locating the horizon, separating the seawater and the seafloor, and then using different visual cues to detect fish both within the water column and over the seafloor. The detector can be applied to video data, or a single image (forgoing the visual motion cues). The precision and recall rates of 86.4% and 94.5% have been achieved.

Keywords

fish detection; underwater video; benthic habitats

Share

COinS