Personal tools
You are here: Home / Posters / York


'Development of an Optical Image Database and Processing Tools for Automated Classification of Benthic Habitat and Enumeration of Scallop Stocks'; Amber York, Richard Taylor, Norman Vine, Steve Lerner, Dvora Hart, Lakshman Prashad, and Scott Gallager - USA

The goal of this project is to develop a generic software package for benthic habitat characterization and measurement of sea scallop (Placopecten magellanicus) size and abundance. The package is designed to allow input of a stream of individual images representing output from a still or video camera mounted on a sled, AUV or other platform. Each image is sequentially processed for color correction, light field illumination, background and foreground segmentation based on texture and color, and extraction of segmented regions of interest (ROI). The ROIs are then passed to a classifier (Support Vector Machine) that has been trained with a library of manually identified targets including scallops, fish, clams, mussels, sea stars, sand dollars, etc. Associations of specific targets with texture category are made through discriminate analysis. Size, shape and density (number of individuals per unit area) of scallops and other organisms are added to a georeferenced database for display and visualization. The software package provides an image and data archive capable of making results available to interested parties over the internet. Automated characterization of bottom substrate and quantification of benthic organism distributions, including invasive species, will provide a better understanding of the benthic components of areas currently classified as essential fish habitat and marine ecosystems in general.

Document Actions

« May 2017 »