“An Automatic Identification and Monitoring System for Coral Reef Fish” by Edward Pavoni, Andrew Rapport, Serena Tsang, Lev Barinov, Jigesh Baxi, James Bibby and Prof. Joseph Wilder (Rutgers), Prof. Gareth Russell (NJIT).
To help gauge the health of ecosystems on coral reefs, we developed a proof-of-concept prototype of an underwater camera module to automatically census coral reef fish populations. Recognition challenges include pose and lighting variations, complicated backgrounds, within-species color variations and small within-family differences between species (Fig. 1). During an early stage of the program, seniors Lev Barinov, Jigesh Baxi, and James Bibby carried out a project involving an epipolar analysis for the special case two cameras at a 45 degree angle looking through the ½” thick polycarbonate walls of an aquarium to determine the depth, and pose of free-swimming fish and disambiguate overlapping fish. At a later stage of the program, when a fully submersible prototype was tested at the New York Aquaium (Fig. 2) Edward Pavoni, Andrew Rapport, and Serena Tsang, carried out a Capstone project involving analysis of similar species that could be confused when overall shape and color features were insufficient for distinguishing them from each other. Using localalized pattern and color features within the similar species improved the recognition rate significantly. During the course of the project ECE graduate students Chetan Tonde and Ganesh Sundar helped in mentoring the undergraduate researchers. All of these students are listed as co-authors in a paper to be delivered in August at the SPIE "Applications of Digital Image Processing XXXV" conference in August, 2012.