Automatic classification and recognition of coral reefs from underwater video images

Abstract

This thesis presents a new approach to automate the classification and estimation of coral reef based on computer vision techniques by performing classification and recognition on digitized video obtained from surrounding reef ecosystem. By knowing the state of the coral reefs, marine scientist can predict fish yield or determine how the reef system is coping with the environment in a less tedious and less time consuming way. In this project, the objective is to classify and recognize a predetermined reef benthos groups. These groups are alive coral, dead coral, sand and rubble. As for the alive coral group, possible higher level of taxonomic classification is performed to discriminate the classes of branching, digitate and tabulate solely based on digital video recording.