Segmentation and Classification of Coral for Oceanographic Surveys: A Semi-Supervised Machine Learning Approach
M. Johnson-Roberson · Suresh Kumar · S. Willams
TL;DR
A two-step method combines camera and acoustic data from underwater vehicles to segment and classify coral, producing coral density statistics during surveys.
BRIEF
Monitoring coral reefs requires knowing which coral species are present and how densely they grow. Autonomous underwater vehicles can collect camera and acoustic data over reefs, but turning that data into usable statistics is slow if done by hand. This work describes a two-phase method that first segments coral from the surrounding scene and then classifies it, combining visual and acoustic measurements. The aim is to gather coral density statistics during an AUV mission, supporting environmental monitoring of reefs.
Abstract
This work presents a technique for the autonomous segmentation and classification of coral through the combination of visual and acoustic data. Autonomous Underwater Vehicles (AUVs) facilitate the live capture of multi-modal sensor information about coral reefs. Environmental monitoring of these reefs can be aided though the autonomous extraction and identification of certain coral species of interest. The technique presented employs a two phase procedure of segmentation and classification to gather statistics about coral density during autonomous missions with an AUV.