Integrated approach to analyse benthic images from an autonomous underwater vehicle deployed at Pemba Island, Tanzania

Abstract

Manual analysis of large amounts of benthic images is time consuming and costly. This challenge has led to the development of automated image analysis techniques such as CoralNet. The technique combines an online repository and machine learning to completely or partially automate classification of benthic images. Here, the integration of Coral Point Count with Excel Extensions (CPCe) and CoralNet is showcased. CPCe was applied to manually annotate images captured by an autonomous underwater vehicle (AUV) deployed at Pemba Island, Tanzania and then to train and build confidence in CoralNet to automatically annotate more images. Further, possible outputs that can be derived from assessing the relationships between the cover of benthic variables and depth are demonstrated.