Cover Picture and Issue Information

Abstract

This month's cover image features a 72 square meter orthomosaic of a coral reef automatically segmented and classified using RapidBenthos. Developed by Remmers et al., this innovative workflow uses machine learning for feature extraction and analyses of photogrammetric data from underwater orthomosaics of coral reefs. The automated workflow integrates the Segment Anything Model (SAM) and ReefCloud point annotation in a two-stage process: (1) employing a pre-trained, open-source machine learning segmentation model, which removes the need for users to manually generate fine-scale segmented training data, and (2) classifying the resulting segments using the underlying survey images from multiple viewpoints, achieving classification at higher taxonomic levels. This study demonstrates that artificial intelligence tools can automatically and reliably extract data from orthomosaics with an unprecedented level of taxonomic detail and accuracy. While the approach has been developed and tested on images from a shallow coral reef, it has the potential to be applied to any ecosystem monitored via photogrammetry and is scalable for large-area applications. Such advancements enable the sustainable scaling of photogrammetric monitoring techniques, offering a more comprehensive understanding of coral reefs community composition and habitat. In turn, this will inspire new research questions, enhance ecosystem models, and support improved ecosystem management.