Abstract
Posidonia oceanica meadows are a species of seagrass highly dependent onrocks for their survival and conservation. In recent years, there has been aconcerning global decline in this species, emphasizing the critical need forefficient monitoring and assessment tools. While deep learning-based semanticsegmentation and visual automated monitoring systems have shown promise in avariety of applications, their performance in underwater environments remainschallenging due to complex water conditions and limited datasets. This paperintroduces a framework that combines machine learning and computer visiontechniques to enable an autonomous underwater vehicle (AUV) to inspect theboundaries of Posidonia oceanica meadows autonomously. The frameworkincorporates an image segmentation module using an existing Mask R-CNN modeland a strategy for Posidonia oceanica meadow boundary tracking. Furthermore, anew class dedicated to rocks is introduced to enhance the existing model,aiming to contribute to a comprehensive monitoring approach and provide adeeper understanding of the intricate interactions between the meadow and itssurrounding environment. The image segmentation model is validated using realunderwater images, while the overall inspection framework is evaluated in arealistic simulation environment, replicating actual monitoring scenarios withreal underwater images. The results demonstrate that the proposed frameworkenables the AUV to autonomously accomplish the main tasks of underwaterinspection and segmentation of rocks. Consequently, this work holds significantpotential for the conservation and protection of marine environments, providingvaluable insights into the status of Posidonia oceanica meadows and supportingtargeted preservation efforts