Shallow Water Seagrass Survey at Studland Bay with the AUV Smarty200

This paper presents the results of an autonomous seafloor imaging survey to map the distribution of seagrass around 10 eco-moorings installed to protect seagrass meadows in the Studland Bay Marine Conservation Zone (MCZ). The survey was carried out using the University of Southampton’s Smarty200 Autonomous Underwater Vehicle (AUV) in July 2022. Approximately 10,000 stereo image pairs were gathered from an altitude of 1 m along transects totaling 2.95 km. Images were classified according to the density of seagrass using a location- regularised semi-supervised deep-learning method developed at the University of Southampton. Three hundred expert-labelled images were used to train the classifier and the accuracy of the results were evaluated on a separate set of two hundred expert-labelled images. The results show the detailed distribution of habitats at the site and qualitative comparisons with high- resolution satellite imagery are made.