Towards Visual Detection, Mapping and Quantification of Posidonia Oceanica using a Lightweight AUV

Abstract

Posidonia Oceanica (P.O.) is a Mediterranean endemic seagrass strongly related to the health of the coastal ecosystems. Monitoring the presence and state of P.O. is essential not only for safeguarding the shallow-water life diversity, but also as an indicator of the water quality. Nowadays, the control of P.O. is done by divers in successive missions of a duration limited by the capacity of the scuba tanks. This paper proposes the application of robotic and computer vision technologies to upgrade these current methods, namely: 1) employing a lightweight Autonomous Underwater Vehicle (AUV) equipped with cameras to survey and image marine areas, 2) the automatic discrimination of P.O. from the rest of the seafloor, using several techniques based on image texture analysis and machine learning, and, 3) the fast computation of 2D maps (photo-mosaics) of the surveyed areas from all the images included in the grabbed video sequences; these mosaics are extremely useful to measure the real extension of the meadows and some of the descriptors needed for a biological analysis. Experiments conducted with an AUV in several marine areas of Mallorca reveal promising results in the discrimination of different patterns of P.O. and in the construction of highly realistic photo-mosaics of the surveyed areas.