Online video mosaicing through SLAM for ROVs

This paper focuses on preliminary results of a system for online video mosaicing for ROVs operating in the proximity of the seabed. The proposed approach relies on a simultaneous localisation and mapping (SLAM) system already tested in typical operating conditions with the Romeo ROV. In particular, previous experimental activity demonstrated the possibility of defining suitable unambiguous features, i.e. image templates, to be extracted automatically by the system and used as SLAM landmarks. The mosaic is now constructed positioning the collected images according to the SLAM-based vehicle motion estimation. It is worth noting that dead-reckoning navigation is performed using data supplied by a laser triangulation optical correlation (LTOC) sensor, for the estimate of the camera/vehicle linear motion on the basis of monocular video data.