Sammendrag
Low-cost ROVs are currently entering the market en masse, and providers face a great challenge in keeping the cost low while providing the best user experience. For this purpose, simple sensors and advanced software is needed. We explore the possibility of monocular camera-based Simultaneous Localization And Mapping(SLAM) on an ROV to achieve dynamic positioning. The main challenge of monocular SLAM is lack of scale, due to a camera being a bearing-only sensor. Thus, it tells us nothing about the scale of our map or the scale of the ROV position within the map. We implement the ORB-SLAM algorithm from (Mur-Artal et al., 2015), along with laser range measurements to estimate scale. The position estimate from SLAM is used for feedback control of uDrone, a low-cost ROV built by students at the Marine Technology department of the Norwegian University of Science and Technology. We present and discuss challenges, methods and results form the work, concluding on the feasibility of underwater monocular SLAM for dynamic positioning.
Vis fullstendig beskrivelse
