The presented review illustrates the growing popularity of vision based UUV navigation methods. The diversity of optical imaging has been demonstrated with regard to applications of cable tracking and inspection, mosaicking, positioning & localisation and sensor fusion. Vision systems are a very useful sensor for the navigation of underwater vehicles and have been the subject of increased interest over the last decade as a result of improved processing capabilities of hardware and the need for more flexible and accurate sensor solutions. The increase in research efforts into vision systems is due to its inexpensive nature and its common inclusion on underwater vehicles as a payload sensor. Vision also has a number of advantages over other types of sensors for underwater applications.
Cameras are light weight and do not possess a minimum operating range unlike their acoustic counterparts (Nolan 2006). Despite these advantages over other sensors, machine vision underwater poses an amount of difficult challenges to be overcome for it to be successfully incorporated into control (Matsumoto & Ito 1995). Marine snow, low contrast, non-uniform illumination and a lack of distinguishable features on the seabed are just some of the inherent difficulties faced when using optics underwater (Cufi et al. 2002). Kinsey et al. indentified that solutions and experimental results for underwater vehicle navigation in the x and y horizontal plane were particularly rare in literature (Kinsey & Whitcomb 2004). This is of particular interest as vision is a very useful sensor for horizontal plane navigation in the correct conditions while near the seafloor. It was later stated by Kinsey et al. that there is a distinct need for improved near seabed (near intervention) UUV navigation systems for the exploitation of scientific data and the near seabed operations (Kinsey et al. 2006). Current systems while sufficiently precise and fast for dynamic positioning remain unconvincing in near intervention operations. Vision systems have inherently fast frame rates (update rates) depending on the capability of both the hardware and software while also having resolutions of sub centimetre accuracy depending on vehicle altitude and resolution. These attributes as well as the other advantages discussed previously make vision based methods ideal for near intervention class missions. Sensor fusion techniques will allow for more complementary synthesis of vision algorithms with DVL, INS, compass and altimeters proven to provide a more robust navigation solution. While online processing of optical data has been an issue in the past improving hardware capabilities should now allow for real-time implementation of vision based algorithms with sufficient accuracy and update rates.
Was this article helpful?