Computer Vision Applications in the Navigation of Unmanned Underwater Vehicles

Abstract

The inquisitive nature of humans has lead to the comprehensive exploration and mapping of land masses on planet earth, subsequently scientists are now turning to the oceans to discover new possibilities for telecommunications, biological & geological resources and energy sources. Underwater vehicles play an important role in this exploration as the deep ocean is a harsh and unforgiving environment for human discovery. Unmanned underwater vehicles (UUV) are utilised for many different scientific, military and commercial applications such as high resolution seabed surveying (Yoerger et al. 2000), mine countermeasures (Freitag et al. 2005), inspection and repair of underwater man-made structures (Kondo & Ura 2004) and wreck discovery and localisation (Eustice et al. 2005). Accurate vehicle position knowledge is vital for all underwater missions for correct registration between sensor and navigation data and also for control and final recovery of the vehicle. The characteristics of the underwater environment pose a plethora of difficult challenges for vehicle navigation and these obstacles differ greatly from the issues encountered in land, air and space based navigation (Whitcomb 2000). The rapid attenuation of acoustic and electromagnetic radiation in water restricts the range of acoustic and optical sensors and also limits communication bandwidth. As a consequence of this severe absorption acoustic and optical sensors require submersion near to the survey mission site to gather accurate high resolution data sets. The limitation on communication bandwidth means that vehicle autonomy can only be achieved when the large majority of computation is performed onboard. Whereas land based vehicles can rely on Global Positioning System (GPS) for accurate 3D position updates, the underwater equivalent acoustic transponder network is limited by range, accuracy, the associated cost and deployment & calibration time.