Abstract
The operation of intelligent or autonomous undersea robotic vehicles will require the employment of various sensors for accurate positioning and localization, detection, and identification of lost objects, targets, and (or) nearby obstacles. While a range of technologies has been developed and has been in operation a decade or more, sensing based on visual cues is a crucial capability, which is yet to be developed. The task involves computer processing of two-dimensional images of a scene in order to extract certain useful information about the environment. This refers to information which will enable the vehicle to perform its task, and (or) interact with the environment successfully. While the development of this technology for land or space applications has motivated nearly three decades of extensive research in computer vision, it has been recognized that the achievement of similar capabilities for undersea applications is a much more challenging task.