On shape and range recovery from images of underwater environments

Abstract

Unmanned underwater vehicles require some sense of perception of the environment so as to carry out their missions with some degree of autonomy. Perception based on visual cues can tremendously enhance the performance of such vehicles. For example, the capability to determine the three-dimensional position or shape of a nearby object from two-dimensional images can be useful for localization, recognition, or obstacle avoidance. In this paper, we consider selected methods for determining the three-dimensional shape of, or range to, objects from two-dimensional images for application in the underwater domain. In contrast to techniques commonly used in land environments based on tracking points, lines, contours, or other isolated scene features, these methods rely on the information in the brightness patterns of the whole image. Selected experimental results are provided to demonstrate the concepts.