Autonomous Underwater Navigation with 3D Environment Modeling Using Stereo Images

Abstract

A method for navigation of an autonomous underwater vehicle (AUV) based on visual odometry is described. Modifications of the method aimed at enhancing the accuracy of AUV localization and reducing computational costs are proposed. An algorithm that provides for a long tracking of image features and increases the accuracy in the calculation of AUV local motion is considered. An adaptive methodology for calculating the trajectory is proposed, as well as a method for AUV visual navigation in local maneuvering conditions, based on the use of a virtual coordinate network. A method for solving the problem of 3D reconstruction of objects from images, intended for underwater inspection operations, is described.