Visual Navigation for the Inspection of Industrial Underwater Structures Using a Virtual Coordinate Network

Abstract

Inspection of underwater industrial infrastructure facilities using autonomous uninhabited underwater vehicles (AUVs) requires high accuracy of AUV navigation relative to the objects being inspected. Along with traditional navigation aids based on the inertial navigation system and sonars, technologies based on video information processing are being actively developed today. Methods based on visual odometry can provide higher navigation accuracy in conditions of local maneuvering at small distances to objects. However, during long-term movements of the AUV, such methods tend to accumulate errors in the calculation of the trajectory. Therefore, this paper proposes a navigation method that can significantly improve the accuracy of AUV movement in the coordinate space of the inspected object due to the generation and use of a virtual network of coordinate reference to an underwater object.