Real-time Monocular Visual Odometry for Turbid and Dynamic Underwater\n Environments

Abstract

In the context of robotic underwater operations, the visual degradations\ninduced by the medium properties make difficult the exclusive use of cameras\nfor localization purpose. Hence, most localization methods are based on\nexpensive navigational sensors associated with acoustic positioning. On the\nother hand, visual odometry and visual SLAM have been exhaustively studied for\naerial or terrestrial applications, but state-of-the-art algorithms fail\nunderwater. In this paper we tackle the problem of using a simple low-cost\ncamera for underwater localization and propose a new monocular visual odometry\nmethod dedicated to the underwater environment. We evaluate different tracking\nmethods and show that optical flow based tracking is more suited to underwater\nimages than classical approaches based on descriptors. We also propose a\nkeyframe-based visual odometry approach highly relying on nonlinear\noptimization. The proposed algorithm has been assessed on both simulated and\nreal underwater datasets and outperforms state-of-the-art visual SLAM methods\nunder many of the most challenging conditions. The main application of this\nwork is the localization of Remotely Operated Vehicles (ROVs) used for\nunderwater archaeological missions but the developed system can be used in any\nother applications as long as visual information is available.\n