Real-time optical SLAM-based mosaicking for unmanned underwater vehicles

Abstract

This article discusses the possibility of building in real-time a mosaic of the seafloor relying on a simultaneous localization and mapping (SLAM) framework. The goal is to provide an unmanned underwater vehicle with a relatively rough visual map of the seafloor to support basic navigation and context awareness. To achieve that goal, an accurate estimation of the location of the visual landmarks and, in particular, the correct data association when a visual landmark is re-visited by the vehicle are the crucial points. Instead of using a global mosaic, this work uses the combination of a set of local mosaics constructed in the vicinity of the SLAM visual landmarks. The contributions of this article are mainly the use of SURF features, the local mosaics approach and the real-time capability. The use of SURF features allows eliminating false positives in the data association of SLAM visual landmarks. The local mosaics approach is an effective way of correcting the effects of the drift on the mosaic in real time. The main contribution is the real-time capability as it will be seen. The algorithm was tested using a batch of experimental data in typical operating conditions and the results prove the effectiveness of the approach. Electronic supplementary materialThe online version of this article (

Binary visual features for ROV motion estimation

“…Future work includes testing the same approaches in more complex applications, namely SLAM and mosaicking. The SLAM-based mosaicking presented in [13] was already updated with the use of BRIEF features and this work is currently under submission. [25].…”

“…This is important if the algorithm is applied in a real time context. In this work, binary features are tested only for motion estimation but this in fact is just a small step of a more complex real-time SLAMbased mosaicking algorithm presented in [13]. Hence, the need of having this small step fast and efficient arises.…”

Challenges and State-of-the-Art Solutions to Underwater Slam

“…The next year brought the work of [29], which discusses the possibility of building a mosaic of the seabed in real-time, dependent on a EKF SLAM framework. The use of visual features through the SURF algorithm allows the elimination of false positives in the data association of visual landmarks.…”

Enhancing autonomous capabilities and human-robot interaction for Unmanned Surface Vehicles

“…As far as the authors could find, there is no other USV with a similar approach. The motivation for this feature has to do with successful work done in real-time mosaicking of the seafloor using a ROV navigating near the seabed described in [13]. Obtaining a rough still usable mosaic of the seabed in real time allows the operator (in the ROV case) or the end-user (in the USV case) to plan and adjust missions during the mission itself.…”

“…More than 75GB of video data were collected during the mission. Moreover, the work described in [13] will need to be improved to deal with some issues that did not arise while working with an ROV near the seabed. The change of illumination is one of the most important because it influences the contrast of the image and therefore the laser spots detection.…”