Both Remotely Operated underwater Vehicles (ROVs) and Autonomous Underwater Vehicles (AUVs) are frequently deployed to acquire geometric bathymetric data. However, it is often discovered post-survey that the acquired data coverage is incomplete. Given the high operational cost associated with underwater deployments, it is essential to incrementally visualize surface coverage in real-time to support informed decision-making by both the operators of ROVs and the AUVs during data collection. In addition, traditional incremental surface reconstruction methods, such as Digital Terrain Models (DTMs), are inherently limited in expressiveness: they represent surfaces as height fields, allows only one elevation value per (x, y) coordinate and thus cannot capture overhangs or vertical structures. To overcome these limitations, we adapt the original Ball Pivoting Algorithm (BPA) into an incremental, real-time, and free-form surface reconstruction method, referred to as Incremental BPA (IBPA). Our method incrementally constructs an orientable, manifold mesh from streaming point cloud data without imposing assumptions regarding point cloud overlap or spatial distribution. Furthermore, we introduce a hole detection mechanism that identifies and highlights incomplete mesh regions. Compared to existing approaches, our method supports more complex surface topologies without prior structural assumptions. The source code of our reference implementation is available: https://github.com/Mauhing/Incremental-BPA

The exploration and documentation of underwater archaeological and historical sites have been significantly enhanced by the utilization of marine robotics of various sizes, propulsion capabilities, sensor configurations, and autonomy levels. Highly flexible and articulated underwater vehicles, also referred to as kinematically hyper-redundant robots, have garnered a lot of interest in subsea operations in the last few years, thanks to their high degree of maneuverability and adaptability to the environment they are navigated into. This article presents the integrated photogrammetric mapping of an underwater cultural heritage site using an underwater snake robot for the first time. The snake robot Eely submerged down to the 20th century wreck Figaro, 35 m deep, in Trygghamna fjord, in the Svalbard archipelago. A multidisciplinary team of engineers, marine archaeologists, and surveyors cooperated on this purpose within the wider framework of a shipwreck mapping campaign in the High Arctic. While the snake robot is capable of fully autonomous untethered operations, for this mission, the 6-m-long vehicle was deployed as a remotely operated vehicle so that real-time mapping and online decision-making were feasible. The Eely platform's photogrammetric concept, that is described in detail in the paper, is based on its ability to change shapes depending on the geometry of the object or area of mapping interest. A set of multiple cameras and lights was mounted along the three segments of the articulated robot aiming for visual data acquisition and photogrammetric documentation, while a set of acoustic sensors was used to enhance global situational awareness on the site and provide real-time information about the progress of the mission to the supervising archaeologist through augmented maps. A multibeam echosounder, mounted on the bottom of the middle segment of the robot, allowed for status updates of the site's coverage, while a forward-looking sonar allowed for obstacle avoidance. In this paper, we provide a detailed description of the methodology used for the wreck survey covering each stage from mission planning (mission's scope, definition of sub-tasks, platform and sensors configuration, and design and simulation of the robot's transects over the wreck), to the real-world operations and data acquisition, and, finally, results. We compare our findings with previous missions at the same wreck site, both qualitatively and quantitively, and finally discuss how such innovative technology, like snake robots, can bring marine archaeological mapping to a new level.

While the use of neural radiance fields (NeRFs) in different challenging settings has been explored, only very recently have there been any contributions that focus on the use of NeRF in foggy environments. We argue that the traditional NeRF models are able to replicate scenes filled with fog and propose a method to remove the fog when synthesizing novel views. By calculating the global contrast of a scene, we can estimate a density threshold that, when applied, removes all visible fog. This makes it possible to use NeRF as a way of rendering clear views of objects of interest located in fog-filled environments. Additionally, to benchmark performance on such scenes, we introduce a new dataset that expands some of the original synthetic NeRF scenes through the addition of fog and natural environments. The code, dataset, and video results can be found on our project page: https://vegardskui.com/fognerf/

This paper presents a novel method for the visual-based 3D mapping of underwater cultural heritage sites through marine robotic operations. The proposed methodology addresses the three main stages of an underwater robotic mission, specifically the planning phase, the mission-time and the offline processing phase. Initially, we approach mission planning through multi-vision sensor configurations and simulations of the underwater medium’s effects. Subsequently, we demonstrate a possibility for real-time 3D surface reconstruction and hole detection by using Poisson Surface Reconstruction (PSR) and the Ball Pivoting Algorithm (BPA), that allows for real-time quality assessment of the acquired data and control of the coverage of the site. Last, an offline photogrammetric workflow is discussed in terms of geometric reliability and visual appearance of the results. The presented three-step methodological framework has been developed and tested in both simulation and real-world environments for three wreck sites in the fjord of Trondheim, Norway, introducing among others novel marine robotic technology like the articulated robot Eelume.

Underwater exploration and related research face significant challenges, including incomplete data collections and the high cost of operations. An underwater robot's capacity to independently identify, target, and address gaps in 3D surface data, termed as 'holes', during a mission has the potential to significantly lower the expenses associated with research endeavors, such as those in underwater archaeology or subsea monitoring, and to improve sea surface mapping for autonomous robots. This study introduces a novel approach that combines the boundary-edge of a triangle mesh with Principal Component Analysis (PCA) to actively gather incomplete data. We propose two distinct techniques for active hole-filling, as well as a method to determine the most appropriate strategy. We introduce the flying-over and following-rail approaches to automatically determine new sensor positions for collecting additional point cloud data, which are then used to reconstruct the surface and fill the gaps. Experiments were conducted using both simulated underwater images with ORB-SLAM3 (Campos et al. (2021)) and real-world Multibeam Echo Sounder data. Copyright (c) 2024 The Authors. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0/)

In this work, we present a boundary and hole detection approach that traverses all the boundaries of an edge-manifold triangular mesh, irrespectively of the presence of singular vertices, and subsequently determines and labels all holes of the mesh. The proposed automated hole-detection method is valuable to the computer-aided design (CAD) community as all boundary-edges within the mesh are utilized and for each boundary-edge the algorithm guarantees both the existence and the uniqueness of the boundary associated to it. As existing hole-detection approaches assume that singular vertices are absent or may require mesh modification, these methods are ill-equipped to detect boundaries/holes in real-world meshes that contain singular vertices. We demonstrate the method in an underwater autonomous robotic application, exploiting surface reconstruction methods based on point cloud data. In such a scenario the determined holes can be interpreted as information gaps, enabling timely corrective action during the data acquisition. However, the scope of our method is not confined to these two sectors alone; it is versatile enough to be applied on any edge-manifold triangle mesh. An evaluation of the method is performed on both synthetic and real-world data (including a triangle mesh from a point cloud obtained by a multibeam sonar). The source code of our reference implementation is available: https://github.com/Mauhing/hole-detection-on-triangle-mesh.

Conventional SLAM methods which work very well in typical above-water situations, are based on detecting key-points that are tracked between images, from which ego-motion and the 3D structure of the scene are estimated. However, in underwater environments with marine snow — small particles of organic matter which are carried by ocean currents throughout the water column — keypoint detectors are prone to detect the marine snow particles. As the vast majority of SLAM front ends are sensitive against outliers, and the marine snow acts as severe "motion noise", failure of the regular egomotion and 3D structure estimation is expected. For this reason, we investigate the structure and appearance of marine snow and developed two schemes which classify keypoints into "marine snow" or "clean" based on either the image patches obtained from usual keypoint detectors or the descriptors computed from these patches. This way the subsequent SLAM pipeline is protected against ’false’ keypoints. We quantitatively evaluate the performance of our marine snow classifier on both real underwater video scenes as well as on simulated underwater footage that contains marine snow. These simulated image sequences have been created by extracting real marine snow elements from real underwater footage, and subsequently overlaying these on "clean" underwater videos. Qualitative evaluation is also done on a night-time road sequence with snowfall to demonstrate applicability in other areas of autonomy. We furthermore evaluate the performance and the effect of marine snow detection & suppression by integrating the snow suppression module in a full SLAM pipeline based on the pySLAM system.

Underwater visual perception requires being able to deal with bad and rapidly varying illumination and with reduced visibility due to water turbidity. The verification of such algorithms is crucial for safe and efficient underwater exploration and intervention operations. Ground truth data play an important role in evaluating vision algorithms. However, obtaining ground truth from real underwater environments is in general very hard, if possible at all.In a synthetic underwater 3D environment, however, (nearly) all parameters are known and controllable, and ground truth data can be absolutely accurate in terms of geometry. In this paper, we present the $VAROS$ environment, our approach to generating highly realistic under-water video and auxiliary sensor data with precise ground truth, built around the Blender modeling and rendering environment. $VAROS$ allows for physically realistic motion of the simulated underwater (UW) vehicle including moving illumination. Pose sequences are created by first defining waypoints for the simulated underwater vehicle which are expanded into a smooth vehicle course sampled at IMU data rate (200 Hz). This expansion uses a vehicle dynamics model and a discrete-time controller algorithm that simulates the sequential following of the waypoints.The scenes are rendered using the raytracing method, which generates realistic images, integrating direct light, and indirect volumetric scattering. The $VAROS$ dataset version 1 provides images, inertial measurement unit (IMU) and depth gauge data, as well as ground truth poses, depth images and surface normal images.

This report presents results obtained in the CageReporter project regarding the development of a 3D vision system to be used for data capture in fish cages. The developed system enables to obtain high-quality data with the overall goal to identify fish conditions and perform cage inspections during daily operations, as well as the robotic vision for an underwater vehicle during the adaptive operation planning in the cage. A compact and robust sensor with optical components and lighting system was developed. In addition, this activity presents development of methods to evaluate the quality of the captured data. Based on defined quality criteria associated with fish conditions and cage inspection operations, algorithms have been developed to evaluate whether the quality criteria are met. The algorithms have been validated using image data obtained from 24/7 video streams from a full-scale fish cage. The work furthermore includes the development of image processing algorithms to estimate the distance and orientation relative to the inspected object of interest, such as the fish or the net. The developed algorithms have been validated based on vision data obtained during tests both in laband full scale. PREPARED BY Eleni Kelasidi CHECKED BY Nina Bloecher APPROVED BY Leif Magne Sunde This document has been approved according to SINTEF's approval procedure, and is digitally secured