underwater mapping

Underwater mapping 2024

Authors: Shuo Pang, Ye Li, Liang Xiao, Francisco Rego, & Teng Ma This paper intends to inform and highlight the importance and advancements of unmanned marine vehicles (USVs and UUVs) in ocean observation, particularly in extreme environments like the deep sea and polar regions. It aims to showcase recent technological advancements and future implications in areas such as sensing, control, navigation, and communication of these vehicles, emphasizing their role in enhancing the quantity and quality of oceanographic data collected.

Authors: Danielle F. Morey, Randall S. Plate, Cherry Y. Wakayama, & Zelda B. Zabinsky This study focuses on optimizing maritime survey operations' topology using unmanned underwater vehicles (UUVs). It employs a multi-fidelity approach to evaluate different configurations for assigning UUVs to data collection sensors or locations. That involves using both low-fidelity models, which simplify assumptions to allow for sensitivity analysis with low computational cost, and a high-fidelity simulation model, which provides detailed performance metrics under more realistic conditions.

Authors: Tianchi Zhang and Yuxuan Liu This text describes a new method for enhancing underwater images, specifically designed for use with Autonomous Underwater Vehicles (AUVs). It outlines the challenges faced with current methods, such as high computational demands and complex network structures, and introduces a novel approach using a multi-teacher knowledge distillation GAN (MTUW-GAN) to improve image quality while being resource-efficient, making it suitable for real-time deployment on AUVs. The text also highlights the advantages of this method over existing approaches in terms of visual quality, model parameters, and performance.

Authors: Jiapeng Liu, Te Yu, Chao Wu, Chang Zhou, Dihua Lu, and Qingshan Zeng This document validates a new underwater integrated navigation system for Autonomous Underwater Vehicles (AUVs) that is cost-effective, lightweight, and compact compared to traditional systems. It describes the development and testing of this system, highlighting its advantages in terms of precision and cost, as well as its potential applications in both military and civilian contexts.

Authors: Rebeca Chinicz, and Roee Diamant This paper describes a research study focused on improving the detection capabilities of Synthetic Aperture Sonar (SAS) in autonomous underwater vehicle (AUV) surveys by combining SAS with optical images. It addresses the challenges of matching multimodal images (optical and acoustic) by proposing an entropic method for recognizing matching images of the same object. It also investigates the probabilistic dependency between the two modalities and presents results that demonstrate improved detection and false alarm rates.

Authors: Dapeng Zhang, Bowen Zhao, Yi Zhang, and Nan Zhou This document intends to explain the process and importance of determining hydrodynamic coefficients for remotely operated vehicles (ROVs) used in marine engineering. It discusses the advantages of ROVs, the necessity of establishing a motion mathematical model, and the use of simulation techniques to predict performance and guide design improvements to convey the significance of accurate hydrodynamic simulations in enhancing the maneuverability and design of ROVs.

$12 - Underwater Camera Calibration Based on Double Refraction$

Authors: Yushan Sun, Tian Zhou, Liwen Zhang and Puxin Chai This document describes advancements in the methodology for calibrating underwater cameras to improve measurement accuracy in underwater environments. It outlines the challenges of double refraction errors and presents a solution involving a double refraction model and a parameter optimization method using genetic algorithms to convey the effectiveness of the proposed method in achieving a more precise determination of underwater camera parameters compared to other algorithms.

Authors: Taoran Lu, Su Qiu, Hui Wang, Shihao Zhu and Weiqi Jin This text presents a research solution for improving underwater imaging using single-photon avalanche diode (SPAD) devices. The text outlines the challenges associated with SPAD devices, such as high production costs and small array areas, and proposes a method to simulate SPAD data and develop a denoising network using deep learning to remove backward-scattering interference. It also highlights the effectiveness of this approach by discussing the experimental results and improvements in specific performance metrics.

Authors: Wenqiang Zhang, Xiaobing Chen, Xiangwei Zhou, Jianhua Chen, Jianguo Yuan, Taibiao Zhao, and Kehui Xu This study introduces a new dataset and a classification model for identifying geomorphic features in dredge pit environments using sidescan sonar (SSS) images and deep learning methods. It aims to highlight the effectiveness and efficiency of the new Effective Geomorphology Classification (EGC) model, discuss the utility of transfer learning and data augmentations, and emphasize the value of the dataset and model for hazard monitoring and other machine learning applications.

Authors: Qiang Fu, Chao Dong, Kaikai Wang, Qingyi He, Xiansong Gu, Jianhua Liu, Yong Zhu, Jin Duan This text describes the development and effectiveness of a new underwater laser polarization detection technology. It outlines the challenges addressed by this technology, the proposed solution, and the results of experiments conducted to validate the device's accuracy and precision in detecting underwater targets.

Authors: Ziming Chen, Jinjin Yan, Ruen Huang, Yisong Gao, Xiuyan Peng, and Weijie Yuan The text describes a new path-planning algorithm for autonomous underwater vehicles (AUVs) that effectively addresses the challenges posed by ocean currents. It outlines the limitations of conventional algorithms and introduces a novel approach that combines elements of the influences and constraints of ocean currents, which leverages the strengths of two widely employed path- planning algorithms and genetic algorithms to improve path planning by considering ocean current influences.

Authors: Yang Zhang, Qingchao Xia, Canjun Yang, Ruiyin Song, Dingze Wu, Xin Zhang, Rui Zhou, and Shuyang Ma The intent of the text is to describe the development and evaluation of a miniaturized underwater profiler system for ocean observation. It outlines the advantages of such profilers, the design and modeling process, and the implementation of a new algorithm (DSRCKF) for dead reckoning. The text also compares this algorithm with others and discusses its effectiveness, particularly in certain directions and attitudes. The intent is to inform about the research and development process, the technical details of the system, and the future plans for real-world testing.

Authors: Zeyang Liang, Kai Wang, Jiaqi Zhang, and Fubin Zhang The intent of this text is to present a study focused on improving the positioning accuracy of autonomous underwater vehicles (AUVs) through the development of a novel method for visual simultaneous localization and mapping (SLAM) in underwater environments. The text outlines the proposed approach, which involves an underwater multisensor fusion SLAM system based on image enhancement, and highlights the effectiveness of this method in reducing errors compared to other algorithms.

Authors: Tong Liu, Sainan Zhang, and Zhibin Yu This paper outlines the challenges current methods face due to irregular illumination in underwater environments, introduces a proposed solution, an underwater self- supervised monocular depth estimation network, and presents a new approach for improving underwater depth estimation. This approach integrates image enhancement and auxiliary depth information to address issues like low-light conditions and overexposure.

Authors: Rupeng Wang, Jiayu Wang, Ye Li, Teng Ma, and Xuan Zhang This document informs researchers about the current state and future developments in the field of underwater terrain-aided navigation (TAN) to provide a comprehensive overview of the application background, operating principles, technical aspects, and relevant algorithms involved in underwater TAN, as well as to summarize cutting-edge issues in the field.

Authors: Quanhong Ma, Shaohua Jin, Gang Bian, and Yang Cui This document describes a proposed solution to improve the accuracy of multi-scale seafloor target detection in side-scan sonar images, which are challenged by high noise and complex background textures. It outlines the development and implementation of a model using the BES-YOLO network, detailing the enhancements made to the YOLOv8 network, such as the introduction of an efficient multi-scale attention mechanism, a bi-directional feature pyramid network, and a Shape_IoU loss function. It also discusses the preprocessing of the dataset to enhance network robustness and presents experimental results that demonstrate improved detection accuracy.

Authors: Qing Li, Hongjian Wang, Yao Xiao, Hualong Yang, Zhikang Chi, and Dongchen Dai This article introduces an unsupervised stereo matching method based on semantic attention, addressing the challenge of insufficient training data. It aims to improve the intelligence of underwater robots and advance scientific research and marine resource development. It highlights the design of an adaptive double quadtree semantic attention model and an unsupervised semantic loss function, emphasizing the method's robustness and improved performance metrics through evaluations.

33 - Enhancing Underwater Object Detection and Classification Using Advanced Imaging Techniques: A Novel Approach with Diffusion Models

Authors: Prabhavathy Pachaiyappan, Gopinath Chidambaram , Abu Jahid, and Mohammed H. Alsharif This study proposes a new approach to improve underwater object detection and classification to address challenges posed by environmental factors such as water turbidity and variable lighting conditions by integrating advanced imaging techniques with diffusion models. This research aligns with Sustainable Development Goal (SDG) 14: Life Below Water, and it introduces a new imaging technique, AIT-YOLOv7, which enhances the accuracy of detecting and classifying underwater objects. It highlights the methodology, the use of the TrashCan dataset for validation, and the significant improvement in detection accuracy compared to existing methods.

Authors: Hao Feng, Yan Huang, Jianan Qiao, Zhenyu Wang, Feng Hu, and Jiancheng Yu This study focuses on improving the tracking of underwater cables using autonomous underwater vehicles (AUVs) equipped with side-scan sonar. It aims to address issues related to AUV motion stability and imaging quality by proposing a new strategy involving non-myopic receding-horizon optimization and the use of a long short- term memory network. Also, it highlights the development of an efficient decision-making framework to execute these strategies in real time, and it reports on the validation and effectiveness of the proposed method through comparative experiments.

Authors: Bo Wang, Jie Wang, Chen Zheng, Ye Li, Jian Cao, and Yueming Li This document describes a proposed solution to improve the detection of seabed-contacting segments during underwater pipeline operations. It outlines the challenges faced due to weak structural features and environmental conditions and introduces a technical solution involving a cascade attention module and a prefusion module integrated with a convolutional neural network. The intent is to convey the effectiveness of these modules in enhancing the performance of neural network models for target detection and instance segmentation, as evidenced by experimental results.

06 - Learning hybrid dynamic transformers for underwater image super-resolution.

Authors: Xin He, Junjie Li, and Tong Jia Underwater image super-resolution is important for making underwater images clearer, which aids in analysis and exploration. This paper presents an effective hybrid dynamic Transformer (HDT-Net) for this task. It uses dynamic local self-attention to capture details and sparse non-local self-attention to avoid irrelevant information. Our method outperforms previous models in both quantitative and qualitative tests.

35 - NeuRSS: Enhancing AUV Localization and Bathymetric Mapping With Neural Rendering for Sidescan SLAM.

Author: Yiping Xie , Jun Zhang, Nils Bore, and John Folkesson Implicit neural representations and neural rendering help estimate bathymetry from sidescan sonar data. The accuracy is affected by the positioning of underwater vehicles. This article proposes a framework, NeuRSS, that improves localization and bathymetric mapping using dead reckoning and loop closures. It includes an initial bathymetric estimate from sidescan sonar to guide further optimization. The approach is validated with large surveys from both surface vessels and AUVs, comparing results to ground truth and multibeam echo sounders.

40 - Monitoring dispersal patterns sea foam injected by offshore platforms using satellite optical multispectral imagery.

Authors: M. Tedeschi, F. Filipponi, M. Picone, A. Grillo, M. Gabellini, G. Trinchera Liquefied natural gas regasification plants can use seawater to convert gas from liquid to gas. An initiative was launched to identify "foams" created during this process using satellite data. The sea foam detection procedure involves cloud masking, mask refinement, detection, and spatial filtering. Results from 2015-2022 demonstrated the procedure's effectiveness in identifying sea foam dispersal patterns, which can be applied to other remote sensing imagery.

41 - RUSNet: Robust fish segmentation in underwater videos based on adaptive selection of optical flow

Authors: Peng Zhang, Zongyi Yang, Hong Yu, Wan Tu, Chencheng Gao, and Yue Wang Fish segmentation in underwater videos is important for tracking fish populations and surveying fishery resources. Researchers have worked on improving this process using underwater optical flow. However, existing methods struggle with non-fish motions, prompting the development of RUSNet, a new segmentation network. This model includes a quality evaluation module for optical flow, a decoder to refine fish localization, and a selective fusion method for predictions. Testing showed RUSNet outperformed current models, offering a reliable solution for fish segmentation in challenging underwater environments.

42 - Infrared and visible image fusion based on multi-level detail enhancement and generative adversarial network.

Authors: Xiangrui Tian, Xiaohan Xianyu, Zhimin Li, Tong Xu, Yinjun Jia Infrared and visible image fusion technology is useful in target detection and tracking. Current methods often miss the connection between local and global features, leading to image blurring and loss of detail. This paper proposes a GAN-based method that enhances edge information and improves feature extraction. Experimental results show better texture detail and superior performance compared to other methods.

Click on the octopus to return to the top of the page

08 - Pose-graph underwater simultaneous localization and mapping for autonomous monitoring and 3D reconstruction by means of optical and acoustic sensors

Authors: Alessandro Bucci, Alessandro Ridolfi, Benedetto Allotta Modern mobile robots require precise localization and navigation systems, especially in underwater environments. A sensor fusion algorithm based on Maximum A Posteriori (MAP) techniques has been developed to improve navigation accuracy and consider landmark points. The algorithm's performance was evaluated in an experimental campaign on Stromboli Island.

09 - A Visual–Inertial Pressure Fusion-Based Underwater Simultaneous Localization and Mapping System

Author: Zhufei Lu, Xing Xu, Yihao Luo, Lianghui Ding, Chao Zhou, and Jiarong Wang This paper proposes a multi-sensor fusion Simultaneous Localization and Mapping (SLAM) system to improve the accuracy of naval mine countermeasures (MCM) operations. This system integrates depth, Inertial Measurement Unit (IMU), and optical sensors into the visual SLAM algorithm, enhancing pose estimation and map scale estimation. The authors enphase that the system reduces scale error, trajectory error, and square root by up to 41.6% compared to other SLAM models.

23 - Underwater Mapping and Optimization Based on Multibeam Echo Sounders

Authors: Feihu Zhang, Tingfeng Tan, Xujia Hou, Liang Zhao, Chun Cao, and Zewen Wang This study presents a multibeam data processing system for underwater mapping accuracy, addressing the issue of weak correlation between sonar data and inter-frame feature matching, including functions for sonar parameter control, data storage, and point cloud conversion. The Iterative Extended Kalman Filter algorithm is used for odometry estimation, and the Generalized Iterative Closest Point algorithm is used for point cloud registration.

34 - Positioning Systems for Unmanned Underwater Vehicles: A Comprehensive Review

Authors: Christos Alexandris, Panagiotis Papageorgas, and Dimitrios Piromalis This paper reviews key technologies used for Unmanned Underwater Vehicle positioning, including acoustic systems, inertial navigation, Doppler velocity logs, and GPS, which are crucial for seabed mapping, marine infrastructure inspection, and search and rescue operations. The review addresses challenges like communication constraints and environmental variability, and discusses future trends in positioning system development.

43 - High-Precision Sub-Wavelength Motion Compensation Technique for 3D Down-Looking Imaging Sonar Based on an Acoustic Calibration System

Authors: Jun Wang, Peihui Liang, Junqiang Song, Pan Xu, Yongming Hu, Peng Zhang, Kang Lou, Rongyao Ren, and Wusheng Tang Three-dimensional hydro-acoustic imaging is a promising research area in underwater signal processing, with applications in marine environmental resource surveying, seabed topography, and underwater early warning. A down-looking 3D-imaging sonar with a linear array structure was designed to address resolution issues and improve along-track resolution. A high-precision sub- wavelength motion compensation method was proposed to overcome motion error interference, leading to higher- quality images.

44 - Bridging Disciplines with Photogrammetry: A Coastal Exploration Approach for 3D Mapping and Underwater Positioning

Authors: Ali Alakbar Karaki, Ilaria Ferrando, Bianca Federici, and Domenico Sguerso This study introduces an innovative method for 3D mapping and underwater positioning, combining terrestrial photogrammetry with underwater photogrammetry by a snorkeler. The method detects snorkeler positions on orthorectified images, enabling a cost-effective and comprehensive tool for 3D coastal mapping. This approach aids in understanding the spatial context of objects on the seabed and supports coastal resilience.

46 - Wake Detection and Positioning for Autonomous Underwater Vehicles Based on Cilium-Inspired Wake Sensor

Authors: Xuanye Hu, Yi Yang, Zhiyu Liao, Xinghua Zhu, Renxin Wang, Peng Zhang, and Zhiqiang Hu This paper presents a passive detection method for autonomous underwater vehicle wakes using a cilium- inspired wake sensor (CIWS)that measures the wake velocities of the vehicle at different positions, establishing a velocity field model. The characteristic frequency of the vehicle's propeller is identified, and a mapping model is established between the spectral amplitude and wake velocity. This provides a technical foundation for underwater detection and tracking missions.

Authors: Tong Liu, Kaiyan Zhu, Xinyi Wang, Wenbo Song, and Han Wang This article presents the Zero-Reference Parameter Estimation Network (Zero-UAE), designed to enhance underwater images by addressing challenges such as color attenuation and contrast reduction caused by poor lighting. Unlike traditional deep learning methods that require extensive training data, the Zero-UAE model leverages a novel adaptive curve model and lightweight parameter estimation to enhance underwater images without the need for additional datasets. The model utilizes specialized non-reference loss functions to address color bias and uneven illumination, demonstrating competitive performance across multiple datasets while maintaining minimal hardware requirements.

Authors: Peng Jiang, Hao Wang, and Zhiwen Xiong This study presents a novel dynamic routing scheme for underwater acoustic sensor networks utilizing autonomous vehicles to collect data. It incorporates source location protection (SLP) technology to safeguard privacy and is validated through MATLAB simulations, demonstrating significant improvements in energy efficiency and stability of the sensor network in complex underwater environments.

22 - Image stitching and target perception for Autonomous Underwater Vehicle-collected side-scan sonar image

Authors: Zhuoyu Zhang, Rundong Wu, Dejun Li, Mingwei Lin, Sa Xiao, and Ri Lin This article discusses the innovative integration of Autonomous Underwater Vehicles (AUVs) and side-scan sonar for enhanced seabed detection and marine monitoring. It presents a new detection method utilizing a sliding window technique, a sophisticated sonar image stitching approach that tackles intensity variations, and a target perception framework based on shadow region segmentation. Experimental results indicate significant improvements in detection accuracy and target height measurement, underscoring the effectiveness of the proposed methodologies.

21 - Photogrammetry of the Deep Seafloor from Archived Unmanned Submersible Exploration Dives - Published by MDPI - Journal of Marine Science and Engineering

Authors: Claudia H. Flores, and Uri S. ten Brink This article presents a workflow for utilizing modern photogrammetry techniques to construct extensive virtual geological outcrops from archived video images collected at deep-sea depths (over 1000 meters). Traditionally, the analysis of these images has required significant effort for individual reviews or small mosaics. The proposed workflow addresses the challenges of non-linear submersible paths, poor navigation, variable lighting, and camera orientation, enabling quantitative measurements that were previously unattainable. This method offers a cost-effective means to analyze deep-sea geological environments through existing video data, especially when integrated with rock sample analysis

37 - A Novel Platform and Workflow for Underwater Photogrammetry Surveys - Published by The International Society for Photogrammetry and Remote Sensing

Authors: Alaa Mufti, Petra Helmholz, David Belton, Amir Allahvirdi-Zadeh, and Iain Parnum This article extensively outlines the advantages of underwater photogrammetry over traditional echosounder surveys, particularly in terms of cost- effectiveness and data richness. The study highlights the critical roles played by various sensors in ensuring accuracy and quality while emphasizing the essential need for reliable camera calibration in achieving precise 3D reconstructions. Overall, the research sets a promising precedent for future innovations in marine mapping, showcasing the potential for improved practices derived from emerging technologies and methodologies.

Authors: Weiwen Chen, Tingting Zhuang, Yuanfang Zhang, Teng Mei, Xiaoyu Tang This study addresses the challenge of accurate and efficient underwater object detection, critical for marine defense, aquaculture, and ocean development, by enhancing the YOLOv7 detector with three key innovations: (1) a Convolutional Omni-Efficient Layer Aggregation Network (CO-ELAN) backbone module that improves feature extraction via multi-dimensional dynamic convolution and attention; (2) an optimized normalized Wasserstein distance loss for better small-target localization through probabilistic bounding-box modeling; and (3) a white balance–blur fusion underwater image enhancement method to improve input quality.

Authors: Charu Manivannan, Jakub Krukar, Angela Schwering Sketch maps are widely used in disciplines like spatial cognition and environmental psychology but are challenging to analyze due to their abstract and generalized nature. Current methods struggle with generalized information, requiring exact matches between sketch and metric maps, despite memory often being generalized. This paper introduces (1) an algorithm to detect generalization in sketch maps and (2) an online tool to generate corresponding generalized metric maps.

14 - Design of Scanning Units for the Underwater Circumferential- Scanning LiDAR Based on Pyramidal-Shaped Refl ectors and a Rapid Detection Method for Target Orientation