OCEANS 2021 Student Poster Competition

EarthzineOceans, Oceans conferences

The coveted OCEANS student poster competition saw good participation in the 2021 edition too.

13 June, 2022

Shyam Kumar Madhusudhana

A flagship event of the MTS/OES OCEANS conferences is the Student Poster Competition (SPC) which is open to undergraduate and graduate students from colleges and universities around the world. The SPC was envisioned and created by Col. Norman Miller and was first implemented at the Seattle OCEANS conference in 1989. It has been a feature of OCEANS conferences ever since. From the pool of aspiring applicants typically 15-20 students are selected to participate in the Competition, based on reviews (two stages) of their abstracts. The selected students get an opportunity to present, at the OCEANS conference, a poster describing their work.

The fusing together of this year’s Porto and San Diego OCEANS conferences resulted in a single SPC event. The pandemic-induced lockdowns and travel restrictions around the globe had forced last year's OCEANS conference to be held as an entirely virtual event. In contrast, this year’s event was held in a hybrid format allowing participants the option to either attend in-person or participate virtually. The final set of participants included 14 virtual and 5 in-person participants.

Related Stories

Sustaining Long-Term Ocean Observations

The challenges of deep-sea exploration

OCEANS 22 – An in-person technical treat in the post-Covid era

OCEANS 22 – An in-person technical treat in the post-Covid era

OSM 2022 Small Island developing states session

Awards and certificates distribution during the Gala. Left-to-right: Vladimir Djapic (LOC SPC Chair); four of the on-site participants; Schmidt Ocean Institute rep. Jyothika Virmani (standing in front); MTS & OES Presidents; six of the judges; Society SPC Chairpersons Josh Kohut (MTS), Shyam Madhusudhana (OES) and Kristina Norman (MTS).

The hybrid format brought its own challenges to conducting the SPC. The in-person participants presented their posters in the traditional form, while the virtual participants’ presentations and interactions with the judges were arranged to be held over a teleconference call - the students presented their posters virtually one after the other to the cohort of judges that were congregated in a hall at the conference venue. I would like to congratulate the Local Organizing Committee (LOC) for their proactive efforts in ensuring that the event was conducted at the desired expectation levels. The LOC SPC Chair, Vladimir Djapic, deserves a special mention for his commendable efforts despite having had to relocate to Europe during the run-up to the conference. We had a total of 8 volunteers that offered to serve as judges for the Competition; and their efforts and contributions to the event’s success are much appreciated. I also take this opportunity to thank the sponsors—Office of Naval Research and Schmidt Ocean Institute—for their support of this year’s SPC and also for their continued support for OCEANS conferences in general.

The list of participants (including the prize winners) together with their affiliation, poster title and an abstract of their poster are given below.

OSM 2022 Small Island developing states session

The first-prize winner, Schuyler Nardelli

First Prize, Norman Miller Award (Certificate & $3000 prize)

Schuyler Nardelli, Rutgers University, USA

Developing a convolutional neural network to classify phytoplankton images collected with an Imaging FlowCytobot along the West Antarctic Peninsula

Abstract—High-resolution optical imaging systems are quickly becoming universal tools to characterize and quantify microbial diversity in marine ecosystems. Automated detection systems such as convolutional neural networks (CNN) are often developed to identify the immense number of images collected. The goal of our study was to develop a CNN to classify phytoplankton images collected with an Imaging FlowCytobot for the Palmer Antarctica Long-Term Ecological Research project. A medium complexity CNN was developed using a subset of manually-identified images, resulting in an overall accuracy, recall, and f1-score of 93.8%, 93.7%, and 93.7%, respectively. The f1-score dropped to 46.5% when tested on a new random subset of 10,269 images, likely due to highly imbalanced class distributions, high intraclass variance, and interclass morphological  similarities of cells in naturally occurring phytoplankton assemblages. Our model was then used

to predict taxonomic classifications of phytoplankton at Palmer Station, Antarctica over 2017-2018 and 2018-2019 summer field seasons. The CNN was generally able to capture important seasonal dynamics such as the shift from large centric diatoms to small pennate diatoms in both seasons, which is thought to be driven by increases in glacial meltwater from January to March. Moving forward, we hope to further increase the accuracy of our model to better characterize coastal phytoplankton communities threatened by rapidly changing environmental conditions.

OSM 2022 Small Island developing states session

The third-prize winning presentation

Second Prize (Certificate and $2000 prize)

Isaac Gerg, Penn State University, USA

A learnable image compression scheme for synthetic aperture sonar imagery

Abstract—Synthetic aperture sonar (SAS) is an imaging modality which produces high and constant resolution images of the seafloor. These sonars are often mounted to a unmanned underwater vehicle (UUV) to autonomously collect imagery of a prescribed survey area. While a survey is underway, UUV communications back to the operator are often limited due to the use of a low-bandwidth acoustic communications (ACOMMS) channel. Because of this, high-quality SAS imagery is rarely sent over this link due to the lack of an efficient compression scheme to send such information. Creating an efficient SAS image compression scheme provides at least two operational benefits: (1) image chips beamformed and tagged by onboard processing algorithms can be quickly communicated to operators while a survey is ongoing, and (2) cooperative UUVs can exchange salient image chips among themselves to reconcile position ambiguity and obtain a shared reference frame. In this work we propose a learned image compression scheme for SAS imagery using deep neural networks (DNNs). DNNs have already been applied to the image compression problem but almost exclusively for optical imagery. We highlight some important differences between SAS imagery and optical imagery which prevents the simple application of off-the-shelf (OTS) methods like JPEG and WebP to SAS imagery. We propose an image compression scheme which specifically addresses the domain-specific properties of SAS imagery to obtain useful image compression performance on a real-world SAS dataset. We show that we can reduce the bitrate by up to thirty-five percent while still maintaining the same perceptual image quality as OTS codecs.

Third Prize (Certificate and $1000 prize)

Diogo Teixeira, University of Porto - LSTS, Portugal

3D tracking of a river plume front with an AUV

Abstract—The problem of the concurrent tracking and mapping of a river plume front with an autonomous underwater vehicle (AUV) is formulated and addressed in the framework of an interdisciplinary approach building on experience in robotics and oceanographic field studies. The problem formulation is targeted at the scientific study of the processes by which the river and the ocean interact. The approach extends previous work in AUV plume tracking to the simultaneous tracking and mapping under different ocean and meteorological conditions. This is done with the help of parameterizable motion control algorithms to enable adaptation to these time-varying conditions. The approach is evaluated in simulation with the help of a high-resolution hydrodynamic model. The test plan covers over 300 test cases exercising the most representative combinations of the ocean and meteorological conditions. Lessons learned and future operational deployments are discussed in the conclusions.

Xinwei Chen, Memorial University of Newfoundland, Canada

A spatial-temporal ensemble network for estimating wave height from X-band marine radar image sequences

Abstract—A spatial-temporal ensemble network is proposed to estimate significant wave heights (SWHs) from X-band marine radar image sequences. After image subarea selection, a 1024-dimension deep spatial feature vector is first extracted from each radar image using deep convolutional layers obtained from GoogLeNet. The temporal behavior of each radar image sequence can be analyzed by inputting the spatial feature vector sequence into a recurrent neural network (RNN) consisting of gated recurrent units (GRUs), followed by the regression layers that generate the estimated SWH. The network is trained using shipborne marine radar data collected during a sea trial, while simultaneous buoy measurements are used as ground truths for calibration and performance evaluation. Compared to the traditional signal-to-noise (SNR)-based method, the proposed network reduces the root mean square error by 0.35 m and 0.36 m for image sequences collected under rainless and rainy conditions, respectively.

Blake Cole, Massachusetts Institute of Technology, USA

AIS-based collision avoidance in MOOS-IvP using a geodetic unscented Kalman filter

Abstract—This paper describes the design and implementation of a low-cost collision avoidance system, designed primarily for use on small and medium-sized autonomous surface vehicles (ASVs). The proposed methodology leverages real-time information broadcast via the Automatic Information System (AIS) messaging protocol, in order to estimate the position, speed, and heading of nearby vessels. The state of each target vessel is recursively estimated in geodetic coordinates using an Unscented Kalman Filter (UKF). Once identified, each vessel is avoided in accordance with the International Regulations for Preventing Collisions at Sea (COLREGs). This capability is enabled by MOOS-IvP, a behavior-based autonomy middleware that is able to make navigation decisions by weighing the relative importance of multiple competing objectives. For the purposes of collision avoidance, each target vessel produces a two-dimensional objective function which increases the cost of heading and speed combinations that will result in a collision or near-miss event. However, the primary mission behaviors remain active, allowing the IvP solver to choose an optimal combination of vessel speed and heading which drive the vehicle toward a desired state while simultaneously minimizing the risk of collision. It is shown through field testing that the proposed framework is an effective, robust means of collision avoidance.

Diogo Duarte, INESC TEC - Faculdade de Engenharia da Universidade do Porto, Portugal

Multiple vessel detection and tracking in harsh maritime environments
Abstract—Recently, research concerning the navigation of Autonomous Surface Vehicles (ASVs) has been increasing. However, a big scale implementation of these vessels is still held back by a plethora of challenges such as multi-object tracking. This article presents the development of a tracking model through transfer learning techniques, based on referenced object trackers for urban scenarios. The work consisted in training a neural network through deep learning techniques, including data association and comparison of three different optimisers, Adadelta, Adam and SGD, determining the best hyper-parameters to maximise the training efficiency. The developed model achieved decent performance at tracking large vessels in the ocean, being successful even in harsh lighting conditions and lack of image focus.

Elizaveta Dubrovinskaya, Universidad Carlos III Madrid, Spain

Validation of localization via wideband acoustic arrays for underwater fauna monitoring at sea

Abstract—We present a technique for simultaneous detection, path tracking and accurate 3D underwater localization using wideband arrays of complex geometry based on acoustic reflections clustering. We have extended our previously proposed algorithm for 3D localization designed for arrays that do not meet typical constraints of one half-wavelength spacing between the closest array elements. Inspired by the scope of SYMBIOSIS, a hybrid opto-acoustic system for pelagic fish species monitoring, we added more functionalities that fit the needs of the project.

The proposed algorithm can automatically discriminate moving targets from stationary environmental features and track them to estimate their possible time of arrival to the system. We test the algorithm in several autonomous deployments including shallow and deep water. The experimental results for marine fauna monitoring have shown a good performance in various environments.

Joana Fonseca, Royal Institute of Technology KTH, Sweden

Algal bloom front tracking using an unmanned surface vehicle: numerical experiments based on Baltic Sea data

Abstract—We consider the problem of tracking moving algal bloom fronts using an unmanned surface vehicle (USV) equipped with a sensor that measures the concentration of chlorophyll a. Chlorophyll a is a green pigment found in plants, and its concentration is an indicator of phytoplankton abundance. Our algal bloom front tracking mission consists of three stages: deployment, data collection, and front tracking. At the deployment stage, a satellite collects an image of the sea from which the location of the front, the reference value for the concentration at this front and, consequently, the appropriate initial position for the USV are determined. At the data collection stage, the USV collects data points to estimate the local algal gradient as it crosses the front. Finally, at the front tracking stage, an adaptive algorithm based on recursive least squares fitting using recent past sensor measures is executed. We evaluate the performance of the algorithm and its sensitivity to measurement noise through MATLAB simulations. We also present an implementation of the algorithm on the DUNE onboard software platform for marine robots and validate it using simulations with satellite model forecasts from Baltic sea data.

Guy Gubnitsky, University of Haifa, Israel

Multispectral target detection in sonar imagery

Abstract—Detection of underwater objects in sonar imagery is a key enabling technique, with applications ranging from mine hunting and seabed characterization to marine archaeology. Due to the non-homogeneity of the sonar imagery, the majority of detection approaches are geared towards detection of features in the spatial domain to identify anomalies in the seabed’s background. Yet, when the seabed is complex and includes rocks and sand ripples, spatial features are hard to discriminate, leading to high false alarm rates. With the aim of detecting man-made objects in complex environments, we utilize, as a detection metric, the expected spectral diversity of reflections to differentiate man-made objects’ reflections from the relatively flat frequency response of natural objects’ reflections, such as rocks. Our solution merges a set of preregistered sonar images, each of which are obtained at a different frequency band. Using the Jain’s fairness as a metric to evaluate the spectral diversity of a suspected object within a low or high resolution sonar imagery, respectively, our solution detects anomalies across the spectrum domain. We tested our algorithm over simulated data and over multispectral data obtained in a designated sea experiment. The results show that, compared to benchmark schemes, our approach obtains better performance in terms of the trade-off between false alarm rate and detection capability.

Miguel Gutierrez Gaitan, CISTER/FEUP-ISEP, Portugal

Wireless radio link design to improve near-shore communication with surface nodes on tidal waters

Abstract—Wireless radio links deployed over aquatic areas (e.g., sea, estuaries or harbors) are affected by the conductive


properties of the water surface, strengthening signal reflections and increasing interference effects. Recurrent natural phenomena such as tides or waves cause shifts in the water level that, in turn, change the interference patterns and cause varying impairments to propagation over water surfaces. In this work, we aim at mitigating the detrimental impact of tides on the quality of a line-of-sight over-water link between an onshore station and a surface node, targeting mission data transfer scenarios. We consider different types of surface nodes, namely, autonomous underwater vehicles, unmanned surface vehicles and buoys, and we use WiFi technology in both 2.4 GHz and 5 GHz frequency bands. We propose two methods for link distance/height design: (i) identifying a proper Tx-Rx distance for improved link quality at each point of the tidal cycle; (ii) defining the height/distance that minimizes the path loss averaged during the whole tidal cycle. Experimental results clearly show the validity of our link quality model and the interest of method (i). Analytical results confirm method (ii) and show that it outperforms, in both frequency bands, the common practice of placing onshore antennas at the largest possible height and/or surface nodes at a short but arbitrary distance.

Yuying Huang, McGill University, Canada

An autonomous probing system for collecting measurements at depth from small surface vehicles

Abstract—This paper presents the portable autonomous probing system (APS), a low-cost robotic design for collecting water quality measurements at targeted depths from an autonomous surface vehicle (ASV). This system fills an important, but often overlooked niche in marine sampling by enabling mobile sensor observations throughout the near-surface water column without the need for advanced underwater equipment. We present a probe delivery mechanism built with commercially available components and describe the corresponding winch control system. and data captured from this system in a field deployment. Finally, we discuss design trade-offs and present areas for future improvement. Project details are available on our website.

Stéphane Imbert, IMT Atlantique, France

Two-dimensional spreading waveform for adaptive rate acoustic underwater communication


Abstract—Autonomous Underwater Vehicles (AUVs) are becoming a part of the ocean navigation ecosystem, and their applications deal with seafloor mapping, channel sounding, surveillance, submarine volcanism survey or mine warfare. During a mission, it is of great importance, through a reliable acoustic wireless communication link, that ships or surface buoys receive live data from an AUV and exchange information data with it. Furthermore, the communication range can vary during a mission. Thus, this paper addresses the challenging topic of designing a modem, to transmit information to an AUV, by emphasizing on the adaptivity and robustness of the link. The designed modem combines multi-carrier signal and 2D spreading and is called MC-SS-2D. A variable length spreading sequence combined with OFDM creates a flexible design that uses frequency diversity, time diversity and adds a processing gain to transmit at various distances. The modem was tested using rayleigh channels as well as experiments in a water tank and at sea using the IROMI platform.

Rodney Itiki, University of North Carolina at Charlotte, USA

Method for estimation of marine hydro-kinetic power based on high-frequency radar data

Abstract—Marine Hydro-Kinetic (MHK) is renewable energy in the moving waters of the oceans. This research proposes a method for estimating the total MHK energy harvested by turbines spatially distributed in the ocean current. The MHK energy is variable in time and space. The method represented by an algorithm is implemented in MATLAB. It reads hourly measurements of seawater speed to estimate the power profile of the generated MHK farms. In the U.S., some speed measurements from high-frequency (HF) radars in the coastal areas are publicized on the internet by the National Oceanic and Atmospheric Administration (NOAA). The algorithm functionality is demonstrated in a case study with the NOAA radar measuring the seawater speed of the Gulf Strem off the coast of North Carolina. The peak value of the power profile sets a reference for the sizing of the MHK platform equipment and cabling. The algorithm is also useful for planning of renewables expansion of utility companies, selection of offshore site with high power output, techno-economic feasibility studies, and subsequent steps of engineering for a proposed MHK farm project. A comprehensive discussion of the implications of the proposed method for the natural, economic, social, and political environment paves the way for a myriad of potential multidisciplinary topics for future blue energy research.

Juhwan Kim, Pohang University of Science and Technology, Republic of Korea

Line laser mounted small agent ROV based 3D reconstruction method for precision underwater manipulation

Abstract—We proposed a line laser based 3D reconstruction method for small agent ROV to perform precision underwater manipulation. Perception of the surrounding environment underwater is essential for manipulation. Especially for small underwater vehicles with relatively limited and weak sensors, it is very difficult to perform 3D reconstruction and manipulation of objects underwater. This paper proposed that a small agent ROV attached with a line laser performs 3D object reconstruction by rotating motion. In general, a line laser scan can be performed when an actuator moves the line laser or when the vehicle’s position can be accurately known. However, in a small underwater vehicle without such an actuator and a high-performance navigation sensor, a laser scan cannot be performed smoothly. Therefore, we devised a method that performs a line laser scan by rotating the vehicle around the object only by detecting the target object with the camera. We confirmed that the target objects could be 3D scanned using the line laser scan simulator. In addition, we carried out experiments in the water tank, and 3D reconstruction of three target objects was possible. As a result, we showed that our line laser scan method could perform underwater 3D reconstruction of target objects even for small underwater vehicles with few sensors.

Peeyush Kumar, University of Bonn, Germany

Combined use of a frame and a linear pushbroom camera for deep-sea 3D hyperspectral mapping

Abstract—Hyperspectral (HS) imaging produces an image of an object across a large range of the visible spectrum, and not just the primary colors (R, G, B) of conventional cameras. It can provide valuable information for object detection, analysis of materials and processes in environmental science in the deep-sea, especially for the study of benthic environments and pollution monitoring.

In this paper, we address the problem of camera calibration towards 3D hyperspectral mapping where GPS is not available, and the platform navigational sensors are not accurate enough to allow direct georeferencing of linear sensors, as is the case with traditional aerial platform methods.

Our approach presents a preliminary method for 3D hyper-spectral mapping that uses only image processing techniques to reduce reliance on GPS or navigation sensors. The method is based on the use of standard RGB camera coupled with the hyperspectral pushbroom camera. The main contribution is the implementation and preliminary testing of a method to relate the two cameras using image information alone.

The experiments presented in this paper analyze the estimation of relative orientation and time synchronization parameters for both cameras through experiments based on epipolar geometry and Monte-Carlo simulation. All methods are designed to work with real world data.

Mathilde Letard, Paris Sciences Lettres University, France

Classification of coastal and estuarine ecosystems using full-waveform topo-bathymetric lidar data and artificial intelligence

Abstract—Coastal and estuarine ecosystems are facing spatio-temporal changes and suffer from the effects of accelerated natural destructive processes due to climate change. Monitoring these areas is crucial to protect them and maintain the ecological balance of shorelines. In this context, full-waveform airborne topo-bathymetric lidar is a reliable tool to collect data seamlessly over land-water continuum zones, thanks to its dual wavelength configuration. It is therefore optimal for coastal habitats monitoring and mapping. However, lidar waveform processing often relies on peak detection and feature extraction that are difficult to configure and often sensitive to noise. In this article, we rather suggest not to rely on hand-crafted features by relying on U-time, a neural network inspired by the well-known UNet convolutional neural network, to identify peaks in waveforms and classify them to discriminate coastal ecosystems efficiently. The network is tested on green waveforms and we evaluate in addition the contribution of infrared intensities. Results show equivalent performances, and obtain over 92% of accuracy when accepting a 2 samples margin of error for peaks location, which does not impact heavily waveform analysis, considering usual peaks widths. Our study shows green waveforms alone allow habitats detection with a F-score of 94%, outperforming previous methods.

Shashank Swaminathan, Franklin W. Olin College of Engineering, USA

Optimal planning with uncertainty for a moored profiler monitoring an aquaculture installation

Abstract—There is burgeoning global demand for marine seafood, and one proposed approach to meeting this is expanding offshore farming practices. When developing such offshore aquaculture farms, it is critical to maintain the health and safety of both the food stock and the environment. To do so, there must be consistent and thorough monitoring of the installations, with data collected over large spatial and temporal ranges, spanning up to hundreds of meters deep for weeks at a time. We focus on using a moored automatic mobile profiler (MAMP) to provide environmental data along the entire spatial range relevant to the farm. However, while MAMPs have much higher spatial range and resolution than fixed sensor arrays, an aquaculture site is dynamic. To capture the shifts in the site’s state, the profiler must take quick, sparse measurements and determine the overall environment from those samples. This paper’s goal is to propose an implementation of a probabilistic algorithm that enables reconstruction and prediction of the farm’s state from sparse samples for use in path planning using Gaussian Regression Processes. The proposed approach is verified against experimental ocean environment data, and shows >95% accuracy in environment reconstruction while requiring only 10% of the number of data samples as the ground truth data set. It is further implemented in conjunction with a Markov Decision Process based planner to demonstrate the algorithm’s use in path planning. This work indicates the high potential of the algorithm for use on dynamic environments like aquaculture farm sites.

Jenny Walker, University of Southampton, UK

Towards observation condition agnostic fauna detection and segmentation in seafloor imagery for biomass estimation

Abstract—The performance of automated object detection and segmentation in marine imaging applications is sensitive to hardware and environmental factors that result in a large variability in the appearance of subjects in images. This paper investigates physics based scale normalisation, lens distortion normalisation, and data augmentation techniques to overcome this, working towards a condition agnostic object detection system. A total of over 700 rockfish in images taken from different altitudes using different camera equipped Autonomous Underwater Vehicles at the Southern Hydrates Ridge (depth 780 m) are used to train and test object detection and segmentation using Mask R-CNN. Images taken from low altitudes of ∼2 m achieve a maximum mean average precision (mAP) score of 97.42%, and images taken from high altitudes of ∼6 m achieve a maximum score of 87.4% when object detection and segmentation is trained and tested on images taken from the same altitudes. When transferring knowledge across different imaging conditions, a mAP score of 87.7% is achieved when transferring knowledge from high to low altitude datasets, and 49.6% when transferring from low to high altitudes. In both cases, significant gains in performance is seen when the images used are scale normalised. The results indicate that increasing the pixel resolution, or the size an object appears within the image, benefits learning regardless of the optical resolution images are taken at, and this should be carefully considered in future object detection and segmentation studies. We also describe a novel method to estimate biomass distribution from the segments output by modern machine learning algorithms that can be easily adapted for different morphospecies.

Jonathan Wallen, University of Hawaii at Manoa, USA

Co-design optimization for underwater vehicle docking systems

Abstract—The design of autonomous underwater vehicles (AUVs) and their docking stations has been a popular research topic for several decades. Although many AUV and dock designs have been proposed, materialized, and commercialized, most of these existing designs prioritize the functionality of the AUV over the dock, or vise versa; there has been limited formal research in analytical optimization for AUV docking systems. In this paper, a multidisciplinary optimization framework is presented with the aim to fill this theoretical gap. We propose a co-design optimization method that optimizes multiple design parameters governing the archetype of an AUV and its docking system. Capturing the user design intents in the optimization process, the proposed method produces a set of optimal design parameters that satisfies a set of predefined bounds, constraints, and initial conditions. Three cases of design optimization are reported for different design intents. Each optimal design found in the three cases is compared to an existing system to show the validity of this design optimization framework.