Seafloor Images Research Articles

Leveraging Metadata in Representation Learning With Georeferenced Seafloor Imagery

Camera equipped Autonomous Underwater Vehicles (AUVs) are now routinely used in seafloor surveys. Obtaining effective representations from the images they collect can enable perception-aware robotic exploration such as information-gain-guided path planning and target-driven visual navigation. This letter develops a novel self-supervised representation learning method for seafloor images collected by AUVs. The method allows deep-learning convolutional autoencoders to leverage multiple sources of metadata to regularise their learning, prioritising features observed in images that can be correlated with patterns in their metadata. The impact of the proposed regularisation is examined on a dataset consisting of more than 30 k colour seafloor images gathered by an AUV off the coast of Tasmania. The metadata used to regularise learning in this dataset consists of the horizontal location and depth of the observed seafloor. The results show that including metadata in self-supervised representation learning can increase image classification accuracy by up to 15% and never degrades learning performance. We show how effective representation learning can be applied to achieve class balanced representative image identification for summarised understanding of imbalanced class distributions in an unsupervised way.

Fauna associated with cold seeps in the deep Colombian Caribbean

Cold-seep chemosynthetic communities are found in patchy areas where methane and other hydrocarbons leak through the seafloor. Although relatively common in continental margins, chemosynthetic communities in the Caribbean region are poorly known. Their existence in Colombian waters has been suggested by the presence of carbonates and specimens of seep- specialist families obtained through trawling surveys, and seismic and multi-beam data. Habitat models incorporating bathymetry and backscatter data have also predicted the presence of hard ground areas (presumed authigenic carbonates formed due to chemosynthetic activity), which are consistent with water column plume anomalies. Here, we present the first characterization of the megafauna found in cold-seep communities in the Colombian Caribbean between 2300 and 3300 m using seafloor images taken with a towed camera system and samples collected through piston coring. Chemosynthetic communities at these sites resemble those found off Trinidad and Tobago and the Gulf of Mexico. Dominant species include tubeworms (Lamellabranchia sp. and Escarpia sp.), mussels (Bathymodiolus sp.), shrimp (Alvinocaris sp.), and squat lobsters (Munidopsis sp.). These results represent foundational knowledge of deep-sea ecosystems in the southern Caribbean.

The megafauna community from an abyssal area of interest for mining of polymetallic nodules

Polymetallic nodules increase habitat heterogeneity in some abyssal benthic ecosystems by providing hard substrate. Besides their important role in structuring ecosystems, polymetallic nodules have high grades of valuable minerals and are a target of likely future exploitation, particularly in the Pacific Clarion-Clipperton Zone (CCZ). Mining activities will remove hard substrate and sediment and cause sediment plumes potentially affecting faunal communities over large areas. Long-lived megafaunal assemblages may be particularly vulnerable but data are lacking on the density, biodiversity and community structure in many areas of the CCZ. This study aims to provide megabenthic community baseline data from two physically similar areas (B6S02 and B4S03) located in the contract area of Global Sea Mineral Resources N.V. (GSR). Seafloor images, obtained by an autonomous underwater vehicle are used to characterise the large areas required for robust evaluation of sparse megafauna. Higher altitude images cover a larger area for the same effort but have reduced resolution compared to images obtained closer to the seafloor, leading to difficulties in detecting and identifying individuals. Our comparison of images obtained at different altitudes shows that images taken above 8m altitude underestimate the megafauna density by almost 50%, so images <8m were used for further analysis. We also used multiobserver agreement analysis to improve the megafaunal annotation consistency, and the quality and robustness of the data in this study. The two GSR areas were significantly different in nodule coverage and megafauna composition and a general positive relationship between nodule coverage and megafauna abundance was observed. Differences in the megafaunal composition were primarily caused by differences in echinoderms (asteroids, echinoids, ophiuroids and holothuroids), representing almost 70% of the megafauna.

Quantitative Expression of the Burial Phenomenon of Deep Seafloor Manganese Nodules

Manganese (polymetallic) nodules on the deep seafloor in the open ocean have attracted great interest because of their economic potential. Visual data on nodules found on the deep seafloor such as photographs and videos have increased exponentially with the recent progress of related technologies. These data are expected to reflect useful information for estimating these mineral resources, as well as understanding their geological origin. Although the size, number, and coverage of manganese nodules have been measured in seafloor images, the burial of such nodules has not been sufficiently examined. This paper focuses on mathematical expression of the burial of the manganese nodules and attempts to quantitatively elucidate relations among burial degree and nodule geological parameters. The results, that is, a dataset obtained by calculations of relations among parameters, are also utilized for considerations of quantitative expression of burial. These considerations are expected to contribute to a better understanding of the geological origin of manganese nodules.

Automatic Overlapping Area Determination and Segmentation for Multiple Side Scan Sonar Images Mosaic

Seafloor image is important for marine scientific research and ocean engineering. With the wide use of side scan sonar (SSS), multiple measured SSS images need to be mosaicked to form a large-scale and high-resolution seafloor image. This article proposes an automatic mosaic method. First, the overlapping areas between adjacent strips are determined automatically by the track lines and swath width. In the overlapping areas, the image matching is conducted using the speeded-up robust features algorithm with the constraint of geographic coordinates of detected feature points. Then, the overlapping area is segmented using the k-means cluster method. Within each segmented region, the coordinate transformation model is established to make the positions of common features in the overlapping areas unique. With regards to multiple SSS images mosaic, the method to mosaic in sequence is also proposed according to the permutation parameter of each strip. Finally, the SSS images in the overlapping areas are fused and a large-scale seafloor image is formed. This proposed method was applied in one water area of Shenzhen, demonstrating good performance in terms of coordinate consistency of the same features in the mosaicked image. The mean coordinate deviations of the same feature points in overlapping areas were nearly zero and the standard deviation of them also decreased. This proposed method is easily transferable to other study areas and provides an objective, repeatable means for multiple SSS images mosaic.

An Optical Image Transmission System for Deep Sea Creature Sampling Missions Using Autonomous Underwater Vehicle

The exploration of oceans using autonomous underwater vehicles (AUVs) is necessary for activities, such as the sustainable management of fishery resources, extraction of seafloor minerals and energy resources, and inspection of underwater infrastructure. As the next step in ocean exploration, AUVs are expected to employ end-effectors to make physical contact with seafloor creatures and materials. We propose a scenario for realizing a sampling mission using an AUV that is equipped to sample marine life. In this scenario, the sampling AUV observes the seafloor while concurrently transmitting the observed images to a surface vessel for inspection by the AUV operators. If the received images show an object of interest, the object is selected as a candidate of sampling target by the operators, who send a sampling command to the AUV. After receiving the command, the AUV returns to the target area and attempts to sample it. In this paper, we propose a system for transmitting images of the seafloor as part of the sampling-mission scenario. The proposed image transmission system includes a process for enhancing images of the deep seafloor, a process for selecting interesting images, and processes for compressing and reconstructing images. The image enhancement process resolves imaging problems resulting from light attenuation, such as color attenuation and uneven illumination. The process for selecting interesting images selects those that contain interesting objects, such as marine life. The selection process prevents the transmission of meaningless images that contain only flat sand on the seafloor. The proposed image compression method, which is based on color depth compression, reduces the amount of data. The combined process of selecting an interesting image and compressing it reduces various problems in acoustic communication, such as low information density and data loss. Instead of an overall image, part of an overall image is transmitted by a set of data packet, and each received data packet is reconstructed onboard the vessel. Because of image compression, the colors of a reconstructed image differ from those of an enhanced image. However, the reconstructed image contains similar colors, and the structural similarity index was found to be 91.4% by evaluating images that were subjected to a 4-b color compression. The proposed image transmission system was tested in the Sea of Okhotsk, and these tests were performed four times in different sea areas (minimum depth 380 m, maximum depth 590 m). The results show that the size of the data for a single image was reduced by a factor of 18 using the proposed image compression process, with each image taking 3.7 s to be transmitted via an acoustic modem (20 kb/s). Of the automatically selected images, 63% contained marine life, and the total transmission success rate was 22%.

Seabed Mapping Using Shipboard Multibeam Acoustic Data for Assessing the Spatial Distribution of Ferromanganese Crusts on Seamounts in the Western Pacific

Cobalt-rich ferromanganese crusts (Fe–Mn crusts), potential economic resources for cobalt, nickel, platinum, and other rare metals, are distributed on the surface of seamounts, ridges, and plateaus. Distribution of Fe–Mn crust deposits and their geomorphological characteristics are prerequisites to selecting possible mining sites and to predicting the environmental impact of deep-sea mining activity. Here, we map the spatial distribution of Fe–Mn crust deposits on seamount summits and flanks in the Western Pacific using shipboard multibeam echo sounder (MBES) data and seafloor images from a deep-towed camera system (DCS) and evaluate the relationship between acoustic backscatter variations and the occurrence of Fe–Mn crusts. We find a positive correlation between high backscatter intensity, steep seabed slope gradients, and the occurrence of Fe–Mn crusts. However, our analysis was not effective to distinguish the spatial boundary between several seabed types that occur over small areas in mixed seabed zones, particularly where transition zones and discontinuous seabed types are present. Thus, we conclude that MBES data can be a valuable tool for constraining spatial distribution of Fe–Mn crust deposits over a large exploration area.

Underwater Object Classification in Sidescan Sonar Images Using Deep Transfer Learning and Semisynthetic Training Data

Sidescan sonars are increasingly used in underwater search and rescue for drowning victims, wrecks and airplanes. Automatic object classification or detection methods can help a lot in case of long searches, where sonar operators may feel exhausted and therefore miss the possible object. However, most of the existing underwater object detection methods for sidescan sonar images are aimed at detecting mine-like objects, ignoring the classification of civilian objects, mainly due to lack of dataset. So, in this study, we focus on the multi-class classification of drowning victim, wreck, airplane, mine and seafloor in sonar images. Firstly, through a long-term accumulation, we built a real sidescan sonar image dataset named SeabedObjects-KLSG, which currently contains 385 wreck, 36 drowning victim, 62 airplane, 129 mine and 578 seafloor images. Secondly, considering the real dataset is imbalanced, we proposed a semisynthetic data generation method for producing sonar images of airplanes and drowning victims, which uses optical images as input, and combines image segmentation with intensity distribution simulation of different regions. Finally, we demonstrate that by transferring a pre-trained deep convolutional neural network (CNN), e.g. VGG19, and fine-tuning the deep CNN using 70% of the real dataset and the semisynthetic data for training, the overall accuracy on the remaining 30% of the real dataset can be eventually improved to 97.76%, which is the highest among all the methods. Our work indicates that the combination of semisynthetic data generation and deep transfer learning is an effective way to improve the accuracy of underwater object classification.

Benthic Foraminifera in Arctic Methane Hydrate Bearing Sediments

Benthic foraminifera have been widely used as proxy for paleo-methane emissions, mainly based on their stable isotopic signature. In cold seeps, the ecology of these organisms remains uncertain, in particular their ability to thrive during active phases of seepage. In this study, we evaluate the benthic foraminiferal response to methane seepage in Arctic sediments. We do so by examining living and dead benthic foraminiferal assemblages (>63 µm) of eleven push cores collected in two of the most active pockmarks (Lunde and Lomvi) along Vestnesa Ridge, offshore western Svalbard. Benthic foraminiferal assemblages are interpreted in the context of sediment geochemistry, seafloor images, and pore water analyses, which are used to characterize the different microhabitats. At the sampling locations, methane is currently being released making these the ideal sites to investigate the connection between the benthic foraminiferal distribution and methane seepage in the Arctic Ocean. Our results show that benthic calcareous foraminifera live in methane charged sediments, even if the faunal density and diversity is low. We note that the eutrophic-tolerant species Melonis barleeanus withstand the methane-induced hostile geochemical conditions and that it seems to prosper on the additional food availability represented by microbial mats growing at methane seeps. We also observe that the methane transport mechanisms affect different species differently. For example, sediments characterized by advective-like conditions are distinguished by a high density of living individuals, dominated by Cassidulina neoteretis, whereas sediments characterized by methane diffusion exhibit a very low faunal density. Agglutinated foraminifera are less abundant in sediments influenced by methane seepage, suggesting that this group of foraminifera does not tolerate the geochemical conditions at seeps. A comparison between the size fractions >63 and >125 µm highlights the importance of studying the finer size fraction for ecological studies in the Arctic Ocean. In the light of our results, we conclude that benthic foraminiferal can thrive at active methane seeps, where assemblages are clearly affected by methane flux.

Upper slope processes and seafloor ecosystems on the Sabrina continental slope, East Antarctica

This study applies detailed seafloor bathymetry data and seafloor images to understand upper slope features and how these influence the distribution of seafloor biota on the East Antarctic margin. The East Antarctic slope on the Sabrina margin has been shaped by diverse processes related to repeated glaciation. Differences in the morphology of gullies on the upper slope enable an understanding of the likely processes that have been active on this margin. Gully morphology varies according to changes in slope gradient, which may have driven variations in sedimentation. Areas of lower slope angles may have led to rapid sediment deposition during glacial expansion to the shelf edge, and subsequent sediment failure. Typically, gullies in these areas are U-shaped, initiate well below the shelf break, are relatively straight and long, and have low incision depths, consistent with formation due to mass wastage. Areas of higher slope angles likely experienced enhanced flow of erosive turbidity currents during glaciations associated with the release of sediment-laden basal meltwaters. Sediment-laden subglacial meltwater flows typically create gullies such as those we observe that initiate at, or near, the shelf break; are V-shaped in profile; and have high sinuosity, deep incision depths and a relatively short downslope extent. The short downslope extent reflects a reduced sediment load associated with increased seawater entrainment as the slope becomes more concave in profile. These differences in gully morphology have important habitat implications associated with differences in the composition and beta-diversity of the seafloor communities. This upper slope region also supports seafloor communities that are distinct from those on the adjacent shelf, highlighting the uniqueness of this environment for biodiversity. Conservation strategies therefore need to consider slope and shelf communities as distinct and equally important components of the Antarctic ecosystem.

Optimal sidescan sonar and subbottom profiler surveying of ancient wrecks: The ‘Fiskardo’ wreck, Kefallinia Island, Ionian Sea

Marine geophysical data collected during an underwater natural and cultural heritage assessment survey along the coastal zone of Kefallinia Island in the Ionian Sea, Greece, showed among other seafloor features, the presence of a Roman shipwreck and its amphorae cargo on the seafloor. The study and analysis of the collected data demonstrated that: (i) sidescan sonar and chirp sub-bottom profiling systems, can successfully detect ancient shipwrecks and their amphorae cargo on the seafloor, (ii) the use of objective computer vision techniques in processing sidescan sonar seafloor images, is a valuable tool for the separation of potential ancient shipwreck targets from other seafloor features with similar acoustic signatures. Furthermore, a guideline for the data acquisition parameters that should be used to obtain optimal seafloor sonar images to maximize the separation of potential shipwreck targets from other seafloor features, is provided. The underwater sonar remote sensing techniques may also provide adequate indication regarding the amphorae hull stowage and its vulnerability to human activity in the area. The shipwreck is dated between 1st century BC and 1st century AD and is one of the largest found so far in the Mediterranean Sea, for that period. It is estimated that it was carrying about 6,000 amphorae. The amphorae cargo, visible on the seafloor, is in very good state of preservation and the shipwreck has the potential to yield a wealth of information about the shipping routes, trading, amphorae hull stowage and ship construction during the relevant period and is therefore considered to be of significant archaeological importance.

Preliminary Observations of the Abyssal Megafauna of Kiribati

We report on preliminary observations of the abyssal megafauna communities in the exclusive economic zone of Kiribati, a huge abyssal area with few previous studies. These observations also provide useful context for marine minerals exploration within the EEZ and for the neighbouring Clarion Clipperton Zone (CCZ), where deep-sea mining operations are planned. Seafloor images collected during seabed mining exploration were used to characterise megafaunal communities (fauna > 1 cm) in three abyssal plain areas in the eastern Kiribati EEZ (study area extending from 1-5°N and 173-156°W). Additionally, hydrographic features in each of the survey locations were inferred by reference to near-seabed current flows modelled using open-sourced oceanographic data. The images showed a dominance of foraminiferal organisms. Metazoan communities were high in morphospecies richness but had low density. These general patterns were comparable to abyssal megabenthic communities in the CCZ. There was evidence of spatial variation between the assemblages in Kiribati, but there was a relatively large pool of shared morphospecies across the entire study area. Low metazoan density constrained detailed assessment of spatial variation and diversity at local scales. This finding is instructive of the levels of sampling effort required to determine spatial patterns in low density abyssal communities. The results of this study are preliminary observations that will be useful to guide future biological survey design and marine spatial planning strategies.

Classification of Coral Reef Benthos around Ganquan Island Using WorldView-2 Satellite Imagery

Xu, H.; Liu, Z.; Zhu, J.; Lu, X., and Liu, Q., 2019. Classification of coral reef benthos around ganquan island using WorldView-2 satellite imagery. In: Guido-Aldana, P.A. and Mulahasan, S. (eds.), Advances in Water Resources and Exploration. Journal of Coastal Research, Special Issue No. 93, pp. 466–474. Coconut Creek (Florida), ISSN 0749-0208.Coral reefs are structures formed by animals known as stony corals, which have an important impact on the physical and ecological environment of their surroundings. Presently, coral reefs are severely damaged by a variety of human activities and by global climate change. This paper is intended to explore the potential of using WorldView-2 (WV2) satellite images for coral reef bottom types classification. The research area for this study is around Ganquan Island in the South China Sea. Sunny sky and cloudless WV2 satellite imagery is selected to reduce atmospheric impact and improve the accuracy of coral reef classification. For the preprocessed imagery data, training samples were manually established, and the seafloor reef bottom types in the sea near Ganquan Island were divided into five substratum categories: coral reef, sand, reef, spray and deep sea water. These are classified using a Support Vector Machine (SVM), a Neural Network and the Maximum Likelihood method. The classification results were verified by comparing the satellite image visual interpretation classification results and the field seafloor image data. Confusion matrixes was compared and the classification accuracy was evaluated. The results show that the classification based on the SVM method is the best, with a total classification accuracy of 93.33% and a Kappa coefficient of 0.89. These results prove that high-resolution images can provide detailed and accurate information for coral reef bottom types classification and thus provide a scientific basis and reliable data for the establishment of a more complete coral reef ecological monitoring system and marine ecological environmental protection system.

Observation of a high abundance aggregation of the deep-sea urchin Chaetodiadema pallidum A. Agassiz and H.L. Clark, 1907 on the Northwestern Hawaiian Island Mokumanamana

AUV images of deep-sea megafauna on the seamount Mokumanamana in the Northwestern Hawaiian Islands revealed a large aggregation of the diadematoid urchin Chaetodiadema pallidum. This species had not been previously recorded on this seamount, but was known from other locations in the Hawaiian Archipelago. A total of 11,360 individuals were counted by eye from 1,369 seafloor images along the 200, 250, and 300 m depth contours on the west side of Mokumanamana. To estimate the full population size, Inverse Distance Weighting and Empirical Bayesian Kriging interpolation calculated a total population size of between 139,552 and 144,063 individuals. These estimates are 1-2 orders of magnitude greater than any previous estimates of urchin aggregation abundances. Little biological data exists for this species; thus, it is difficult to know if this is a spawning aggregation, a feeding front, a type of “schooling” behavior, or some combination thereof, but a food source did seem to be present in the form of mesophotic green algal detritus.

Tidal inlets in the Anthropocene: Geomorphology and benthic habitats of the Chioggia inlet, Venice Lagoon (Italy)

AbstractAdopting a multidisciplinary approach, we mapped with unprecedented detail the seafloor morphology, sediment distribution and benthic habitats of a tidal inlet in the Venice Lagoon, Italy, which has been greatly impacted by human activity. Thanks to very high‐resolution multibeam data, we identified ebb and flood‐tidal deltas, a tidal point bar, active dune fields, pools and scour holes.The seafloor substrate of the inlet was investigated by integrating automatically classified multibeam backscatter data with sediment samples and underwater seafloor images. The sediment composition comprises four textural classes with 75% overall thematic accuracy. The particle size distribution of each morphological feature was assessed distinguishing, in particular, sediments over crests and troughs of small‐dune fields with wavelengths and heights of less than 4 m and 0.2 m, respectively.Adopting state‐of‐the‐art benthic habitat mapping procedures, we found seven distinctive benthic habitats that reflect spatial variability in hydrodynamics and sediment transport pathways. The dominant classes were Sand with sparse shell detritus (46%) and Bare sand (32%). The rip‐rap revetment used for the inlet jetties and for the hard structures, built in the inlet channel to protect Venice from flooding, created a new habitat that covers 5.5% of the study area. This study shows how combining geomorphological and ecological analyses is crucial for the monitoring and management of tidal inlets and coastal infrastructures. © 2019 John Wiley &amp; Sons, Ltd.

Characteristics and Deposit Stratigraphy of Submarine-Erupted Silicic Ash, Havre Volcano, Kermadec Arc, New Zealand

Submarine eruptions dominate volcanism on Earth, but few are observed or even identified. Knowledge of how they operate is largely based on inference from ancient deposits, lagging by a decade or more our understanding of subaerial eruptions. In 2012, the largest wholly deep-subaqueous silicic eruption with any observational record occurred 700–1220 m below sea level at Havre volcano, Kermadec Arc, New Zealand. Pre- and post-eruption shipboard bathymetry surveys, acquisition by autonomous underwater vehicle of meter-scale-resolution bathymetry, and sampling by remote-operated vehicle revealed 14 seafloor lavas and three major seafloor clastic deposits. Here we analyze one of these clastic deposits, an Ash with Lapilli (AL) unit, which drapes the Havre caldera, and interpret the fragmentation and dispersal processes that produced it. Seafloor images of the unit reveal multiple subunits, all ash-dominated. Sampling destroyed layering in all but two samples, but by combining seafloor imagery with granulometry and componentry, we were able to determine the subunits’ stratigraphy and spatial extents throughout the study area. Five subunits are distinguished; from the base these are Subunit 1, Subunit 2a, Subunit 3, Subunit 4 (comprising the coeval Subunit 4 west and Subunit 4 east), and Subunit 2b. The stratigraphic relationships of the four AL unit subunits to other seafloor products of the 2012 Havre eruption, coupled with the wealth of remote-operated vehicle observations and detailed AUV bathymetry, allow us to infer the overall order of events through the eruption. Ash formed by explosive fragmentation of a glassy vesicular magma and was dispersed by a buoyant thermal plume and dilute density currents from which Subunits 1 and 2 were deposited. Following a time break (days/weeks?), effusion of lava along the southern caldera rim led to additional ash generation; first by syn-extrusive ash venting, quenching, brecciation, and comminution (S3 and S4e) and then by gravitational collapse of a dome (S4w). Slow deposition of extremely fine ash sustained S2 deposition across the times of S3 and S4 emplacement, so that S2 ash was the last deposited. These thin ash deposits hold information critical for interpretation of the overall eruption, even though they are small in volume and bathymetrically unimpressive. Ash deposits formed during other submarine eruptions are similarly likely to offer new perspectives on associated lavas and coarse pumice beds, both modern and ancient, and on the eruptions that formed them. Submarine ash is widely dispersed prior to deposition, and tuff is likely to be the first product of eruption identified in reconnaissance exploration; it is the start of the trail to vent hydrothermal systems and associated mineralized deposits of submarine volcanoes, as well as a sensitive indicator of submarine eruptive processes.

Ocean Floor Observation and Bathymetry System (OFOBS): A New Towed Camera/Sonar System for Deep-Sea Habitat Surveys

Towed camera systems are commonly used to collect photo and video images of the deep seafloor for a wide variety of purposes, from pure exploratory research to the development of management plans. Ongoing technological developments are increasing the quantity and quality of data collected from the deep seafloor. Despite these improvements, the area of seafloor, which towed systems can survey, optically remains limited by the rapid attenuation of visible wavelengths within water. We present an overview of a new towed camera platform integrating additional acoustical devices: the ocean floor observation and bathymetry system (OFOBS). The towed system maintains continuous direct communication via fiber optic cable with a support vessel, operational at depths up to 6000 m. In addition to collecting seafloor photo and video data, OFOBS gathers sidescan data over a 100-m swath width. OFOBS functionality is further augmented by a forward looking sonar, used to aid in hazard avoidance and real-time course correction. Data collected during the first field deployments of OFOBS, at a range of seamounts on the Langseth Ridge/Gakkel Ridge intersection (86° N, 61° E) in the high Arctic in September 2016, are presented to demonstrate the functionality of the system. Collected from a location with near continuous ice cover, this explanatory data set highlights the advantages of the system for deep-sea survey work in environments currently difficult to access for the majority of subsurface research platforms.

Compositional Differences in the Habitat-Forming Bryozoan Communities of the Antarctic Shelf

In some areas of the Antarctic shelf, bryozoans are abundant, acting as ecosystem engineers creating secondary structures with wide benthic coverage and harboring numerous other species. As the combined forces of global warming and ocean acidification threaten these habitats, we measured the composition of habitat-forming bryozoan communities using two techniques for imaging the sea floor, a YoYo-camera system and the AWI Ocean Floor Observation System (OFOS). YoYo-camera transects of the Bellingshausen, Amundsen, and Ross Seas were conducted during a research cruise on the R/V Nathaniel B. Palmer in 2013. OFOS transects included sites in the northern Palmer Archipelago where it borders the Scotia Sea and the Weddell Sea as part of the DynAMO project during the PS81 and PS96 cruises of R/V Polarstern in 2013 and 2015-16, respectively. Areas of bryozoan colonies were measured from the sea floor images using machine-learning algorithms available through the Trainable Weka Segmentation plugin developed for FIJI software. Habitat-forming bryozoan communities in the Palmer Archipelago and Ross Sea were largely composed of anascan flustrid species with finely mineralized skeletons, and to a lesser extent by other ascophoran lepraliomorph and umbonulomorph species having more robustly mineralized skeletons. Although habitat-forming bryozoan communities in the shallower (200 m) sites of the Weddell Sea also contained flustrid species, percent area and composition of flustrid bryozoans declined with increasing depth. Lepraliomorph and umbonulomorph bryozoan morphotypes were more abundant in the Weddell Sea, maintaining their relative percent area and increasing their percent composition between 200 - 400 m. Moreover, our analyses of species composition based on externally gathered datasets show similar trends among sites, depths, and degrees of colony mineralization to our seabed imaging study. Variation present in the bryozoan species compositions of the Amundsen and Bellingshausen Seas suggest that these areas potentially represent divergent bryozoan communities requiring further validation via remote imaging surveys. Overall, compositional differences among Antarctic habitat-forming bryozoan communities are likely influenced by the combined effects of seasonal ice scour and carbonate chemistry, which in an increasingly acidified and warming ocean may put the communities of the eastern Weddell Sea at greater risk.

Seafloor Image Color Enhancement Method based on Retinex model and Experiment Report in the undersea environment

Seafloor Image Color Enhancement Method based on Retinex model and Experiment Report in the undersea environment

Deep Semantic Segmentation in an AUV for Online Posidonia Oceanica Meadows Identification

Recent studies have shown evidence of a significant decline of the Posidonia oceanica (P.O.) meadows on a global scale. The monitoring and mapping of these meadows are fundamental tools for measuring their status. We present an approach based on a deep neural network to automatically perform a high-precision semantic segmentation of P.O. meadows in sea-floor images, offering several improvements over the state of the art techniques. Our network demonstrates outstanding performance over diverse test sets, reaching a precision of 96.57% and an accuracy of 96.81%, surpassing the reliability of labelling the images manually. Also, the network is implemented in an Autonomous Underwater Vehicle (AUV), performing an online P.O. segmentation, which will be used to generate real-time semantic coverage maps.