This PDF file contains the front matter associated with SPIE Proceedings Volume 12543, including the Title Page, Copyright information, Table of Contents and Conference Committee lists.

In the underwater environment, scattering due to optical turbulence can degrade radio frequency (RF) information encoded on a laser beam’s intensity profile and can limit the effectiveness of free-space optical communications. However, for sensing applications, changes in the RF subcarrier due to turbulence could be used to characterize the underwater environment. An alternative form of modulation can be created through heterodyne detection by interfering two co-propagating beams with different optical frequencies on an optical detector. This form of modulation was used to sense spatial properties of optical turbulence using beams carrying orbital angular momentum (OAM). Upon detection, each modulation method results in an oscillating photocurrent; however, it is not clear if the photocurrent produced by each modulation method responds differently to the effects of turbulence. To address how these modulation schemes may be affected by turbulence, a series of experiments are conducted. The results are analyzed to identify the impact of different charges of OAM relative to a Gaussian beam.

Standard transmissometer techniques for measuring beam attenuation are unable to distinguish unperturbed, transmitted light from light scattered in the forward direction. Recent research efforts have proposed methods to distinguish forward-scattered photons by exploiting the spatial incoherence caused by scattering. We revisit the approach of Alley et al. (2018) and further explore the changes to coherence structure caused by particle scattering. We consider methods to decrease the data processing time, motivated by the goal of an in-situ sensor that can be evaluated in real time. Additionally, we consider how experimental devices could be optimized to perform in high-noise, marine environments.

We consider the design and generation of spatially partially coherent (SPC) beams carrying orbital angular momentum (OAM) propagating through complex random media. It has been theoretically shown that spatial coherence can be controlled through a prescribed linear superposition of Laguerre-Gaussian (LG) modes. Experimentally the SPC beams are obtained by randomly cycling the phase screens of the coherent modes, with each mode contributing a weight that is proportional to its eigenvalue in the coherent mode decomposition equation. The spectral degree of coherence, ξ , theoretically varies from 0 (fully coherent) to 1 (incoherent). Experimentally, it is suggested that we can reach the highest level of incoherence when the modes are combined where LG mode orders are of equal weights. Preliminary measurements indicate a reduced coherence corresponding to increasing ξ. Our experimental design imposes turbulence on the beam to examine the effects of its spatial partial coherence on the scintillation index (SI). It has been shown that benefits to communication system performance, specifically underwater, can be achieved through the control of spatial coherence properties of laser light propagation.

The use of light with orbital angular momentum (OAM) is being investigated in a wide range of fields that include material interrogation, light propagation, sensing and communication1 . The defining characteristic of the electric field of OAM light is an angular term described by an azimuthal component such as e ^-imφ , which produces helical phasefronts and an angular momentum equal to mħ² . For the cases where m ≠ 0 the beams can have an intensity minima along the central axis creating a beam of light which is tube-like in structure. This, together with the angular component of the energy flux, can cause the scattering interactions with materials to be different than that from plane waves. Theoretical angular scattering calculations show that the light scattering maxima can occur at different angles from the forward direction of zero degrees. In this work we investigate the scattering properties of OAM light from single, micron sized spherical particles that are suspended in a linear electrodynamic trap. Using phase plates we generate OAM beams (wavelength of 532 nm) that are incident on a single suspended particle. Using three separate CCDs we capture the scattered light intensity over a total range across 40 degrees in the forward, back and side scattering planes. Comparisons between angular scattering measurements from Gaussian beams and OAM mode equal to 3 is presented.

Submersible vehicles, such as Uncrewed Underwater Vehicles (UUVs), can be lost underwater and it is desirable to locate them before they go missing or collide with other vehicles. These objects generate inherent signatures when they move in the ocean. With a better understanding of these signatures, more information regarding the motion of underwater vehicles can be inferred using LiDAR technology. In this paper, the authors review existing literature on various non-acoustic signatures of a submerged body, namely, the electromagnetic, biological and thermal signatures, turbulent wake patterns, internal waves and vortex structures. The authors discuss how these signatures evolve both spatially and temporally. Furthermore, the review investigates how environmental and operational parameters such as the stratification of the medium, vehicle speed, shape and depth affect the non-acoustic signatures. Underwater vehicles operating at low speeds and large depths have bioluminescent, Kelvin wake and Bernoulli hump signatures that are difficult to detect on the surface. Thermal signatures, vortex wakes and far wakes are only likely to be detected at the vehicle’s depth. This is primarily due to the ocean stratification which suppresses the vertical motion of underwater turbulent signatures. Thermal signatures appear to be most likely to be detected. The study concludes that relying on a single signature to detect submersibles is not advisable and future methods for underwater vehicle detection should use multiple sensors to detect complementary signatures.

In this work, the radar processing technique of Range-Doppler processing is investigated for its potential to enhance performance of the radar-encoded laser system in rangefinding applications. One of the challenges experienced by this system is in discriminating between the returns from underwater objects and environmental clutter in highly-scattering and/or low signal-to-noise ratio conditions. The intention of this work is to investigate whether the addition of a new dimension, velocity, will improve ranging performance. This work presents the application of the Range-Doppler processing technique to transform data collected in a laboratory water tank. Results and performance improvements using the radar-encoded laser system are compared against those obtained with a conventional, short-pulse laser.

This paper addresses the role of uncertainty in spatial point-process models, such as those that might arise in modelling ship traffic. We consider a doubly stochastic Poisson point process where the intensity function is uncertain. To assess the role of uncertainty, we conduct a large set of numerical trials where we estimate a doubly stochastic Poisson point-process model from historical target data, and the evaluate the model by assessing the target detection performance of a set of sensors whose locations are selected using the model. Our work is motivated by seabed sensors that detect ship traffic, and we conduct numerical trials using historical ship traffic data near the mouth of the Chesapeake Bay, Virginia, USA, that was recorded by the Automated Identification System.

Optical sensors play a significant role in maritime sensing. The high spatial resolution of cameras is vital to object identification, but the performance of flood illuminated CCD/CMOS cameras quickly deteriorates in degraded visual environments (DVE) such as turbid coastal waters and murky harbors. Several laser-based sensors have been developed over the past few decades to enhance optical imaging in DVEs. However, since conventional approaches require the laser and receiver be located on the same platform, the size, weight, and power (SWaP) requirements are incompatible with small remotely operated vehicles (ROVs). Researchers at the Naval Air Warfare Center – Aircraft Division in Patuxent River, MD (NAWCAD PAX) developed [6] and patented [8] an optical imager utilizing a bistatic geometry where the laser and receiver are mounted on separate, smaller platforms. This allows the SWaP of the sensor components to be optimized for small platforms, like ROVs. The transmitter uses a modulated laser beam and a microelectromechanical system (MEMS) scanner to sequentially illuminate an underwater object in a raster fashion. A distant receiver collects the object reflected laser light and reconstructs the imagery. The transmitter scans the object at frame rates comparable to video cameras, producing a laser video stream. This paper will report on laboratory tests comparing the spatial resolution performance, as a function of attenuation length, of the bistatic frame-based laser imager (FLI) to a conventional flood illuminated camera typical of small ROVs.

The ctenophore Mnemiopsis leidyi is an opportunistic species that can be extremely abundant and invasive in many parts of the world. It is well known for its bright bioluminescence, but its light emission response to flow stimulation has not been rigorously quantified. The objective of this study was to determine the luminescent response of cydippid larvae of M. leidyi to two types of mechanical stimuli, including an impeller pump within the UBAT bathyphotometer and stirring as the stimulus within an integrating sphere. Tests were conducted with less than one week old cydippid larvae, analyzing flash parameters of rise time, peak intensity, decay slope, decay time, total integrated emission (TMSL), integrated flash emission, and flash duration. Cydippid larval size had a positive correlation with peak intensity. There were four patterns of bioluminescent responses from the UBAT but they did not have statistically different flash kinetics. For the integrating sphere, the average peak intensity and TMSL were much greater than for the UBAT, possibly due to the two forms of stimulation. However, a constant phosphorescent emitter was 2.6 times brighter when measured with the integrating sphere compared to the UBAT, suggesting inaccurate photon calibration of the UBAT perhaps due to light measurement geometry. This study provides a well-defined baseline of cydippid larvae flash responses that can be used for interpreting field measurements made with bathyphotometers and to determine their contribution to the bioluminescence potential of waters where they are present.

Traditional acoustic and magnetic techniques for detecting underwater vehicles are becoming less reliable due to advances in underwater technology such as composite materials, miniaturized electronics, and more space efficient battery technologies. Optical remote sensing technologies, such as Light Detection and Ranging (LiDAR) systems, are promising alternatives due to their high measurement accuracy, independence of operating environments, and simple integration onto airborne platforms. However, the penetration depth of direct detection methods is limited by the strong attenuation of light by the water column. Detection techniques that rely on monitoring changes to the inherent optical properties (IOPs) and other remotely sensed properties of water are thus being considered. Effects of underwater vehicles on ocean properties such as salinity and temperature have been well studied, but a stronger understanding of their effect on IOPs and optical constituents of the ocean is required for these new detection techniques. In this paper, the authors develop a system to measure the effect of underwater turbulence equivalent to an inspection-class Remotely Operated Vehicle (ROV) on the IOPs and other physical properties of the water column. Measurements are taken in an indoor water tank, freshwater reservoir, coastal waters and oceanic waters. Four vertical thrusters are used as a turbulence generator. Four commercially available sensors monitor the changes in IOPs and optical constituents of water above the turbulence generator. Preliminary results are presented on the effect of underwater turbulence on the optical and ocean properties measured. We conclude that the turbulence was able to be detected via changes in the IOPs at a distance of 10m under most conditions, with caveats and qualifiers discussed.

The total non-water absorption coefficient of seawater, 𝑎_𝑝𝑔(𝜆) (𝑚⁻¹ ) (light absorption coefficient after subtraction of pure water contribution) provides information about the amount of light absorbed by various optically significant substances in natural waters (𝜆 is wavelength). Partitioning 𝑎_𝑝𝑔^(𝜆) into phytoplankton, 𝑎_𝑝ℎ^(𝜆) and colored detrital matter, 𝑎_𝑑𝑔(𝜆) is useful to understand the light interaction with the distribution and variability of constituent matter, light availability at various depths, and ecological and biogeochemical cycles, as these constituents represent pools of carbon and other elements. Some of the existing partitioning methods either require ancillary inputs or assume limited shapes for constituents absorption in deriving the 𝑎_𝑝ℎ(𝜆) and 𝑎_𝑑𝑔(𝜆) from 𝑎_𝑝𝑔(𝜆). In this study, we propose a decomposition method of 𝑎_𝑝𝑔(𝜆) using a spectral optimization routine utilizing a spectral library consisting of various shapes for both phytoplankton and colored detrital matter absorption components. The proposed method does not require any ancillary inputs in deriving absorption of the constituent subcomponents. Performance of the proposed method is evaluated using two dataset compilations covering a very wide range of water types with sampling locations. Among various parameterizations tested in the decomposition method, the parameterization with the phytoplankton shape model of Ciotti et al. (2005) combined with exponential, stretched exponential, and hyperbolic shapes for colored detrital matter resulted in lower Mean Absolute Percentage Errors (MAPE) consistently across all sites. The good performance of the proposed method is characterized by average MAPE values of 17% and 13% and average percentage absolute errors (%AE) of 15.5% and 11.5% for the derived 𝑎_𝑝ℎ(443) and 𝑎_𝑑𝑔(443) respectively. The proposed method exhibited better performance with 7 - 10% lower average spectral MAPE (MAPE averaged over all wavelengths) values compared to two other existing partitioning algorithms in optically complex waters. The proposed method can be used for deriving the constituents absorption from input data of 𝑎_𝑝𝑔(𝜆) collected from various oceanographic and remote-sensing platforms. Since apg is a core product of several semi-analytical ocean color inversion algorithms, this approach has relevance to the future hyperspectral NASA PACE ocean color imager, as it is directly adaptable to hyperspectral reflectance data.

The design and the calibration of the Beam-attenuation, b and bb Laser Underwater Environment Sensor (4BLUES), which borrows from advanced designs developed by the authors’ individual groups in the course of the last two decades. The sensor features unparalleled accuracy in determining the optical scattering and backscattering coefficients, which are critical parameters for the remote sensing of water constituents and in-water bio-optical applications. Calibration procedures performed in the laboratory with spheres of known properties are presented together with an assessment of the performance in two field experiments in coastal waters. The more complex calibration of the scattering channel includes a novel inversion approach, which accounts for scattering and absorption losses along the laser-beam path. Strong agreement was observed with current advanced sensors for extinction and the backscattering measurement for one of the field experiments. Total scattering comparison with the current state of the art showed reasonable agreement (within 25%) for all stations except one, where the interpretation for the higher discrepancies is unresolved. These preliminary results suggest further assessment is warranted.

A rigorous view of the uncertainties in Sunstone PSICAM (Point Source Integrated Cavity Absorption Meter) measurements are investigated. Multiple sources of error can greatly influence the accuracy of PSICAM absorption measurements: measurement noise, spectrometer stability, light source stability, calibration and external verification of the standard are a few of the sources of errors that were investigated. Measurement noise can be reduced through repetitive measurements. Using more than 300 averages for each measurement provided repeatability of measurements with a mean absolute difference less than 0.001 m^-1 across the visible spectra. Like with other tube and cuvette style absorption meters, bubbles are a potent contaminant when determining absorption. The stability of the spectrometer with time and temperature are explored. Non-linearity and wavelength registration are also considered in reducing errors. Light source stability and output with time are also presented. Due to its small size the PSICAM is used both in the lab and at sea. Accurate absorption values require calibration to be conducted to account for the reflectivity (𝜌) of the sphere. Currently nigrosine dye is used to measure the reflectivity and requires a priori knowledge of the absorption of the dye. An examination of the stability of nigrosine dye is conducted for both filtered and non-filtered dye. A solid standard is proposed to improve the determination of reflectivity in the sphere.

The vulnerable loggerhead sea turtle Caretta caretta is the most common sea turtle to nest on the beaches of the Georgia coast. A growing concern to the conservation of these long-lived marine reptiles is from non-point-source inland community lights creating shoreward sky glow that impacts turtle nesting. Calibrated quantum or radiometric light measuring devices often have limited dynamic range and resolution and are not able to effectively measure incoming irradiance, while consumer cameras images are altered to adjust for human light perception. In order to study the disorienting effect of light, low cost, portable, quickly-acquired spectral radiometric measurements of coastal light pollution are needed to fully characterize its effect on marine macrofauna, e.g., sea turtle hatchlings. Radiometric devices are expensive, have low sensitivity, low resolution, and are impractical for the large-area spatial collection of light pollution measurements in nighttime beachfront environments. Consumer cameras have large dynamic range with a variety of available mating optics, but are intended for photometric imaging of scenes in accordance with the CIE standard characterizing the human visual system. We present a recently-patented method to associate relatively low-cost CCD/CMOS RGB sensor data directly to nm-precision, visual-light irradiance spectra. Calibration uses well-characterized solar imagery at local apparent noon passing through an inexpensive apparatus comprised of multiple neutral density and narrowband filters as well as a transparent diffraction grating to simultaneously capture irradiance and spectral data for generation of calibration values. Measurements have been collected near sea turtle nesting sites in order to evaluate potential beachfront lighting experienced by the nesting female turtles and emerged hatchlings.

Accurate estimation of atmospheric wind velocity plays an important role in weather forecasting, flight safety assessment and cyclone tracking. Atmospheric data captured by infrared and microwave satellite instruments provide global coverage for weather analysis. Extracting wind velocity fields from such data has traditionally been done through feature tracking, correlation/matching or optical flow means from computer vision. However, these recover either sparse velocity estimates, oversmooth details or are designed for quasi-rigid body motions which over-penalize vorticity and divergence within the often turbulent weather systems. We propose a texture based optical flow procedure tailored for water vapor data. Our method implements an L1 data term and total variation regularizer and employs a structure-texture image decomposition to identify key features which improve recoveries and help preserve the salient vorticity and divergence structures. We extend this procedure to a multi-fidelity scheme and test both flow estimation methods on simulated over-ocean mesoscale convective systems and convective and extratropical cyclone datasets, each of which have accompanying ground truth wind velocities so we can qualitatively compare performances with existing optical flow methods.

Cyprus can be found in a critical location, a geostrategic location among Asia, North Africa, and Europe. This is particularly important, thus, can affect the marine environment with the vessels that pass through the sea. A severe form of pollution can be an oil spill. An oil spill is a liquid, usually petroleum hydrocarbon, released to the sea and can affect the marine ecosystem and humans. Oil spills can be recorded as marine incidents or accidents. In August 2021, one of the essential accidental oil spills came from the northeast coast of Syria. In the present study, we will try to implement a spatial analysis of oil spill detection by combing remote sensing techniques and monitoring, using the Sentinel Application Platform (SNAP) and the consequences that follow this procedure. The implementation was made using two different polarization; vertical polarization (VV) and horizontal comparison (VH). The preliminary results show that the Sentinel-1 SAR data would give effective results and spatial information on oil spill detection to decision-makers.

Image analysis in scenes with low contrast in the infrared is difficult in general. At sea, this can be further challenging due to varying levels of sea clutter. Ideally, adaptive algorithms should be able to consider the clutter to determine how sensitive they can be, and to what degree success is possible. This would allow greater control over the balance between Type I and Type II errors,1 and improve performance. To achieve this, we need measures for sea clutter. This paper is a search for features that seem useful for discriminating between different levels of sea clutter. We investigate spatial features, such as texture energy measures, properties of the Fourier coefficients, and the statistical features of segments in a thresholded image. In addition, inspired by the field of radar, where sea clutter is well studied, we will look at the statistical distribution of the IR sea clutter itself. Results indicate that these features, when paired with classification algorithms, can be useful to discriminate between scenes with various levels of sea clutter. In addition, they might even be suitable as a quantitative measure of the amount of clutter in the scene.

Recent incidences of sabotage of Oil and Gas pipelines in the Baltic Sea have made it clear that sub-sea pipelines are vulnerable and will require surveillance for protection. Harbors and ports are noisy. Pressure measurements from hydrophones allow only measurements of signal arrival times. With only this information, a dense, spatially dispersed sensor array is required for detection and localization. Using vector sensors allow additional estimation of direction of propagation at each measurement point, allowing separating signals from multiple sources. A smaller number of vector sensors can replace a larger number of scalar pressure sensors. Static sea-floor deployment of the sensor array, allows accurate mapping after deployment, reducing inaccuracies in sensor locations found in mobile systems. We describe a method that allows using multi-component sea-floor sensors to identify and track any source of acoustic energy even when the orientations of the multi-component sensors are unknown. Sources that satisfy certain simple assumptions can be used to find the absolute orientation of the multi-component sensors. The methodology allows using a simplified sensor array where a short-interval fiber-optic DAS array is combined with a sparse array of two-component, transverse-only, point sensors. We describe optical multi-component accelerometers, that are easily combinable with DAS sensor arrays. Without electronics, there is no need for power - and the risk of failure of electronics left for a long period of time is eliminated, significantly increasing the lifetime of the installation. The acquisition system for the hybrid optical array will be simpler and enable “always-on” monitoring.

During summer and fall of 2022, multiple flights were conducted with an airborne blue wavelength scanning lidar west of San Diego and in waters surrounding Iceland. Geo-registered lidar 𝑘 profiles reveal multiple ocean parameters such as mixed layer depth variations and dense plankton layers over scales of meters to kilometers. These measurements can be compared to both historical measurements of lidar 𝑘 profiles conducted with in situ instrumentation, as well as to satellite-derived measurements of ocean parameters. The main challenge of in situ oceanographic measurements is the difficulty in achieving efficient coverage of a wide area. Meanwhile, satellites cover a wide area but may not provide sufficient resolution for oceanographic studies; for example NASA’s Aqua/MODIS satellite pixel spacing resolution is on the order of 10 kilometers. The airborne lidar measurements provide larger coverage area than an in situ instrument while also providing higher resolution and greater depth penetration than a satellite measurement. This paper provides an overview of the airborne blue wavelength scanning lidar and demonstrates measurements of two ocean water properties, the average diffuse attenuation coefficient in the mixed layer and the mixed layer depth. The airborne lidar measurements of these properties show reasonable agreement with relevant satellite and in situ databases.

The Stokes vector components and the degree of linear polarization of light reflected from the air-water interface contain information about the roughness of the ocean surface, which is correlated with the wave slope statistics and may be used to retrieve it using the Polarization Slope Sensing (PSS) method (Zappa et al., 2008). This statistic is a part of the radiative transfer simulations in the atmospheric correction of the ocean color satellites and other applications. A modification of the method, which minimizes the impact of upwelling light on polarimetric measurements of the reflected light was applied by using Teledyne DALSA camera equipped with a Sony sensor, where each of 1232x1028 pixels had four subpixels with 0-, 90-, 45- and 135-degrees orientation of polarization. In addition, a filter wheel with several color filters was installed in front of the camera, allowing to measure wave slope characteristics at several spectral bands. Shipborne measurements during VIIRS Cal/Val cruises in the Gulf of Mexico and in Hawaii and from a helicopter at several heights during the CCNY cruise in the Chesapeake Bay showed the advantage of the proposed modified polarimetric slope sensing technique. Measured variances of the wave slopes were mostly in the range predicted by Cox-Munk relationships with corresponding standard deviations.

Video surveys are commonly used to monitor the abundance and distribution of managed species to support management. However, considerable effort, time, and cost are required for human review and automated fish species recognition provides an effective solution to remove the bottleneck of post-processing. Implementing fish species detection techniques for underwater imagery is a challenging task. In this work, we present the Multiple Instance Active-learning for Fish-species Recognition (MI-AFR), which is formulated as an object detection-based approach to perform localization and classification of fish species. It can select the most informative fish images from unlabeled sets by estimating the uncertainty of unlabeled images by using adversarial classifiers trained on labeled sets. Moreover, we have analyzed the improved performance of MI-AFR by considering different backbone networks as a trade-off between speed and accuracy. For experiments, we have used the fine-grained and large-scale reef fish dataset obtained from the Gulf of Mexico – the Southeast Area Monitoring and Assessment Program Dataset 2021 (SEAMAPD21). The experimental results illustrate that the superiority of the proposed method can establish a solid foundation for active learning in fish species recognition, especially with a small number of labeled sets.

Species recognition is an important aspect of video based surveys, which support stock assessments, inspecting the ecosystem, handling production management, and protecting endangered species. It is a challenging task to implement fish species detection algorithms in underwater environments. In this work, we introduce the YOLOv5 model for the recognition of fish species that can be implemented as an object detection model for analyzing multiple fishes in a single image. Moreover, we have modified the depth scale of different layers in the backbone of the YOLOv5 model to obtain improved results on fish species recognition. In addition, we have implemented a transformer block in the backbone network and introduced a class balance loss function to obtain enhanced performance. It can perform fish species recognition as an object detection approach by classifying each of the fish species in addition to localizing for the estimation of the position and size of the fish in an image. Experiments are conducted on the fine-grained and large-scale reef fish dataset that we have obtained from the Gulf of Mexico – the Southeast Area Monitoring and Assessment Program Dataset 2021 (SEAMAPD21). The experimental results demonstrate that an enhanced YOLOv5 model can yield better detection results in comparison to YOLOv5 for underwater fish species recognition.

Fish species recognition and detection are essential for fishery industries. Accurate and robust species classification and detection play a vital role in monitoring fish activities and identifying the distribution of a specific species, which is vital to know the endangered species. It is also essential for controlling production and overall ecosystem control and management. However, the role of current artificial intelligence technologies, such as deep learning, is limited in the ocean system compared to other areas like robotics and security. The major challenge in building a deep learning network is data availability, time, and cost of annotation and labeling. In this work, we build a semi-supervised deep-learning network to recognize fish species. The model is based on a student-teacher network where the teacher network generates pseudo-labels, and the student network is trained with the generated pseud-labels and the labeled data simultaneously. The student network updates the teacher network via an exponential moving average method. The model consists of a faster R-CNN with a feature pyramid network detector. The experimental result of the model on the challenging fish dataset shows a promising result for building semi-supervised object detection models.

Characterization of the optical turbulence of complex media is important to designing resilient free-space optical communication systems. Previous studies have used machine learning algorithms to characterize optical turbulence in the atmospheric environment, but we propose to extend this concept to the underwater medium. Our experimental design propagates a Gaussian beam ~1.25 meters through a Rayleigh-Bénard (RB) turbulence tank, which creates realistic optical turbulence that is fully controllable and repeatable. The intensity and phase distortions of the Gaussian beam after propagation will be collected and used to train a convolutional neural network (CNN), for the purpose of the underwater optical turbulence characterization. The CNN will be trained to classify turbulence levels based on both intensity and phase measurements in varied levels of optical turbulence.

Recently, with the development of high-resolution remote sensing technology and artificial intelligence-based image decoding capability, several detection methods have been studied for shore debris. This study used pan-sharpened KOMPSAT-3A images (spatial resolution: 0.55 m), and the atmospheric correction was performed using the COST model. In order to obtain input data corresponding to Styrofoam, 12 pieces with a size of 0.9 to 3.6 m were installed on the sand, vegetation, and rock, respectively, and 96 pixels were selected through a random sampling method. The classification was performed on four regions of interest. Styrofoam-classified pixels were displayed on the satellite image. As a result of the SVM Linear model, the accuracy was 0.68%, which was very high compared to other models. It was found that the calculated area was underestimated by about 186m2 compared to the drone, which causes microscopic coastal debris due to the relatively low satellite resolution.

Marine pollution is a major environmental hazard and a serious healthcare, economic, and social issue. Machine learning (ML) and deep learning (DL) techniques can be used to automate marine waste removal and make the cleanup process more efficient. The proposed study uses image classification to help categorize the level of marine pollution in ocean underwater regions. The performance of two deep convolutional neural networks (VGG19 and ResNet50) is investigated in this study and VGG19 reported an accuracy of 98.1%.

Active illumination with underwater laser imaging has unique advantages for the identification of underwater objects, especially in shallow waters, complex marine environments and inaccessible locations. Laser intensity images embody valuable information that can be utilized for object recognition; however, backscattered light from the water column and other particulates blur the resulting laser images, rendering the objects in the images unintelligible. Although over the years a variety of deblurring and other image restoration and enhancement algorithms have been proposed, these works primarily consider optical images of scenery, not monotone underwater images of objects, for which contours are more critical. This work proposes the utilization of edge metrics to evaluate the efficacy of image restoration and enhancement algorithms for underwater laser images. Our results provide insight into the best methods for improving underwater laser image quality for object recognition.

Plastic pollution has emerged as one of the biggest environmentally threatening issues. Using image classification, the proposed study aids in categorizing the level of marine pollution in ocean underwater regions. This study classified the amount of pollution in the ocean using the two variants of Inception Convolutional Neural Network (CNN) models i.e., Inception- ResNet V2, and InceptionV3. High accuracies of up to 96.4% have been reported. This study will help researchers working in the field of water quality detection.

Seagrasses are one of the most valuable ecosystems in the sea. Seagrasses provide shelter and protection to newborn fish, and food, it is an indicator of the health of coastal ecosystems and plays a key role in the protection of erosion. Seagrasses provide many ecosystem services. In order to act helpfully for the health of this kind of valuable ecosystem, it is very important to map the seagrass beds. Seagrass mapping allows us to detect, monitor, and finally protect them from different sources of pressure. Seagrasses are very sensitive to changes, and this can affect the health of this type of ecosystem. For the first step, we have to detect what exists in Cyprus's area of interest. There are many ways we can implement the detection and mapping of seagrass. We will provide an estimation with pixel-based image analysis. The purpose of the paper proceedings is to detect seagrass with pixel-based image analysis, in Cyprus's area of interest, and especially the broader area of Paphos.

Maritime surveillance is critical for threat prevention, national security, and safety maintenance. Maritime traffic involves the worldwide navigation of millions of vessels. Satellite missions like Sentinel-1 made it possible to collect systematic data for vessel tracking. Using the Copernicus Open Access Hub service, it is now feasible to access satellite data in fully automated and near real-time mode and deliver vessel information through a web portal interface. This paper aims to detect vessels in exclusive economic zone (EEZ) of the Republic of Cyprus, Southwest of Cyprus using freely available Sentinel- 1 SAR imagery data. Comparison of results obtained using Sentinel Application Platform (SNAP) and Arc GIS Pro - Deep Learning. The preliminary results show that both tools can be used to ship detection and give satisfactory results.