ScienceWatch@KIOST
Not a member yet
29648 research outputs found
Sort by
Vessel detection for maritime traffic management using U-Net with backbone networks
No full textEnsuring maritime safety in South Korea's heavily trafficked waters requires vessel monitoring systems capable of comprehensively and simultaneously surveying the entire maritime domain. Existing systems often monitor only localized regions, highlighting the critical need for nationwide maritime surveillance capabilities to support traffic network management. Therefore, this study proposes a deep learning–based approach for large-scale vessel detection using Sentinel-2 satellite imagery and U-Net architectures. Four convolutional neural network backbones—Visual Geometry Group Net, Residual Net, Dense Convolutional Network, and EfficientNet—were integrated into the U-Net encoder–decoder framework. These models were trained on manually annotated vessel data and evaluated using the Dice coefficient to measure detection accuracy. Among the tested architectures, the EfficientNetB2 backbone achieved superior performance with a Dice coefficient of 0.9919. The proposed methodology demonstrates the potential of combining high-resolution optical satellite imagery with optimized deep learning models for reliable, wide-area vessel detection for maritime traffic management.11Nsciescopu
Multiscale Window Analysis for Enhanced Detection of Sea Ice in Liaodong Bay Using Geostationary Ocean Color Imager-II and Sentinel-2 Data
No full textSea ice should be detected accurately to mitigate risks to marine industries, protect coastal infrastructure, and understand the dynamic interactions between sea ice and the marine environment. In this study, we developed and validated a sea ice detection algorithm based on a multilayer perceptron (MLP) for the Bohai Sea using Geostationary Ocean Color Imager-II (GOCI-II) and Sentinel-2 data. High-resolution Sentinel-2 imagery generates the sea ice truth data, which are then resampled to align with the GOCI-II grid. Rayleigh-corrected reflectance data from GOCI-II are used as input variables for the MLP algorithm. Multiscale Window Analysis enhances detection accuracy and reduces the false detections inherent in pixel-based approaches. The performance of the detection algorithm is evaluated using different window sizes (1 × 1, 3 × 3, 5 × 5, 7 × 7, and 9 × 9 pixels). The results demonstrate that increasing the window size enhances performance metrics, underscoring the importance of incorporating the surrounding reflectance variations in sea ice detection. Additionally, the developed algorithm was applied to hourly GOCI-II images for further analysis. A qualitative evaluation confirmed that sea ice was successfully detected, and the algorithm effectively identified the temporal movement of sea ice.11Ysciescopu
Optimized preprocessing methods for marine RNA metavirome studies
No full textThe pathogenicity of marine RNA viruses has been extensively studied, as they infect a wide range of hosts. Recently, interest in the diversity and ecological roles of marine RNA viruses has been increasing. Marine RNA viruses are generally smaller than DNA viruses, making them more challenging to concentrate, and their inherent instability leads to rapid degradation. Therefore, optimizing an efficient RNA virus concentration system is crucial. In this study, we conducted 95 experiments with seawater volumes ranging from 60 L to 300 L to evaluate the capture efficiency of RNA viruses using three concentration methods: membrane filtration, ultrafiltration, and ultracentrifugation. Filtering 100 L of seawater through a 0.2 μm nitrocellulose disc yielded a 97.3 % library QC pass rate, whereas an equivalent 0.2 μm TFF cassette and Fe3+ flocculation each yielded 0 % QC success. The optimal method to concentrate marine RNA viruses involved filtering 100 L of seawater through a 0.2 μm membrane, followed by a primary concentration using a 30-kDa tangential flow filtration system and finally ultracentrifugation. Ultracentrifugation enriched RNA-virus contigs to 81.5 % of the virome, which was three-fold higher than 30 kDa centrifugal ultrafiltration (25.7 %; p < 0.05), and recovered 763 high-quality contigs spanning 27 families and 66 genera. This approach demonstrated high reproducibility. Our results present an effective method for capturing and analyzing RNA viruses in marine environments, providing a valuable tool for further investigating their diversity and ecological roles.33Nscopu
Annual distribution and deposition of atmospheric 210Pb in Busan, the largest port city in Korea
No full textThis study examined the annual and seasonal variations of 210Pb in aerosols collected in Busan, a major port city in South Korea. Aerosol samples were obtained during two sampling campaigns: (i) total suspended particles (TSP) from April 2019 to February 2020, and (ii) particulate matter with diameters ≤10 μm (PM10) and ≤2.5 μm (PM2.5) from March 2020 to February 2021. The activity concentrations of 210Pb ranged from 0.10 to 1.98 mBq m−3 (TSP), 0.09–1.97 mBq m−3 (PM10), and 0.02–2.07 mBq m−3 (PM2.5), showing clear seasonal trends. Activities peaked in autumn and winter (October–February) and were lowest in summer (July–September), with up to two-fold seasonal variation. These patterns were mainly attributed to meteorological variability and increased anthropogenic emissions during colder months due to seasonal wind shifts. No significant differences in 210Pb activity appeared among size-fractionated samples (TSP, PM10, and PM2.5), indicating a preferential association of 210Pb with fine aerosols. Dry deposition fluxes of 210Pb in Busan were comparable to those reported in other global cities, despite only including the dry component. 210Pb activities in TSP correlated with 40K in fallout dust, likely from resuspended soil, while 210Pb in PM2.5 paralleled 7Be variations in air. Moreover, 210Pb/Pb ratios showed seasonal changes, notably in PM2.5 during the dry season. The results suggest that aerosol behavior in Busan is notably influenced by anthropogenic sources during dry months, especially in fine dust (e.g., PM2.5). Overall, 210Pb proves useful as a radioactive tracer for understanding aerosol dynamics.11Nsciescopu
Analysis of the Influence of Input Conditions on the Diffuse Attenuation Coefficient for Downwelling Irradiance Based on the Radiative Transfer Model
No full textThe diffuse attenuation coefficient for downwelling irradiance (Kd) represents the exponential decrease of light with depth due to scattering and absorption in seawater, and is one of the key products for describing ocean optical properties. To estimate Kd over broad oceanic regions, semi-analytical models utilizing remote-sensing reflectance (Rrs) and inherent optical properties (IOPs) derived from satellite observations are commonly employed. Among these, Kd models based on radiative transfer simulations often consider only the solar zenith angle (SZA) among various atmospheric and water quality variables, which can lead to discrepancies between estimated and actual values. In this study, we investigate the effects of four atmospheric variables—SZA, wind speed, aerosol type, and aerosol optical depth (AOD)—on Kd using radiative transfer simulations. To reflect varying water quality, simulations were conducted for three water types: clear, intermediate, and turbid. By analyzing the change in Kd values as each environmental variable was changed from its minimum to maximum condition, we found that SZA had the greatest impact, followed by AOD. In contrast, wind speed and aerosol type showed minimal influence and may be excluded from future algorithm improvement considerations. When analyzing Kd estimated from the semi-analytical model, SZA-induced errors reached up to 28.17% in turbid waters, and AOD-related errors were as high as 17.93%. Although the semi-analytical model theoretically accounts for SZA, significant discrepancies were still observed, especially under turbid conditions. Therefore, to improve the accuracy of Kd estimation, the influence of SZA should be prioritized, and additional corrections for AOD and turbidity should also be incorporated into the Kd model development. This study identifies key environmental variables affecting Kd retrieval and suggests directions for improving the model, contributing to the development of more accurate Kd algorithms applicable under diverse marine conditions.33Nscopuskc
Comparison of False Positive Case in Coastal Debris Using Deep Learning-Based Object Detection Models
No full textDeep learning-based object detection models, such as YOLO and DETR, have been actively studied for monitoring coastal debris. While recent models exhibit minimal differences in quantitative accuracy and performance, the underlying algorithms and methodologies for object detection vary across models. Consequently, detection outcomes can differ based on the type of the debris and the characteristics of the coastal environment. Nonetheless, there is a notable lack of studies that provide a quantitative analysis of these findings. Therefore, this study analyzed the false positives of coastal debris using the YOLOv10 and RT-DETR models to identify the detection characteristics of each model. To ensure comparable performance between the two models, hyperparameters were fine-tuned to achieve a mean Average Precision (mAP) exceeding 0.9. A dataset of approximately 350,000 coastal debris images (sourced from https://www.aihub.or.kr/) was utilized to train both models, with an 8:2 split between training and validation sets. Coastal debris was classified into 11 categories: Glass, Metal, Net, PET Bottle, Plastic Buoy, Plastic ETC, Plastic Buoy of China, Rope, Styrofoam Box, Styrofoam Buoy, and Styrofoam Piece. To analyze the detection characteristics of the trained models, images of coastal with various types of debris were collected using UAVs. False positive objects were classified and systematically analyzed based on the detection results of the collected coastal debris images using the two model. The analysis of false positives revealed that the YOLOv10 model exhibited a 72% false positive rate for Styrofoam buoys, attributed primarily to the significant impact of object color and shape. In the RT-DETR model, false positive rates were observed at 22% for seaweed and 20% for Styrofoam buoys, with object color and surface composition as key contributing factors. Based on these findings, it is recommended to consider the characteristics of the coastal and the distributed debris when selecting a deep learning model for coastal debris detection. Future studies on precise classification of coastal debris and diverse environmental data will facilitate the selection of optimal deep learning models for specific field conditions.1
Northernmost record of Valenciennea wardii (Actinopterygii, Gobiiformes, Gobiidae) from Ulleung Island, Northwest Pacific
No full textValenciennea wardii (Playfair in Playfair et G & uuml;nther, 1867) often exhibits color variations depending on its habitat. However, owing to its reclusive behavior and low abundance, it is challenging to collect V. wardii using fishing gear, and ecological and population genetic studies remain limited. On 11 November 2024, a single specimen of the tropical gobiid fish V. wardii, which was previously unrecorded in the waters around the Korean Peninsula, was collected from the coastal waters off Ulleung Island via SCUBA diving. It was analyzed using detailed morphometric and meristic measurements in conjunction with mitochondrial cytochrome c oxidase subunit I (COI) sequencing to confirm its taxonomic identity. Morphological analysis revealed distinct characteristics, particularly brown transverse body bands and a black blotch with a white margin on the first dorsal fin, which distinguish it from other closely related species. Genetic analysis showed high similarity to V. wardii specimens collected from the South China Sea and Sri Lanka but indicated notable divergence from those collected in the Philippines, suggestive of potential cryptic diversity within the species. This study reports the poleward range expansion of V. wardii and proposes the new Korean name "gal-saeg-tti-bok-gi-mang-duk" based on the characteristics of this species having dark brown transverse bands. These records provide concrete evidence for the poleward expansion of this tropical marine fish species and play a crucial role in monitoring the impacts of climate change on marine ecosystems.11Ysciescopu
Optimal Mixing Design and Field Application Protocol of Lightweight-Foamed Soils with Waste Fishing Nets
No full textLightweight-foamed soils are mixed soils with foam and cement to enhance the solidity and lightness of soils. Marine wastes, especially waste fishing nets, can be additives to reinforce the engineering properties of lightweight-foamed soils. In this paper, lightweight-foamed soils reinforced with waste fishing nets were investigated. Dredged soil and waste fishing nets were collected and pre-processed for testing. For optimization, the water content, foam ratio, cement ratio, net ratio, net conditions, and curing days were evaluated with respect to workability, unit weight, and strength. The variables were narrowed down based on the performance criteria. The results found that a water content of around 100%, cement ratio of 20%, foam ratio of 5%, and net ratio of 4% with shredded nets provide the best engineering performance of lightweight-foamed soils. The use of nets presented a superior increase in critical strength rather than an obvious increase in peak strength. A normalized factor was used to predict the required strength of lightweight-foamed soils. Finally, this study proposes field implementation methods in terms of the initial conditions of soils and optimal conditions of soils, resulting in the depletion of waste fishing nets.11Ysciescopu
Application of quality-controlled sea level height observation at the central East China Sea: Assessment of sea level rise
No full textThis study presents a state-of-the-art quality control (QC) process for the sea level height (SLH) time series observed at the Ieodo Ocean Research Station (I-ORS) in the central East China Sea, a unique in situ measurement taken in the open sea for over 2 decades at a 10 min interval. The newly developed QC procedure, named Temporally And Locally Optimized Detection (TALOD), has two notable differences in characteristics from the typical ones: (1) spatiotemporally optimized local range check based on the high-resolution tidal prediction model TPXO9 and (2) consideration of the occurrence rate of a stuck value over a specific period. In addition, the TALOD adopts an extreme event flag (EEF) system to provide SLH characteristics during extreme weather. A comparison with the typical QC process, satellite altimetry, and reanalysis products demonstrated that the TALOD method could provide reliable SLH time series with few misclassifications. A budget analysis suggested that the sea level rise at the I-ORS was primarily caused by the barystatic effect, and the trend differences between observations, satellite, and physical processes were related to vertical land motion. It was confirmed that ground subsidence of −0.89 ± 0.47 mm yr−1 is occurring at I-ORS. As a representative of the East China Sea, this qualified SLH time series makes dynamics research possible spanning from a few hours of nonlinear waves to a decadal trend, along with simultaneously observed environmental variables from the air–sea monitoring system at the research station. This TALOD QC method is designed to process SLH observations in the open ocean, but it can be generally applied to SLH data from tidal gauge stations in the coastal regions.11Ysciescopu
Classification of ISAR Images using 2D PCA and Neural Network Classifier
No full textWe propose an efficient method for classifying Inverse Synthetic Aperture Radar (ISAR) images with robustness to both translation and rotation. To achieve rotational invariance, we estimate the Relative Rotation Angle (RRA) by applying a two-Dimensional (2D) Fourier Transform (FT) followed by polar mapping of the spectral magnitude. The angle that maximizes the correlation between the polar-mapped spectra of the training and test images is selected as the RRA. The test image's 2D FT spectrum is then rotated by the estimated RRA, and a 2D inverse FT is performed to restore spatial domain data with corrected orientation. Classification is subsequently conducted using two-Dimensional Principal Component Analysis (2D-PCA) and a simple neural network classifier with two hidden layers. The proposed method demonstrates high classification accuracy under low signal-to-noise ratio conditions, even with a limited training dataset, and shows greater resilience to image defocus compared to conventional techniques.
본 논문에서는 병진 및 회전에 대한 불변성을 갖는 역합성 개구면 레이더(ISAR, Inverse Synthetic Aperture Radar) 영상 식별을 위한 효과적인 방법을 제안한다. 제안된 방법은 학습 영상과 시험 영상 간의 상대 회전각을 추정하기 위해 ISAR 영상에 대한 이차원 푸리에 변환을 수행한 후, 해당 스펙트럼을 극좌표계로 사상한다. 학습 영상과 시험 영상의 사상 영상 간 상관계수가 최대가 되는 각도를 회전각으로 설정한 후, 시험 영상의 2D 푸리에 스펙트럼을 회전각 만큼 회전시키고, 이를 이차원 역푸리에 변환하여 회전에 대한 불변성을 확보한다. 영상 식별은 이차원 주성분 분석과 2개 은닉층의 간단한 신경망 분류기를 통해 수행된다. 제안된 방법은 적은 수의 학습 데이터와 낮은 신호대잡음비 환경에서도 높은 식별 정확도를 나타내며, 기존 방법에 비해 영상의 흐림 현상에 덜 민감한 특성을 보인다.22Nkc