Coastal debris is a global environmental issue that requires systematic monitoring strategies based on reliable statistical data. Recent advances in remote sensing and deep learning-based object detection have enabled the development of efficient coastal debris monitoring systems. In this study, two state-of-the-art object detection models—RT-DETR and YOLOv10—were applied to UAV-acquired images for coastal debris detection. Their false positive characteristics were analyzed to provide guidance on model selection under different coastal environmental conditions. Quantitative evaluation using mean average precision (mAP@0.5) showed comparable performance between the two models (RT-DETR: 0.945, YOLOv10: 0.957). However, bounding box label accuracy revealed a significant gap, with RT-DETR achieving 80.18% and YOLOv10 only 53.74%. Class-specific analysis indicated that both models failed to detect Metal and Glass and showed low accuracy for fragmented debris, while buoy-type objects with high structural integrity (Styrofoam Buoy, Plastic Buoy) were consistently identified. Error analysis further revealed that RT-DETR tended to overgeneralize by misclassifying untrained objects as similar classes, whereas YOLOv10 exhibited pronounced intra-class confusion in fragment-type objects. These findings demonstrate that mAP alone is insufficient to evaluate model performance in real-world coastal monitoring. Instead, model assessment should account for training data balance, coastal environmental characteristics, and UAV imaging conditions. Future studies should incorporate diverse coastal environments and apply dataset augmentation to establish statistically robust and standardized monitoring protocols for coastal debris.

The increasing intensity of typhoons associated with climate change has significantly elevated the risk of anchor dragging of vessels sheltering in coastal anchorages, leading to collisions, groundings, and secondary maritime accidents. Despite its operational importance, anchor dragging under extreme weather conditions remains poorly understood and is rarely addressed as a predictive problem. This study proposes a data-driven framework to detect and predict anchor dragging of anchored vessels during typhoon events by integrating vessel motion data with meteorological and oceanographic forcing. Automatic Identification System (AIS) data were combined with typhoon track and intensity information, high-resolution marine weather fields, and bathymetric data for Typhoon Kong-Rey (2018), which directly affected Jinhae Bay, one of the largest typhoon shelter areas in Korea. Anchored vessels were identified using speed-based criteria, and vessel-specific anchor circles were constructed by estimating anchor positions from AIS heading information, anchor chain length, and vessel dimensions. Anchor dragging events were labeled based on deviations from the anchor circle, supported by visual verification. To predict dragging occurrence, a genetic algorithm-based automated machine learning framework (TPOT) was applied to optimize preprocessing steps, feature selection, model structure, and hyperparameters. The explanatory variables included vessel kinematics, wind speed and direction, atmospheric pressure, and local water depth. The resulting model successfully distinguished high-risk vessels during peak typhoon influence, demonstrating strong predictive performance and robustness across vessel types. The proposed approach provides a probabilistic early-warning capability for anchor dragging, enabling prioritized monitoring of high-risk vessels rather than uniform risk management. This framework supports proactive decision-making for Vessel Traffic Services (VTS), port authorities, and emergency response agencies, and contributes to reducing cascading maritime accidents under intensifying extreme weather conditions.

Bo-Ram Kim, Ye-Been Do, Jun-Hwa Chi, Tae-Hoon Kim. Korean J. Remote Sens. 2025;41:327-39. https://doi.org/10.7780/kjrs.2025.41.2.1.7

Recent studies have increasingly utilized drones for remote sensing, driven by the widespread distribution of marine debris along coastal areas. When monitoring coastal debris using drones, flight altitude is a critical factor that directly impacts both the quality of image data and the monitoring duration. However, designing monitoring systems based solely on altitude may lead to variations in spatial resolution (GSD) caused by differences in camera specifications across various drone models. Such variations in GSD levels impact the interpretability of debris within the imagery. This study evaluates the interpretability of coastal debris at different GSD levels determined by drone specifications and flight altitudes. Based on prior studies, we collected data at four altitudes by GSD: 18.6 m (GSD: 0.5 cm/pixel), 27.9 m (GSD: 0.75 cm/pixel), 37.2 m (GSD: 1.0 cm/pixel), and 46.5 m (GSD: 1.25 cm/pixel). Coastal debris types were categorized into eight classes, defined based on the top 10 most frequently identified debris types over a four-year period in Korea. We also assessed the quality and interpretability of coastal debris data under varying spatial resolutions of drone imagery, with a particular focus on the eight defined categories. Interpretability was assessed based on the National Image Interpretability Rating Scales (NIIRS), developed by Image Intelligence, which defines four interpretability levels: I (Identify), B (Distinguish), D (Detect), and N (Not Detect). The results demonstrated that the interpretability of coastal debris varies depending on debris type, color, and size with changes in GSD. Furthermore, the detectable categories of debris were defined for each GSD level. Through this study, it is expected to support decisions on appropriate GSD settings and monitoring methods for different coastal debris survey objectives and conditions. The findings may also help in developing national policies for managing coastal debris.

The contamination of sediments with hazardous substances in the coastal waters of South Korea adversely affects the health of marine ecosystems, causing water quality deterioration, hypoxia, red tides, foul odors, and ecological toxicity. This necessitates rapid identification of pollution sources and their resolution. However, the diversity of pollutant types and sources across regions poses challenges to pollution mitigation and management. To address this, there is a need for technology capable of quantitatively evaluating the contributions of complex pollution sources, as well as a system that supports user-friendly data querying, filtering, and visualization. Such a system should also facilitate information and data sharing among users, fostering collaboration within organizations and with external partners.In response, we developed an automated program for pollution source attribution, encompassing the entire process from data analysis algorithms to result visualization. The program integrates multiple source attribution methods, including the Chemical Mass Balance (CMB) model, Non-Negative Matrix Factorization (NMF) model, and Bayesian Isotope Mixture (BIM) model, to enhance the identification of pollution origins. To ensure accessibility and ease of use, the program was implemented using R Shiny, a web-based platform built on the R programming language.The automated program accepts csv files as input to estimate pollution contributions and provides visualization of modeling results through graphs, matrices, and geospatial point contributions on maps. This approach enhances researchers' understanding and facilitates efficient utilization of the results.

Deep 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.