The increase in maritime traffic and vessel size has strengthened the need for economical and safe maritime transportation networks. Currently, ship path planning is based on past experience and shortest route usage. However, the increasing complexity of the marine environment and the development of autonomous ships require automatic shortest path generation based on maritime traffic networks. This paper proposes an efficient shortest path planning method using Dijkstra’s algorithm based on a maritime traffic network dataset created by extracting maritime traffic routes through a spatial-temporal density analysis of large-scale AIS data and Delaunay triangulation. Additionally, the depth information of all digital charts in Korea was set as a safety contour to support safe path planning. The proposed network-based shortest path planning method was compared with the path planning and sailing distance of a training ship, and compliance with maritime laws was verified. The results demonstrate the practicality and safety of the proposed method, which can enable the establishment of a safe and efficient maritime transportation network along with the development of autonomous ships.

The majority of ocean pollution stems from terrestrial sources, with more than 80% attributed to chemicals, industrial waste, toxic metals, plastics, sewage, and other land-based materials. Therefore, the management of ocean pollution must initiate from terrestrial interventions, necessitating the control of pollutants discharged from land sources and the crucial pursuit of identifying the roots of ocean pollution. Identifying the causes of ocean pollution enables the exploration of mitigation strategies for the entities responsible for pollution generation and facilitates the formulation of relevant policies. Furthermore, it can contribute to imposing costs for purification and raising awareness about the seriousness of ocean pollution. However, Rivers and streams that lead to the ocean are influenced by various changing factors as they pass through the land, making it a globally challenging task to pinpoint the sources of pollution. Accordingly, in this study, we collected water quality observation data measured at five points in a specific estuary area and water quality data from five areas near the source of pollution, and conducted a study to calculate the contribution of pollution from the source. We performed statistical analysis and machine learning-based pollution source analysis, and developed an improved artificial intelligence model proposed in this study that complements the limitations of existing analysis methods. The variables used in the analysis were POC average concentration, POC δ13C, PN average concentration, PN δ15N, average concentration, and DOC δ13C. For basic data analysis, data distribution analysis, similarity/discrimination analysis, and clustering analysis of variables by branch were performed. Basic data analysis allowed for dividing the data's characteristics into four groups, but discrepancies in similarities emerged among items based on each analysis method, limiting the meaningfulness of the data analysis. For this, we analyzed which pollutants contaminated the five points in the river estuary using machine learning techniques such as XGBoost and a deep learning neural network, an artificial neural network model. The XGBoost analysis categorized pollution sources for each location into 1 to 2 categories, showing accuracies ranging from 51.04% to 99.92%. However, due to the intrinsic nature of machine learning, predicted values tend to maximize similarity to the most similar pollutant source, resulting in extreme values exceeding 99%. The artificial neural network analysis resulted in the classification of 2 or more pollution sources for each location, with accuracies ranging from 33.23% to 62.45%. This is considered a result of relatively lower accuracy due to the unique characteristics of each location. To overcome the limitations of each model, this study created an integrated model that aggregates results from multiple models to determine the similarity. The analysis using the integrated model effectively identified pollution sources excellently without encountering extreme accuracy issues. To ensure the reliability of future pollution contribution assessment models, determining pollution contribution through isotopic fraction analysis will be necessary.

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.

Recently, interest in maritime accidents and safety-related research, such as preventing collisions between marine vessels, detecting illegal vessels, and predicting vessel routes, is increasing. Vessel location databased vessel distribution map can support decision-making for maritime safety management, and if the vessel distribution can be predicted, it is possible to take a preemptive response for maritime security such as fishing safety management and illegal fishing prevention. In this study, a training dataset for vessel distribution prediction was constructed by collecting V-Pass data, weather warnings, and marine environment data. The result of resampling of reporting interval of vessel location data was mapped to grid data to evaluate the vessel density, and a total of 1,314,000 of training data were constructed for the study area. In the future, research to evaluate the accuracy by performing vessel distribution prediction modeling should be conducted.