Natalia Wawrzyniak

Time and accuracy are key elements of underwater search and rescue operations. This article proposes a system with a modified U-net architecture, leveraging spatial and channel attention techniques for binary segmentation of drowned victims from side-scan sonar imagery. The base architecture is enhanced by a novel spatial skip-connection attention (SSCA) and multihead attention (MHA). The SSCA improves feature representation by highlighting relevant spatial information in the shallow layers. At the same time, the MHA module captures abstract channel-wise relations, improving the information flow and feature representation deeper in the network. In our experiments, we used original samples collected from the Oder River in Szczecin, using a drowned dummy and an Edgetech 4125 for sonar imaging, to produce a novel sonar drowned victims data set. The results obtained on the test set (99.54% accuracy and 75.76% mean dice coefficient) show that the model can adapt to the task, achieving significant results.

Sonar imaging provides representative real-time mapping of the water bottom, which is particularly valuable in search-and-recovery operations, as it helps quickly locate drowned victims. However, practical analysis of sonar data is often challenging due to its complexity and automated recognition demanding substantial training data. In this paper, we present an extension to SDVD (Sonar Drowned Victims Dataset), a sonar image dataset with annotated binary masks marking drowned victims for training segmentation models. To improve segmentation processing, we propose Multi-Channel Input Augmentation (MCIA). This method augments the segmented image with additional channels, increasing it from one to seven, increasing the model’s information input. To achieve this, each channel represents a processed original image by applying: Sobel edge detection, Haar wavelet, High-Pass features, pixelated map, CLAHE and distance transform. To evaluate segmentation performance, we tested different models on all SDVD subsets, alongside various data augmentation techniques, such as RICAP. Among the models, MCIA-U2Net demonstrated the highest accuracy, achieving an Intersection over Union (IoU) of 75.89% and a Dice Score of 86.30% on base SDVD, 70.12% IoU and 82.44% Dice Score on CleanedSDVD, 68.55% IoU and 81.33% Dice

The paper presents collaborative research conducted by European researchers, mariners, and cybersecurity experts on identifying and developing preventive procedures for maritime cybersecurity within the CyberSEA project. The project's objective is to equip seafarers with the necessary knowledge to detect and respond effectively to cyber threats, minimizing the risk of disruptions to maritime operations. The research included several key tasks aimed at developing the proposed cybersecurity protocols. The initial phase involved identifying cyber vulnerabilities in the maritime sector and defining the main risk areas. Subsequently, 34 real-life scenarios were constructed to illustrate these threats. Comprehensive desk research was conducted on best cybersecurity practices in other industries that were transferable to the maritime sector to establish a set of preventive measures. The final phase resulted in the development of 10 detailed protocols for seafarers, specifying the procedural steps to be taken in response to distinct cyber threats. The study highlights the essential role of human factors and education in maritime cyber resilience, offering a transferable framework for integration into Safety Management Systems and vocational training curricula.

Epidemic outbreaks on sea-going vessels pose a significant health and safety risk, particularly in isolated maritime environments lacking professional medical staff. The COVID-19 pandemic exposed major deficiencies in the preparedness and response protocols of maritime operations, where vessels became de facto quarantine units without access to structured medical decision support. This study presents the development and evaluation of DESSEV, a decision support system designed to assist maritime personnel in identifying and managing infectious disease outbreaks onboard vessels. A comprehensive non-SQL knowledge base was constructed using international epidemiological guidelines (WHO, CDC, ECDC) and medical literature, encompassing 22 infectious diseases and 35 symptoms categorized into 8 clinical domains. To support probabilistic diagnostics, artificial patient profiles were generated reflecting real-world symptom distributions. Three predictive models—Decision Tree, Naive Bayes, and Random Forest were trained. Their performances were assessed through cross-validation and random sampling techniques. Evaluation on a holdout test set of real patient cases showed that the Random Forest model achieved superior performance across all The Random Forest algorithm was thus selected as the core prediction engine for the DESSEV application. While not a substitute for professional medical care, DESSEV improves situational awareness and supports early risk mitigation actions aboard vessels. Future work will focus on integrating real-time health telemetry and user feedback to further refine diagnostic accuracy and usability.

Defects in the source code that affect security are one of the main elements used to carry out cyber attacks. Examining source code for vulnerabilities is a difficult and expensive process. As a result, specialized software is needed for this. Due to the development of various artificial intelligence methods, improving existing vulnerability detection methods is possible. In particular, it is possible to reduce the number of false positives and enable the detection of complex vulnerabilities that require understanding the broader context of the code. The article presents the concept of a system for automatic analysis of vulnerabilities in source code, along with the challenges and problems related to its design and use.

Side-scan sonar (SSS) images are difficult for automatic analysis due to the acoustic measurement parameters as well as the number of different objects that can be distant. In addition, there is a risk that the seabed analysis application may be attacked. For this purpose, we propose a solution based on convolutional neural networks with bilinear pooling in order to achieve higher values of classification accuracy. Bilinear pooling merge data from two networks and return classification results. The first network’s branch receives the original image and the second one after applying the superpixel method. This approach allows to focus on different types of features. In addition, we introduced a mechanism of poisoning detection that analyze images and results from the network. For the evaluation process, we used the real SSS images obtained between two water channels in Szczecin city in north-western Poland. The importance of scientific research indicates the accuracy of the analysis as well as the safety of the measurements performed.

The identification of ships plays a crucial role in security and managing vessel traffic for ports and onshore facilities. Existing video monitoring systems help visually identify a vessel where other systems are not present or sufficient. Readable vessel plates and hull inscriptions of detected ships in the video stream allow using text location and recognition methods to obtain ships’ identification names or numbers. The obtained information can be then matched with available ship registers. The automation of the process has met many challenges related to the often-low quality of available video streams, heterogeneous regulations on the marking of ships, and the specifics of natural scene text recognition, such as quickly alternating imaging conditions or the interference of the background. The main contribution of this research is a method that can identify any type of vessel in an image that has visible inscriptions (name, registration number) placed on the hull and must be registered in a public registry. The proposed method works with low-quality images with inscriptions placed under different angles and different, readable sizes. Our method recognised 91% of vessels from our test dataset. Obtained identification times have not exceeded 1s. The quality and efficiency of the proposed solution indicate that it is suitable for practical implementation in onshore monitoring systems.

Side-scan sonar generates an image based on the reflection of sound waves. The reflection takes place with the first encountered object, which, in practice, may be a passing fish or even a wave from the boat turbine. Therefore, such data are sensitive to all noises. In this article, we propose a real-time automatic system of side-scan sonar analysis that can detect and classify different objects. Our proposition is based on the processing of a given image to quickly verify whether there is anything other than the bottom of the river/sea. If so, this image is analyzed in terms of regions of interest (ROIs) by the histogram module. This action allows the extraction of only objects of interest, which are then classified by convolutional neural networks (CNNs). A proposed model also contains an automatic mechanism of adding the sample to the database in order to later guarantee the accuracy of the classifiers. The model is a hybridization of image processing techniques with a machine learning approach to analyze difficult images. The presented system has been tested on the bottom of a river in Szczecin, Poland. We reached 90% of accurate classification in case scenario and 88% in simulation on used datasets. The obtained results were presented and discussed in terms of the advantages of practical application in analyzing side-scan sonar images.