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, thereby enhancing 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-U 2 Net demonstrated the highest accuracy, achieving an Intersection over Union (IoU) of 75.89% and a Dice Score of 86.30% on base SDVD, 71.57% IoU and 83.43% Dice Score on CleanedSDVD, and 68.55% IoU and 81.33% Dice Score on ExtendedSDVD. Additionally, we utilized a channel-level Local Interpretable Model-Agnostic Explanation (LIME) method for detailed evaluation underlining the efficiency of the proposal. These results highlight the effectiveness of the proposed approach in training segmentation models and indicate significant potential for further enhancements in sonar image recognition methods for victim recovery applications.

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 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.