Objectives. We propose a modern method for semantic segmentation of ultra-high-resolution (4K) video frames in the field of Earth remote sensing using a modification of the DeepLabV3+ convolutional neural network.

Methods. The method is aimed at solving two critical problems: the limited receptive field of the model when working with local image fragments due to GPU memory constraints, and the poor separation of classes with similar color characteristics and texture. The baseline architecture was supplemented with two attention mechanism modules to improve object localization, and the standard cross-entropy loss function was supplemented with a Supervised Contrastive Learning (SupCon) algorithm for better separation of similar classes. Feature-wise Linear Modulation (FiLM) was embedded into the Atrous Spatial Pyramid Pooling (ASPP) module to introduce global context in the form of features extracted from the original image.

Results. The results of the experiments showed that the proposed approach successfully minimizes artifacts at class boundaries, has a high ability to distinguish visually similar classes, and improves segmentation accuracy under the conditions of a specific dataset. The proposed architecture outperforms the baseline by 0,45 % in terms of mean Intersection over Union.

Conclusion. The developed modification of DeepLabV3+ effectively addresses the challenges of semantic segmentation of ultra-high-resolution images in the field of Earth remote sensing, achieving high accuracy with a negligible increase in computational complexity.

Objectives. When replacing diesel buses with electric buses, the problem of selecting the composition of the electric bus fleet, assigning service trips from a given set to electric buses, and determining charging infrastructure and charging schedules becomes relevant. The aim of the study is to create models and methods for obtaining cost-effective solutions to this problem for heterogeneous electric buses and charging stations under different passenger flow volumes for different trips.

Methods. Methods of set theory, graph theory, operations research, and mathematical programming are used.

Results. A mathematical model has been developed for the problem of optimizing a fleet of various-type electric buses, their routing, assignment to depots at the beginning and end of the planning interval, determination of the locations and number of charging stations of different types at route terminals, and charging schedules. The objective function is the total daily capital and operating costs of electric buses, charging stations, and consumed electricity. A two-level decomposition scheme for solving the problem is proposed, at the upper level of which daily assignments for electric buses are determined (assigning different types of electric buses to service trip sequences with varying passenger volumes). The lower level distributes electric buses among depots, locates charging stations at route terminals, and determines charging schedules for electric buses with fixed trip assignments. A heuristic local search algorithm is proposed for solving the upper-level subproblem, and a mixed integer linear programming model is developed for the lower-level subproblem.

Conclusion. The solution to the upper-level subproblem consists of assigning different types of electric buses to service trips with varying passenger volumes. Standard solvers, such as Gurobi Optimizer, can be used to solve the formulated lower-level subproblem.

Objectives. The objectives of the work are to demonstrate the possibility of applying computer technology for modeling residual welding phenomena and compare them with the results of analytical calculations for classic welded structures. A software package project for automating the calculation of residual welding phenomena is presented.

Methods. The principles of applying the force-based method for calculating the stress-strain state (SSS) of general-purpose welded structures are demonstrated using the example of beam and shell welded structures. A description of existing approaches for calculating welding phenomena is provided, as well as a software package implementing the force-based method for calculating the SSS of welded structures. Prospects for transitioning to parallel architectures and further development and application of the force-based method are analyzed.

Results. Good convergence of the results of calculating deformations of welded structures in comparison with analytical methods and the possibility of automating the creation of computer models for calculating residual welding phenomena are shown, using the example of a software complex project. Conclusion. The main directions of development and advantages of the technology are highlighted: reduction of labor intensity, CAD integration, optimization of assembly processes, and scalability. Prospects for transitioning to parallel architectures are discussed.

Objectives. The aim of the work is the development and experimental verification of a hybrid postquantum image encryption algorithm based on a 3D chaotic system and the Rubik's cube principle. Methods. Issues of the vulnerability of traditional encryption methods to attacks using quantum computers are considered. To create a cryptographically robust algorithm, it is proposed to combine the advantages of chaotic systems and quantum computing. A new 3D chaotic system based on a modified Lorenz system is used for generating pseudorandom keys. The Rubik's cube principle is applied to perform image scrambling operations (cyclic shifting of rows and columns). The NEQR (Novel Enhanced Quantum Representation) model is used for quantum representation and data processing. The algorithm is implemented in Python using the Quiskit, OpenCV, and NumPy libraries, as well as the AerSimulator quantum computing simulator. Results. A hybrid encryption algorithm has been developed and implemented as a computer program. The algorithm was tested on images of various sizes and formats. Resistance to statistical and differential attacks has been experimentally confirmed: the correlation coefficient of adjacent pixels in the encrypted images is close to zero, the information entropy is close to the ideal value of 8, and the NPCR and PSNR values exceed 99,25 % and are less than 13 dB, respectively. Conclusion. The proposed hybrid algorithm demonstrates high cryptographic strength and efficiency. Key sensitivity, the use of quantum principles, and performance optimization through multithreading make it practical for use in security-critical areas such as IoT, telemedicine, and cloud storage. The software implementation operates locally without requiring an internet connection, which ensures data security.

Objectives. The purpose of the work is to develop and programmatically implement a three-vector algorithm for detecting Russian-language neural network text fragments based on probabilistic analysis. The object of the study is Russian-language texts, the subject is statistical signs of their origin. Special attention is paid to the quantitative comparison of the proposed approach with the basic single-vector detection methods the perplexy detector and the GLTR rank analysis method.

Methods. The proposed algorithm combines three feature vectors: token predictability, the proportion of rare lexemes, and the information density of the text, estimated via the algorithmic compression ratio. The author's regression language model rugpt3small is used as a probabilistic core. The implementation is made in Python using the transformers, PyTorch, NLTK, and zlib libraries. The experimental validation was conducted on a sample of 100 mixed documents (16,000 sentences) of five thematic groups, the neural network fragments of which were generated by four language models DeepSeek, GPT, Qwen and YandexAI. The quality was assessed by the deviation from the reference proportion of the neural network text and by classification metrics.

Results. After applying the calibration adjustment, the average index for the sample was 45.2 points against a reference level of 50, and the values for all four models were in the narrow range of 44.445.9 points, which indicates the algorithm's stability to the choice of a generative model. The deviation of the average index from the standard was 4.8 points versus 11.1 points for the method based on perplexity and 7.6 points for GLTR; the F1 measure at the sentence level reached 0.76 versus 0.63 and 0.66, respectively.

Conclusion. The three-vector algorithm has proven its effectiveness in the task of detecting neural network fragments and can serve as the basis for tools for verifying academic texts and anti-plagiarism systems. Further development of the research is related to the expansion of the sample, testing the resistance to paraphrasing and editing, as well as conducting ablative analysis

Objectives. The aim of this study is to develop an approach for predicting the timing of several key stages in multi-stage processes.

Methods. A method is proposed based on processing event sequences taking into account stages, time intervals, and information about process participants, as well as a consistent consideration of the relationships between the predicted stages.

Results. Experiments conducted on editorial document flow data demonstrated improved accuracy compared to recurrent neural networks and a significant advantage in the early stages compared to gradient boosting models.

Conclusion. The method allows obtaining of consistent interval forecasts for several stages of a process and can be used to quickly update the estimates without the use of specialized equipment.