Browsing by Author "Sineglazov, Victor"

Now showing 1 - 20 of 31

A Comprehensive Framework for Underwater Object Detection Based on Improved YOLOv8
(National Aviation University, 2024-03-29) Sineglazov, Victor; Синєглазов, Віктор Михайлович; Savchenko, Mykhailo; Савченко, Михайло Володимирович
Underwater object detection poses unique challenges due to issues such as poor visibility, small densely packed objects, and target occlusion. In this paper, we propose a comprehensive framework for underwater object detection based on improved YOLOv8, addressing these challenges and achieving superior performance. Our framework integrates several key enhancements including Contrast Limited Adaptive Histogram Equalization for image preprocessing, a lightweight GhostNetV2 backbone, Coordinate Attention mechanism, and Deformable ConvNets v4 for improved feature representation. Through experimentation on the UTDAC2020 dataset, our model achieves 82.35% precision, 80.98 % recall, and 86.21 % mean average precision at IoU = 0.5. Notably, our framework outperforms the YOLOv8s model by a significant margin, while also being 15.1% smaller in terms of computational complexity. These results underscore the efficiency of our proposed framework for underwater object detection tasks, demonstrating its potential for real-world applications in underwater environments.
About Algorithms of Target Positioning
(National Aviation University, 2022-06-24) Hley, Oleg; Глей, Олег Сергійович; Sineglazov, Victor; Синєглазов, Віктор Михайлович
Software has been developed for the system most suitable for wind shear detection. The formulation of the target designation problem is presented. The features of modern radars are considered. The aim of the work is to find the most suitable system for detecting wind shear and to develop software for this system that offers the pilot a turnkey solution for flying through the danger zone of wind shear. The paper presents the definitions of wind shear, statistics of air crashes caused by this phenomenon. The systems for determining wind shear are described and the most suitable system is selected. The principle of operation and the algorithm of operation of the Doppler radar station are given. A simulation of the detection of a dangerous wind shear zone was carried out. The pilot uses the proposed software solution. Thus, the task of determining the wind shear zone and the decisions proposed on this basis to fly around the dangerous zone have been completed. The results show that introducing random noise into labeled samples degrades the accuracy of the model, while introducing random noise into unlabeled data can, on the contrary, increase the accuracy of the model.
Adaptive Control of Manipulator Robots in a Dynamic Environment Using Neural Networks
(National Aviation University, 2024-06-26) Sineglazov, Victor; Синєглазов, Віктор Михайлович; Khotsyanovsky, Volodymyr; Хоцянівський, Володимир Петрович
The purpose of the study is to develop an approach to planning the trajectory of the manipulator robot using an intelligent system based on neural networks. For this purpose, the work considered the processes of planning and deploying the movement of the robot. The analysis of existing methods of planning the movement of manipulator robots and the review of intelligent control systems made it possible to obtain a complete picture of the current state of this issue. A system is proposed that can perceive the environment and control the movement of the robot by generating the correct control commands. For this, 3 tasks were solved, namely: analysis of the environment in order to determine its features, determination of the trajectory in order to neutralize the collision and determination of controlled influences for the executive authorities in order to implement the movement. The functionality and structure of the neural network for solving each of the tasks are proposed.
Algorithms for the formation of recommendations in the information system
(National Aviation University, 2021-10-21) Sineglazov, Victor; Синєглазов, Віктор Михайлович; Oliinuk, Yuriy; Олійник, Юрій Іванович
The article deals with the problem of scalability and dimension reduction of data in the algorithms of recommendations. It is proposed to improve the item-to-item algorithm by excluding from the user-item matrix elements that that do not have enough estimates. Thus more denser data are used that allows to receive more exact results. Also due to the fact that the dimension of the user-item matrix decreases, the execution time of the algorithm decreases. To solve the problem, the Tachimoto coefficient, the cosine measure, the Pearson correlation coefficient and the Euclidean distance are used to calculate the degree of similarity of the elements. The efficiency of the usual item-to-item algorithm and the algorithm were compared using only the active values in the user-item matrix. The obtained results confirm the efficiency of the item-to-item algorithm based on a dense matrix. The obtained results can be used to optimize the operation of any recommendation system.
Automatic system for UAV identification in multi-position sound monitoring systems
(2020-12-12) Sineglazov, Victor; Синєглазов, Віктор Михайлович; Pantyeyev, Roman; Пантєєв, Роман Леонідович; Balov, Oleksandr; Балов, Олександр Володимирович
The paper deals with methods and algorithms for increasing the accuracy of unmanned aerial vehicle identification. Accumulation of data on radiation parameters during the session of observation of an object allows to make continuous specification of the made decisions by the set moment of the end of a session of observation the reliability of its identification would be sufficient for acceptance of necessary prompt actions. If to consider a problem of identification taking into account the specified specifics, then in its structure it is necessary to allocate several stages of decision-making depending on volume available for the considered moment of observation of data, conditions of their receiving and processing by the technical means of the information-measuring system. Observation system and control of objects of a sound emission represent difficult sound engineering complexes in which work not only equipment rooms and software, the knowledge base and data, but also intellectual resources of experts are involved. Decisions of different levels in such systems are made not only automatically technical means on the basis of formal algorithms, but also operators, on the basis of the and borrowed experience. Taking into account it is necessary to consider, both bases of creation of formal algorithms of decision-making, and feature of decision-making by the expert for the purpose of clarification of conditions of ensuring the maximum reliability of identification.
Automation disign of hybrid vertical-axial rotors
(National Aviation University, 2022-01-05) Sineglazov, Victor; Синєглазов, Віктор Михайлович; Stanislavchuk, Oleksandr; Станіславчук, Олександр Валентинович
The paper proves the need to create vertical-axial rotors of wind power plants in the city strip, which can be placed on roofs, which makes it possible to increase their energy efficiency by 60-70%. It is shown that the placement of such rotors on the roofs has its own characteristics, which is the need to take into account the relief of the roof of the house, its surface area, rose and wind speed over it and others. Examples of wind farms are considered and it is proved that their energy efficiency can be increased by using hybrid vertical-axial rotors, which consist of a combination of Darrieus and Savonius rotors, where the Darrieus rotor is the main source (s) of wind energy conversion. in electric, while the rotor (s) of Savonius provide acceleration of Darrieus rotors. In order to improve the quality of design, an automated design system was developed, which includes the following blocks: determining the forces affecting the rotor, choosing the type of main and accelerating rotors, determining the optimal number of blades, optimal rotor placement, calculation of dynamic rotor characteristics, analysis of probable wind speed characteristics and strength calculation.
Camera Image Processing on ESP32 Microcontroller with Help of Convolutional Neural Network
(National Aviation University, 2022-09-26) Sineglazov, Victor; Синєглазов, Віктор Михайлович; Khotsyanovsky, Volodymyr; Хоцянівський, Володимир Петрович
This paper analyzes a common ESP32 microcontroller with a built-in camera for image classification tasks using a convolutional neural network. ESP32 is commonly used in IoT devices to read data and control sensors, so its computing power is not significant, which has a positive effect on the cost of the device. The prevalence of ultra-low power embedded devices such as ESP32 will allow the widespread use of artificial intelligence built-in IoT devices. The duration of photographing and photo processing is obtained in the paper, as this can be a bottleneck of the microcontroller, especially together with machine learning algorithms. Deployed convolutional neural network, pre-trained on another device, MobileNet architecture on microcontroller and proved that ESP32 capacity is sufficient for simultaneous operation of both the camera and convolutional neural network.
Computer-aided design of hybrid power systems of Ukraine
(National Aviation University, 2019-10-25) Sineglazov, Victor; Синєглазов, Віктор Михайлович
It is shown that the problem of hybrid power systems creation is very important for Ukraine. The high quality of this problem solution can be achieved with help of computer-aided design systems. It is considered the modern. computer-aided design systems and executed their analysis. It is shown that they have some disadvantages. To overcome them it is developed a new hybrid genetic algorithm of unconditional multicriteria optimization, which is based on a new approach of equivalent transformation of conditional optimization problem into unconditional with further “curering” the Pareto optimal points which are located outside of area formed by restrictions. It is considered the examples of computer-aided design of hybrid power systems for different regions of Ukraine.
Determination of Marketing Parameters for Building a Demand Forecasting Model using Neural Networks
(National Aviation University, 2023-12-27) Sineglazov, Victor; Синєглазов, Віктор Михайлович; Novikov, Mikhaylo; Новіков, Mихайло Сергійович
This article is devoted to finding marketing parameters for building a demand forecasting model using neural networks using real data. The work deals with the problem of modeling product demand on the market in marketing using artificial intelligence and machine learning methods. The main features of existing approaches to building models of products on the market, their advantages and disadvantages are shown. The need for their improvement has been identified. A new methodology for solving the problem is presented. The model's demonstrated ability to predict consumer demand based on a variety of marketing parameters helps businesses plan inventory, production, and personnel more effectively and can lead to significant cost savings and improved efficiency.
Generation of UAV-based Training Dataset using Semi-supervised Learning
(National Aviation University, 2022-09-26) Sineglazov, Victor; Синєглазов, Віктор Михайлович; Kalmykov, Vadym; Калмиков, Вадим Віталійович
The paper considers the problem of constructing a training sample based on the use of semi-supervised learning a teacher. The problem statement related to the problem posed is substantiated. It is shown that obtaining a training sample in some cases is a difficult task that requires significant computational and financial costs. The use of semi-supervised learning made it possible to label unlabeled data and thus ensure the creation of a labeled sample of sufficient size. The paper gives examples of generating a training sample, as well as its use for training neural networks, which are used to solve the problem of multiclass classification. Using this approach, you can get a robust data set consisting of a small amount of manually labeled images and a huge amount of pseudo-labeled or augmented data. Using this approach, one can train a classifier to detect and classify any objects in images with bounding boxes and label them accordingly.
Hand Gestures Recognition and Tracking Within Virtual Reality using Hybrid Convolutional Neural Networks
(National Aviation University, 2022-09-26) Sineglazov, Victor; Синєглазов, Віктор Михайлович; Boryndo, Illia; Бориндо, Ілля Олександрович
In this paper analysis of modern virtual reality algorithms based on mobile devices was done. As a result, algorithmic shortcomings were identified and the usage of convolutional neural networks was proposed. Within the research the qualitative analysis of modern architectures of convolutional neural networks was carried out and their separate shortcomings at use in systems on the basis of processor architecture аdvanced RISC machine was shown. As a result of this research it was found that to achieve the target accuracy and speed of the system it is important to use a hybrid convolutional neural network, which significantly improves the quality criteria of the system. The optimal structure and parameters for initialization and training of a hybrid convolutional neural network system used for virtual reality are obtained. The optimal training sample was formed and the use of pre-trained hybrid convolutional neural network on another device of аdvanced RISC machine architecture was described.
Hybrid energy storage system design
(National Aviation University, 2020-07-06) Sineglazov, Victor; Yudenko, Igor; Синєглазов, Віктор Михайлович; Юденко, Ігор Олександрович
The necessity of hybrid energy storage system design is shown. It is considered the main components of hybrid energy storage systems and their features. It is shown that these systems are characterized by a beneficial coupling of two or more energy storage technologies with supplementary operating characteristics. The hybrid energy storage systems may include super capacitors, flywheels, superconducting magnetic energy storage, lithium-ion batteries, fuel cells. It is analyzed the use of every possible component as a part of the hybrid energy storage system in different operating conditions. It is considered a sizing methodology optimization for hybrid energy storage systems based on blast algorithm and hybrid neural networks.
Hybrid neural network optimization system based on ant algorithms
(National Aviation University, 2020-07-06) Sineglazov, Victor; Chumachenko, Olena; Omelchenko, Dmytro; Синєглазов, Віктор Михайлович; Чумаченко, Олена Іллівна; Омельченко, Дмитро Валерійович
The ant multi-criteria algorithm for feed forward neural networks training is proposed. It is used two criteria: the error of generalization and complexity. It is represented a review of neural network learning using swarm algorithms. As a result of training it is determined a structure of neural network (a number of layers and neurons in then) and the values of weight coefficients and biases. Modification of well-known algorithms consists in using the concept of Pareto optimality. It is done the research of proposed algorithm on the example of multilayer perceptron for the approximation problem solution.
Image processing from unmanned aerial vehicle using modified YOLO detector
(National Aviation University, 2021-12-21) Sineglazov, Victor; Синєглазов, Віктор Михайлович; Kalmykov, Vadym; Калмиков, Вадим Віталійович
Identifying objects from drone images is a state-of-the-art task for artificial neural networks. Since drones are always moving at different altitudes, the scale of the object varies greatly, making it difficult to optimize the networks. Moreover, flying at high speeds and low altitudes leads to blurred images of densely populated objects during movement, which is a problem when solving the problem of recognizing and classifying small sized objects. This paper addresses the above problem solutions and solves them by applying an additional prediction model to identify objects of different scales. We also modify the loss function to penalize larger objects more and vice versa to encourage recognition of smaller objects. To achieve improvements, we use advanced techniques such as multiscale testing, image blurring, object rotation, and data distortion. Experiments with a large data set show that our model has good performance in drone images. Compared to the baseline model (YOLOv5), our model shows significant improvements in object recognition and classification.
Integrated Computer-Aided Design System Software of Navigation Complex
(Osvita Ukrainy, Kyiv, 2016-10) Sineglazov, Victor; Godny, Andrew
Presented scheduling mechanisms for computer-aided design system for unmanned aerial vehicles with an integrated environment introduces a new approach to managing hardware resources and design time management. Usage of this mechanism allows to improve scheduling efficiency of multi-tasking applications with shared resources by allowing simultaneous operations with multiple shared data task-readers.
Intelegence Design of Hybrid Vertical-axis Rotors
(National Aviation University, 2023-12-27) Sineglazov, Victor; Синєглазов, Віктор Михайлович; Stanislavchuk, Oleksandr; Станіславчук, Олександр Валентинович
The work is devoted to the necessity of creating the vertical-axis rotors of wind power stations in the urban area, which can be placed on roofs and makes it possible to increase their energy productivity by 60-70%. It is shown that the locations of such rotors on roofs has its own characteristics, which consists in the need to take into account the shape of the topography of the house, its storey, the direction and speed of the winds above it, and others. Examples of implementation of wind farms are considered and it is proven that their energy efficiency can be increased due to the use of hybrid vertical-axis rotors, which consist of a combination of Darrieus and Savonius rotors, where the Darrieus rotor is the main source(s) of wind energy conversion into the electric one, while the Savonius rotor(s) provide the acceleration of the Darrieus rotors. For the implementation it has been used the genetic algorithm. An inelegance design system has been developed. An example of the application of this system for the design of a hybrid rotor is given.
Intelegence diagnostic system of liver fibrosis stages
(National Aviation University, 2020-07-06) Sineglazov, Victor; Синєглазов, Віктор Михайлович; Shevchenko, Maksim; Шевченко, Максим Валерійович
The necessity of constructing an intelligent system for diagnosing stages of liver fibrosis is determined, for which the values of the parameters characterizing the functioning of the liver are determined. Magnetic resonance imaging is considered as the main medical equipment used for diagnosis. A structural diagram of the diagnostic system is developed, which includes a tomogram processing subsystem and a decision-making subsystem. As a basic element of the tomogram processing subsystem, a convolutional neural network (Residual Network) is used, the training of which is carried out using the Transfer Learning algorithm. As the parameter that determine the stage of liver fibrosis, the image texture is used. The decision support subsystem is built on the basis of fuzzy neural networks. Examples of the system when determining the stages of fibrosis are given.
Intellectual system for printed circuit board manufacture dased on Mirae Mx-200
(National Aviation University, 2021-10-21) Sineglazov, Victor; Синєглазов, Віктор Михайлович; Plodystyy, Bogdan; Плодистий, Богдан Олександрович
It is considered the main disadvantages of printed circuit boards manufacture based on the Mirae Mx-200 system. In order to reduce the level of manufacturing defects and increase productivity, it is proposed to include an intelligent unit based on the YOLO neural network in the system, which is implemented by an additional Raspberry controller included in the system. The YOLO neural network is used to process images obtained from an additionally installed video camera, which monitors the production process. In this work, based on the use of the solution to the classification problem, the problem of decision support is formulated and solved. As a result, the operations (actions) that need to be taken are determined: automatic centering, reset, cancel, etc. Using emulation with additional microcontroller connections, the problem of limited installer resources and the implementation of more complex algorithms in the installer's work is solved.
Intellectual system of preparation of images from computer tomographs
(National Aviation University, 2021-01-05) Sineglazov, Victor; Синєглазов, Віктор Михайлович; Kharchuk, Yaroslav; Харчук, Ярослав Володимирович
Artificial neural networks can be trained on useful signals of the source data, but can not be taught on noisy data, so it is usually performed noise reduction or error compensation. This paper implements a noise reduction model based on artificial neural networks to suppress high-noise components, which is important for optimizing pre-filtering methods. The process of cleaning computers’ tomography scans in medical examinations of patients with tuberculosis is considered as an given problem in which the suppression of noise present in the image is required.. In order to reduce the level of radiation due to it is quite harmful to human. the power of the radiation is reduced. As a result, the ratio of the useful signal to noise is reduced, which causes noise, which contaminates the image and complicates its processing. Additional shadows appears on the image that no objects exist, which can provide false diagnosis. An algorithm for structural-parametric synthesis of convolutional neural networks used in image noise suppression has been developed. Computer tomograms of tuberculosis patients provided by the Research Institute of Pulmonology and Tuberculosis of the National Academy of Medical Sciences of Ukraine were used as a training sample.
Intelligent System for Visual Navigation
(Osvita Ukrainy, 2016-10) Sineglazov, Victor; Ischenko, Vitaliy
In this article it is considered the problem of algorithmic maintenance for visual navigation system. It is analyzed the characteristic features of different methods of intelligent system synthesis for visual navigation. The main stages and algorithms of work visual navigation system are encouraged to develop algorithmic maintenance for intellectual visual navigation system based on convolution neural networks.