Khorolska K. Information technology for graphic information recognition based on neural networks

Українська версія

Thesis for the degree of Doctor of Philosophy (PhD)

State registration number

0823U101091

Thesis Registration Form

0823U101091.pdf

Applicant for

Specialization

  • 122 - Комп’ютерні науки

Specialized Academic Board

ДФ 26.055.050

State University of Trade and Economics

Essay

The thesis is a comprehensive study on the modeling, development, and application of information technology for graphic information recognition based on neural networks. The relevance of the research topic is determined by the importance of recognizing graphical information and the practical application of the theoretical foundations for recognizing graphical information based on highly effective solutions. At this stage of practical application development, several unresolved issues remain in the theoretical foundations of graphical information recognition: quality contour extraction; automated object extraction; variability in spatial placement and recognition of objects; practical application of methodological developments in recognizing graphical information; and high-quality classification of recognition objects. Visual image recognition is a crucial component of management and information processing systems, automated systems, and decision-making systems. Tasks associated with classifying and identifying objects, phenomena, and signals, characterized by a finite set of certain properties and features, arise in fields such as robotics, information retrieval, monitoring and analysis of visual data, and artificial intelligence research. Algorithmic processing and classification of images are applied in security systems, access control and management, virtual reality systems, and information search systems. Moreover, with the broad implementation of virtual reality systems and the development of the gaming industry, and considering that a 3D model is initially created as its 2D version, an acute need arises for the rapid conversion of two-dimensional images into three-dimensional models. A significant role in solving the mentioned problems is played by the creation of software complexes and mathematical tools based on the application of neural networks, expert systems, and cybernetics. Particular attention is given to artificial neural networks (or simply neural networks) - computing systems that learn from observational data through an optimization process, wherein model parameters are iteratively adjusted to minimize the difference between the predicted outcome and the actual result. Neural networks, especially convolutional neural networks (CNN), are widely recognized for their effectiveness in analyzing visual images. CNNs are a class of deep learning models specifically designed for processing 2D images. They consist of multiple layers of small neuron collections that process parts of the input image called receptive fields. The outputs from these collections are arranged so they overlap for a better representation of the output image, a key feature of CNNs. Moreover, they are translation-invariant, meaning they can identify an object as the same when it appears in different representations. These functions allow CNNs to capture intricate patterns in spatial and temporal domains – a critical aspect in the task of 3D reconstruction from 2D images. Artificial neural networks also excel in processing noisy, incomplete, or ambiguous data – a scenario commonly encountered in image processing tasks. They can extract significant features even from flawed data (noise, missing values, duplicates, contradictions), ensuring the model's reliability. Additionally, neural networks, specifically CNNs, can recognize hierarchical patterns in the data. For instance, in image processing tasks, they might identify edges and color gradients at a lower level and shapes or parts of an object at a higher level. This functionality is paramount in tasks like 3D reconstruction, which require the model to detect high-level features and relationships in 2D images. Practical significance of the scientific results. The information technology for recognizing graphical information based on neural networks, drawing recognition, and transformation, designed in the study using CASE technology ERwin, considering external influences on it and the interaction of processes within the system, provides the opportunity to implement software applications to solve tasks of graphical information recognition based on neural networks and converting data from bi-vector space to tri-vector space. The developed algorithms and software application architecture, grounded on the created model of graphical image classifiers based on class coverages and elementary classifiers (EC) of primitives to enhance CNN training efficiency, allow a 1.5-2 times reduction in computational costs for CNN training and up to a 2-fold decrease in the overall training error of CNN. This ensures a reduction in resource intensity and error in recognizing graphical information based on neural networks and transforming data from a bi-vector space to a tri-vector one.

Read more

Research papers

1. Khorolska K., Artificial intelligence face recognition for authentication./ Kryvoruchko, O., Bebeshko, B., Khorolska, K., Desiatko, A., Kotenko, N. (2020). Technical Sciences and Technologies, 2 (20), 139-148. https://stu.cn.ua/wp-content/uploads/2021/04/technical-sciences-and-technologies2.pdf

2. Khorolska, K. (2022). Потенціал застосування різних методів штучного інтелекту у задачі розпізнавання креслень та трансформації 2D→3D. Електронне фахове наукове видання "Кібербезпека: освіта, наука, техніка";, 1(17), 21-30. https://doi.org/10.28925/2663-4023.2022.17.2130

3. Khorolska, K. (2022). Аналіз основних методів розпізнавання креслень та можливостей трансформації з 2D У 3D. Електронне фахове наукове видання "Кібербезпека: освіта, наука, техніка" 4(16), 185-193. https://doi.org/10.28925/2663-4023.2022.16.185193

4. Хорольська К. Аналіз основних підходів до вирішення задачі конвертації двовимірних зображень в тривимірну модель Вісник КрНУ імені Михайла Остроградського. Випуск 3 / 2022 (134) DOI https://doi.org/10.32782/1995-0519.2022.3.7

5. Bebeshko, B., Khorolska, K., Kotenko, N., Desiatko, A., Sauanova, K., Sagyndykova, S., Tyshchenko, D. 3D modelling by means of artificial intelligence (2021) Journal of Theoretical and Applied Information Technology, 99 (6), pp. 1296-1308. http://www.jatit.org/volumes/Vol99No6/5Vol99No6.pdf

6. Khorolska K., Lazorenko V., Bebeshko B., Desiatko A., Kharchenko O., Yaremych V. (2022) Usage of Clustering in Decision Support System. In: Raj J.S., Palanisamy R., Perikos I., Shi Y. (eds) Intelligent Sustainable Systems. Lecture Notes in Networks and Systems, vol 213. Springer, Singapore. https://doi.org/10.1007/978-981-16-2422-3_49

7. Lakhno, V., Akhmetov, B., Smirnov, O., Chubaievskyi, V., Khorolska, K., Bebeshko, B. (2023). Selection of a Rational Composition of İnformation Protection Means Using a Genetic Algorithm. In: Rajakumar, G., Du, KL., Vuppalapati, C., Beligiannis, G.N. (eds) Intelligent Communication Technologies and Virtual Mobile Networks. Lecture Notes on Data Engineering and Communications Technologies, vol 131. Springer, Singapore. https://doi.org/10.1007/978-981-19-1844-5_2

8. Khorolska, K., Bebeshko, B., Desiatko, A., & Lazorenko, V. (2021). 3D models classification with use of convolution neural network. Paper presented at the CEUR Workshop Proceedings, 3179 25-34. http://ceur-ws.org/Vol-3179/Paper_3.pdf

9. Khorolska K. , Skladannyi P., Sokolov V., Korshun N., Bebeshko B., Lakhno V., Zhumadilova M (2022) Application of a convolutional neural network with a module of elementary graphic primitive classifiers in the problems of recognition of drawing documentation and transformation of 2D to 3D models. Journal of Theoretical and Applied Information Technology 31st December 2022. Vol.100. No 24 http://www.jatit.org/volumes/Vol100No24/18Vol100No24.pdf

10. Khorolska K. Use of AI in data protection/ Kryvoruchko O., Bebeshko B., Khorolska K. // Безпека ресурсів інформаційних систем : збірник тез I Міжнародної науково-практичної конференції (м. Чернігів 16-17 квітня 2020 р.). – Чернігів : НУЧП, 2020. – с.15-18

11. Бебешко Б.Т., Лазоренко В.В., Хорольська К.В. Безпека інтелектуальної системи управління цифровими активами за допомогою методу k-means при дослідженні видобутку даних // Кібергігієна. Кібербезпека. Безпека держави: матеріали наукових семінарів (Київ, 27 листопада 2020 р.)/відп. ред. АМ Десятко.–Київ: Київ. нац. торг.-екон. ун-т, 2020.–с.34-36 https://knute.edu.ua/file/MjExMzA=/d8e24930571c0d91476be247343bb902.pdf

12. Лазоренко В.В., Бебешко Б.Т., Хорольська К.В. Аналіз методів прогнозування кібератак // Комплексне забезпечення якості технологічних процесів та систем (КЗЯТПС – 2021) : матеріали тез доповідей XІ Міжнародної науково-практичної конференції (м. Чернігів, 26–27 травня 2021 р.) : у 2 т. / Національний університет «Чернігівська політехніка» [та ін.] ; відп. за вип.: Єрошенко Андрій Михайлович [та ін.]. – Чернігів : НУ «Чернігівська політехніка», 2021. – Т. 2. – 236 с. ISBN 978-617-7932-16-0

13. Khorolska K. 3D Model reconstruction using convolutional neural networks for 2D image processing. Proceedings of the VI International Scientific and Practical Conference. Osaka, Japan. 2023. Pp.457-459 DOI: 10.46299/ISG.2023.1.6

14. Khorolska Karyna VR-technology as a modern architecture tool. Management of Development of Complex Systems / Tsiutsiura, Svitlana, Bebeshko, Bohdan, & Khorolska, Karyna, (2020). Management of Development of Complex Systems, 42, 69 – 74, dx.doi.org\10.32347/2412-9933.2020.42.69-74

15. Хорольська К. В. UX-дизайн інформаційної системи підприємства торгівлі. / Котенко Н.О., Жирова Т.О., Десятко А.М., Хорольська К.В., Бебешко Б.Т., Тогжанова К.О. // Вісник Кременчуцького національного університету імені Михайла Остроградського. 2020. Вип. № 3 (122). С. 107–112. DOI: 10.30929/1995-0519.2020.3.67-74

16. Tetiana Zhyrova, Nataliia Kotenko, Volodymyr Tokar, Karyna Khorolska, Bohdan Bebeshko, (2021) Testing the Accessibility of Web-applications The International Scientific Journal «Computer Systems and Information Technologies» 2021, #3 DOI: https://doi.org/10.31891/CSIT-2021-5-12

17. Lakhno V., Akhmetov B., Ydyryshbayeva M., Bebeshko B., Desiatko A., Khorolska K. (2021) Models for Forming Knowledge Databases for Decision Support Systems for Recognizing Cyberattacks. In: Vasant P., Zelinka I., Weber GW. (eds) Intelligent Computing and Optimization. ICO 2020. Advances in Intelligent Systems and Computing, vol 1324. Springer, Cham. https://doi.org/10.1007/978-3-030-68154-8_42

18. Lakhno, V., Mazaraki, A., Kasatkin, D., Kryvoruchko, O., Khorolska, K., Chubaievskyi, V. (2023). Models and Algorithms for Optimization of the Backup Equipment for the Intelligent Automated Control System Smart City. In: Ranganathan, G., Fernando, X., Rocha, Á. (eds) Inventive Communication and Computational Technologies. Lecture Notes in Networks and Systems, vol 383. Springer, Singapore. https://doi.org/10.1007/978-981-19-4960-9_57

19. Zhyrova, T., Kotenko, N., Bebeshko, B., Khorolska, K., Shevchenko, S.(2022) Benchmarking between the DQL Index and the Web Application Accessibility Index using Automatic Test Tools CEUR Workshop Proceedings, 2022, 3288, pp. 110–116 https://ceur-ws.org/Vol-3288/short8.pdf

20. Khorolska K. Cyberattacks prediction with incomplete data/ Bebeshko B., Khorolska K. // Безпека соціально-економічних процесів в кіберпросторі: зб. матеріалів Всеукр. наук.-практ. конф. (Київ, 27 бер. 2019 р.). – Київ : Київ. нац. торг.-екон. ун-т, 2019. – с.123-125

21. Khorolska K. Usage of neural networks in image recognition / O. Kryvorychko, K. Khorolska, V. Chubaevskyi. // Зовнішня торгівля: економіка, фінанси, право. – 2019. – №3 (104). – С. 83–85. http://zt.knute.edu.ua/files/2019/03(104)/9.pdf

22. Bebeshko, B., Khorolska, K., Kotenko, N., Kharchenko, O., & Zhyrova, T. (2021). Use of neural networks for predicting cyberattacks. Paper presented at the CEUR Workshop Proceedings, 2923 213-223. http://ceur-ws.org/Vol-2923/paper23.pdf

Read more

Files

autoreferat-АНОТАЦІЯ_Хорольська.pdf

editrequest-Рішення_Хорольська.pdf

дисертація_Хорольська фінал 18-09-2023 (1).pdf

Similar theses

0824U001153

Oleksandr Tkachyk

Methods and tools for clustering heterogeneous data

0824U001143

Jie Chen

Models and Methods for Assessing the Competitiveness of Universities' Educational Activities in the Context of Digital Transformation.

0824U001146

Bohdan Khudoba

Methods and means of analyzing operator parameters using computer simulators.

0824U001134

Huan Zhou

Models and methods of competency-based human resource management of educational projects.

0824U001104

Ming Li

Models, methods and information technology for choosingthemanagement methodology of international educational projects