CONVOLUTIONAL NEURAL NETWORK HYBRID ARCHITECTURE DEVELOPMENT USING SPECTRAL TRANSFORMATIONS
DOI:
https://doi.org/10.18522/2311-3103-2026-1-%25pKeywords:
Hybrid SNS, Walsh transformation, visual localization, Siamese networks, visual search, WalsPoolingAbstract
The hybrid convolutional neural network architecture with combining spectral and spatial layers, as well as new methods of subsampling (WalsPooling) and convolution (ConvWals) are proposed. The developed system is used to geographical proximity assess of images pair based on their visual similarity. A pair of different sensors obtained images visual similarity determination is complicated by different scales and sensor tilt angles shooting conditions. Based on the low-altitude image fragment, a search in the database of underlying surface images is performed. The search is performed in the surrounding area of a given route based on the vector of image features, which is formed on the last layer of the convolutional neural network. The system uses the Siamese architecture, since a pair of images must be submitted to the input. The relevance of this problem stems from the need to ensure UAV navigation in the absence or unreliability of a GPS signal. The approach to data set formation and its preprocessing is also considered. The database search is performed in the surrounding area of the route, which reduces computational costs. The experiments include an analysis of the applicability of the proposed layers (WalsPooling, ConvWals) and a comparison with traditional pooling and convolution methods. The paper also presents a linear approximation method with trainable parameters for reducing the dimensionality of the convolutional layer. The main advantage of the approach is its resistance to changes in the scale and angle of shooting due to a combination of spectral and spatial features. The results demonstrate the applicability of the method for UAV navigation in conditions of loss of GPS signal is lost or unreliable. The experiment demonstrated that using images reconstructed after spectral transformation yields the best neural network convergence and mean square error. The developed architecture demonstrates robustness to geometric and brightness distortions, and its quality metrics (Precision = 0.728, Recall = 0.800, F1 = 0.872) confirm the effectiveness of the approach for visual localization tasks based on images from a surface database.
References
1. Gaihua Wang, Guoliang Yuan, Meng Lv, WenZhou Liu. Application of Interpolation Pooling in Convo-lutional Neural Networks, Helix, 2018, Vol. 8 (4), pp. 3465- 3469. DOI 10.29042/2018-3465-3469
2. Zhuo Chen, Jiyuan Zhang, Ruizhou Ding. ViP. Virtual Pooling for Accelerating CNN-based Image Classification and Object Detection, WACV, 2020, pp. 1180-1189. Available at: https://openaccess.thecvf.com/content_WACV_2020/papers/Chen_ViP_Virtual_Pooling_for_Accelerating_CNN-based_Image_Classification_and_Object_WACV_2020_paper.pdf.
3. Travis Williams, Robert Li. Wavelet Pooling for Convolutional Neural Networks, Published as a con-ference paper at ICLR, 2018, pp. 1-12. Available at: https://openreview.net/forum?id=rkhlb8lCZ.
4. Juan Manuel Fortuna-Cervantes, Marco Tulio Ramírez-Torres, Marcela Mejía-Carlos, José Salomé Murguía, José Martinez-Carranza, Carlos Soubervielle-Montalvo, César Arturo Guerra-García. Tex-ture and Materials Image Classification Based on Wavelet Pooling Layer in CNN, the Special Issue Re-cent Advances in Deep Learning for Image Analysis. Available at: https://doi.org/10.3390/app12073592.
5. Image-Net. Available at: https://www.image-net.org/.
6. Pengju Liu, Hongzhi Zhang, Wei Lian, Wangmeng Zuo. Multi-Level Wavelet Convolutional Neural Networks, Digital Object Identifier. 10.1109/ACCESS.2019.2921451. https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=8732332.
7. Vershkov N.A., Babenko M.G., Chernykh A.N., Kuchukov V.A., Kucherov N.N., Kuchukova N.N Opti-mizatsiya iskusstvennykh neyronnykh setey s pomoshch'yu veyvlet-preobrazovaniy [Optimization of ar-tificial neural networks using wavelet transforms], Programmirovanie [Programming], 2022,
No. 6, pp. 22-30. DOI: 10.31857/S0132347422060073.
8. Spitsyn V.G., Bolotova Yu.A., Phan N.H., Bui T.T.T. Using a Haar wavelet transform, principal compo-nent analysis and neural networks for OCR in the presence of impulse noise. Available at: https://computeroptics.ru/eng/KO/Annot/KO40-2/400216.html.
9. Mitina O.A., Bazanova D.V. Razrabotka modeli neyronnoy seti na osnove ryada Fur'e [Development of a neural network model based on the Fourier series], NAU [National Association of Scientists], 2023, No. 90-2. Available at: https://cyberleninka.ru/article/n/razrabotka-modeli-neyronnoy-seti-na-osnove-ryada-furie (accessed 12 August 2025).
10. Fortuna-Servantes Kh.M., Ramires-Torres M.T., Martines-Karransa Kh., Murguia-Ibarra Kh.S., Mekhia-Karlos M. Obnaruzhenie ob"ektov v aeronavigatsii s ispol'zovaniem veyvlet-preobrazovaniya i svertochnykh neyronnykh setey: pervyy podkhod [Object Detection in Air Navigation Using Wavelet Transform and Convolutional Neural Networks: A First Approach], Tr. ISP RAN [Proceedings of ISP RAS], 2021, Vol. 33, Issue 2. DOI: 10.15514/ISPRAS–2020–33(2)–9.
11. Belyaev P.Yu., Zikratov I.A. Issledovanie avtonomnoy navigatsii bespilotnykh letatel'nykh apparatov na osnove korrelyatsionnykh metodov sravneniya izobrazheniy [Study of Autonomous Navigation of Un-manned Aerial Vehicles Based on Correlation Methods of Image Comparison], Tr. uchebnykh zavedeniy svyazi [Proceedings of Communications Educational Institutions], 2024, Vol. 10, No. 5, pp. 109-118.
12. Davydova A.A., Chechikov Yu.B. Problema realizatsii sovremennoy GPS-navigatsii dlya BPLA [The problem of implementing modern GPS navigation for UAVs], Perspektivnye nauchnye issledovaniya: opyt, problemy i perspektivy razvitiya: Sb. nauchnykh statey po materialam X Mezhdunarodnoy nauch-no-prakticheskoy konferentsii [Advanced scientific research: experience, problems and development pro-spects: Collection of scientific articles based on the materials of the X International scientific and practical conference]. Ufa, 2023, pp. 100-106.
13. Korikov A.M., Tran V.T. Primenenie neyronnykh setey i korrelyatsionno-ekstremal'nogo koordinatora dlya navigatsii i upravleniya BPLA [Application of neural networks and correlation-extreme coordinator for navigation and control of UAVs], Prirodnye i intellektual'nye resursy Sibiri (SIBRESURS-28-2022): Doklady (materialy konferentsii) 28-y mezhdunarodnoy nauchno-prakticheskoy konferentsii [Natural and intellectual resources of Siberia (SIBRESURS-28-2022): Reports (conference materials) of the 28th international scientific and practical conference]. Tomsk, 2022, pp. 29-33.
14. Lesiv E.A., Vytovtov K.A., Barabanova E.A. Neyrosetevoy algoritm analiza videoinformatsii dlya rezervnoy sistemy lokal'noy navigatsii privyaznykh BPLA s obucheniem na sinteticheskikh dannykh [Neural network algorithm for analyzing video information for a backup local navigation system for teth-ered UAVs with training on synthetic data ], Informatsionnye tekhnologii i tekhnicheskie sredstva uprav-leniya (ICCT- 2023). Mater. VII Mezhdunarodnoy nauchnoy konferentsii [nformation technology and technical means of control (ICCT-2023). Proceedings of the VII International Scientific Conference]. Moscow, 2023, pp. 159-162.
15. Ustinova V.E., Lutsenko A.S., Shpak A.V., Mironenkov G.V., Ivlev V.A. A method for finding the corre-spondence between a railway station model and its visual representation based on graphs, Computing, Telecommunications and Control, 2024, Vol. 17, No. 4, pp. 64-77.
16. Ustinova V.E., Lutsenko A.S., Shpak A.V., Mironenkov G.V., Ivlev V.A. A method for finding the corre-spondence between a railway station model and its visual representation based on graphs, Computing, Telecommunications and Control, 2024, Vol. 17, No. 4, pp. 64-77.
17. Kocheturov V.V., Fedyaev O.I. Siamskie svertochnye neyronnye seti dlya raspoznavaniya lits [Siamese convolutional neural networks for face recognition], Programmnaya inzheneriya: metody i tekhnologii razrabotki informatsionno-vychislitel'nykh sistem (PIIVS-2024): Sb. materialov i dokladov V Mezhdu-narodnoy nauchno-prakticheskoy konferentsii [Software Engineering: Methods and Technologies for Developing Information and Computing Systems (PIIVS-2024): Collection of materials and reports of the V International scientific and practical conference]. Donetsk, 2024, pp. 178-184.
18. Ladanova E.O. Razrabotka metodiki i novoy struktury siamskoy neyronnoy seti dlya analiza geopros-transtvennykh dannykh [Development of a methodology and a new structure of a Siamese neural net-work for analyzing geospatial data], Nauchno-tekhnicheskiy vestnik Povolzh'ya [Scientific and Technical Bulletin of the Volga Region], 2024, No. 12, pp. 144-147.
19. Sumina E.A. Siamese networks as a possible improvement of standard algorithms, Artificial Intelligence: Challenges, Essence, Communication, Creativity and other aspects: Collection of materials from stu-dent conferences. Мoscow, 2024, pp. 104-108.
20. Kurtyanik D. V., Sergeev A.M., Samikov A.V., Semenov A.A. O vizual'noy otsenke rezul'tatov poiska matrits Adamara [On the visual assessment of the results of the search for Hadamard matrices], Nauka. Tekhnika. Tekhnologii (politekhnicheskiy vestnik) [Science. Engineering. Technologies (Polytechnic Bul-letin)], 2018, No. 4, pp. 19-26.
21. Kumkov S.I., TSvetkov A.V. Osnovy teorii signalov: ucheb. posobie [Fundamentals of signal theory: a tutorial]. Ekaterinburg: Izd-vo Ural. un-ta, 2023, 114 p. ISBN 978-5-7996-3636-4.
22. Kostrov B.V., Grinchenko N.N., Baranova S.N., Trushina E.A., V'yugina A.A. Ortogonal'noe kodiro-vanie binarnykh izobrazheniy [Orthogonal coding of binary images], Vestnik Yaroslavskogo vysshego voennogo uchilishcha protivovozdushnoy oborony [Bulletin of the Yaroslavl Higher Military School of Air Defense], 2023, No. 2 (21), pp. 82-87.
