Russian Federation
Russian Federation
Binary image classification tasks are widely employed in engineering and manufacturing systems, including automated control, machine vision, and object monitoring. As the complexity of imaging conditions increases and data volumes expand, it becomes essential to evaluate classical computer vision algorithms alongside deep learning neural network methods to identify the most effective method. Purpose: to perform a practical comparison of the effectiveness of a traditional image processing algorithm and the YOLO neural network model for addressing the binary classification challenge. Methods: traditional image processing techniques, including threshold filtering, morphological operations, and geometric feature analysis, as well as the YOLO detection model trained on a labeled dataset. Results: the classical algorithm has demonstrated high processing speed and adequate accuracy under stable lighting conditions; however, it has exhibited a pronounced decline in performance when imaging conditions changed. In contrast, the YOLO model has demonstrated enhanced accuracy, resilience to both photometric and geometric variations, and consistent performance even in the presence of extraneous noise. Practical significance: the results can guide the development of computer vision systems, aid in the selection of the optimal algorithm for specific operational scenarios, and facilitate the creation of hybrid systems that combine the strengths of both traditional and neural network methods. Discussion: this study confirms that while traditional methods are effective in resource-constrained environments, they are vulnerable to external influences. Conversely, neural network approaches offer superior generalization and stability, making them more advantageous in fluctuating imaging conditions. The novelty of the research lies in the comparative analysis of these methods under identical processing parameters, with a particular focus on practical indicators. This approach facilitates an objective assessment of the applicability domain inherent to each method.
computer vision, deep learning, YOLO, binary classification, image analysis, image processing, neural network methods
1. Senichev A. V., Novikova A. I., Vasil'ev P. V. Sravnenie glubokogo obucheniya s tradicionnymi metodami komp'yuternogo zreniya v zadachah identifikacii defektov // Molodoy issledovatel' Dona. 2020. № 4 (25). S. 64–67.
2. Polkovnikova N. A. Issledovanie metodov i algoritmov komp'yuternogo zreniya na osnove svertochnyh i rekurrentnyh neyronnyh setey // Ekspluataciya morskogo transporta. 2020. № 3 (96). S. 154–168. DOI: 10.34046/ aumsuomt96/21.
3. Saksonov P. V., Bauman A. A. Obzor metodov mashinnogo obucheniya // Sovremennye tendencii i innovacii v nauke i proizvodstve: materialy XII Mezhdunarodnoy nauchno-prakticheskoy konferencii (Mezhdurechensk, Rossiya, 26 aprelya 2023 g.). Mezhdurechensk: Kuzbasskiy gos. tehnicheskiy un-t im. T. F. Gorbacheva, 2023. S. 444-1–444-6.
4. Gonsales R. S., Vuds R. E. Cifrovaya obrabotka izobrazheniy = Digital Image Processing. Third Edition / per. s angl. L. I. Rubanova i P. A. Chochia. 3-e izd., ispr. i dop. M.: Tehnosfera, 2012. 1104 s.
5. Senichev A. V., Novikov S. S. Metody komp'yuternogo zreniya i ih primenenie v analize izobrazheniy. M.: IPRAN RAN, 2020. 142 s.
6. Bradski G., Kaehler A. Learning OpenCV: Computer Vision with the OpenCV Library. Sebastopol (CA): O’Reilly Media, 2008. 575 p.
7. Redmon J., Farhadi A. YOLOv3: An Incremental Improvement // ArXiv. 2018. Vol. 1804.02767. 6 p. DOI: 10.48550/ arXiv.1804.02767.
8. Caballar R. D., Stryker C. What is Computer Vision? URL: http://www.ibm.com/think/topics/computer-vision (data obrascheniya: 07.11.2025).
9. Patel M. The Complete Guide to Image Preprocessing Techniques in Python. URL: http://readmedium.com/thecomplete-guide-to-image-preprocessing-techniques-in-python-dca30804550c (data obrascheniya: 07.11.2025).
10. Thresholding (Image Processing) // Wikipedia. Last update 12 November 2025. URL: http://en.wikipedia.org/wiki/ Thresholding_(image_processing) (data obrascheniya: 16.11.2025).
11. Mathematical Morphology // Wikipedia. Last update 08 November 2025. URL: http://en.wikipedia.org/wiki/Mathematical_ morphology (data obrascheniya: 16.11.2025).
12. Bradski G. The OpenCV Library // Dr. Dobb’s Journal Software Tools. 2000. Vol. 25, Iss. 11. Pp. 120–125.
13. PyTorch: An Imperative Style, High-Performance Deep Learning Library / A. Paszke, S. Gross, F. Massa [et al.] // Advances in Neural Information Processing Systems 32: Proceeding of the 33rd Conference on Neural Information Processing Systems (NeurIPS 2019), (Vancouver, Canada, 08–14 December 2019). NeurIPS Foundation, 2019. Pp. 8024–8035.
14. Hunter J. D. Matplotlib: A 2D Graphics Environment // Computing in Science and Engineering. 2007. Vol. 9, Iss. 3 Pp. 90–95. DOI:https://doi.org/10.1109/MCSE.2007.55.
15. Turay T., Vladimirova T. Toward Performing Image Classification and Object Detection with Convolutional Neural Networks in Autonomous Driving Systems: A Survey // IEEE Access. 2022. Vol. 10. Pp. 14076–14119. DOI: 10.1109/ ACCESS.2022.3147495.
16. Powers D. M. W. Evaluation: From Precision, Recall and F-measure to ROC, Informedness, Markedness and Correlation // Journal of Machine Learning Technologies. 2011. Vol. 2, Iss.1. Pp. 37–63.
17. Klette R. Komp'yuternoe zrenie. Teoriya i algoritmy = Concise Computer Vision. An Introduction into Theory and Algorithms / per. s angl. A. A. Slinkina. M.: DMK Press, 2019. 506 s.
