Том 4 № 2 (2026): Промышленная кибернетика
DOI https://doi.org/10.18799/29495407/2026/2/126
Влияние стратегий аннотации и масштаба входных данных на эффективность семантической сегментации океанических вихрей
Проведен аналитический обзор публикаций по семантической сегментации океанических вихрей. Показано, что архитектуры глубокого обучения, основанные на трансформерах, благодаря механизму самовнимания – наилучший вариант определения типа и общего контура вихря. Для оценки влияния стратегий аннотации и масштаба входных данных на качество сегментации с использованием модели SegFormer-B4 были сформированы наборы данных с масштабами 256×256, 512×512 и 1024×1024 пикселей. В экспериментах сравнивались два принципиально разных подхода к разметке изображений – детальная ручная аннотация (с филаментами) и упрощенная эллиптическая маска (без филаментов). Установлено, что модель, обученная на данных с включением спиральных филаментов, демонстрирует существенно более высокое качество сегментации по таким ключевым метрикам, как коэффициент Дайса (Dice Score) и индекс Жаккара (IoU). Выявлен оптимальный пространственный масштаб входных данных. Наилучшие показатели были достигнуты при использовании изображений размером 512×512 пикселей, что соответствовало исходному разрешению использованной предобученной модели SegFormer-B4. Сформулированы рекомендации по выбору стратегии аннотирования в зависимости от цели анализа – обнаружение наличия вихрей или точное восстановление их морфометрических характеристик.
Ключевые слова:
семантическая сегментация, океанический вихрь, искусственные нейронные сети, машинное обучение, обработка изображений
Библиографические ссылки:
СПИСОК ЛИТЕРАТУРЫ
1. Rosenfeld A., Pfaltz J.L. Sequential operations in digital picture processing. Journal of the ACM, 1966, Vol. 13, Iss. 4,
P. 471–494. DOI: https://doi.org/10.1145/321356.321357
2. Zucker S.W. Region growing: childhood and adolescence. Computer Graphics and Image Processing, 1976, Vol. 5, Iss. 3, P. 382–399. DOI: https://doi.org/10.1016/S0146-664X(76)80014-7
3. Otsu N. A Threshold selection method from gray-level histograms. IEEE Transaction on Systems, Man and Cybernetics, 1979, Vol. 9, Iss. 1, P. 62–66. DOI: 10.1109/TSMC.1979.4310076
4. Long J., Shelhamer E., Darrell T. Fully convolutional networks for semantic segmentation. IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Boston, MA, USA, 2015. P. 3431–3440. DOI: 10.1109/CVPR.2015.7298965
5. Калинин М.А., Вереземская П.С., Криницкий М.А., Борисов М.А., Тилинина Н.Д. Байесовская оптимизация метода идентификации мезомасштабных океанских вихрей в море Лабрадор в данных вихреразрешающего моделирования. Океанологические исследования, 2024, Т. 52, № 4, С. 56–73. DOI: 10.29006/1564-2291.JOR-2024.52(4).4 EDN: FSRROH
6. Khachatrian E., Sandalyuk N., Lozou P. Eddy detection in the marginal ice zone with Sentinel-1 data using YOLOv5. Remote Sensing, 2023, Vol. 15, Iss. 9, art. 2244. DOI: https://doi.org/10.3390/rs15092244 EDN: GGCCEI
7. Kalinin M.A., Krinitskiy M.A., Verezemskaya P.S. Detection of Irminger rings in high resolution ocean hydrodynamic modeling data using artificial neural networks. Moscow University Physics Bulletin, 2025, Vol. 80, Suppl 3, P. S1270–S1278. DOI: https://doi.org/10.3103/S0027134925703102 EDN: PMSWYV
8. Liu Y., Zheng Q., Li X. Detection and analysis of mesoscale eddies based on deep learning. Artificial intelligence oceanography. Eds. Li X., F. Wang. Singapore: Springer, 2023. P. 209–225. DOI: https://doi.org/10.1007/978-981-19-6375-9_10
9. Liu Y., Li X., Ren Y. A deep learning model for oceanic mesoscale eddy detection based on multi-source remote sensing imagery. IGARSS 2020-2020 IEEE International Geoscience and Remote Sensing Symposium, 2020, P. 6762–6765. DOI: 10.1109/IGARSS39084.2020.9323716
10. Geng J., Gao H., Huang B., Radenkovic M., Chen G. ARU2-net: a deep learning approach for global-scale oceanic eddy detection. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2024, Vol. 17, P. 11997–12007. DOI: 10.1109/jstars.2024.3419175 EDN: AQBZOB
11. Sun X., Zhang M., Dong J., Lguensat R., Yang Y., Lu X. A deep framework for eddy detection and tracking from satellite sea surface height data. IEEE Transactions on Geoscience and Remote Sensing, 2021, Vol. 59, № 9, P. 7224–7234. DOI: 10.1109/TGRS.2020.3032523 EDN: BZLKWB
12. Zhao Y., Fan Z., Li H., Zhang R., Xiang W., Wang S., Zhong G. SymmetricNet: end-to-end mesoscale eddy detection with multi-modal data fusion. Frontiers in Marine Science, 2023, Vol. 10, art. 1174818. DOI: https://doi.org/10.3389/fmars.2023.1174818 EDN: ERMINB
13. Huo J., Zhang J., Yang Ju., Li Ch., Liu G., Cui W. High kinetic energy mesoscale eddy identification based on multi-task learning and multi-source data. International Journal of Applied Earth Observation and Geoinformation, 2024, Vol. 128, art. 103714. DOI: 10.1016/j.jag.2024.103714 EDN: QKEYEO
14. Liu Y., Liu Q., Li X. A deep learning-based model for cold anticyclonic eddies and warm cyclonic eddies detection in the Kuroshio extension. 2022 Photonics & Electromagnetics Research Symposium (PIERS). Hangzhou, China, 2022. P. 258–263. DOI: 10.1109/PIERS55526.2022.9792954
15. Xu G., Xie W., Dong C., Gao X. Application of three deep learning schemes into oceanic eddy detection. Frontiers in Marine Science, 2021, Vol. 8, art. 672334. DOI: https://doi.org/10.3389/fmars.2021.672334 EDN: LSUWBJ
16. Gao X., Zhang W., Chen R. Mesoscale eddy identification in the North Pacific based on DeepLabV3+. Journal of Physics: Conference Series, 2025, Vol. 3007, № 1, art. 012052. DOI: 10.1088/1742-6596/3007/1/012052 EDN: QFMTLQ
17. Hou M., Fang L., Wu K., Yang J., Chen G. Multi-scale eddy identification and analysis based on deep learning method and ocean color data. International Journal of Digital Earth, 2025, Vol. 18, № 1, art. 2505624. DOI: https://doi.org/10.1080/17538947.2025.2505624 EDN: WSTHFT
18. Liu Y., Gao L., Liu Q., Li X. Dual-branch neural network for mesoscale eddy identification based on multi-variables remote sensing data. 2023 Photonics & Electromagnetics Research Symposium (PIERS). Prague, Czech Republic, 2023. P. 275–279. DOI: 10.1109/PIERS59004.2023.10221436
19. Sun H., Li H., Xu M., Xia T., Han Z., Liu C. A data-fused lightweight CNN-Transformer model for ocean eddy detection. Science of Remote Sensing, 2026, Vol. 13, art. 100412. DOI: https://doi.org/10.1016/j.srs.2026.100412
20. Xie H., Xu Q., Dong C. Deep learning for mesoscale eddy detection with feature fusion of multisatellite observations. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2024, Vol. 17, P. 18351–18364. DOI: 10.1109/jstars.2024.3468457 EDN: NXKZZS
21. Li B., Tang H., Ma D., Lin J. A dual-attention mechanism deep learning network for mesoscale eddy detection by mining spatiotemporal characteristics. Journal of Atmospheric and Oceanic Technology, 2022, Vol. 39, Iss. 8, P. 1115–1128. DOI: https://doi.org/10.1175/JTECH-D-21-0128.1 EDN: CUVQCO
22. Sun H., Li H., Xu M., Xia T., Yu H. Detecting ocean eddies with a lightweight and efficient convolutional network. Remote Sensing, 2024, Vol. 16, Iss. 24, art. 4808. DOI: https://doi.org/10.3390/rs16244808 EDN: AJFWWY
23. Zhao N., Huang B., Zhang X., Ge L., Ge Chen. Intelligent identification of oceanic eddies in remote sensing data via Dual-Pyramid UNet. Atmospheric and Oceanic Science Letters, 2023, Vol. 16, Iss. 4, art. 100335. DOI: https://doi.org/10.1016/j.aosl.2023.100335 EDN: LNARHY
24. Liu C., Lin X., Xu G., Han G., Liu Y. Improved identification and tracking of three-dimensional eddies in the Southern Ocean utilizing 3D-U-Res-Net. Frontiers in Marine Science, 2024, Vol. 11, art. 1482804. DOI: https://doi.org/10.3389/fmars.2024.1482804 EDN: VBSFJC
25. Xu G., Xie W., Lin X., Liu Y., Hang R., Sun W., Liu D., Dong C. Detection of three-dimensional structures of oceanic eddies using artificial intelligence. Ocean Modelling, 2024, Vol. 190, art. 102385. DOI: https://doi.org/10.1016/j.ocemod.2024.102385 EDN: MXSAHF
26. Liu X., Wang M. Detection of ocean eddies from satellite ocean color and SST measurements using a deep learning approach. International Journal of Applied Earth Observation and Geoinformation, 2025, Vol. 144, art. 104929. DOI: 10.1016/j.jag.2025.104929 EDN: FBVPBK
27. Dong Ch., You Zh., Dong J., Ji J., Sun W., Xu G., Lu X., Xie H., Teng F., Liu Yu., Xu A., Wang Q., Xia Q., Lin X., Fu M., Wang J., Cao Yu., Han G. Oceanic mesoscale eddies. Ocean-Land-Atmosphere Research, 2025, vol. 4, art. 0081. DOI: 10.34133/olar.0081 EDN: OILJRV
28. Bearman A., Russakovsky O., Ferrari V., Fei-Fei L. What's the point: semantic segmentation with point supervision. Computer Vision – ECCV 2016. ECCV 2016. Lecture Notes in Computer Science. Eds. B. Leibe, J. Matas, N. Sebe, M. Welling. Cham: Springer, 2016. Vol. 9911, P. 549–565. DOI: https://doi.org/10.1007/978-3-319-46478-7_34
29. Chelton D.B., Schlax M.G., Samelson R.M. Global observations of nonlinear mesoscale eddies. Progress in Oceanography, 2011, Vol. 91, Iss. 2, P. 167–216. DOI: https://doi.org/10.1016/j.pocean.2011.01.002 EDN: YBYDIJ
30. Xie E., Wang W., Yu Z., Anandkumar A., Alvarez J.M., Luo P. SegFormer: simple and efficient design for semantic segmentation with transformers. arXiv:2105.15203 [cs.CV]. 2021. DOI: https://doi.org/10.48550/arXiv.2105.15203
31. Guangjun X., Yucheng S., Yang Y., Huarong X., Wenhong X., Jingyuan L., Xiayan L., Yu L., Changming D. Recent developments in ai-based oceanic eddy identification. Journal of Marine Sciences, 2024, Vol. 42, № 3, P. 38–50.
REFERENCES
1. Rosenfeld A., Pfaltz J.L. Sequential operations in digital picture processing. Journal of the ACM, 1966, vol. 13, Iss. 4,
pp. 471–494. DOI: https://doi.org/10.1145/321356.321357
2. Zucker S.W. Region growing: childhood and adolescence. Computer Graphics and Image Processing, 1976, vol. 5, Iss. 3, pp. 382–399. DOI: https://doi.org/10.1016/S0146-664X(76)80014-7
3. Otsu N. A Threshold selection method from gray-level histograms. IEEE Transaction on Systems, Man and Cybernetics, 1979, vol. 9, Iss. 1, pp. 62–66. DOI: 10.1109/TSMC.1979.4310076
4. Long J., Shelhamer E., Darrell T. Fully convolutional networks for semantic segmentation. IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Boston, MA, USA, 2015. pp. 3431–3440. DOI: 10.1109/CVPR.2015.7298965
5. Kalinin M.A., Verezemskaya P.S., Krinitskiy M.A., Borisov M.A., Tilinina N.D. Bayesian optimization of the method of identification of mesoscale ocean eddies in the Labrador sea in eddy-resolving modelling data. Journal of oceanological research, 2024, vol. 52, no. 4, pp. 56–73. (In Russ.) DOI: 10.29006/1564-2291.JOR-2024.52(4).4 EDN: FSRROH
6. Khachatrian E., Sandalyuk N., Lozou P. Eddy detection in the marginal ice zone with Sentinel-1 data using YOLOv5. Remote Sensing, 2023, vol. 15, Iss. 9, art. 2244. DOI: https://doi.org/10.3390/rs15092244 EDN: GGCCEI
7. Kalinin M.A., Krinitskiy M.A., Verezemskaya P.S. Detection of Irminger rings in high resolution ocean hydrodynamic modeling data using artificial neural networks. Moscow University Physics Bulletin, 2025, vol. 80, Suppl 3, pp. S1270–S1278. DOI: https://doi.org/10.3103/S0027134925703102 EDN: PMSWYV
8. Liu Y., Zheng Q., Li X. Detection and analysis of mesoscale eddies based on deep learning. Artificial intelligence oceanography. Eds. X. Li, F. Wang. Singapore, Springer, 2023. pp. 209–225. DOI: https://doi.org/10.1007/978-981-19-6375-9_10.
9. Liu Y., Li X., Ren Y. A deep learning model for oceanic mesoscale eddy detection based on multi-source remote sensing imagery. IGARSS 2020-2020 IEEE International Geoscience and Remote Sensing Symposium, 2020, pp. 6762–6765. DOI: 10.1109/IGARSS39084.2020.9323716
10. Geng J., Gao H., Huang B., Radenkovic M., Chen G. ARU2-net: a deep learning approach for global-scale oceanic eddy detection. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2024, vol. 17, pp. 11997–12007. DOI: 10.1109/jstars.2024.3419175 EDN: AQBZOB
11. Sun X., Zhang M., Dong J., Lguensat R., Yang Y., Lu X. A deep framework for eddy detection and tracking from satellite sea surface height data. IEEE Transactions on Geoscience and Remote Sensing, 2021, vol. 59, no. 9, pp. 7224–7234. DOI: 10.1109/TGRS.2020.3032523 EDN: BZLKWB
12. Zhao Y., Fan Z., Li H., Zhang R., Xiang W., Wang S., Zhong G. SymmetricNet: end-to-end mesoscale eddy detection with multi-modal data fusion. Frontiers in Marine Science, 2023, vol. 10, art. 1174818. DOI: https://doi.org/10.3389/fmars.2023.1174818 EDN: ERMINB
13. Huo J., Zhang J., Yang Ju., Li Ch., Liu G., Cui W. High kinetic energy mesoscale eddy identification based on multi-task learning and multi-source data. International Journal of Applied Earth Observation and Geoinformation, 2024, vol. 128, art. 103714. DOI: 10.1016/j.jag.2024.103714 EDN: QKEYEO
14. Liu Y., Liu Q., Li X. A deep learning-based model for cold anticyclonic eddies and warm cyclonic eddies detection in the Kuroshio extension. 2022 Photonics & Electromagnetics Research Symposium (PIERS). Hangzhou, China, 2022. pp. 258–263. DOI: 10.1109/PIERS55526.2022.9792954
15. Xu G., Xie W., Dong C., Gao X. Application of three deep learning schemes into oceanic eddy detection. Frontiers in Marine Science, 2021, vol. 8, art. 672334. DOI: https://doi.org/10.3389/fmars.2021.672334 EDN: LSUWBJ
16. Gao X., Zhang W., Chen R. Mesoscale eddy identification in the North Pacific based on DeepLabV3+. Journal of Physics: Conference Series, 2025, vol. 3007, no. 1, art. 012052. DOI: 10.1088/1742-6596/3007/1/012052 EDN: QFMTLQ
17. Hou M., Fang L., Wu K., Yang J., Chen G. Multi-scale eddy identification and analysis based on deep learning method and ocean color data. International Journal of Digital Earth, 2025, vol. 18, no. 1, art. 2505624. DOI: https://doi.org/10.1080/17538947.2025.2505624 EDN: WSTHFT
18. Liu Y., Gao L., Liu Q., Li X. Dual-branch neural network for mesoscale eddy identification based on multi-variables remote sensing data. 2023 Photonics & Electromagnetics Research Symposium (PIERS). Prague, Czech Republic, 2023. pp. 275–279. DOI: 10.1109/PIERS59004.2023.10221436
19. Sun H., Li H., Xu M., Xia T., Han Z., Liu C. A data-fused lightweight CNN-Transformer model for ocean eddy detection. Science of Remote Sensing, 2026, vol. 13, art. 100412. DOI: https://doi.org/10.1016/j.srs.2026.100412
20. Xie H., Xu Q., Dong C. Deep learning for mesoscale eddy detection with feature fusion of multisatellite observations. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2024, vol. 17, pp. 18351–18364. DOI: 10.1109/jstars.2024.3468457 EDN: NXKZZS
21. Li B., Tang H., Ma D., Lin J. A dual-attention mechanism deep learning network for mesoscale eddy detection by mining spatiotemporal characteristics. Journal of Atmospheric and Oceanic Technology, 2022, vol. 39, Iss. 8, pp. 1115–1128. DOI: https://doi.org/10.1175/JTECH-D-21-0128.1 EDN: CUVQCO
22. Sun H., Li H., Xu M., Xia T., Yu H. Detecting Ocean Eddies with a Lightweight and Efficient Convolutional Network. Remote Sensing, 2024, vol. 16, Iss. 24, art. 4808. DOI: https://doi.org/10.3390/rs16244808 EDN: AJFWWY
23. Zhao N., Huang B., Zhang X., Ge L., Ge Ch. Intelligent identification of oceanic eddies in remote sensing data via Dual-Pyramid UNet. Atmospheric and Oceanic Science Letters, 2023, vol. 16, Iss. 4, art. 100335. DOI: https://doi.org/10.1016/j.aosl.2023.100335 EDN: LNARHY
24. Liu C., Lin X., Xu G., Han G., Liu Y. Improved identification and tracking of three-dimensional eddies in the Southern Ocean utilizing 3D-U-Res-Net. Frontiers in Marine Science, 2024, vol. 11, art. 1482804. DOI: https://doi.org/10.3389/fmars.2024.1482804 EDN: VBSFJC
25. Xu G., Xie W., Lin X., Liu Y., Hang R., Sun W., Liu D., Dong C. Detection of three-dimensional structures of oceanic eddies using artificial intelligence. Ocean Modelling, 2024, vol. 190, art. 102385. DOI: https://doi.org/10.1016/j.ocemod.2024.102385 EDN: MXSAHF
26. Liu X., Wang M. Detection of ocean eddies from satellite ocean color and SST measurements using a deep learning approach. International Journal of Applied Earth Observation and Geoinformation, 2025, vol. 144, art. 104929. DOI: 10.1016/j.jag.2025.104929 EDN: FBVPBK
27. Dong Ch., You Zh., Dong J., Ji J., Sun W., Xu G., Lu X., Xie H., Teng F., Liu Yu., Xu A., Wang Q., Xia Q., Lin X., Fu M., Wang J., Cao Yu., Han G. Oceanic mesoscale eddies. Ocean-Land-Atmosphere Research, 2025, vol. 4, art. 0081. DOI: 10.34133/olar.0081 EDN: OILJRV
28. Bearman A., Russakovsky O., Ferrari V., Fei-Fei L. What's the point: semantic segmentation with point supervision. Computer Vision – ECCV 2016. ECCV 2016. Lecture Notes in Computer Science. Eds. B. Leibe, J. Matas, N. Sebe, M. Welling. Cham, Springer, 2016. Vol. 9911, pp. 549–565. DOI: https://doi.org/10.1007/978-3-319-46478-7_34
29. Chelton D.B., Schlax M.G., Samelson R.M. Global observations of nonlinear mesoscale eddies. Progress in Oceanography, 2011, vol. 91, Iss. 2, pp. 167–216. DOI: https://doi.org/10.1016/j.pocean.2011.01.002 EDN: YBYDIJ
30. Xie E., Wang W., Yu Z., Anandkumar A., Alvarez J.M., Luo P. SegFormer: simple and efficient design for semantic segmentation with transformers. arXiv:2105.15203 [cs.CV]. 2021. DOI: https://doi.org/10.48550/arXiv.2105.15203
31. Guangjun X., Yucheng S., Yang Y., Huarong X., Wenhong X., Jingyuan L., Xiayan L., Yu L., Changming D. Recent developments in ai-based oceanic eddy identification. Journal of Marine Sciences, 2024, vol. 42, no. 3, pp. 38–50.
