NEURAL NETWORK-BASED DETERMINATION OF COARSE-GRAINED MATERIAL SIZE IN DEBRIS-FLOW DEPOSITS OF THE SOUTHERN BAIKAL REGION

This study is devoted to automation of the assessment of the dimensional structure of coarse-grained material on orthophotomaps obtained from unmanned aerial vehicles (UAVs). Orthomosaics with a Ground Sample Distance (GSD) of ~2–4 cm were used on the material of ten debris-flow deposits of the Baikal region; an annotated dataset was prepared in SedimentAlyzer. The algorithm combines object detection by computer vision methods with subsequent classification and Feret diameter estimation using convolutional neural networks (CNN; architectures: residual convolutional network (ResNet) and densely connected convolutional network (DenseNet), transfer learning). The output is performed on overlapping cells; preliminary determinations are merged (non-maximum suppression, NMS / weighted boxes fusion, WBF), probabilities are calibrated (temperature scaling) and shape filters are applied. The implementation is integrated into SedimentAlyzer. The software is registered (state registration certificate for computer program No. 2025616929). The following values were achieved in the retained areas: F1-score = 0.84±0.03; mean absolute Feret error = 4.8 cm; RMSE=7.9 cm; correlation with field measurements R=0.89; agreement with GOST size classes averaged 82 % across basins. Compared to manual interpretation, labor time is reduced by a factor of 6–8. Typical patterns include dominance of clasts 10–100 cm (≈30–40 %) and a variable share of boulders >200 cm (1–13 % with maxima in small tributaries); in transit zones, the proportion of very large fragments is higher than in accumulation zones. The approach enables reproducible mapping of fractions and is suitable for regular monitoring of debris-flow cones and support of engineering decisions.

1. Agafonov B.P., Makarov S.A., 1996. Debris Flow Areas in Pribaikalye. Geoecology. Engineering Geology. Hydrogeology. Geocryology 2, 65–71 (in Russian)

2. Agafonov B.P., Rogozin A.A., 1987. Quantitative Assessment of Debris Flow Drift Into Lake Baikal. In: Problems of Geology and Paleogeography of Siberia and the Far East. ISU Publishing House, Irkutsk, p. 53–63 (in Russian)

3. Aref’eva E.V., Babusenko M.S., Baryshev E.M., Bobreshov D.A., Boreiko V.Ya., Vereskun A.V., Glebov V.Yu., Efimova A.A. et al., 2017. Protecting Populations and Territories in Emergencies Under Current Challenges and Threats: A Reference Manual. VNII GOCHS, Moscow, 452 p. (in Russian)

4. Bezgodova O.V., Istomina E.A., Ovchinnikova E.V., 2018. Assessment of Hazardous Exogenous Processes of the Mondy Depression Based on Morphometric and Landscape Analysis. Geodesy and Cartography 79 (8), 28–37 (in Russian) https://doi.org/10.22389/0016-7126-2018-938-8-28-37.

5. Davis J., Goadrich M., 2006. The Relationship Between Precision-Recall and ROC Curves. In: W.W. Cohen, A. Moore (Eds), ICML’06. Proceedings of the 23rd International Conference on Machine Learning (June 25–29, 2006, Pittsburgh, Pennsylvania, USA). Association for Computing Machinery, New York, p. 233–240. https://doi.org/10.1145/1143844.1143874.

6. Deng J., Dong W., Socher R., Li L.J., Li K., Fei-Fei L., 2009. ImageNet: A Large-Scale Hierarchical Image Database. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (June 20–25, 2009, Miami, FL, USA). IEEE, p. 248–255. https://doi.org/10.1109/CVPR.2009.5206848.

7. Fawcett T., 2006. An Introduction to ROC Analysis. Pattern Recognition Letters 27 (8), 861–874. https://doi.org/10.1016/j.patrec.2005.10.010.

8. Fedorov Yu.A., Kyul E.V., Dzhappuev D.R., 2012. Features of Debris Flow Formation Under Complex Geological and Hydrological Conditions. Bulletin of Higher Education Institutes. North Caucasus Region. Natural Sciences 2 (168), 101–104 (in Russian)

9. Gal Y., Ghahramani Z., 2016. Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning. In: M.F. Balcan, K.Q. Weinberger (Eds), ICML’16. Proceedings of the 33rd International Conference on Machine Learning (June 20–22, 2016, New York, USA). Vol. 48. JMRL, p. 1050–1059.

10. Girshick R., 2015. Fast R-CNN. In: Proceedings of the IEEE Conference on Computer Vision (December 7–13, 2015, Santiago, Chile). IEEE, p. 1440–1448. https://doi.org/10.1109/ICCV.2015.169.

11. GOST 25100-2020, 2020. Grounds. Classification. Standartinform, Moscow, 38 p. (in Russian)

12. Guo C., Pleiss G., Sun, Y., Weinberger K.Q., 2017. On Calibration of Modern Neural Networks. In: D. Precup, Ye.W. Teh (Eds), ICML’17. Proceedings of the 34th International Conference on Machine Learning (August 6–11, 2017, Sydney, Australia). Vol. 70. JMRL, p. 1321–1330.

13. He K., Zhang X., Ren S., Sun J., 2016. Deep Residual Learning for Image Recognition. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (June 27–30, 2016, Las Vegas, Nevada, USA). IEEE, p. 770–778. https://doi.org/10.1109/CVPR.2016.90.

14. Huang G., Liu Z., Van Der Maaten L., Weinberger K.Q., 2016. Densely Connected Convolutional Networks. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (July 21–26, 2017, Honolulu, HI, USA). IEEE, p. 4700–4708. https://doi.org/10.1109/CVPR.2017.243.

15. Ioffe S., Szegedy C., 2015. Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift. In: F. Bach, D. Blei (Eds), ICML’15. Proceedings of the 32nd International Conference on Machine Learning (July 7–9, 2015, Lille, France). Vol. 37. JMRL, p. 448–456.

16. Ivanko A.F., Ivanko M.A., Sizova Yu.A., 2019. Neural Networks: General Technological Characteristics. Scientific Review. Technical Sciences 2, 17–23 (in Russian)

17. H., 2016. Debris Flow Event of 2014 and Its Impact on the Accumulation of the Solid Fraction in the Kyngarga River Channel, Tunka Valley, Southwestern Cisbaikalia, Russia. Geodynamics & Tectonophysics 7 (2), 329–335. https://doi.org/10.5800/GT-2016-7-2-0210.

18. Kedich A.I., Golosov V.N., Kharchenko S.V., 2022. Exogenous Processes in Proglacial Mountain Zones: Quantitative Assessments and Their Accuracy. Proceedings of Kazan University. Natural Sciences Series 164 (1), 109–134 (in Russian) https://doi.org/10.26907/2542-064X.2022.1.109-134.

19. Kharchenko S.V., Fedin A.V., Golosov V.N., 2021. Denudation Rates in the Mountain Periglacial Regions: Research Methods and Results. Geomorphology and Paleogeography 52 (1), 3–18 (in Russian) https://doi.org/10.31857/S0435428121010065.

20. Kingma D.P., Ba J., 2015. Adam: A Method for Stochastic Optimization. In: Proceedings of the 3rd International Conference on Learning Representations (May 7–9, 2015, San Diego). https://doi.org/10.48550/arXiv.1412.6980.

21. Laperdin V.K., Levi K.G., Lekhatinov A.M., Kadetova A.V., Pellinen V.A., Rybchenko A.A., 2014. Causes and Consequences of a Catastrophic Mudflows on 28 June 2014 near Arshan Village in the Republic of Buryatia, Russia. Geodynamics & Tectonophysics 5 (3), 799–816 (in Russian) https://doi.org/10.5800/GT-2014-5-3-0156.

22. Laperdin V.K., Rybchenko A.A., 2016. Estimated Parameters of Mudflow Forming Components of the Environment of the South of Lake Baikal. Sustainable Development of Mountain Territories 8 (1), 52–58 (in Russian) https://doi.org/10.21177/1998-4502-2016-8-1-52-57.

23. Lin T.Y., Goyal P., Girshick R., He K., Dollár P., 2017. Focal Loss for Dense Object Detection. In: Proceedings of the IEEE International Conference on Computer Vision (October 22–29, 2017, Venice, Italy). IEEE, p. 2999–3007. https://doi.org/10.1109/ICCV.2017.324.

24. Loginov G.N., Petrov A.M., 2019. Automatic Detection of Geoelectric Boundaries According to Lateral Logging Sounding Data by Applying a Deep Convolutional Neural Network. Russian Geology and Geophysics 60 (11), 1319–1325. https://doi.org/10.15372/RGG2019134.

25. Makarov S.A., 2012. Debris-Flows of Pribaikalje. Publishing House of the Institute of Geography SB RAS, Irkutsk, 111 p. (in Russian).

26. Mamedov J.H., 2013. Gradation Structure of Flow Deposits and Alluvial Fans and Their Regularities (Based on Torrents of Azerbaijani Part of the Greater Caucasus). Geographical Bulletin 4 (27), 40–48 (in Russian)

27. Mikhailov P.G., Lomtev E.A., Analiyeva A.U., Gusmanova M.S., 2016. Development of the Concept of Creation and Functioning of Monitoring Mudflow in the Mountainous and Foothill Areas. Measuring. Monitoring. Management. Control 1 (15), 15–23 (in Russian)

28. Nosov K.N., 2010. Parameters of Mudflows of the River Baksan Basin. Prirodoobustrojstvo 4, 50–55 (in Russian)

29. Oliferov A.N., 2004. Patterns of Debris Flow Formation in Crimea and the Carpathians. Scientific Notes of the V.I. Vernadsky Taurida National University. Geography Series 171 (4), 66–72 (in Russian)

30. Puritskis Ya.V., Vershinin V.E., 2023. Development of the Concept of a Multiphase Flowmeter. Tyumen State University Herald. Physical and Mathematical Modeling. Oil, Gas, Energy 9 (4 (36)), 76–90 (in Russian) https://doi.org/10.21684/2411-7978-2023-9-4-76-90.

31. Rybchenko A.A., Kadetova A.V., Kozyreva E.A., Yuriev A.A., 2019. Experience of Using Non-Specialized Unmanned Aerial Vehicles for Aerial Surveys in the Studies of Exogenous Geological Processes. Geodynamics & Tectonophysics 10 (4), 1045–1058 (in Russian) https://doi.org/10.5800/GT-2019-10-4-0457.

32. Srivastava N., Hinton G., Krizhevsky A., Sutskever I., Salakhutdinov R., 2014. Dropout: A Simple Way to Prevent Neural Networks from Overfitting. The Journal of Machine Learning Research 15 (1), 1929–1958.

33. Stepanov B.S., Stepanova T.S., 1991. Debris Flow Mechanics: Experiment, Theory, and Calculation Methods. Gidrometeoizdat, Moscow, 379 p. (in Russian)

34. Strom A.L., 2022. Grain‑Size Composition and Structure of Stony‑Avalanche Deposits as a Key to Understanding Their Motion Mechanism. In: Sergeev Readings. Fundamental and Applied Issues of Modern Soil Science: Proceedings of the Annual Session of the RAS Scientific Council on Problems of Geoecology, Engineering Geology and Hydrogeology (March 31 – April 1, 2022). GeoInfo, Moscow, p. 211–217 (in Russian)

35. Stupin V.P., Plastinin L.A., Olzoev B.N., 2020. Use of Remote Sensing Data for Medium-Scale Geoinformation Mapping of Debris-Flow Hazard in the Baikal Mountain Country. In: S.S. Chernomorets, K.S. Viskhadzhieva (Eds), Debris Flows: Disasters, Risk, Forecast, Protection. Proceedings of the 6th International Conference (Dushanbe – Khorog, Tajikistan). Vol. 1. "Promotion" LLC, Dushanbe, p. 541–552 (in Russian) [

36. Velikanova L.I., 2014. Short-Term Forecasting of Debris Flow Probability Based on Neural Network Technologies. Problems of Automation and Control 1 (26), 47–56 (in Russian)

37. Yosinski J., Clune J., Bengio Y., Lipson H., 2014. How Transferable Are Features in Deep Neural Networks? In: Z. Ghahramani, M. Welling, C. Cortes, N. Lawrence, K.Q. Weinberger (Eds), Advances in Neural Information Processing Systems. Proceedings of the 28th Annual Conference on Neural Information Processing Systems (December 8–13, 2014, Montreal, Canada). Vol. 27. NeurlPS, p. 3320–3328. https://doi.org/10.48550/arXiv.1411.1792.