«Automation and Informatization of the fuel and energy complex»

Neural networks application for semantic segmentation of pores in rock sections images

UDC: 004:553.982.23.05
DOI: -

Authors:

¹ National University of Oil and Gas "Gubkin University", Moscow, Russia

Keywords: slots, core marking, augmentation, model training, segmentation maps

Annotation:

Neural networks allow solving semantic segmentation problems. At the preliminary stage of the study, the authors of the article analyzed the image of deposits panoramas and cores, the images obtained are balanced by classes: rock; inter-form, inter-granular pores, micro-pores, cracks; intra-form pores. To augment the data, changes of brightness, contrast, color saturation, color halftones and sharpness were used. The authors of the article used the UNet, UNet++ and MANet models as the selected neural network architectures. To improve the accuracy of neural networks, the ResNet-34 model was used as an encoder. Stochastic gradient descent and error back propagation algorithms were used to train the neural network. To determine the quality of the model, the authors chose the Jaccard coefficient. The weights of the models were preserved when their highest arithmetic mean was reached for all types of pores. Analysis of the results of the built neural network models operation showed that they identify voids well, but do not cope well with their classification.

Bibliography:

1. Sravnitel'nyy analiz sushchestvuyushchikh metodov otsenki kollektorskikh svoystv porod prizaboynoy zony / R.A. Akhtyamov, I.R. Safiullin, R.R. Zaripov, V.S. Khakimov // Karotazhnik. – 2008. – № 10(175). – S. 37–47.
2. Shestakov V.V. Metod i algoritmy obrabotki informatsii dlya postroeniya ob"emnykh modeley geologicheskoy sredy: dis. … kand. tekhn. nauk: 2.3.1. – Tomsk, 2023. – 148 s.
3. Shatyrov A.K. Otsenka geomekhanicheskikh svoystv kollektorov Prisakhalinskogo shel'fa po rezul'tatam modelirovaniya // Izv. vuzov. Geologiya i razvedka. – 2024. – T. 65, № 6. – S. 53–65. – DOI: 10.32454/0016-7762-2023-65-6-53-65
4. Nikonov A.I. Tekhnogennye faktory vozdeystviya na produktivnye otlozheniya mestorozhdeniy uglevodorodov i ikh svyaz' s potentsial'nym osedaniem zemnoy poverkhnosti // Aktual'nye problemy nefti i gaza. – 2020. – № 4(31). – S. 77–92. – DOI: 10.29222/ipng.2078-5712.2020-31.art5
5. Shubin A.V., Klyazhnikov D.V., Ryzhkov V.I. Petrouprugaya inversiya: prognoz kollektorskikh svoystv tyumenskoy svity // Geofizika. – 2018. – № 6. – S. 2–10.
6. Rezul'taty izucheniya kollektorskikh svoystv porod metodom identifikatsionnogo modelirovaniya / O.M. Muravina, A.A. Auzin, V.N. Glaznev, I.A. Ponomarenko // Karotazhnik. – 2018. – № 8(290). – S. 71–80.
7. Potekhin D.V., Galkin S.V. Primenenie tekhnologii mashinnogo obucheniya pri modelirovanii raspredeleniya litotipov na permokarbonovoy zalezhi nefti Usinskogo mestorozhdeniya // Zap. Gornogo in-ta. – 2023. – T. 259. – S. 41–51. – DOI: 10.31897/PMI.2022.101
8. Tupysev A.M. Fraktal'nyy analiz tsifrovykh izobrazheniy obraztsov gornykh porod // Avtomatizatsiya, telemekhanizatsiya i svyaz' v neftyanoy prom-sti. – 2017. – № 1. – S. 11–15.
9. Kitaev A.P., Sineva I.S. Primenenie metodov mashinnogo obucheniya dlya analiza izobrazheniy v zadachakh neftegazovoy litologii // Telekommunikatsii i inform. tekhnologii. – 2022. – T. 9, № 1. – S. 116–126.
10. Obzor metodov semanticheskoy segmentatsii sputnikovykh snimkov Zemli s primeneniem neyrosetevykh tekhnologiy / Ya.V. Makhon'ko, A.A. Petryaeva, V.A. Podmar'kova [i dr.] // Sistemnyy analiz v proektirovanii i upravlenii: sb. nauch. tr. XXVI Mezhdunar. nauch.-prakt. konf., SPb., 13–14 okt. 2022 g.: v 3 ch. Ch. 2. – SPb.: S.-Peterb. politekhn. un-t Petra Velikogo, 2023. – S. 258–266.
11. Ronneberger O., Fischer P., Brox T. U-net: Convolutional Networks for Biomedical Image Segmentation // Medical Image Computing and Computer-Assisted Intervention – MICCAI 2015: 18th Int. Conf., Munich, Germany, Oct. 5–9, 2015: Proceedings: Part III. – Springer Int. Publishing, 2015. – P. 234–241. – DOI: 10.1007/978-3-319-24574-4_28
12. UNet++: A Nested UNet Architecture for Medical Image Segmentation / Zhou Zongwei, M.R. Siddiquee, N. Tajbakhsh, Liang Jianming // Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support: 4th Int. Workshop, DLMIA 2018, and 8th Int. Workshop, ML-CDS 2018, Held in Conjunction with MICCAI 2018, Granada, Spain, Sept. 20, 2018: Proceedings. – Springer Int. Publishing, 2018. – P. 3–11. – DOI: 10.1007/978-3-030-00889-5_1
13. Ma-net: A Multi-Scale Attention Network for Liver and Tumor Segmentation / Fan Tongle, Wang Guanglei, Li Yan, Wang Hongrui // IEEE Access. – 2020. – Vol. 8. – P. 179656–179665. – DOI: 10.1109/ACCESS.2020.3025372
14. Focal Loss for Dense Object Detection / Lin Tsung-Yi, P. Goyal, R. Girshick [et al.] // 2017 IEEE Int. Conf. on Computer Vision (ICCV), Venice, Italy, Oct. 22–29, 2017. – 2017. – P. 2999–3007. – DOI: 10.1109/ICCV.2017.324
15. Smith L.N. Cyclical Learning Rates for Training Neural Networks // 2017 IEEE Winter Conf. on Applications of Computer Vision (WACV), Santa Rosa, CA, USA, March 24–31, 2017. – IEEE, 2017. – P. 464–472. – DOI: 10.1109/WACV.2017.58
16. Deep Residual Learning for Image Recognition / He Kaiming, Zhang Xiangyu, Ren Shaoqing, Sun Jian // 2016 IEEE Conf. on Computer Vision and Pattern Recognition (CVPR), Las Vegas, NV, USA, June 27–30, 2016. – IEEE, 2016. – P. 770–778. – DOI: 10.1109/CVPR.2016.90