ВЫЧИСЛИТЕЛЬНЫЙ МЕТОД БИНАРНОЙ СЕМАНТИЧЕСКОЙ СЕГМЕНТАЦИИ ДЕФЕКТОВ СТРОИТЕЛЬНЫХ КОНСТРУКЦИЙ НА ОСНОВЕ СВЕРТОЧНЫХ НЕЙРОННЫХ СЕТЕЙ
ВЫЧИСЛИТЕЛЬНЫЙ МЕТОД БИНАРНОЙ СЕМАНТИЧЕСКОЙ СЕГМЕНТАЦИИ ДЕФЕКТОВ СТРОИТЕЛЬНЫХ КОНСТРУКЦИЙ НА ОСНОВЕ СВЕРТОЧНЫХ НЕЙРОННЫХ СЕТЕЙ
DOI:
https://doi.org/10.22337/2587-9618-2026-22-2-176-188Ключевые слова:
компьютерное зрение, глубокое обучение, семантическая сегментация, дефекты строительных конструкций, трещины, архитектура U Net, энкодер MobileNetV2, функция потерь ДайсаАннотация
Проблема автоматизированной диагностики дефектов строительных конструкций обусловлена высокой трудоемкостью и субъективностью визуальных обследований. В статье предложен прототип системы компьютерного зрения на основе архитектуры с облегченным блоком извлечения визуальных признаков (энкодером), предназначенный для бинарной семантической сегментации трещин на изображениях строительных поверхностей. Новое предложение заключается в адаптации указанной архитектуры к задаче выделения тонких протяженных объектов на неоднородном фоне с применением комбинированной функции потерь, объединяющей бинарную кросс‑энтропию и коэффициент Дайса. Приведены математическая постановка задачи, формальное описание модифицированного критерия оптимизации, методика подготовки данных с учетом артефактов JPEG‑сжатия, а также результаты вычислительных экспериментов. Полученные в работе на независимой тестовой выборке метрики значения коэффициента Дайса и индекса Жаккара (IoU) подтверждают работоспособность предложенного подхода и определяют направления дальнейшего совершенствования технологии.
Библиографические ссылки
GOST 31937-2024. Buildings and Structures. Rules for Surveying and Monitoring Technical Condition: Interstate Standard. Moscow: Standartinform, 2024. (in Russian)
GOST 31937-2011. Buildings and Structures. Rules for Surveying and Monitoring Technical Condition: Interstate Standard. Moscow: Standartinform, 2011. (in Russian)
SP 13-102-2003. Rules for Surveying Load-Bearing Building Structures of Buildings and Structures. Moscow: Gosstroy of Russia, 2004. (in Russian)
GOST R 53778-2010. Buildings and Structures. Rules for Surveying and Monitoring Technical Condition: National Standard of the Russian Federation. Moscow: Standartinform, 2010. (in Russian)
Pichueva A.S. Methods for Surveying Buildings and Structures // Paradigm. – 2025. – No. 5-2. – pp. 355–360. – EDN BMNNNI. (in Russian)
Grushkovsky P. A., Shchelnikov V. N., Sitnikov A. V. Assessment of the Technical Condition of Building Structures and Facilities by Visual and Instrumental Methods // Bulletin of Tula State University. Technical Sciences. – 2021. – No. 6. – pp. 208–212. – DOI 10.24412/2071-6168-2021-6-208-212. – EDN KKJSSV. (in Russian)
Petrova I. Yu., Mostovoy O. O. Review of the Building and Structure Inspection Process. Problems and Solutions // Caspian Engineering and Construction Bulletin. – 2021. – No. 1 (35). – pp. 69–75. – EDN KGUCUU. (in Russian)
Bogomolov S. I. Analysis of methods for assessing the current technical condition of buildings and structures // Bulletin of Tula State University. Technical sciences. - 2021. - No. 10. - pp. 287-290. - DOI 10.24412/2071-6168-2021-10-287-290. - EDN WFCCXA. (in Russian)
Snegireva A. I., Murashkin V. G. On the issue of surveying building structures, buildings and structures // Expert: theory and practice. - 2021. - No. 6 (15). - pp. 45-51. - DOI 10.51608/26867818_2021_6_45. - EDN JKHDHA. (in Russian)
Semenov A. S. Organization of technical survey of housing stock buildings // Housing construction. – 2010. – No. 12. – pp. 23–25. – EDN NQUBOF. (in Russian)
Kunin, Yu. S. Diagnostics of a unique structure – a 100-meter-high radio tower / Yu. S. Kunin, T. V. Potapova, S. G. Muzychenko // Construction production. – 2023. – No. 2. – pp. 54–61. – DOI 10.54950/26585340_2023_2_54. – EDN QODNCJ. (in Russian)
Kunin, Yu. S. Vibroacoustic methods of quality control of building structures of buildings and structures / Yu. S. Kunin, A. S. Perunov // Industrial and civil engineering. – 2023. – No. 3. – pp. 55–61. – DOI 10.33622/0869-7019.2023.03.55-61. – EDN XRKUCN (in Russian)
Kashevarova G. G., Tonkov Yu. L. Expert system for practical diagnostics of building structures // Academia. Architecture and Construction. – 2022. – No. 2. – pp. 58-91. – DOI 10.22337/2077-9038-2022-1-85-91. (in Russian)
Kashevarova G. G., Tonkov Yu. L. Intelligent technologies in the inspection of building structures // Academia. Architecture and Construction. – 2018. – No. 1. – pp. 92-99. – DOI: 10.22337/2077-9038-2018-1-92-99. (in Russian)
Kashevarova G. G. "Artificial intelligence", or "logical reasoning and reasonable decisions" in technical diagnostics of construction projects // Academia. Architecture and Construction. – 2023. – No. 4. – pp. 166-180. – DOI: 10.22337/2077-9038-2023-4-166-180. (in Russian)
Kuksov, A. S. Prerequisites for the Development of Expert Systems in the Field of Inspection and Diagnostics of Building Structures / A. S. Kuksov, P. I. Andreeva, R. R. Kazaryan // BST: Bulletin of Construction Equipment. – 2020. – No. 5(1029). – pp. 47-49. – EDN VMHOYL (in Russian)
Shesterikov, Yu. A. Information Systems and Technologies for Inspection of Buildings and Structures / Yu. A. Shesterikov, O. A. Stifeeva // Russian journal of transport engineering. – 2020. – Vol. 7, No. 1. – P. 4. – 10.15862/05SATS120. (in Russian)
Nikolyukin, A. N. Self-learning intelligent systems in construction / A. N. Nikolyukin, P. V. Monastyrev. – Tambov: Tambov State Technical University, 2025. – 167 p. – ISBN 978-5-8265-2908-9. – EDN FTSUTJ (in Russian)
Krylov S. A., Kashevarova G. G. Automation of the process of assessing the technical condition of external walls of brick buildings using machine learning technology // Bulletin of DSTU. Technical sciences. – 2025. – No. 3. – pp. 61-70. – DOI: 10.21822/2073-6185-2025-52-3-61-70. (in Russian)
Soldatenko T. N. Model for identification and prediction of defects in a building structure based on the results of its inspection // Magazine of Civil Engineering. – 2011. – No. 7. – pp. 52-61. (in Russian)
Samigullina A. G. The need to develop a model of an expert system for diagnosing the technical condition of building structures // Young scientist. – 2017. – No. 51 (185). – pp. 88–90. – EDN ZXXDMN. (in Russian)
Bugakova T. Yu., Sharapov A. A. Improving the methods of visual inspection of buildings and engineering structures by introducing computer vision and intelligent data processing technologies // Bulletin of SSUGiT (Siberian State University of Geosystems and Technologies). – 2022. – No. 6. – pp.108–119. – DOI 10.33764/2411-1759-2022-27-6-108-119. – EDN MFBIIV. (in Russian)
Hamed G. M. An expert system for concrete diagnosis: диссертация / G. M. Hamed; King Fahd University of Petroleum and Minerals. – Dhahran, 1983.
Bernat E., Gil L. Aided diagnosis of structural pathologies with an expert system // Advances in Structural Engineering. – 2013. – Vol. 16, № 2. – pp. 379–393. – DOI 10.1260/1369-4332.16.2.379.
Moselhi O., Hegazy T., Fazio P. Neural networks as tools in construction // Journal of Construction Engineering and Management. – 1991. – Vol. 117, № 4. – pp. 606–625. – DOI 10.1061/(ASCE)0733-9364(1991)117:4(606).
Nuhu B. K. [et al.] Distributed network-based structural health monitoring expert system // Building Research & Information. – 2021. – Vol. 49, № 1. – pp. 144–159. – DOI 10.1080/09613218.2020.1854083. – EDN BKCFCV.
Arora S. Deep learning-based structural crack detection framework for structural health monitoring of reinforced concrete buildings // Asian Journal of Civil Engineering. – 2026. – pp. 1–11. – DOI 10.1007/s42107-026-01713-8.
de Brito J. [et al.] Expert knowledge-based inspection systems // Inspection, Diagnosis and Repair of the Building Envelope. – Cham: Springer, 2020. – DOI 10.1007/978-3-030-42446-6.
Ronneberger O., Fischer P., Brox T. U Net: Convolutional Networks for Biomedical Image Segmentation // Medical Image Computing and Computer Assisted Intervention – MICCAI 2015. – Springer, 2015. – pp. 234–241.
Siva Rama Krishnan S., Nalla Karuppan M.K., Khadidos A.O. et al. Inception V3 for automated binary classification of construction surface cracks // Scientific Reports. – 2025. – DOI: 10.1038/s41598-025-85983-3.
Bhowmick S., Nagarajaiah S., Veeraraghavan A. Automated crack detection and quantification from UAV videos using U-Net // Sensors. – 2020. – Vol. 20(21). – Art. 6299. – DOI: 10.3390/s20216299.
Kumar P., Batchu S., Swamy N.S., Kota S.R. Real-time structural health monitoring using multi-copter network and edge computing // IEEE Access. – 2021. – Vol. 9. – pp. 1234–1245. – DOI: 10.1109/ACCESS.2021.3102647.
Gharehbaghi V., Noroozinejad Farsangi E., Yang T.Y. et al. FastCrackNet: Identification of concrete cracks under noise and shadows using a 12-layer deep network // Sensors. – 2022. – Vol. 22(22). – Art. 8986. – DOI: 10.3390/s22228986.
Gürer B., Karslıgil M.E. Earthquake damage assessment using deep learning-based segmentation on drone imagery // Proc. SIU 2024. – IEEE, 2024. – DOI: 10.1109/SIU61531.2024.10601138.
Chen H., Song J., Dietrich O. et al. BRIGHT: A globally distributed multimodal dataset for automated building damage assessment after disasters // Earth System Science Data. – 2025. – DOI: 10.5194/essd-2025-269.
Russo L., Tapete D., Ullo S.L., Gamba P. Post-event SAR-only building damage assessment framework integrating geospatial data // arXiv preprint. – 2025. – arXiv:2506.22338.
Hoier C., Ahmed K.M. UAV video super-resolution and VLM-based building damage classification // arXiv preprint. – 2025. – arXiv:2508.17130.
Manzini T., Perali P., Murphy R.R. Operational deployment of sUAS-based AI for building damage assessment during hurricanes Debby and Helene // arXiv preprint. – 2025. – arXiv:2511.03132.
Sandler M., Howard A., Zhu M., Zhmoginov A., Chen L.C. MobileNetV2: Inverted Residuals and Linear Bottlenecks // Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR). – 2018. – pp. 4510–4520.
Khanh Ha. Crack Segmentation: [program code] / Khanh Ha. – Text: electronic // GitHub: [site]. – URL: https://github.com/khanhha/crack_segmentation
Загрузки
Опубликован
Выпуск
Раздел
Лицензия
Copyright (c) 2026 International Journal for Computational Civil and Structural Engineering
Это произведение доступно по лицензии Creative Commons «Attribution-NonCommercial-ShareAlike» («Атрибуция — Некоммерческое использование — На тех же условиях») 4.0 Всемирная.
