Automated post-earthquake pavement damage detection using deep learning on aerial digital images
DOI:
https://doi.org/10.21754/tecnia.v35i2.2512Palabras clave:
Earthquake damage, Pavement, Deep Learning, PhotogrammetryResumen
Perú is located in a highly seismic zone, making it vulnerable to infrastructure damage caused by earthquakes. For this reason, an early evaluation after a severe earthquake is critical for mitigating impacts, particularly for decision-makers. Pavements often sustain significant damage among the most affected infrastructure, leading to crack formation. These cracks not only disrupt transportation networks but also pose safety hazards and hinder economic activities. Identifying pavement cracks is an important step in post-earthquake assessment; however, traditional inspection methods are typically slow, error-prone, labor-intensive, and often inaccessible in high-risk zones, limiting their effectiveness. This study applies deep learning techniques for automated pavement crack detection in post-earthquake scenarios using aerial images to address this issue. A DeepLabv3+ convolutional neural network was trained on 5600 labeled pavement crack images. The model achieved a segmentation Intersection Over Union (IoU) of 65%. Finally, it was applied to aerial photogrammetry data from the 2007 Pisco earthquake, where a comparison with visual inspection methods demonstrated the potential to improve post-earthquake infrastructure assessments significantly.
La detección de grietas en pavimentos es un paso clave en este proceso; sin embargo, los métodos de inspección tradicionales suelen ser lentos, propensos a errores, demandan un alto esfuerzo manual y, en muchas ocasiones, son inaccesibles en zonas de alto riesgo, lo que limita su efectividad. Para abordar este problema, en este estudio se aplican técnicas de aprendizaje profundo para la detección automatizada de grietas en pavimentos a partir de imágenes aéreas en escenarios post-sismo.
Se entrenó una red neuronal convolucional DeepLabV3+ con aproximadamente 5,600 imágenes etiquetadas de grietas en pavimentos, alcanzando el modelo un IoU de 65%. Finalmente, se aplicó a datos fotogramétricos aéreos del terremoto de Pisco de 2007, donde una comparación visual demostró su potencial para mejorar significativamente la evaluación de infraestructuras tras un sismo.
Descargas
Citas
[1] F. Kadri, B. Birregah, and E. Châtelet, “The Impact of Natural Disasters on Critical Infrastructures: A Domino Effect-based Study,” J. Homel. Secur. Emerg. Manag., vol. 0, no. 0, Jan. 2014, doi: 10.1515/jhsem-2012-0077.
[2] H. McCullough, “Earthquake damage to transportation systems,” National Oceanic and Atmospheric Administration, USA, 1994.
[3] S. Karimzadeh, M. Ghasemi, M. Matsuoka, K. Yagi, and A. C. Zulfikar, “A Deep Learning Model for Road Damage Detection After an Earthquake Based on Synthetic Aperture Radar (SAR) and Field Datasets,” IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens., vol. 15, pp. 5753–5765, 2022, doi: 10.1109/JSTARS.2022.3189875.
[4] M. Sathurshan, A. Saja, J. Thamboo, M. Haraguchi, and S. Navaratnam, “Resilience of Critical Infrastructure Systems: A Systematic Literature Review of Measurement Frameworks,” Infrastructures, vol. 7, no. 5, p. 67, May 2022, doi: 10.3390/infrastructures7050067.
[5] A. Ragnoli, M. R. De Blasiis, and A. Di Benedetto, “Pavement Distress Detection Methods: A Review,” Infrastructures, vol. 3, no. 4, p. 58, Dec. 2018, doi: 10.3390/infrastructures3040058.
[6] D. Blackman, H. Nakanishi, and A. M. Benson, “Disaster resilience as a complex problem: Why linearity is not applicable for long-term recovery,” Technol. Forecast. Soc. Change, vol. 121, pp. 89–98, Aug. 2017, doi: 10.1016/j.techfore.2016.09.018.
[7] F. Yamazaki and W. Lui, “Remote sensing technologies for post-earthquake damage assessment: a case study on the 2016 kumamoto earthquake,” presented at the 6th Asia Conf. on Earthquake Engg., Asia, Sept. 2016.
[8] V. Croce and I. Zaragoza, “UAV-based 3D Photogrammetry for post-Earthquake Studies on Seismic damaged Cities – A Case Study: Castelluccio di Norcia,” presented at the International Conference on Education and Information Systems, Technologies and Applications EISTA, Orlando, USA, July 2018.
[9] N. Kerle, F. Nex, M. Gerke, D. Duarte, and A. Vetrivel, “UAV-Based Structural Damage Mapping: A Review,” ISPRS Int. J. Geo-Inf., vol. 9, no. 1, p. 14, Dec. 2019, doi: 10.3390/ijgi9010014.
[10] V. Hoskere, Y. Narazaki, T. A. Hoang, and B. F. S. Jr, “Towards Automated Post-Earthquake Inspections with Deep Learning-based Condition-Aware Models,” Sept. 24, 2018, arXiv: arXiv:1809.09195. doi: 10.48550/arXiv.1809.09195.
[11] N. S. Gulgec, M. Takáč, and S. N. Pakzad, “Structural Damage Detection Using Convolutional Neural Networks,” in Model Validation and Uncertainty Quantification, Volume 3, R. Barthorpe, R. Platz, I. Lopez, B. Moaveni, and C. Papadimitriou, Eds., in Conference Proceedings of the Society for Experimental Mechanics Series. , Cham: Springer International Publishing, 2017, pp. 331–337. doi: 10.1007/978-3-319-54858-6_33.
[12] A. Banerji, P. Topdar, and A. Datta, “A Critical Review of Road Failures due to Major Earthquakes in Asian Continent,” presented at the Sixth International Conference on Advances in Design and Construction of Structures, Bangalore, India, Oct. 2017.
[13] P. Dondi et al., “Improving Post-Earthquake Crack Detection using Semi-Synthetic Generated Images,” 2025, pp. 19–35. doi: 10.1007/978-3-031-91907-7_2.
[14] Y. J. Lee, H. G. Jung, and J. K. Suhr, “Semantic Segmentation Network Slimming and Edge Deployment for Real-Time Forest Fire or Flood Monitoring Systems Using Unmanned Aerial Vehicles,” Electronics, vol. 12, no. 23, p. 4795, Nov. 2023, doi: 10.3390/electronics12234795.
[15] J. Alarcón, F. Taucer, and E. So, “THE 15 AUGUST 2007 PISCO, PERU, EARTHQUAKE - POST-EARTHQUAKE FIELD SURVEY,” presented at the The 14 th World Conference on Earthquake Engineering, Beiking, China, Oct. 2008.
[16] J. O’Connor, L. Mesa, and M. Nykamp, “Damage to the Highway System from the Pisco, Peru Earthquake of August 15, 2007.”
[17] S. Han, K. Phasouk, J. Zhu, and Y. Fong, “Optimizing deep learning-based segmentation of densely packed cells using cell surface markers,” BMC Med. Inform. Decis. Mak., vol. 24, no. 1, p. 124, May 2024, doi: 10.1186/s12911-024-02502-6.
[18] L. Hou, D. Samaras, T. M. Kurc, Y. Gao, J. E. Davis, and J. H. Saltz, “Patch-based Convolutional Neural Network for Whole Slide Tissue Image Classification,” Mar. 09, 2016, arXiv: arXiv:1504.07947. doi: 10.48550/arXiv.1504.07947.
[19] Y. Chen, C. Zhou, Z. Yan, T. Huang, G. Wang, and J. Hu, “Lightweight Semantic Segmentation Network Based on DeepLabV3+,” in 2022 International Conference on Artificial Intelligence and Computer Information Technology (AICIT), Yichang, China: IEEE, Sept. 2022, pp. 1–5. doi: 10.1109/AICIT55386.2022.9930215.
Descargas
Publicado
Cómo citar
Número
Sección
Licencia
Derechos de autor 2025 TECNIA
Esta obra está bajo una licencia internacional Creative Commons Atribución 4.0.
Los artículos publicados por TECNIA pueden ser compartidos a través de la licencia pública internacional Creative Commons: CC BY 4.0. Permisos lejos de este alcance pueden ser consultados a través del correo tecnia@uni.edu.pe
