Debris Detection from Drone Imagery in Earthquake Zones Using Artificial Intelligence

Öz

In this study, a debris detection system is proposed using the YOLOv8 deep learning algorithm. The system is trained on a custom dataset composed of drone footage captured following the 2023 Hatay-Maraş earthquake, supplemented with general debris images. Aimed at enhancing the efficiency of post-earthquake search and rescue operations, the system leverages high-resolution drone imagery to provide real-time detection capabilities. The dataset was annotated using the Roboflow platform, enriched through various data augmentation techniques, and the model training process was optimized accordingly.The effectiveness of the model was evaluated both through performance metrics on the validation dataset and in a physical test environment simulating real-world conditions. A high-resolution image of a debris scene, printed at 2×3 meters, was placed on the ground and scanned by a drone. The YOLOv8 model successfully identified the debris regions in real-time. In addition, the model was tested by capturing images using a camera module connected to a Raspberry Pi 4B device. The captured images were processed separately, and the model successfully performed debris detection despite the limited computational capacity.These results demonstrate that the proposed system, with its portable and energy-efficient structure, can be practically deployed in disaster zones. Overall, the findings indicate that the YOLOv8-based model delivers high performance in terms of both detection accuracy and operational feasibility, revealing strong potential for integration into disaster management systems.

Anahtar Kelimeler

• Debris
• YOLO
• Image Processing

Yapay Zeka Kullanılarak Deprem Bölgelerinden Drone ile Alınan Görüntülerden Enkaz Tespitinin Gerçekleştirilmesi

Öz

Bu çalışmada YOLOv8 derin öğrenme algoritması kullanılarak enkaz tespitine yönelik bir sistem önerilmektedir. 2023 yılında meydana gelen Hatay-Maraş depremi sonrasında drone ile elde edilen görüntüler ile birlikte genel enkaz görüntülerinin de eklenmesiyle oluşturulan bir veri seti kullanılmıştır. Deprem sonrası arama-kurtarma operasyonlarının etkinliğini artırmayı hedefleyen bu sistem, yüksek çözünürlüklü drone görüntülerinden elde edilen verilerle eğitilmiş ve gerçek zamanlı tespit kabiliyeti sunmaktadır. Çalışma kapsamında oluşturulan veri setleri Roboflow platformu ile etiketlenmiş, çeşitli veri artırma teknikleri uygulanmış ve model eğitim süreci optimize edilmiştir. Modelin etkinliği, hem doğrulama veri kümesindeki performans metrikleriyle hem de saha koşullarına benzer şekilde oluşturulan fiziksel bir test ortamında gözlemlenmiştir. 2×3 metre boyutlarında basılmış enkaz görüntüsünün zemin üzerine yerleştirilerek drone ile taranması sonucunda modelin başarıyla tespit yaptığı görülmüştür. Ayrıca, YOLOv8 modeli Raspberry Pi 4B cihazına bağlı kamera modülü ile görüntü alındıktan sonra bu görüntüler işlenmiş ve düşük donanım kapasitesine rağmen model, enkaz tespitini başarıyla gerçekleştirmiştir. Bu durum, geliştirilen sistemin taşınabilir ve enerji verimli yapısıyla afet sahalarında pratik olarak kullanılabileceğini göstermektedir. Elde edilen bulgular, YOLOv8 tabanlı sistemin hem doğruluk hem de uygulama açısından enkaz tespiti görevinde başarılı sonuçlar verdiğini ve afet yönetimi süreçlerine entegre edilebilecek potansiyele sahip olduğunu ortaya koymaktadır.

Anahtar Kelimeler

• Enkaz
• YOLO
• Görüntü İşleme

Kaynakça

1. Abdi G, Jabari S (2021). A multi-feature fusion using deep transfer learning for earthquake building damage detection. Canadian Journal of Remote Sensing, 47(2): 337–352.
2. Fernandez Galarreta J, Kerle N, Gerke M (2015). Uav-based urban structural damage assessment using object-based image analysis and semantic reasoning. Natural Hazards and Earth System Sciences, 15(6): 1087–1101.
3. Fujita A, Sakurada K, Imaizumi T, Ito R, Hikosaka S, Nakamura R (2017). Damage detection from aerial images via convolutional neural networks. IEEE IAPR international conference on machine vision applications (MVA), 5–8
4. Hasanlou, M., Shah-Hosseini, R., Seydi, S. T., Karimzadeh, S., & Matsuoka, M. (2021). Earthquake damage region detection by multitemporal coherence map analysis of radar and multispectral imagery. Remote Sensing, 13(6), 1195.
5. Hong Z, Zhong H, Pan H, Liu J, Zhou R, Zhang Y, Han Y, Wang J, Yang S, Zhong C (2022). Classification of building damage using a novel convolutional neural network based on post-disaster aerial images. Sensors, 22(15): 5920.
6. Joshi AR, Tarte I, Suresh S, Koolagudi SG (2017). Damage identification and assessment using image processing on post-disaster satellite imagery. IEEE Global Humanitarian Technology Conference (GHTC).
7. Kalantar B, Ueda N, Al-Najjar HA, Halin AA (2020). Assessment of convolutional neural network architectures for earthquake-induced building damage detection based on pre-and post-event orthophoto images. Remote Sensing, 12(21): 3529.
8. Khodaverdi Zahraee N, Rastiveis H (2017). Object-oriented analysis of satellite images using artificial neural networks for post-earthquake buildings change detection. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, 42: 139–144.

9. Kıratlı R, Eroğlu A (2024). Monitoring Post-Earthquake Search and Rescue Operations through UAVs Vision: Teams, Equipment and Structural Damage Detection. 8th International Artificial Intelligence and Data Processing Symposium (IDAP'24), 21–22 Eylül 2024, Malatya, Türkiye. IEEE.
10. Rao A, Jung J, Silva V, Molinario G, Yun SH (2023). Earthquake building damage detection based on synthetic-aperture-radar imagery and machine learning. Natural Hazards and Earth System Sciences, 23(2): 789–807.
11. Ünlü R, Kiriş R (2022). Detection of damaged buildings after an earthquake with convolutional neural networks in conjunction with image segmentation. The Visual Computer, 38(2): 685–694.
12. Xiong C, Li Q, Lu X (2020). Automated regional seismic damage assessment of buildings using an unmanned aerial vehicle and a convolutional neural network. Automation in Construction, 109: 102994.
13. Zhang J, Gong L (2013). Sar images before and after earthquake change detection based on object oriented method and damage evaluation. IEEE International Geoscience and Remote Sensing Symposium-IGARSS.