ISEP - DM – Engenharia Civil
URI permanente para esta coleção:
Navegar
Percorrer ISEP - DM – Engenharia Civil por assunto "3D models"
A mostrar 1 - 1 de 1
Resultados por página
Opções de ordenação
- Deep Learning-based Damage Identification for the Remote Inspection of BridgesPublication . FONSECA, JOÃO LUÍS PAULA DIAS RIBEIRO DA; Ribeiro, Diogo Rodrigo Ferreira; Santos, Ricardo Manuel Pereira; Jorge, Tomás SimõesBridges play a vital role in keeping people, goods, and economies connected. However, their constant exposure to the aggressive environment, intensified by climate change, accelerates deterioration and demands more frequent inspections. Traditional inspection methods are expensive, time-consuming, subjective, and sometimes unsafe, showing the clear need for smarter and safer alternatives. This thesis explores a remote inspection approach for reinforced concrete bridges using photogrammetry, laser scanning, deep learning, and projection techniques. A dedicated application, based on the YOLO11-seg algorithm, was developed to automatically detect common damage instances such as cracks, efflorescences, and exposed rebars. To go further, a ray-casting method was implemented to project these detections onto the photogrammetryand laser scanning–based 3D models, allowing precise mapping and measurement of structural damage. To assess the effectiveness of this approach, two case studies were conducted. The results demonstrated the complementarity of photogrammetry and laser scanning, while also highlighting the limitations of each method. Deep learning achieved promising results (mAP50 = 61%), not yet sufficient for full automation but very useful for semi-automated workflows that save time while retaining human oversight. The ray-casting projection proved particularly effective, ensuring consistent and reliable 3D mapping of the damage instances. Together, these results show that combining data acquisition, 3D reconstruction, AI-driven detection, and projection-based mapping can deliver a unique pipeline from raw data to actionable insights. In the context of rising demands for resilient infrastructure, this work offers a step toward safer, more efficient and intelligent bridge inspections.