Pasar al contenido principal

Resultados relacionados

Resultados totales (Incluyendo duplicados): 2
Encontrada(s) 1 página(s)

Investigo. Repositorio Institucional de la Universidade de Vigo
Publicación Artículo científico (JournalArticle)

Automatic identification of pavement subsurface distresses from ground penetrating radar data using few-shot data augmentation and YOLOv11

  • Elseicy , Ahmed Mossad Ibrahim
  • Arias Sánchez, Pedro
  • Solla Carracelas, Maria Mercedes
Accurate identification of pavement distress using ground penetrating radar (GPR) data is essential for nondestructive maintenance and safety assessment of roadway structures. However, the scarcity of annotated GPR datasets limits the performance of deep learning models. This study presents a method for automated pavement distress detection using few-shot data augmentation and YOLOv11 (you only look once, version 11). A generative framework based on an improved explainable single-image generative adversarial network (ExSinGAN) is used to generate realistic GPR images from a small number of original samples. The generated data are combined with the urban roadway disease dataset (URDD) to train the YOLOv11 detector to identify subsurface distresses, such as cavities and looseness. Comparative experiments with gprMax-simulated data and field validation confirm the reliability and adaptability of the proposed approach. The results show an increase in mean average precision (mAP) from 84.69% to 89.31%, precision from 77.73% to 83.13%, and recall from 81.30% to 83.37% compared to the baseline trained on URDD alone. These findings demonstrate that few-shot generative augmentation effectively mitigates data scarcity and improves the robustness of GPR-based distress detection. The developed model was further validated on field data, confirming its generalization to real-world GPR measurements. The proposed method supports non-destructive monitoring of pavement structures. It provides practical value to civil engineers by facilitating early detection of underground defects and reducing the risk of pavement collapse from sinkholes., Agencia Estatal de Investigación | Ref. TED2021-130183B-I00, Agencia Estatal de Investigación | Ref. PID2022-138526OB-I00, Agencia Estatal de Investigación | Ref. PREP2022-000030, Agencia Estatal de Investigación | Ref. CNS2024-154766
Investigo. Repositorio Institucional de la Universidade de Vigo
Publicación Artículo científico (JournalArticle)

InSAR-derived subsidence susceptibility mapping using open-access geospatial data and random forest

  • Alvarez Dieguez, Belén Lara
  • Balado Frías, Jesús
  • Solla Carracelas, Maria Mercedes
Ground instability presents an increasing challenge to the safety and resilience of transport infrastructure, particularly in relation to processes such as subsidence and sinkhole-related hazards. This study presents a reproducible, open-access machine learning framework for mapping subsidence susceptibility along the AG-64 highway in Spain, integrating satellite-based Interferometric Synthetic Aperture Radar (InSAR) data with geospatial predictors. Using a Random Forest classifier, the methodology incorporates five primary environmental variables and four engineered interaction features (specifically designed to capture the complex synergy between lithological conditions and climatic forcing). The class-weighted model achieved a high recall of 0.92 and precision of 0.83 using a conventional random 70/30-pixel split, indicating strong pixel-level discrimination in identifying subsiding areas while minimizing false negatives. To assess spatial robustness, the model was further evaluated using spatial block cross-validation based on an estimated autocorrelation range of 4.48 km. Evaluation on an independent spatial test set comprising 61 geographically excluded 5-km blocks yielded more conservative performance (ROC–AUC = 0.591), with a recall of 0.671 and precision of 0.187 at the spatially calibrated threshold. It is noteworthy that the mapped susceptibility patterns were contextually supported using Falling Weight Deflectometer (FWD) measurements and maintenance records from the official inspection campaign. Results show a statistically significant, positive and moderate association between historical pavement deflection and predicted susceptibility in road segments without documented maintenance. This workflow provides a transparent, data-driven tool for infrastructure managers to prioritize field inspections and enhance the resilience of road corridors in geohazard-prone environments., Agencia Estatal de Investigación | Ref. CNS2024-154766, Agencia Estatal de Investigación | Ref. TED2021-130183B-I00, Agencia Estatal de Investigación | Ref. RYC2022-038100-I, Agencia Estatal de Investigación | Ref. RYC2022-038100-I, Universidade de Vigo/CISUG