Monocular Computer Vision-Based Simultaneous Pothole Segmentation and 3-D Volume Prediction Using 3DpredicNet
IEEE Transactions on Industrial Informatics · 2026-11 · 22 (11) · 10138-10149
DOI: 10.1109/tii.2026.3706348 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Industrial Informatics, 2026-11. Verfasst von Rahmat Ali und 4 weiteren Autor:innen. Die Arbeit nennt 51 Literaturangaben. Gefördert von: Research Manitoba Innovation Proof-of-Concept, CFI JELF, NSERC Discovery Horizons.
Abstract (Originalsprache)
A new 3DpredicNet is proposed to segment potholes and predict their 3-D volume using monocular computer vision (red, green, and blue (RGB) images) simultaneously. The network produces pixelwise segmentation masks and depth maps from a single RGB image. These outputs are combined with camera parameters such as height, viewing angle, focal length, and sensor dimensions using geometric formulas to calculate pothole volume. This approach allows quantitative damage assessment without specialized depth sensors. To develop a state-of-the-art deep learning network, various advanced modules were extensively modified and integrated, including convolution blocks, residual intensive convolution modules, depthwise separable convolution blocks, dual attention modules, downsampling and upsampling operations, and residual criss-cross attention modules. To train and test the 3DpredicNet, a dataset containing 440 images named Pothole440 was developed by collecting RGB images ($6000 \times 4000 \times 3$), and its corresponding 3-D scan data collected were used as the ground truth of pothole 3-D depth. The 3DpredicNet outperformed other recent networks, such as UNet, Deeplabv3+, AttentionUNet, MobileNetV3, YOLOv11, U-Mamba, ScSegamba, and SegResNet, for the pothole segmentation and its depth prediction on the Pothole440 dataset. 3DpredicNet achieved an average error of 0.036 across absolute relative error, squared relative error, root-mean-squared error, and root-mean-squared logarithmic error metrics, outperforming all the other networks by 37.93%. 3DPredicNet demonstrated strong segmentation performance, achieving the highest average across all evaluation metrics—including precision, recall, accuracy, F1-score, and mean intersection over union—although its accuracy was slightly lower than that of a competing method. Furthermore, its processing speed of 44.50 frames per second (FPS) comfortably meets the requirements for real-time video processing.
Förderung
- Research Manitoba Innovation Proof-of-Concept
- CFI JELF
- NSERC Discovery Horizons
Zitieren
APA
Ali, R., Saeed, Q. B., Buyukozturk, O., Lee, S., & Cha, Y. (2026). Monocular Computer Vision-Based Simultaneous Pothole Segmentation and 3-D Volume Prediction Using 3DpredicNet. IEEE Transactions on Industrial Informatics, 22(11), 10138-10149. https://doi.org/10.1109/tii.2026.3706348
Vancouver
Ali R, Saeed QB, Buyukozturk O, Lee S, Cha Y. Monocular Computer Vision-Based Simultaneous Pothole Segmentation and 3-D Volume Prediction Using 3DpredicNet. IEEE Transactions on Industrial Informatics. 2026;22(11):10138-10149. doi:10.1109/tii.2026.3706348
BibTeX
@article{Ali2026,
author = {Ali, Rahmat and Saeed, Qasim Bin and Buyukozturk, Oral and Lee, SangHyun and Cha, YoungJin},
title = {Monocular Computer Vision-Based Simultaneous Pothole Segmentation and 3-D Volume Prediction Using 3DpredicNet},
journal = {IEEE Transactions on Industrial Informatics},
year = {2026},
volume = {22},
number = {11},
pages = {10138--10149},
doi = {10.1109/tii.2026.3706348}
}
EUVIMED