Deep Learning for Underwater Crack Detection: Integrating Physical Models and Uncertainty-Aware Semantic Segmentation
Open Access
- 23 September 2025
- journal article
- research article
- Published by MDPI AG in Infrastructures
- Vol. 10 (10) , 255
Abstract
Underwater crack detection is critical for ensuring the safety and longevity of submerged infrastructures, yet it remains challenging due to water-induced image degradation, limited labeled data, and the poor generalization of existing models. This paper proposes a novel deep learning framework that integrates physical priors and uncertainty modeling to address these challenges. Our approach introduces a physics-guided enhancement module that leverages underwater light propagation models, and a dual-branch segmentation network that combines semantic and geometry-aware curvature features to precisely delineate irregular crack boundaries. Additionally, an uncertainty-aware Transformer module quantifies prediction confidence, reducing the number of overconfident errors in ambiguous regions. Experiments on a self-collected dataset demonstrate State-of-the-Art performance, achieving 81.2% mIoU and 83.9% Dice scores, with superior robustness in turbid water and uneven lighting. The proposed method introduces a novel synergy of physical priors and uncertainty-aware learning, advancing underwater infrastructure inspection beyond the current data-driven approaches. Our framework offers significant improvements in accuracy, robustness, and interpretability, particularly in challenging conditions like turbid water and non-uniform lighting.Keywords
Funding Information
- National Natural Science Foundation of China (52508178, 52208471, 2025A1515010155, 2023KQNCX217)
- Guangdong Basic and Applied Basic Research Foundation (52508178, 52208471, 2025A1515010155, 2023KQNCX217)
- Youth Innovation Talent projects in Ordinary Higher education institutions (52508178, 52208471, 2025A1515010155, 2023KQNCX217)
This publication has 35 references indexed in Scilit:
- Crack detection with minimal labels: a mixed supervision approach with multiscale transformersJournal of Real-Time Image Processing, 2025
- MDCCM: a lightweight multi-scale model for high-accuracy pavement crack detectionSignal, Image and Video Processing, 2025
- Remotely operated vehicle (ROV) underwater vision-based micro-crack inspection for concrete dams using a customizable CNN frameworkAutomation in Construction, 2025
- Plug-and-play method for segmenting concrete bridge cracks using the segment anything model with a fractal dimension matrix promptAutomation in Construction, 2024
- Transformer-based semantic segmentation and CNN network for detection of histopathological lung cancerBiomedical Signal Processing and Control, 2024
- Method for detecting surface defects of underwater buildings: Binocular vision based on sinusoidal grating fringe assistanceAlexandria Engineering Journal, 2023
- Detecting large-scale underwater cracks based on remote operated vehicle and graph convolutional neural networkFrontiers of Structural and Civil Engineering, 2022
- Underwater crack pixel-wise identification and quantification for dams via lightweight semantic segmentation and transfer learningAutomation in Construction, 2022
- Uncertainty-aware consistency regularization for cross-domain semantic segmentationComputer Vision and Image Understanding, 2022
- Uncertainty quantification using Bayesian neural networks in classification: Application to biomedical image segmentationComputational Statistics & Data Analysis, 2020