Answers are grounded in this paper and cite their sources.
Computational study demonstrates reliable defect segmentation and calibrated uncertainty quantification in underwater bridges, highlighting specific environmental interference mechanisms.
: Underwater bridge inspection is often degraded by environmental interference such as turbidity and illumination distortion, resulting in unreliable defect predictions and potentially misleading decisions. To improve reliability, an uncertainty-aware Bayesian segmentation framework integrating uncertainty quantification and attribution analysis is proposed for underwater bridge defect detection. The framework quantifies epistemic and aleatoric uncertainties to evaluate prediction reliability under underwater interference. An attribution-based interpretation strategy was further developed to identify physical sources of uncertainty and investigate the effects of environmental interference on detection reliability. Experiments on mixed underwater datasets demonstrated that the framework achieved high segmentation MIoU while providing well-calibrated uncertainty estimation. Attribution analysis reveals relationships between environmental interference and defect detection reliability: water turbidity and biological attachments primarily affect exposed rebar detection, whereas rough concrete textures predominantly interfere with spalling identification. The proposed framework enhances uncertainty-aware underwater inspection by providing interpretable reliability information, supporting maintenance decisions based on risk assessment.
No takes yet. Share an insight, caveat, or question.
Sun et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: