Answers are grounded in this paper and cite their sources.
Inherent trade-offs exist between data transfer efficiency and recognition accuracy, the latter being a critical factor for reliable structural damage recognition and condition assessment. Our previous work systematically investigated deep learning-based compression and reconstruction of damage images, with a particular focus on the relationship between compression efficiency and reconstruction accuracy. Building upon this foundation, this study proposes a novel compressive sensing approach for structural damage images, with a primary focus on investigating the influence of reconstructed images under different compression ratios on refined crack identification and the extraction of multiple damage indicators. Experimental results demonstrate significant differences between original images and nine types of compressed–reconstructed images, while showing remarkable similarity among the nine compressed–reconstructed variants despite their varying compression ratios. To further extract crack metrics, we developed a spline interpolation-based maximum inscribed circle algorithm capable of measuring crack width at subpixel precision. Evaluation results indicate that compared to total crack length and area measurements, the compression-reconstruction process has the least impact on the quantitative accuracy of maximum crack width. This paper provides comprehensive analysis of how the compression-reconstruction process influences precise crack identification.
No takes yet. Share an insight, caveat, or question.
Zhang et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: