Answers are grounded in this paper and cite their sources.
Proposed method shows 96.89% accuracy in assessing compaction in rockfill roadbeds, utilizing image analysis and neural networks.
Compaction serves as a fundamental criterion for evaluating the structural integrity and performance of rockfill roadbeds. To facilitate an efficient and non-invasive assessment of compaction, this paper introduces a rapid, non-destructive, and image-based methodology. The proposed approach capitalizes on the strong correlation between the void ratio derived from surface imagery and the three-dimensional void content of coarse aggregates (VCA). Initially, greyscale image analysis was utilized to extract salient image features pertinent to compaction. Subsequently, a fully connected neural network (FCNN) was deployed to encode both image-based and construction-related features into vectors of uniform dimensionality. Ultimately, a refined convolutional neural network (CNN) was leveraged to model the intricate relationship between the extracted feature vectors and compaction levels. To empirically validate the efficacy and robustness of the proposed framework, controlled laboratory experiments were conducted on 173 samples, culminating in a remarkable accuracy of 96.89%. The experimental findings substantiate the effectiveness of this approach and underscore the critical role of its underlying principles in advancing image-based compaction assessment for rockfill roadbeds.
No takes yet. Share an insight, caveat, or question.
Wu et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: