Surface cracks are crucial for structural health monitoring of various types of buildings. Despite substantial advancements in crack detection through deep neural networks, their reliance on pixel-level crack annotation escalates labeling costs and renders the labeling procedure time-intensive. Consequently, academics have suggested multiple Explainable Artificial Intelligence (XAI) methodologies to enhance the efficacy of pseudo-labeling. However, fractures’ slender, continuous, and inconspicuous characteristics render current XAI approaches ineffective in adequately gathering feature information. This work examines the characteristics of many XAI strategies through extensive experimentation. It synthesizes the advantages of each strategy to mitigate the uncertainty error associated with a singular model in the fracture region. Moreover, we formulate and implement various integration strategies to mitigate and enhance the discrepancies across distinct XAI algorithms across two separate datasets. The experimental results indicate that the proposed method provides more accurate basic annotations for weakly supervised crack segmentation.
25 September 2025 in Measurement Science and Technology
Detection and segmentation of concrete cracks are critical tasks in structural health monitoring. Although existing fully-supervised learning methods have made significant progress, data acquisition and precise annotation remain costly and challenging, especially in the continuous monitoring of large-scale infrastructure. The high cost and inconsistency of annotations have become limiting factors. To reduce reliance on manual labeling, weakly-supervised semantic segmentation using image-level labels has emerged as an active research focus. However, the pseudo-labels generated by these methods generally exhibit lower quality compared to fully-supervised approaches. To bridge this gap, this study proposes FGR-Grad, a weakly-supervised semantic segmentation method based on the FullGrad algorithm. The approach incorporates a gradient correction strategy (via guided backpropagation (GBP) or rectified backpropagation (RBP) mechanisms) to refine gradient propagation: On one hand, it introduces GBP layer-wise within the FullGrad algorithm, replacing raw gradients with guided gradients to prioritize the propagation of crucial gradient information towards important network parameters, thereby enhancing the model’s focus on key features. On the other hand, it employs the RBP mechanism to prune redundant information in FullGrad saliency maps. This replaces raw gradients with rectified gradients filtered by importance scores, blocking information flow from irrelevant features and eliminating redundancies to highlight the most discriminative target regions. Experimental validation on three public datasets demonstrates that the FGR-Grad algorithm significantly narrows the performance gap between weakly-supervised and fully-supervised methods, with its inference accuracy and visual results outperforming existing weakly-supervised techniques.
23 March 2026 in Pattern Analysis and Applications
To support campus security governance, especially campus traffic safety management, many ternary extraction techniques in knowledge graphs rely on character-level text analysis; however, the differences between word semantics and overall word meanings often lead to ambiguity and overlap in ternary extraction. Traditional methods are difficult to effectively manage overlapping relationships in text, which seriously damages the flexibility and extraction accuracy of the dataset. In addition, entity separation significantly affects entity relationship triplet extraction, and weak associations between remote entities often blur entity boundaries and reduce recall rates. This study offers the MGRel entity relationship extraction model, which integrates dependent syntactic analysis with a graph neural network to address the issues above and enhance knowledge acquisition for campus security scenarios. Firstly, by incorporating the dependent syntactic parser alongside the dual analysis mechanism of global semantic dependency and syntactic dependency, it effectively captures long-distance semantic associations and enhances entity relationship recognition accuracy; secondly, it devises the architecture of a hierarchical semantic graph convolutional neural network to facilitate the fine-grained extraction of deep implied semantic features among entities; finally, the attention-driven multi-feature fusion module is presented to improve the discriminative capacity of the ternary classifier via a noise filtering approach. The experimental results on three core general-purpose benchmark datasets—NYT, WebNLG, and DuIE—show that the F1 score of this model increases by 1.3, 0.4 and 3.2%, respectively, compared with the current optimal model, demonstrating a considerable advantage over the comparative techniques and potential value for campus security–oriented campus traffic safety applications.
12 November 2026 in Pattern Analysis and Applications