As the service life of metro tunnels increases, the lining and associated components often sustain damage from various factors, jeopardizing operational safety and structural integrity. Traditional manual inspections and digital image processing methods fall short owing to the challenging internal environment, suboptimal lighting conditions, and the highly similar textures shared by components and tunnel linings. To address these challenges, we focus on linear array images that capture the tunnel’s inner wall and propose FESS-SAM for the semantic segmentation of all elements without the need for additional training. We introduce an innovative iterative segmentation mechanism utilizing hierarchical stacking thresholds to extract instance-level masks. It alleviates issues posed by mask redundancy, complex features, and the inherent limitations of the Segment Anything Model (SAM). Furthermore, with the establishment of the tunnel component ontology, we address the pain point of semantic deficiency by employing a random forest classifier trained on high-dimensional image features. The experimental results obtained from a dataset of 996 images show a mean Intersection over Union (mIoU) of 0.819 and an F1 score of 0.895 across six representative tunnel elements. These results indicate that FESS-SAM outperforms existing supervised segmentation models in terms of visualization and accuracy metrics. This advancement addresses the data dependency issues in complex scenarios and achieves millimeter-level precision in segmenting critical tunnel components. By innovatively introducing a cutting-edge vision foundation model into real-world infrastructure management, it offers a robust and scalable solution for large-scale, automated, and precise metro tunnel maintenance.

Due to the lack of effective attack detection measures, cyberattacks may cause strong damage to industrial cyber–physical systems (CPSs). The embedding of attack categories learned by the existing attack detection methods is highly coupled to each other with fuzzy boundaries and overlapped neighborhood, leading to weak robustness and high false positive rates. To address these issues, in this article, we propose a few-shot attack detection method based on decoupled prototype learning (DPL-FSAD), aiming to enhance the detection accuracy and generalization capabilities for malicious attacks in CPS. Specifically, we first introduce feature contrastive learning to extract differentiated features from highly similar samples, achieving compact intraclass and sparse interclass feature embedding space. To solve the problem of fuzzy boundaries of different attack categories, prototype contrastive learning is then employed to reduce the coupling degree among prototypes and enhance their discriminability. A regularization term is exploited to mitigate the overfitting problem by reducing the gap between the feature embedding and prototypes. Furthermore, an orthogonal constraint is employed to separate prototypes of different attack types, generating a decoupled prototype embedding space. The experimental results on three public cyberattack datasets show that, compared with the suboptimal model a few-shot learning model with Siamese convolutional neural network (FSL-SCNN), the proposed DPL-FSAD can improve the precision by 5.53%,F1-score by 3.3%, and reduce the false positive rate by 2.37% in average, which proves that the space decoupled prototype learning is effective for improving the generalization and robustness of industrial CPS attack detection in few-shot scenario.

Accurate extraction of tunnel lining is crucial for deformation monitoring, damage detection, and tunnel modeling. However, the extraction of tunnel lining is still challenging due to the different tunnel shapes and complicated construction scenes. As for tunnel point clouds, inconsistent local density and non-structural features also pose challenges. Previous research has shown that utilizing the centerline of a tunnel is the preferred method for extracting tunnel lining. However, imprecision often arises due to noise points and point cloud hole. To address these challenges, we propose a novel extraction framework for various tunnel shapes that seldom depends on the centerline. First, the structure detection module is designed to detect the lining structure from the raw point cloud. Following that, multi-scale clustering filtering is applied for fine extraction. The global clustering filtering procedure is then implemented to process the projected point cloud. Finally, the local filtering approach is created for confusion clustering. Experimental results show that the highest Kappa coefficient of the proposed method is 95.5%. Compared with the elliptic cylinder method, the angle threshold method, and the template method, the extraction accuracy of our method is improved by 6.2%, 10.9%, and 9.7%, respectively.

TOR (The Onion Router) network is a widely used open source anonymous communication tool, the abuse of TOR makes it difficult to monitor the proliferation of online crimes such as to access criminal websites. Most existing approches for TOR network de-anonymization heavily rely on manually extracted features resulting in time consuming and poor performance. To tackle the shortcomings, this paper proposes a neural representation learning approach to recognize website fingerprint based on classification algorithm. We constructed a new website fingerprinting attack model based on convolutional neural network (CNN) with dilation and causal convolution, which can improve the perception field of CNN as well as capture the sequential characteristic of input data. Experiments on three mainstream public datasets show that the proposed model is robust and effective for the website fingerprint classification and improves the accuracy by 12.21% compared with the state-of-the-art methods.

The maintenance of subway tunnels is vital to ensure the safety of their daily operation. Issues experienced by shield subway tunnels, especially the water leakages, require rapid and accurate detection and diagnosis. Due to the large number of disturbances in the tunnels, conventional algorithms face limitations when extracting discriminative features. To solve this problem, we propose a novel and efficient deep learning model for extracting multiscale and discriminative features of water leakages based on mobile laser scanning (MLS) point cloud intensity images. A new residual network module (Res2Net) is integrated with a cascade structure to form a unified model to extract the multiscale features of water leakages. The model can fully consider geometric characteristics of water leakages and grade the residual connections in a single residual block. This expands the size of receptive field in each network layer and can better facilitate the extraction of geometric characteristics of water leakages. Finally, we verify the advantages of the proposed method via experiments on five water leakage datasets of tunnel intensity images converted from point clouds obtained by a self-developed MLS system and compare its performance with other methods.

Tunnel dislocation affects the stability and waterproof of the structure and endangers its service life. This paper presents a dislocation calculation method of shield tunnel based on cross sectional point clouds, which contain three main steps. First, a longitudinal joint detection method is proposed based on the circumferential joints detected by the existing gradient accumulation method. The point clouds containing capping block in each ring are converted into gray image to obtain the center line position of the block, and the longitudinal joints are calculated through combining with the tunnel design data. Second, tunnel appendages are removed quickly through unfolding the tunnel point clouds by cylinder projection and adjusting the parameters of the cloth simulation filtering algorithm. Finally, the circumferential dislocations are calculated by selecting multiple denoised sections on both sides of the joint and clustering the points at each angle. Meanwhile, the longitudinal dislocations are calculated through fitting the segments in each ring separately. Experimental results show that the automatic extraction rate of longitudinal joints is higher than 94%. The point clouds filtering method can quickly separate the very long tunnel lining from the close appendages attached to it. Meanwhile, the RMSE of the repeated circumferential and longitudinal dislocation are evaluated to be 1.56 mm and 0.57 mm respectively. Compared with the existing state of art methods, these values reach 1.75 mm and 0.63 mm. Through the method proposed, the dislocation value at any mileage and any angle of tunnel can be displayed intuitively.

In the maintenance work of tunnels, images are often used to detect diseases, but collections of tunnel images are limited by the tunnel environment and working time. Three-dimensional laser scanning technology can acquire high-precision tunnel information efficiently, and the main problem to be solved by using this technology to collect tunnel inner wall images is the dimensionality reduction of the laser tunnel point cloud data. This letter proposes a tunnel surface parameterization algorithm based on a harmonic map, where a 3-D tunnel point cloud is used as a data source to reconstruct a triangle mesh model of the tunnel and then generate a harmonic map depth map of the tunnel inner wall on the triangle mesh. We can obtain the spatial distribution and position information of the appendages and detect whether there are cracks, water leakage, falling pieces, and other diseases by the depth images. The results of this study indicate that the proposed algorithm is suitable for tunnels of various shapes and has low area distortion, which can better avoid the loss of information during dimensionality reduction. Compared with other existing methods, the algorithm has higher efficiency and applicability.

Due to errors in sensors and positioning, there exist mismatches between different phases of mobile laser scanning point clouds, which impede the application of point cloud, such as changing detection and deformation monitoring. To rectify such mismatches, we designed a 3-D deep feature construction method for point cloud registration. The proposed method combines two 3-D convolutional neural networks into a uniform deep learning model to extract 3-D deep features. First, the corresponding points and noncorresponding points are set to train the deep learning model to minimize the distance between corresponding points’ features and maximize the distance between features of noncorresponding points. Second, in the test phase, the 3-D deep feature for each keypoint was extracted by the trained deep learning model. This could be used to determine the corresponding points by the$k$-dimensional tree and random sample consensus (RANSAC) algorithm. Finally, a transformation matrix was calculated based on the corresponding points and was then applied to point cloud registration. The experimental results illustrated that the proposed method of using 3-D deep features is more efficient at a corresponding point search than representatives of three existing methods. It also improved registration accuracy.