Ground-based astronomical images are often degraded by atmospheric turbulence and deterministic optical aberrations introduced by telescope design and manufacturing processes. Joint mitigation of these distortions remains challenging due to the lack of reliable ground-truth data. To address this issue, a physics-based atmospheric–optical imaging model is developed to generate a large-scale, physically consistent simulated dataset, enabling supervised learning without real paired observations. Based on this, an attention-enhanced generative adversarial network (AE-GAN) is proposed for astronomical image restoration. The network incorporates a Channel Attention Block (CAB) and a Semantic Attention Module (SAM) within a feature pyramid architecture to enhance multi-scale representation and suppress turbulence-induced distortions. Experimental results show that the proposed method achieves consistent restoration performance under varying turbulence strengths, aberration amplitudes, and noise levels. Compared with recent Transformer-based methods, it maintains competitive performance across different aberration types while achieving significantly higher computational efficiency (1.21 s per image, 3.5× faster). In addition, the model trained on simulated data generalizes effectively to real astronomical observations.

Frequency-domain plane-wave imaging (FPWI) enables ultrafast frame rates for ultrasonic nondestructive evaluation. However, image quality can still be degraded by noise and artifacts. Most coherence-weighting schemes are implemented in the time domain using sliding-window statistics, which introduce substantial computation and memory-access overhead. In this work, we compute coherence weights directly in the frequency domain using the fast Fourier transform (FFT). Inspired by spatial coherence principles related to the van Cittert–Zernike theorem (VCZ), we propose a normalized amplitude coherence factor (NACF) as an amplitude-domain alternative to the normalized energy coherence factor (NECF). Experiments on representative defect datasets show that FPWI with NACF improves contrast ratio (CR), signal-to-noise ratio (SNR), and array performance indicator (API) relative to baseline FPWI, while further sharpening defect boundaries and enhancing lateral resolution. These results demonstrate a favorable balance between artifact suppression and resolution enhancement, making the proposed method a promising solution for ultrasonic NDE scenarios requiring both high frame rates and high imaging quality.

Phase-shifting profilometry (PSP) plays a dominant role in the field of three-dimensional (3D) surface-shape measurement due to its high accessible measurement accuracy and spatial resolution. However, PSP relies on multiple projections, which makes it prone to motion-induced errors and thus its performance is limited in dynamic scenes. In this paper, we introduce a novel single-shot 3D measurement method based on stereo color PSP. This method requires no additional images, benefiting from geometric constraints and colored phase-shifted fringes. The phase errors caused by color crosstalk and gamma nonlinearity are corrected by the chord distribution equalization method and the cross-correlation-based global phase offset correction method. The point cloud obtained using the corrected phase is reprojected onto the second camera plane, and the corresponding matching points are found on the epipolar lines within a limited search range. After stereo correspondence, the final 3D point cloud of the measured object is reconstructed based on stereo technology. Three dynamic scenes (a changing palm, a rotating Beethoven statue on a platform, and a rotating electric fan) are assessed to verify the robustness and accuracy of the proposed technique in a range of dynamic and complex environments.

The Total Focusing Method (TFM) in ultrasonic phased array imaging offers high resolution and promises improved detection of near-surface flaws. However, the irregular structure of wheel rim treads presents challenges for detecting such small volumetric defects. In this study, an Adaptive Total Focusing Method (ATFM) is proposed under water immersion phased array testing conditions. The technique reconstructs the interface using a pulse-echo approach and applies band-pass filtering to Full Matrix Capture (FMC) data via a Kaiser window. Based on Delay-and-Sum (DAS) beamforming, the effectiveness of six different coherence factor weighting methods is compared. The results demonstrate that the ATFM approach incorporating the Circular Coherence Factor (CCF) offers significant advantages in near-surface noise suppression and defect resolution. It successfully detects flat-bottom hole (FBH) defects with a diameter of 0.5 mm located 3 mm beneath the rim surface, achieving a signal-to-noise ratio (SNR) of 28.93 dB and a resolution of 45.79 dB.

Reconstruction models with Transformer architectures have achieved state-of-the-art performance in anomaly detection, yet the issue of learning shortcuts persists. We address this problem in reconstruction tasks by eliminating residual connections in the Transformer structure. While traditional Transformers rely on residual connections to alleviate gradient vanishing, such designs in reconstruction tasks tend to allow models to bypass deep feature learning and directly copy shallow-layer features, resulting in performance degradation. By removing residual connections, we compel the model to learn deeper features more comprehensively, thereby improving reconstruction quality. Additionally, we design a novel feature processing mechanism based on features extracted by convolutional neural networks (CNNs): First, global contextual information is extracted through average pooling (AvgPool) for feature dimensionality reduction. Then, 1D convolution (Conv1D) is applied along the channel dimension to further fuse features. Finally, attention weights are generated via the Sigmoid activation function to highlight critical features. Under unified training and inference conditions, our method outperforms previous state-of-the-art approaches on the MVTec-AD dataset and demonstrates exceptional performance on a train component dataset.

Fringe projection profilometry (FPP) has been widely used in the field of 3D measurement with advantages of high precision, non-contact, and low cost. However, measuring objects with shiny surfaces or large variations in reflectivity remains challenging, because image saturation causes loss of encoded information and reconstruction failure. This paper proposes an adaptive fringe projection method to avoid image saturation. Compared with existing AFP methods, the proposed method compensates for the estimated surface reflectivity without additional projected patterns. First, two uniform grayscale patterns were projected to estimate initial surface reflectivity and determine a low projection intensity. Then, orthogonal low intensity phase-shifted patterns and a white uniform pattern were projected to determine the camera-projector coordinate mapping and calculate saturated regions. Meanwhile, the background intensity calculated using the phase-shifting algorithm was used to compensate for the initial reflectivity which used to calculate the best projection intensities. Finally, the adaptive fringe patterns were generated. Experiments demonstrate that the proposed method improves measurement accuracy for highly reflective surfaces.

Automatic fault detection based on machine vision technology is crucial for the operational safety of trains. However, when imaging moving trains, system errors may induce localized geometric distortions in the captured images, altering the shapes of critical train components. This, in turn, undermines the precision of subsequent diagnostic algorithms. Therefore, image registration prior to anomaly detection is essential. To address this need, we redefine the horizontal registration of line-scan images as a disparity estimation problem on rectified stereo pairs, which is solved using a proposed dense matching network. The disparity is iteratively refined through a GRU-based update module that constructs a multi-scale cost volume with positional encoding and self-attention. To overcome the absence of real-world disparity ground truth, we generate a physics-based simulation dataset by analytically modeling the nonlinear relationship between train velocity variations and line-scan image distortions. Extensive experiments on diverse real-world train image datasets under varied operational conditions demonstrate that our method consistently outperforms alternatives, achieving 5.8% higher registration accuracy and a fourfold increase in processing speed over state-of-the-art approaches. This advantage is particularly evident in challenging scenarios involving repetitive patterns or texture-less regions.

With the ease of acquiring RGB-D images from line-scan 3D cameras and the development of computer vision, anomaly detection is now widely applied to railway inspection. As 2D anomaly detection is susceptible to capturing condition, a combination of depth maps is now being explored in industrial inspection to reduce these interferences. In this case, this paper proposes a novel approach for RGB-D anomaly detection called Dual-Branch Cross-Fusion Normalizing Flow (DCNF). In this work, we aim to exploit the fusion strategy for dual-branch normalizing flow with multi-modal inputs to be applied in the field of track detection. On the one hand, we introduce the mutual perception module to acquire cross-complementary prior knowledge in the early stage. On the other hand, we exploit the effectiveness of the fusion flow to fuse the dual-branch of RGB-D inputs. We experiment on the real-world Track Anomaly (TA) dataset. The performance evaluation of DCNF on TA dataset achieves an impressive AUROC score of 98.49%, which is 3.74% higher than the second-best method.

This study presents an unsupervised deep learning framework aimed at improving defect detection in A-scan ultrasonic signals from high-speed train axles. By addressing the challenges associated with manual interpretation in traditional ultrasonic testing, we introduce a dual-encoder model with attention mechanisms that can analyze both global waveform patterns and localized anomalies without the need for labeled training data. The proposed architecture aims to isolate defect-induced signal variations from environmental noise through parallel feature processing. Evaluations on real-world industrial data indicate that the model performs well, with promising results (F1: 0.94, AUC: 0.98) in comparison to threshold-based methods, especially in low signal-to-noise scenarios. This approach has the potential to support more reliable large-scale axle inspection, reducing the reliance on annotated datasets and contributing to the improvement of rail safety through automated anomaly detection.

Fringe projection profilometry (FPP) technique is a widely used method in three-dimensional measurement and has broad applications in industrial measurement. This method typically requires projecting multiple phase-shifted fringe patterns to achieve high-precision imaging. However, in industrial contexts, the measured objects are often metallic components with high reflectivity, which would adversely affect the reconstruction quality of FPP. High Dynamic Range (HDR) technology can effectively address such issues; however, it requires an increased number of projected fringe patterns, which reduces measurement efficiency. This paper proposes a single-shot 3D measurement method that combines traditional HDR techniques with deep learning to measure objects with highly reflective surfaces. The method employs two network models: the Phase Recovery Network (PRNet) and the Coarse Phase Unwrapping Network (CPUNet). PRNet receives a single fringe image with saturated pixels and outputs the numerator and denominator terms for phase retrieval. CPUNet takes the wrapped phase data processed through an arctangent function from PRNet's output and predicts the coarse unwrapped phase to assist in phase unwrapping, yielding high-precision phase results. Experiments show that this approach requires only one shiny fringe image as input and effectively restores phase errors caused by high reflectivity, achieving a high reconstruction accuracy for the complete 3D shape of the object.

As a critical operational component, railway wheels undergo ultrasonic testing, one of the most prevalent non-destructive testing methods. However, most ideal defects seldom manifest in real-world scenarios, leading to a scarcity of available cases. In light of this challenge, we propose a novel dual-network approach in this paper, designed for both data augmentation and the detection of extreme data. This approach aims to enhance defect detection performance. To validate the feasibility of our algorithm, we assembled a substantial dataset comprising simple ultrasonic B-scan defects. Our algorithm demonstrates high detection accuracy, surpassing 95% with augmented data, about a 15% improvement over the original dataset. Furthermore, employing augmented data for detecting real data yields commendable results. This paves the way for integrating deep learning with traditional nondestructive testing in future industrial inspections, ensuring efficient defect detection and bolstering railway safety and personnel training efforts.

The lack of real defect data samples has become a challenging problem for the effective application of deep learning networks in ultrasound target detection. This paper proposes a data augmented generative adversarial network (DCSGAN) aimed at overcoming the scarcity of welding ultrasonic defect data in training target detection networks. This network utilizes bilinear interpolation to expand the real data sample space, facilitating the extraction of high-dimensional defect spatial features through deeper networks. By obtaining a mixed dataset of generative data and real data, training and testing experiments are conducted on the object detection network. The experimental results demonstrate that the data augmentation method proposed in this paper effectively enhances the detection rate of ultrasonic welding defects in the target detection network, which has reference significance for similar application scenarios of ultrasonic defect detection.

During the ultrasonic defect detection process of train wheel, electromagnetic interference and poor coupling of equipment can lead to disturbances or frame losses in the data. This results in some omissions and incorrect detections in subsequent defect target detection tasks, posing significant risks. The efficiency of classifying abnormal data through manual and traditional algorithms is not high. To address this issue, this study introduces for the first time the application of a deep learning classification network to the quality classification of ultrasonic B-scan data. Based on the size of our ultrasonic B-scan dataset, we selected ResNet34 as our classification network. Additionally, we augmented the ultrasonic B-scan dataset with enhanced data. The detection results demonstrate that the proposed network successfully differentiated between different types of samples in the test set, achieving classification accuracy of over 95%. This method effectively classifies abnormal data from normal data, thereby improving the efficiency of subsequent damage detection tasks. It holds significant reference value for image quality classification in similar applications.

When observing astronomical targets from ground-based telescopes, sequential short-exposure astronomical image acquisition is usually performed, where the astronomical images exhibit constant distortion and quality changes, and atmospheric turbulence makes the captured astronomical images seriously affected by noise and severely reduces the resolution. In this paper, we use a deep neural network-based image distortion registration technique to eliminate atmospheric blurring by registering the sequential short-exposure astronomical images and superimposing the registered images to improve image quality. This paper introduces an attention technique based on the Cyclemorph network, which identifies three attention units on the feature extraction network and uses a self-gating soft attention technique to generate a trainable gating signal that allows the network to associate useful local information and then combines the feature vectors obtained from it to make the final prediction. Additionally, this soft attention method can help with accurate target localisation and boost the network's overall performance. To show how well the method works, it is applied to a collection of image datasets of really photographed astronomical targets as well as a set of image datasets of simulated astronomical targets built by a telescope atmospheric imaging model. The built networks go through independent training and testing. According to the experimental findings, the proposed method can successfully register consecutive astronomical photographs, neutralize the impact of atmospheric turbulence, and raise the resolution of astronomical images.

Components with complex surfaces can pose a challenge for achieving phased array ultrasonic testing. Although coupling issues can be addressed by immersing or adding plexiglass wedges, obtaining high-quality images in a short time remains difficult under such conditions. This paper proposes an adaptive plane wave imaging scheme to address these challenges. First, the interface is reconstructed using a plane wave priority echo estimation method. Then, a beamforming algorithm adapted to the irregular interface is used to generate low-resolution (LR) images, which are then reconstructed into high-resolution (HR) images in a short time with convolu-tional neural network (CNN). The CNN is trained with simulation data, and the defects of test blocks and wheel rim are tested. Experimental results show that the reconstruction method combined with CNN significantly improves imaging quality and speed compared to the conventional adaptive plane wave beamforming algorithm based on the time-domain physical model.

In the field of computer vision such as target detection and 3D positioning, point cloud registration has always been one of the key problems, which requires alignment two point clouds through rigid spatial transformation, accuracy, robustness, speed and other factors. Point cloud registration based on deep learning has received a lot of research in recent years. Compared with the conventional methods, They show a great advantage in their registration performance, To improve the performance of deep learning on point-cloud registration, This paper uses Kernel correlation to compute and store the neighborhood information of point clouds, While extracting local geometric features by convolutional neural network and Offset_Attention module aggregation, Then use Singular value decomposition SVD to predict the final rigid transformation matrix, and finally achieve high-quality registration. In this paper, by training our model on the ModelNet40 dataset, And the source and target point cloud are sampled independently, And also extract the non-axisymmetric targets for additional tests, Achievalized a more equitable registration network experiment, the Root-mean-square deviation (RMSE) is 12.58% higher than before, which verifies the effectiveness of our network.

Total focusing method imaging is a phased array (PA) post-processing imaging method based on full matrix acquisition data with high resolution and high signal-to-noise ratio. However, the placement of the probe is largely limited by the shape of the detection surface. In this paper, a water membrane PA probe that can perform in-service detection of unknown interface objects in indirect water immersion mode is used, and a covariance correlation weighting factor for optimizing adaptive Total focusing method (ATFM) imaging results is proposed. Experiments of artificial test block and actual wheel rim have shown that the ATFM can not only identify the profile of the unknown interface, but also image the interior of the test object. Furthermore, the contrast signal-to-noise ratio of the defect signal is increased by 40 dB on average, and the defect resolution is also Improved. This method, meanwhile, can truly reflect the morphology of internal defects in actual inspection objects, which is conducive to non-destructive evaluation of defects.

防松铁丝用于防止螺栓松动,若其发生断裂可能导致螺栓松动或丢失,影响关键部件的正常运行,威胁动车组运行安全.由于防松铁丝发生断裂时,断裂部位特征信息少,特征区域较小,使基于传统卷积神经网络的分类方法难以有效提取到断裂特征信息,导致分类精度不高,容易漏判误判.基于以上原因,利用深度学习方法,建立基于卷积神经网络的孪生网络模型,通过距离度量学习对防松铁丝进行分类.同时,为进一步提高模型分类性能,提出一种双边界损失函数.实验表明,基于双边界损失函数的模型较传统的基于交叉熵损失函数的分类模型性能更优.通过测试,使用的方法能够较好地克服光照、油渍、水渍、部件移动等带来的伪异常,鲁棒性更强.

Three-dimensional (3D) shape reconstruction based on structured light technique is one of the most crucial techniques in the field of optical measurement due to the nature of non-contact and high-precision. In recent years, researchers found that the deep-learning method can significantly improve the quality and efficiency of 3D reconstruction. So far, the research object of structured light 3D reconstruction technology related to deep-learning methods usually adopts plaster models. However, the surface of the reconstructed objects always contains rich colour texture information in a real-world scenario, which affects the accuracy of the reconstruction, especially for some single-shot applications. To solve the above problem, we propose a deep-learning-based method single-shot 3D reconstruction method (SSR) for colour textured objects, and it can reconstruct colour textured objects with high precision. The experimental results show that the proposed method provides good guidance for the practical application and scientific research of fringe projection profilometry based on colour textured objects.