Metal surface defect segmentation can play an important role in dealing with the issue of quality control during the production and manufacturing stages. There are still two major challenges in industrial applications. One is the case that the number of metal surface defect samples is severely insufficient, and the other is that the most existing algorithms can only be used for specific surface defects and it is difficult to generalize to other metal surfaces. In this work, a theory of few-shot metal generic surface defect segmentation is introduced to solve these challenges. Simultaneously, the Triplet-Graph Reasoning Network (TGRNet) and a novel dataset Surface Defects- 4 ⁱ are proposed to achieve this theory. In our TGRNet, the surface defect triplet (including triplet encoder and trip loss) is proposed and is used to segment background and defect area, respectively. Through triplet, the few-shot metal surface defect segmentation problem is transformed into few-shot semantic segmentation problem of defect area and background area. For few-shot semantic segmentation, we propose a method of multi-graph reasoning to explore the similarity relationship between different images. And to improve segmentation performance in the industrial scene, an adaptive auxiliary prediction module is proposed. For Surface Defects- 4 ⁱ , it includes multiple categories of metal surface defect images to verify the generalization performance of our TGRNet and adds the nonmetal categories (leather and tile) as extensions. Through extensive comparative experiments and ablation experiments, it is proved that our architecture can achieve state-of-the-art results.

Nondestructive testing of welding images is still a significant challenge due to the imaging characteristics of radiographic images and the extremely random distribution of welding defect types. The practical application of supervised methods for welding nondestructive testing encounters significant challenges due to the limited availability of densely annotated samples and the absence of prior information regarding unknown defects. Hence, this article first proposes a rapid screening method using only defect-free image training to screen suspected defect images and normal images of welds in real time. Specifically, we first combine the correspondence mechanism and the representation mechanism, aiming to: 1) alleviate the smoothing reconstruction behavior caused by small defects and weak texture defects in weld; and 2) mitigate the offset learning behavior resulting from the differences between natural images and industrial weld images in the normalizing flow method. We propose a memory-aware transformer-based encoder, thus improving representations of complex defect-free images. Moreover, a dual-decoder strategy is introduced, which remaps the latent dependencies generated by the encoder through a semantic correspondence mechanism and reconstruction-guided normalizing flow, enabling effective learning of knowledge from weld images. We apply this framework to an industrial case of weld images, the experimental results demonstrate that our method outperforms other existing approaches.

The input data of photometric stereo have three dimensions: one photometric dimension of different lights and two spatial dimensions, i.e., rows and columns in the image coordinate. Recent deep-learning-based photometric stereo algorithms usually use 2-D/3-D convolutions to process the input with three dimensions. Assumptions that violate the natural characters of photometric stereo problem, e.g., spatial pixel interdependence and light permutation invariance, have to be made due to the dimension mismatch. In this article, we propose a self-attention photometric stereo network (SPS-Net), which can exploit the information in all three dimensions without violating these natural characters. In SPS-Net, the spatial information is extracted by convolutional layers and the photometric information is aggregated by the proposed photometric fusion blocks based on the self-attention mechanism. Extensive experiments on both synthetic and real-world data sets are conducted. The proposed SPS-Net achieved higher performance than the state-of-the-art algorithms photometric stereo task with dense lightings. Without any changes, the proposed algorithm also outperformed the benchmarks in sparse and light-information-robust photometric stereo tasks.

With the growing demand for higher product quality in manufacturing, X-ray non-destructive testing has found widespread application not only in industrial quality control but also in a wide range of industrial applications, owing to its unique capability to penetrate materials and reveal both internal and surface defects. This paper presents a systematic review of recent advances and current applications of X-ray-based defect detection in industrial components. It begins with an overview of the fundamental principles of X-ray imaging and typical inspection workflows, followed by a review of classical image processing methods for defect detection, segmentation, and classification, with particular emphasis on their limitations in feature extraction and robustness. The focus then shifts to recent developments in deep learning techniques—particularly convolutional neural networks, object detection, and segmentation algorithms—and their innovative applications in X-ray defect analysis, which demonstrate substantial advantages in terms of automation and accuracy. In addition, the paper summarizes newly released public datasets and performance evaluation metrics reported in recent years. Finally, it discusses the current challenges and potential solutions in X-ray-based defect detection for industrial components, outlines key directions for future research, and highlights the practical relevance of these advances to real-world industrial applications.

Studs are small screws welded on a car's body. Measuring the key point positions of the studs within production cycle time, also known as online measurement, is crucial for the quality management of the cars. However, the existing stud measuring systems can only measure one stud in a camera position, which is not efficient enough for online measurement. Accurately measuring multiple studs in one camera position is still a challenge. In this article, we propose a stud online measuring system (SOMS), which can measure multiple studs with high accuracy based on photometric stereo and deep learning techniques. To achieve accurate measurements, the SOMS uses normal maps, which are obtained by photometric stereo. Conventional photometric-stereo-based inspection methods reconstruct the surface from the normal maps; however, the reconstruction is inaccurate when the object has a large depth variation range. Instead of reconstruction, the SMOS directly uses a convolutional neural network (CNN) to localize the key points of studs from the normal maps. To adapt CNN to normal map input, we proposed a fusion block which extracts the most significant features among three dimensions of the normal maps. A test based on a collection of different types of studs was conducted to evaluate the advantage of the SOMS. Our system outperformed benchmarks in a series of experiments. Because the SOMS was developed without any prior knowledge on the geometry of the stud, it also has the potential for other optical challenging geometric features' online measurements.

Grain size plays a fundamental role in the mechanical properties of materials. Recently, automatic measurement of average grain size attacks more and more attention based on computer vision. However, low contrast, twin grains, thin boundary, and low connectivity limit the achievement of automatic and accurate image analysis. Inspired by the calculation procedure of grain size, we propose a center-guided and connectivity-preserving network for grain boundaries segmentation. On the one hand, the proposed center feature extraction module and center-guided feature recalibration mechanism make the network pay more attention to the center area. On the other hand, a connectivity-preserving loss function is integrated with the network, which forces the network to converge towards high connectivity. Benefit by above aspects, our network can segment the grain boundary with high structural integrity and avoid the complex post-processing process. Experiments on the SRIF-GSM dataset reveal that our method achieves 85.98 mIoU and 95.60 clDice scores, demonstrating significant advantages compared with the state-of-the-art semantic segmentation methods.