Design of lightweight metal surface defect detection technology for YOLOv7-tiny using Anchor-Free algorithm.

Computer vision and deep learning techniques are the most emerging technologies in this era. Both of these can greatly raise the rate at which defects on metal surfaces are identified while performing industrial quality checks. The identification of faults over metal surfaces can be viewed as a significant challenge since they are easily impacted by ambient factors including illumination and light reflections. This paper proposes novel metal surface defect detection network called as YOLOv-5s-FRN in response to the problems of ineffective detection brought by the conventional manual inspection system. The proposed system is developed through the integration of a novel architectural module called as Feature Recalibration Network (FRN) to the You Only Look Once-version-5 small network )YOLOv-5s(. In order to extract the global feature information from the provided image, FRN is able to evaluate the interdependencies between the channels. This improves the feature discrimination capability and prediction accuracy of the defect detection system. The incorporation of FRN structure makes YOLOv-5s architecture to selectively enhance the necessary features and discard the unwanted ones. Therefore, the proposed novel method will efficiently detect and classify the metal surface defects such as crazing, patches, inclusions, scratches, pitted surfaces and rolled in scale. North Eastern University surface defect database (NEU-DET) has been used to train and test the proposed architectural model. The suggested system has been compared with alternative models based on several performance matrices such as precision, recall and Mean Average Precision (mAP). It is observed that the proposed YOLOv-5s-FRN architecture provides significant performance improvement than state-of-the-art methods. The proposed system has been provided satisfactory results by means of improvement in mAP and time consumption. The proposed model has delivered value of mAP_0.5 as 98.05% and that of mAP_0.5:0.95 as 89.03%.

Defects in various products are unavoidable because of measurement errors and equipment accuracy limitations in the production process. Recent advances in metal surface defect detection have focused on optimizing traditional methods, developing new detection techniques, and exploring deep learning-based algorithms, providing technological support to improve metal manufacturing quality and production efficiency. To ensure the highest yield rate and meet production requirements, all products must undergo defect inspections before leaving the factory. However, Traditional methods for detecting metal surface defects require a lot of manual involvement, are difficult to accurately detect small defects, are susceptible to environmental interference, and lack stability and reliability. To address this issue, we propose the MeDERT model for metal surface defect detection. Our approach involves a new Span-sensitive Texture Fusion (STF) module structure that focuses on multi-headed attention modules to recover lost details and boost inspection speed and on top of that use the Jump-sensitive detail recovery feature fusion module to ensure the validity of the extracted textures. Additionally, we introduce singular value decomposition and pretzel noise to model the noise and enhance model robustness through data augmentation. Our MeDERT model achieved state-of-the-art (SOTA) results on a specified dataset, demonstrating its effectiveness in enhancing inspection efficiency and accuracy.

Metal surface quality control is significant in the production line of metal products. Detecting metal surface defects is challenging due to the various types and morphological patterns. Recent advances have witnessed deep learning-based automated optical inspection systems as a promising solution. This paper presents an enhanced YOLOv4 model for metal surface defect detection (MSDD). Specifically, we integrate three boosting components into the original YOLOv4, including 1) a self-dependent attentive fusion (SAF) block, placed within the model neck, to enhance inter-path and cross-layer feature fusion, 2) a component randomized Mosaic augmentation (CRMA) scheme to strategically discourage an over-transformed image to participate in training, and 3) a perturbation agnostic (PA) label smoothing method to keep the model from making over-confident predictions and thus act as a means of regularization. The proposed method has been validated on a self-developed MSDD dataset. It is shown that each boosting component can lead to an impressive mAP gain, and the final model outperforms the baselines, namely, Faster R-CNN, YOLOv4, YOLOv5, and YOLOX, by 7.85%, 6.51%, 3.76%, and 3.57%, respectively.

A cascaded twin dense network based metallic surface defect detection algorithm SCSEG-Net is proposed to address the problems of low accuracy and difficulty in locating defects in real-time metal surface defect detection. The method incorporates atrous spatial pyramid pooling structure to obtain feature maps with different sampling rates to capture multi-scale information. Meanwhile, in order to enhance the classification accuracy, the low-level texture and boundary features obtained by shallow convolution and the complex high-level feature information obtained by deep convolution are fused during training to better optimize the training parameters through the cascade network. SCSEG-Net can convert defect images into pixel-level prediction masks and quickly acquire real defect classes. The SCSEG-Net algorithm is trained, evaluated and verified on the industry standard steel surface defect data set. The results show that compared with similar algorithms, proposed method can segment the contour of steel surface defects more accurately and complete the classification with an <italic>F</italic>₁-score of 97.8% and a recall rate of 98.81%.

Metallic surface defect detection is of great significance in quality control for production. However, this task is very challenging due to the noise disturbance, large appearance variation, and the ambiguous definition of the defect individual. Traditional image processing methods are unable to detect the damaged region effectively and efficiently. In this paper, we propose a new defect detection method, Attention Cascade R- CNN with Mix-NMS (ACRM), to classify and locate defects robustly. Three submodules are developed to achieve this goal: 1) a lightweight attention block is introduced, which can improve the ability in capture global and local feature both in the spatial and channel dimension; 2) we firstly apply the cascade R-CNN to our task, which exploits multiple detectors to sequentially refine the detection result robustly; 3) we introduce a new method named Mix Non-Maximum Suppression (Mix- NMS), which can significantly improve its ability in filtering the redundant detection result in our task. Extensive experiments on a real industrial dataset show that ACRM achieves state-of-the- art results compared to the existing methods, demonstrating the effectiveness and robustness of our detection method.

Research on rapid detection of cross-scale defects in surface based on deep learning

The defect inspection in traditional metal processing usually adopts the manual visual inspection, which is easy to cause problems such as false detection and leakage. Besides, the detection rate of existing metal defect inspection has been affected by the high reflectivity of metal surface. For avoiding the problems of existing detection algorithms, a metal defect detection system based on laser triangulation and laser vision is given. The system is composed of industrial camera and strip light source. In order to avoid the interference of metal reflection on defect detection, the green strip light is used as the light source of the system. Based on the proposed system, this paper presents a metal surface detection algorithm with three primary color channels, that is, metal surface defects are identified by identifying the RGB values of metal surface pixels. Finally, through static experiments, the detection rate of the metal defect surface detection system based on laser vision is 100%. The defect detection method proposed in this paper can effectively avoid the influence of high reflectivity of metal surface on the detection rate, and can be widely used in the field of metal processing.

To address the challenges of low localization accuracy, weak feature discrimination, and real-time constraints in small-scale metallic surface defect detection, this study proposes a novel detection framework named FF-MDE (fine-grained multi-branch diversified encoder). The proposed model enhances detection capability through three key modules: a direction-aware multi-branch convolutional block (BBDES) that strengthens orientation-sensitive feature extraction via re-parameterization; a multi-scale fusion network (MFFN) incorporating a location-offset heterogeneous kernel selection strategy and a channel cross-embedding mechanism to align cross-resolution features and expand receptive fields adaptively; and a lightweight attention-guided up-sampling that improves fine-detail recovery while suppressing irrelevant responses. A large number of experiments on the NEU-DET and GC10-DET datasets show that FF-MDE has achieved excellent performance. The indicators of mAP50 are 70.3 and 57.9% respectively, which are 4.1 and 3.3% higher than the benchmark model and 7.6 and 6.5% higher on average than the existing methods. Moreover, the real-time inference speed of this network exceeds 60 FPS, providing a powerful and deployable solution for high-precision defect detection in industrial vision inspection systems.

Aiming at the problems of inefficient detection caused by traditional manual inspection, unclear features and few defect sample in metal surface defect detection, an improved metal surface defect detection technology the siamese network is proposed. The proposed method constructs a deep siamese network with maximum interval, which consists of 4 convolutional layers, 4 max pooling layers and the top fully connected layer to improve the accuracy of the recognition of difficult small targets while ensuring the detection efficiency of the model. The performance of the improved metal surface defect detection technology is compared with other detection methods on the NEU surface defect database. The experimental result verify that the mean average precision of the proposed method is 96.1%, which is higher that of support vector machine, region convolutional neural network, faster region convolutional neural network, single shot multibox detector, you only look once neural network v3.

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 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">i</sup> 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 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">i</sup> , 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.

Small defects on the surface of copper strips have a significant impact on key properties such as electrical conductivity and corrosion resistance, and existing inspection techniques struggle to meet the demand in terms of accuracy and generalisability. Although there have been some studies on metal surface defect detection, there is a relative lack of research on highly reflective copper strips. In this paper, a lightweight and efficient copper strip defect detection algorithm, SC-AttentiveNet, is proposed, aiming to solve the problems of the large model size, slow speed, insufficient accuracy and poor generalisability of existing models. The algorithm is based on ConvNeXt V2, and combines the SCDown module and group normalisation to design the SCGNNet feature extraction network, which significantly reduces the computational overhead while maintaining excellent feature extraction capability. In addition, the algorithm introduces the SPPF-PSA module to enhance the multi-scale feature extraction capability, and constructs a new neck feature fusion network via the HD-CF Fusion Block module, which further enhances the feature diversity and fine granularity. The experimental results show that SC-AttentiveNet has a mAP of 90.11% and 64.14% on the KUST-DET and VOC datasets, respectively, with a parameter count of only 6.365 MB and a computational complexity of 14.442 GFLOPs. Tests on the NEU-DET dataset show that the algorithm has an excellent generalisation performance, with a mAP of 76.41% and a detection speed of 78 FPS, demonstrating a wide range of practical application potential.

Defect size distributions play an important role in process characterization and yield prediction. To determine efficient procedures to extract defect size distributions, a database was set up that consists of hundreds of defect images to provide defect size distributions also reflecting irregular outlines of defects. Further investigation shows that even a single extension value per defect may provide a precise size distribution. Furthermore, the proposed Harp Test Structure (HTS) containing hundreds of parallel lines will also provide a defect size distribution, but it is based on electrical measurements only. The comparison of defect size distributions using both measurement procedures shows that not only optical measurements, but also electrical measurements at a Harp Test Structure are sufficient to obtain a precise defect size distribution that also enables size distribution modeling for yield prediction.

The detection of metal surface defects is crucial in the field of industrial production. However, in practical applications, challenges such as ambiguous defect directions, large scale differences, and strong background interference are often encountered. This paper proposes an improved multi-scale fine-grained object detection framework based on YOLOv11, referred to as “DA-YOLO.” Firstly, the 3D attention module (3DAM) was used to enhance the model’s ability to model spatial direction features and improve the model’s ability to perceive fine-grained structures. Secondly, a feature enhancement module (AMFEM) employing multi-scale convolution and a spatial-channel attention mechanism was constructed, significantly boosting the model’s recognition accuracy for multi-scale targets and blurry boundary defects. Furthermore, an intersection and union ratio Aware Joint Loss function (IoU-Aware joint loss, IAJ-Loss) was proposed and designed, which further enhanced the quality perception ability and stability of the model in complex detection scenarios. The experimental results show that the DA-YOLO model improved mAP@0.5 by 5.42% and 4.05% respectively on the GC10-DET and NEU-DET datasets compared to the baseline YOLOv11 model, demonstrating superior defect detection performance.

Defect density and defect size distributions (DDSDs) are key parameters used in IC yield loss predictions. Traditionally, memories and specialized test structures have been used to estimate these distributions. In this paper, we propose a strategy to accurately estimate DDSDs for shorts in metal layers using production IC test results.

In-line defect inspection is of crucial importance in controlling the quality of VLSI manufacturing processes. Wafer inspection tools such as KLA 21XX coupled with the review/classification stations, such as KLA25XX can be extremely effective in detecting problems in the processed batches of wafers. However, there is a need to properly allocate these machines among several process steps (levels) to maximize their utilization. Moreover, results of this in- line inspection such as defect distributions, both spatial and size, could be used to predict the yield impact of the detected defects. Up to now, the so-called 'kill ratios' were used to indicate the yield impact of defects per layer (e.g., polysilicon, metal1, metal2). The typical accuracy of 'kill ratio' based prediction was around 25%. In this paper, we demonstrate the methodology of yield impact prediction based on critical area concept. Critical area is extracted from the actual product layout by a software tool called MAPEX developed at CMU. The information about defect densities and size distributions is obtained from KLA inspection equipment. A prototype of a software system was developed that combines the information about the critical area with the defect size distributions, declustering of defect data into a yield model. This methodology and software system were verified on the experimental data from AMD Austin fabrication facility. The 'yield impact' prediction (equivalent to 'kill ratio' prediction) was performed for 7 different products ranging from SRAM to microprocessor to telecommunication products. The data was collected from more than 2000 wafers. The accuracy of our prediction was verified by comparing our functional yield estimate to the actual yield values from the probe testing measurements. We obtained an accuracy of 1% for the defects found by KLA in h polysilicon layer for dies affected by single defects only. Moreover, we were able to normalize the defect inspection data for the entire product family. This was possible due to the availability of information about critical area for each of the product. Hence, we were able to demonstrate the advantage of this methodology of determining the 'kill ratios' for the defect inspection data.