LSDM-PCB: A Lightweight Small Defect Detection Model for Printed Circuit Board

For Printed Circuit Board (PCB) surface defect detection, traditional detection methods mostly focus on template matching-based reference method and manual detections, which have the disadvantages of low defect detection efficiency, large errors in defect identification and localization, and low versatility of detection methods. In order to further meet the requirements of high detection accuracy, real-time and interactivity required by the PCB industry in actual production life. In the current work, we improve the You-only-look-once (YOLOv4) defect detection method to train and detect six types of PCB small target defects. Firstly, the original Cross Stage Partial Darknet53 (CSPDarknet53) backbone network is preserved for PCB defect feature information extraction, and secondly, the original multi-layer cascade fusion method is changed to a single-layer feature layer structure to greatly avoid the problem of uneven distribution of priori anchor boxes size in PCB defect detection process. Then, the K-means++ clustering method is used to accurately cluster the anchor boxes to obtain the required size requirements for the defect detection, which further improves the recognition and localization of small PCB defects. Finally, the improved YOLOv4 defect detection model is compared and analyzed on PCB dataset with multi-class algorithms. The experimental results show that the average detection accuracy value of the improved defect detection model reaches 99.34%, which has better detection capability, lower leakage rate and false detection rate for PCB defects in comparison with similar defect detection algorithms.

Detecting small defects against a complex surface is highly challenging but crucial to ensure product quality in industry sectors. However, in the detection performance of existing methods, there remains a huge gap in the localization and segmentation of small defects with limited sizes and extremely weak feature representation. To address the above issue, this paper presents a weighted matrix decomposition model (WMD) for small defect detection against a complex surface. Firstly, a weighted matrix is constructed based on texture characteristics of RGB channels in the defect image, which aims to improve contrast between defects and the background. Based on the sparse and low-rank characteristics of small defects, the weighted matrix is then decomposed into low-rank and sparse matrices corresponding to the redundant background and defect areas, respectively. Finally, an automatic threshold segmentation method is used to obtain the optimal threshold and accurately segment the defect areas and their edges in the sparse matrix. The experimental results show that the proposed model outperforms state-of-the-art methods under various quantitative evaluation metrics and has broad industrial application prospects.

Along with the development of electronic technology, the integration of numerous components on printed circuit board (PCB) boards has resulted in increasingly complex and intricate layouts. Small defects in traces can lead to failures in electronic functions, making the inspection of PCB surface layouts a critical process in quality control. Given the limitations of manual inspection, which struggles to detect such defects due to their size and complexity, there is a growing need for a PCB inspection system that utilizes automated optical inspection (AOI) based on deep learning detection. This research develops and compares two deep learning algorithms, faster region-based convolutional neural networks (R-CNN) and YOLOv8, to identify the most effective algorithm for detecting defects on PCB layouts. The findings of this study indicate that the YOLOv8 algorithm outperforms faster R-CNN, with the YOLOv8x variant emerging as the best model for defect detection. The YOLOv8x model achieved performance scores of 0.962 (mAP@50), 0.503 (mAP@50:95), 0.953 (Precision), 0.945 (Recall), and 0.949 (F1-score). These results provide a strong foundation for further research into the application of AOI for PCB defect detection and other quality control processes in manufacturing, using optimized deep learning models.

Nowadays, defect detection technology based on deep learning continuously increases the surface quality requirements of hot-rolled strip steel. However, due to limitations in industrial production, defect datasets often suffer from insufficient training samples and imbalanced categories. This paper proposes effective solutions, namely the GT-CutMix offline data augmentation algorithm and lightweight small sample defect detection models. The GT-CutMix augmentation algorithm significantly improves defect utilization by accurately sampling defect locations and integrating them into the original data set. We design the S-deconvolutional single shot detector (DSSD) defect detection model by constructing a lightweight SI-MobileNet to replace the ResNet101 backbone of the DSSD network. This can reduce the resource parameters and consumption. At the same time, it can speed up training and inference. To further improve the detection accuracy, we integrate the pyramid split attention (PSA) mechanism into the prediction module of DSSD and construct the SA-DSSD model. Under the GT-CutMix augmentation algorithm, the mAP of S-DSSD and SA-DSSD models on X-SDD dataset are 76.83% and 78.63%, respectively. Meanwhile, the corresponding detection speeds are 45 FPS and 40 FPS, respectively. In addition, on the NEU-DET cross-dataset experiment, the mAP of the SA-DSSD model reaches 74.88%. Our methods are highly effective and generalized for small sample defect detection, which can provide selective solutions for specific needs such as high speed and precision in different industrial production scenarios.

PCB (printed circuit board) is an extremely important component of all electronic products, which has greatly facilitated human life. Meanwhile, tons of PCBs in the waste streams become a waste of resources, which puts the recycling and reuse of PCBs in urgent need. In the manufacturing and recycling of electronic products, the classification of PCBs, recognition of sub-components, and defect detection have been the key technology. Traditional manual detection and classification are subjective and rely on individuals’ experience. With the development of artificial intelligence, lots of research efforts have been dedicated to the automated detection and recognition of PCBs. In this paper, we propose a transformer-based model, LPViT, for defect detection and classification of PCBs. We conduct the defect detection task on the dataset DeepPCB, which consists of six different types of PCB defects. Defect detection benefits both manufacturing and recycling of PCBs. Among many electronic products, a group of affordable, general-purpose, and small-size PCBs is very popular, which are referred to as micro-PCBs. The classification and recognition of those PCBs will greatly facilitate the recycling and reuse process. We conduct the classification task on a dataset called <bold xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">micro-PCB</b> , which includes 12 types of popular, general-purpose, affordable, and small PCBs. Through comparative experiments, our system demonstrates its advantage in both classification and defect detection tasks.

This research aims to propose an effective model for the detection of defective Printed Circuit Boards (PCBs) in the output stage of the Surface-Mount Technology (SMT) line. The emphasis is placed on increasing the classification accuracy, reducing the algorithm training time, and a further improvement of the final product quality. This approach combines a feature extraction technique, the Principal Component Analysis (PCA), and a classification algorithm, the Support Vector Machine (SVM), with previously applied Automated Optical Inspection (AOI). Different types of SVM algorithms (linear, kernels and weighted) were tuned to get the best accuracy of the resulting algorithm for separating good-quality and defective products. A novel automated defect detection approach for the PCB manufacturing process is proposed. The data from the real PCB manufacturing process were used for this experimental study. The resulting PCALWSVM model achieved 100 % accuracy in the PCB defect detection task. This article proposes a potentially unique model for accurate defect detection in the PCB industry. A combination of PCA and LWSVM methods with AOI technology is an original and effective solution. The proposed model can be used in various manufacturing companies as a postprocessing step for an SMT line with AOI, either for accurate defect detection or for preventing false calls.

In order to address the challenges of deployment difficulties and low small-object detection efficiency in current deep learning-based defect detection models on terminal devices with limited computational capacity, this paper proposes a lightweight steel surface defect detection model, Pyramid-based Small-target Fusion YOLO (PSF-YOLO), based on an improved YOLOv11n object detection framework. The model employs a low-parameter Ghost convolution (GhostConv) to substantially reduce the required computational resources. Additionally, the traditional feature pyramid network structure is replaced with a Multi-Dimensional-Fusion neck (MDF-Neck) to enhance small-object perception and reduce the number of model parameters. Moreover, to achieve multi-dimensional integration in the neck, a Virtual Fusion Head is utilized, and the design of an Attention Concat module further improves target feature extraction, thereby significantly enhancing overall detection performance. Experimental results on the GC10-DET+ dataset demonstrate that PSF-YOLO reduces model parameters by 25% while achieving improvements of 3.2% and 3.3% in mAP_{50} and mAP_{50-95}, respectively, compared to the baseline model. This approach offers valuable insights and practical applicability for deploying defect detection models on terminal devices with limited computational resources.

Lithium batteries are pivotal in new energy vehicle (NEV) applications. This study proposes a novel defect detection model, Small-object Detection-YOLO (SD-YOLO), based on YOLOv11, to overcome the limitations of conventional methods in detecting small surface defects in lithium batteries, particularly addressing the challenge of high miss rates for low-contrast targets. The Similarity-Aware Activation Module (SimAM) is integrated into the backbone, enabling the model to prioritize small-scale defects and enhance feature extraction. The C3k2 GhostDynamicConv module is introduced to replace conventional convolution layers, enhancing feature representation while reducing computational complexity through dynamic convolutions and Ghost feature integration. To further refine model performance, the Focaler-Iou loss function is employed, replacing the original CIoU loss function, optimizing bounding box regression, and accelerating model convergence. Experimental results show that the proposed model significantly outperforms the original YOLOv11 in key performance metrics, with a 1.3% increase in precision, a 0.5% increase in recall, and a 2.2% increase in mean average precision (mAP). Furthermore, the model outperforms other state-of-the-art methods in both detection speed (FPS) and small-target detection capabilities.

Surface defect detection of printed circuit boards (PCBs) is a critical stage in ensuring product quality on production lines in electronics manufacturing. The excellent performance of defect detection methods using deep learning models such as convolutional neural networks (CNNs) and autoencoders is limited by image uncertainty under uneven ambient light or unstable transmission channels and label uncertainty due to human perception errors or lack of expert knowledge. To overcome these difficulties, a novel collaborative learning classification model is proposed for surface defect detection on PCBs. An auxiliary inference model is designed to deal with label uncertainty. Then, a symmetric residual filter is set up based on a multiscale symmetric convolutional network to remove image uncertainty in the dataset. At the same time, knowledge-transfer-based probabilistic classification is used to improve the efficiency and performance of the model in different defect detection. Furthermore, a cooperative joint probabilistic inference engine is used to improve the efficiency of the model effectively. Results on both public datasets and self-collected datasets show that the proposed model achieves excellent performance on various quantitative metrics and is suitable for defect detection on datasets collected in complex industrial environments.

Eddy Current is a powerful mean of detection of detects located in conductive parts. This technique has already proved great performances and brought solutions to different industrial issues in nuclear or aeronautics domains for instance. Probes used in Non Destructive Testing (NDT) are mainly based on winding coils. This technology has shown good efficiency and gave good results in a lot of applications. Nonetheless, it reveals some limits in some cases, when the part has a complex shape for instance or when the defect is deeply buried. Therefore, other technologies have been developed at CEA LIST. An original scheme, optimized using the NDT platform CIVA, led to the development of a 32‐elements flexible probe, based on micro‐coils. Experimental testing reveals its efficiency in the detection of small surface defects. In other hand, magnetic sensors are very attractive for NDT. Thanks to their small size, a 22‐GMRs arrays probe and a 96‐AMRs array probes have been achieved. Their high spatial resolution and efficiency in the detection of small defects are demonstrated. The high sensitivity of magnetic sensors at low frequency has been used to design a flexible probe dedicated to the detection of deep defects. Its design and experimental testing are given.

LDD-Net: Lightweight printed circuit board defect detection network fusing multi-scale features

Printed circuit boards (PCBs) are primarily used to connect electronic components to each other. It is one of the most important stages in the manufacturing of electronic products. A small defect in the PCB can make the final product inoperable. Therefore, careful and meticulous defect detection steps are necessary and indispensable in the PCB manufacturing process. The detection methods can generally be divided into manual inspection and automatic optical inspection (AOI). The main disadvantage of manual detection is that the detection speed is too slow, resulting in a waste of human resources and costs. Thus, in order to speed up the production speed, AOI techniques have been adopted by many PCB manufacturers. Most current AOI mechanisms use traditional optical algorithms. These algorithms can easily lead to misjudgments due to different light and shadow changes caused by slight differences in PCB placement or solder amount so that qualified PCBs are judged as defective products, which is also the main reason for the high misjudgment rate of AOI detection. In order to effectively solve the problem of AOI misjudgment, manual re-judgment is currently the reinspection method adopted by most PCB manufacturers for defective products judged by AOI. Undoubtedly, the need for inspectors is another kind of labor cost. To reduce the labor cost. of manual re-judgement, an accurate and efficient PCB defect reinspection mechanism based on deep learning algorithm is proposed. This mechanism mainly establishes two detection models, which can classify the defects of the product. When both models have basic recognition capabilities, the two models are then combined into a main model to improve the accuracy of defect detection. In the study, the data provided by Lite-On Technology Co., Ltd. were implemented. To achieve the practical application value in the industry, this research not only considers the problem of detection accuracy, but also considers the problem of detection execution speed. Therefore, fewer parameters are used in the construction of the model. The research results show that the accuracy rate of defect detection is about 95%, and the recall rate is 94%. Compared with other detection modules, the execution speed is greatly improved. The detection time of each image is only 0.027 s, which fully meets the purpose of industrial practical application.

Phased Array Ultrasonic Testing (PA-UT) has been proved to be the most feasible way to inspect defects in electrofusion (EF) joints of polyethylene (PE) pipes. The recognition of defects in PA-UT results relies on the experience of operators, and results in inconsistent defective detection rate and low recognition speed. In this paper, an automatic defect recognition model for PA-UT inspection images was proposed based on convolutional neural network (CNN), realizing recognition of four typical defects in EF joint, i.e. wire dislocation, holes, lack of fusion and cold welding. The proposed recognition model consisted of anomaly detection model and defect detection model. The anomaly detection model recognized anomalies of PA -UT inspection images. The Defect detection model classified and located the defects of abnormal PA -UT inspection images. The anomaly detection model was assigned with a large anchor scale, to realize fast recognition, so as to meet the requirement of real-time recognition practical inspection. And defect detection model was designed to achieve high accuracy of defects recognition. By optimizing parameters of learning rate and dataset augmentation, the anomaly detection model and the defect detection reached a good combination of robustness and accuracy. Experiments showed that the proposed recognition model can improve the recognition speed and accuracy compared to single defect detection model.

Objectives:This study aimed to compare the diagnostic accuracy of inverted and unprocessed digitized periapical radiographs for detection of peri-implant defects.Materials and Methods:A total of 30 osteotomy sites were prepared in three groups of control, study group 1 with 0.425 mm defects and study group 2 with 0.725 mm defects using the SIC and Astra Tech drill systems with 4.25mm and 4.85mm diameters. Small and large defects were randomly created in the coronal 8mm of 20 implant sites; implants (3.4mm diameter, 14.5mm length) were then placed. Thirty periapical (PA) radiographs were obtained using Digora imaging system (Soredex Corporation, Helsinki, Finland), size 2 photostimulable storage phosphor (PSP) plate sensors (40.0mm×30.0mm) and Scanora software. Unprocessed images were inverted using Scanora software by applying image inversion and a total of 60 images were obtained and randomly evaluated by four oral and maxillofacial radiologists. Data were analyzed using the t-test.Results:Significant differences were observed in absolute and complete sensitivity and specificity of the two imaging modalities for detection of small and large defects (P<0.05). Unprocessed digital images had a higher mean in terms of absolute sensitivity for detection of small defects, complete sensitivity for detection of large peri-implant defects and definite rule out of defects compared with inverted images.Conclusion:Unprocessed digital images have a higher diagnostic value for detection of small and large peri-implant defects and also for definite rule out of defects compared with inverted images.

High-speed railway catenaries are vital components in railway traction power supply systems. To ensure stable contact between the pantograph and the catenary, droppers are positioned between the messenger wire and contact line. The failure of one or more droppers will affect the power supply of the catenary and the operation of the railway. In this paper, we modify the You Only Look Once version five (YOLOv5) model in several ways and propose a method for improving the identification of dropper status and the detection of small defects. Firstly, to focus on small target features, the selective kernel attention module is added to the backbone. Secondly, the feature graphs of different scales extracted from the backbone network are fed into the bidirectional feature pyramid network for multiscale feature fusion. Thirdly, the YOLO head is replaced by a decoupled head to improve the convergence speed and detection accuracy of the model. The experimental results show that the proposed model achieves a mean average precision of 92.9% on the dropper dataset, an increase of 3.8% over the results using YOLOv5s. The detection accuracy of small dropper defects reaches 79.2%, representing an increase of 10.8% compared with YOLOv5s and demonstrating that our model is better at detecting small defects.