CLIP-MDC: CLIP encoder based multimodal defect classification with synthetic anomaly generation for real-time surface defect detection

With the development of integrated circuit packaging technology, the layout of printed circuit boards has become complicated. Moreover, the traditional defect detection methods have been difficult to meet the requirements of high precision. Therefore, in order to solve the problem of low efficiency in defect detection of printed circuit boards, a defect detection method based on pseudo-inverse transform and improved YOLOv5 is proposed. Firstly, a defect image restoration model is constructed to improve image clarity. Secondly, Transformer is introduced to improve YOLOv5, and the batch normalization and network loss function are optimized. These methods improve the speed and accuracy of PCB defect detection. Experimental verification showed that the restoration speed of the image restoration model was 37.60%-42.38% higher than other methods. Compared with other models, the proposed PCB defect detection model had an average increase of 10.90% in recall and 12.87% in average detection accuracy. The average detection accuracy of six types of defects in the self-made PCB data set was over 98.52%, and the average detection accuracy was as high as 99.1%. The results demonstrate that the proposed method can enhance the quality of image processing and optimize YOLOv5 to improve the accuracy of detecting defects in printed circuit boards. This method is demonstrably more effective than existing technology, offering significant value and potential for application in industrial contexts. Its promotion could facilitate the advancement of industrial automation manufacturing.

Several image processing methods have been implemented over recent years to assist and partially replace on-site technician visual inspection of both manufactured parts and operational equipments. Convolutional neural networks (CNNs) have seen great success in their ability to both identify and classify anomalies within images, in some cases they do this to a higher degree of accuracy than an expert human. Several parts that are manufactured for various aspects of turbomachinery operation must undergo a visual inspection prior to qualification. Machine learning techniques can streamline these visual inspection processes and increase both efficiency and accuracy of defect detection and classification. The adoption of CNNs to manufactured part inspection can also help to improve manufacturing methods by rapidly retrieving data for overall system improvement. In this work a dataset of images with a variety of surface defects and some without defects will be fed through varying CNN set-ups for the rapid identification and classification of the flaws within the images. This work will examine the techniques used to create CNNs and how they can best be applied to part surface image data, and determine the most accurate and efficient techniques that should be implemented. By combining machine learning with non-destructive evaluation methods component health can be rapidly determined and create a more robust system for manufactured parts and operational equipment evaluation.

The wood-based furniture manufacturing industries prioritize quality of production to meet higher market demands. Identifying various types of edge-glued wooden panel defects are a challenge for a human worker or a camera. Several studies have shown that the detection of edge-glued defects with low, high, normal, overlong, short is identified but detection of residue and bluntness is highly challenging. Thus, the present model identifies defects of low, high, normal, overlong, short by computer vision and/or deep learning, whereas defects of residue and bluntness by deep learning based decide by pass for having better performance. The goal of this paper is to provide an improved defect detection solution for wood-based furniture manufacturing industries by process automation. Therefore, a system was designed that takes defect input images from a camera as raw image and laser-aligned image for defect detection of the edge-glued wooden panel. The process automation then performs computer vision-based image features extraction with deep learning for defect detection. The aim of this paper is to solve edge-glued defect detection problems by using design and implementation of edge-glued wooden defect detection, that can be stated as edge-glued wooden panel defect detection using deep learning (WDD-DL) for process automation by artificial intelligence and Automated Optical Inspection (AOI) consolidation. Possibly there exist several types of defects on the edges while edge-banding on the wooden panel in furniture manufacturing. Therefore, the scope is to achieve higher accuracy by raw image and laser-aligned image feature extraction using deep learning algorithms for final result defect classification in WDD-DL by AOI. The WDD-DL system uses Gabor, Harris corner, morphology, structured light detection and curvature calculation for pre-processing and InceptionResnetV2 Convolutional Neural Network algorithm to attain the best results. The applications of this work can be found in quality control of the furniture manufacturing industry for an edge, corner, joint defect detection of the wooden panels. The WDD-DL achieves best results as the precision, recall and F1 score are 0.97, 0.90 and 0.92, respectively. The experiments demonstrate higher accuracy achievement as compared to other methods with overkill and escape rate analysis. Ultimately, the discussion section provides an interesting experience sharing about the necessary factors for implementing the WDD-DL in real-time industrial operations.

Machine learning (ML)-powered vision for robotic inspection has accelerated with smart manufacturing, enabling automated defect detection and classification and real-time process optimization. This review provides insight into the current landscape and state-of-the-art practices in smart manufacturing quality control (QC). More than 50 studies spanning across automotive, aerospace, assembly, and general manufacturing sectors demonstrate that ML-powered vision is technically viable for robotic inspection in manufacturing. The accuracy of defect detection and classification frequently exceeds 95%, with some vision systems achieving 98-100% accuracy in controlled environments. The vision systems use predominantly self-designed convolutional neural network (CNN) architectures, YOLO variants, or traditional ML vision models. However, 77% of implementations remain at the prototype or pilot scale, revealing systematic deployment barriers. A discussion is provided to address the specifics of the vision systems and the challenges that these technologies continue to face. Finally, recommendations for future directions in ML-powered vision for robotic inspection in manufacturing are provided.

In this study, a machine vision system is developed to achieve fabric inspection and defect classification processes automatically. The system consists of an image acquisition hardware and an image processing software. A simple and portable system was designed so that it can be adapted easily to all types of the fabric inspection machines. The software of the system consists of defect detection and classification algorithms. The defect detection algorithm is based on wavelet transform, double thresholding binarization, and morphological operations. It was applied real time via a user interface prepared by using MATLAB® program. The defect classification approach is based on gray level co-occurrence matrix and feed forward neural network. Five commonly occurring defect types, warp lacking, weft lacking, soiled yarn hole, and yarn flow, were detected and classified. The defective and defect-free regions of the fabric were detected with an accuracy of 93.4% and the defects are classified with 96.3% accuracy rate.

Industrial automatic fabric inspection system, a critical technology in the industry, enhances both total production quantity and quality compared to conventional inspection techniques. This study aims to create a reliable and effective real-time automated visual inspection system for fabrics, focusing on defect detection. The goals of the study can be stated as; installing a system with advanced technology for capturing and processing images swiftly, the development and deployment of a system capable of autonomously learning and scanning fabrics in use, and the creation of a smart framework for accurate fabric defect detection and classification. We focus on the development of unsupervised fabric defect detection using a convolutional autoencoder model, and defect classification using a convolutional neural network model, which takes input as the feature vector generated by the convolutional autoencoder. The experimental outcomes have displayed significant success rates in both detecting defects and classifying them, confirming the effectiveness of the framework in real-time visual inspection systems.

Silicon Carbide continues to become more mainstream and enjoy adoption in various high voltage applications. With the growth of volume and variety of devices shipping to customers, it is important to understand the yield and reliability implications of all the defects that are present in the substrates and epitaxial layers. In our previous works [1,2] we published ways of detecting and screening reliability causing crystal defects. There are however a variety of visible defects present both in the bare substrates and after epitaxial growth. In this work we present the detection and classification of these defects into killer and non-killer defects as measured by electrical tests after device fabrication. This methodology enables the accurate prediction of yield before starting the wafers for device fabrication. We then extend this method to predict yield on various product dies sizes and validate those results experimentally. We have worked to detect all the visible defects present in our epitaxial layers. However due to the large variety of defects present and their varying effect on device performance, just detection is not enough. The defects have to be finely classified into various defect types to gauge their electrical impact. We have classified the defects into various types of triangular defects, fall-down particles, carrots, scratches, pits, step bunching, V-type defects [2,3], large and small topographic defects and defects related to manual handling. Through an iterative process involving electrical characterization of fabricated Schottky barrier diodes (SBD) we have determined which of these classes are killer defects. It is found that mainly fall-down particles, a subset of triangular defects, carrots and large topographic defects are the main causes of electrical failure at wafer sort. Other defects like fainter obtuse triangular defects [4], V-type defects, scratches, step-bunching and pits do not cause outright fails. Second degree effects are observed for many of these non-killer defect types. Other purely crystal defects having no visible defect signatures like basal plane dislocations, stacking faults, grain boundaries were also detected and classified. However these were also determined not to have any influence on SBD device yield at wafer sort. Once the killer defects are determined after epitaxial growth, we created a yield model based on the actual die grid of our actual physical device layout. The yield prediction of this model was tested experimentally and validated for different wafers across multiple lots. A very close agreement of ±2% was observed across all the wafers. This model enables us to predict wafer yield right after epitaxy and before starting the wafers in the fabrication line very accurately. As product lines involve multiple current ratings with different die sizes, a further enhancement of this model was done to predict the yield on a wide variety of die sizes corresponding to device current ratings ranging from 2 Amps to 20 Amps. The predicted yield was validated experimentally and was found to be accurate within ±0.5%. The accuracy of both these models enables very efficient binning of epitaxial wafers to the correct device die size to ensure optimum yield at electrical wafer sort. It is also important to gauge the impact of the non-killer defects on the electrical properties of various devices. Our analysis shows that some of these defects cause slightly increased leakage currents both in the forward conduction and reverse blocking conditions without causing the devices to fail. More extensive measurements and reliability testing is needed to understand some of the longer term impacts of the presence of these classes of defects in the SiC material. Both the above yield models along with the validation against experimental data will be presented. A detailed description of the various classified defect types will be presented. Second order effects of the non-killer defect types on electrical tests will also be discussed. [1] H. Das et al. ECS Transactions, 69 (11) 29-32 (2015) [2] H. Das et al. ECS Transactions, 75 (12) 233-237 (2016) [3] H. Wang et al. Materials Science Forum, Vols. 778-780, 332 (2014) [4] G. Chung et al. Materials Science Forum Vols. 679-680, 123 (2011)

Sieve mesh defect detection is a critical yet challenging task in industrial quality control. Traditional manual inspection methods suffer from low efficiency and limited accuracy due to reliance on human vision. To overcome these limitations, this paper proposes an automated defect detection system for high-precision 325-mesh sieves based on machine vision. The system utilizes a high-resolution line-scan camera to acquire detailed images of the sieve mesh. A defect detection method based on an improved YOLOv4 architecture is then developed. Specifically, the original CSPDarknet53 backbone is replaced with an enhanced VGG-16 network to reduce computational complexity. Additionally, a multiscale spatial pyramid pooling (SPP) module is introduced to improve the detection of defects with varying sizes. The proposed method is evaluated on a self-constructed sieve defect dataset. Experimental results demonstrate that the system significantly improves both the accuracy and efficiency of defect detection, making it well-suited for practical applications in automatic sieve inspection.

As a crucial component of the photovoltaic (PV) power generation system, effective defect detection of PV cells can help to ensure the power generation efficiency and the system safety. The photo-thermal-radiation infrared thermography experimental system has been built, and the infrared images of normal and defective PV cells were collected by an IR camera. An effective convolutional neural network (CNN) based model with the residual architecture is designed here to detect and classify the defects. An offline date augmentation method is performed to overcome the insufficient image dataset in the step of dataset establishment, so as to improve the defects classification capability of the designed CNN model. The defects classification accuracy of the proposed model is obtained as 96.96% with 3-class. In order to verify the defect classification ability of the designed CNN model, other four models were adopted for comparative analysis. The results indicate that the proposed CNN model has advantages in defect classification of PV cells. The completed work can provide meaningful exploration in the industrial application of defect detection of PV cells.

Purpose. The main goal of this study is to review and systematize the materials and defects of the cylinder block and cylinder liners of internal combustion engines (ICE), as well as to analyze the issues and current trends in defect detection during repair. Methodology. The main problems currently associated with the defect detection of cylinder blocks and liners of ICEs include the following: the rapid growth of motorization, which leads to an increase in the number of cylinder blocks and liners requiring defect inspection; the use of manual labor, which makes the defect detection process labor-intensive and costly; the need for classification of defect types and their automatic recognition; the consideration of operating conditions influencing defect development; the necessity of implementing new defect detection processes; the need for quantitative wear assessment and prediction of component residual life; the adaptation of inspection methods and evaluation criteria to new materials. The study employs the following research methods: theoretical analysis (classification, systematization), review and synthesis of known scientific results, and empirical methods (comparison, observation). Findings. The authors conducted a review and systematization of structural materials and the main defects of cylinder blocks and liners of ICEs. The main issues in modern defect detection were identified as follows: insufficient process automation, low accuracy of visual and instrumental inspection methods, limited adaptability of existing inspection techniques to new materials, and the need for quantitative evaluation of component residual life. The development trends in defect detection systems are defined as the transition to automated and computerized inspection technologies, the creation of databases of typical defects, and the development of algorithms for predicting the residual life of components. Originality. The scientific novelty of this work lies in a comprehensive approach to analyzing structural materials, defects, and modern trends in defect detection of cylinder blocks and liners of ICEs. Materials used in component manufacturing have been systematized, a defect classification has been proposed, and the factors influencing wear intensity and defect formation have been identified. The trends in the development of defect detection systems have been established toward automation, increased sensitivity, and higher inspection accuracy. Practical value. The conducted analysis contributes to improving the efficiency of defect detection and repair processes of cylinder blocks and liners of ICEs, as well as to the development of informational databases of typical defects and the conditions under which they occur.

Sanitary ceramic products, such as toilet and wash basin, are widely used in our daily life. Sanitary ceramics are expected to have some excellent physical properties, such as corrosion resistance, easy cleaning, and low water absorption. However, surface defects in sanitary ceramics are inevitable due to complex production processes and changing production environment. Therefore, surface defect detection must be performed in the manufacturing process of sanitary ceramics. There are many types of surface defects in sanitary ceramics, and different types of defects have large differences in characteristics and scales. Traditional detection methods with artificially designed features and classifiers are difficult to apply in this context. In addition, there are few studies on surface defect detection methods of sanitary ceramics based on deep neural networks. In this article, a lightweight real-time defect detection network based on the lightweight backbone MobileNetV3 is presented. The proposed network achieves multi-scale detection of surface defects in sanitary ceramics with a multi-layer feature pyramid. Combining region proposal network and anchor-free method, real-time defect detection is achieved. Finally, a detection head with channel attention structure and a low-level mixed feature classification strategy is used to perform defect classification with higher accuracy. Experimental results show that the proposed approach achieves at least 22.9% detection speed improvement and 35.0% average precision improvement while reducing memory consumption by at least 8.4% compared with the classic one-stage SSD, YOLO V3 and two-stage Faster R-CNN methods.

As a key component in the hydraulic brake system of automobiles, the brake joint directly affects the braking performance and driving safety of the vehicle. Therefore, improving the quality of brake joints is crucial. During the processing, due to the complexity of the welding material and welding process, the weld seam is prone to various defects such as cracks, pores, undercutting, and incomplete fusion, which can weaken the joint and even lead to product failure. Traditional weld seam detection methods include destructive testing and non-destructive testing; however, destructive testing has high costs and long cycles, and non-destructive testing, such as radiographic testing and ultrasonic testing, also have problems such as high consumable costs, slow detection speed, or high requirements for operator experience. In response to these challenges, this article proposes a defect detection and classification method for laser welding seams of automotive brake joints based on machine vision inspection technology. Laser-welded automotive brake joints are subjected to weld defect detection and classification, and image processing algorithms are optimized to improve the accuracy of detection and failure analysis by utilizing the high efficiency, low cost, flexibility, and automation advantages of machine vision technology. This article first analyzes the common types of weld defects in laser welding of automotive brake joints, including craters, holes, and nibbling, and explores the causes and characteristics of these defects. Then, an image processing algorithm suitable for laser welding of automotive brake joints was studied, including pre-processing steps such as image smoothing, image enhancement, threshold segmentation, and morphological processing, to extract feature parameters of weld defects. On this basis, a welding seam defect detection and classification system based on the cascade classifier and AdaBoost algorithm was designed, and efficient recognition and classification of welding seam defects were achieved by training the cascade classifier. The results show that the system can accurately identify and distinguish pits, holes, and undercutting defects in welds, with an average classification accuracy of over 90%. The detection and recognition rate of pit defects reaches 100%, and the detection accuracy of undercutting defects is 92.6%. And the overall missed detection rate is less than 3%, with both the missed detection rate and false detection rate for pit defects being 0%. The average detection time for each image is 0.24 s, meeting the real-time requirements of industrial automation. Compared with infrared and ultrasonic detection methods, the proposed machine-vision-based detection system has significant advantages in detection speed, surface defect recognition accuracy, and industrial adaptability. This provides an efficient and accurate solution for laser welding defect detection of automotive brake joints.

The textile sector is crucial to India's economy, and one of its most significant facets is the management of fabric quality. In computer vision, texture analysis is used for the purposes of defect detection, classification, and segmentation. In order to save manufacturing time and costs, this paper explains a fundamental method for identifying various fabric flaws in the textile industry. An essential part of quality control, automated fabric inspection systems help find textile flaws quickly and accurately while also cutting down on human labor. In this work, we assess two classifiers—the NN classifier and the SVM classifier—on a 5000 fabric image samples, TILDA dataset for the purpose of recognizing six distinct defects: holes, horizontal defects, reed markings, burls, slubs, stains, and double end marks. One of the most fundamental and significant components of modern fashion is textile. We can't fathom a world devoid of textiles. Another essential problem in the textile production sector is fabric quality monitoring. When it comes to finding various types of fabric defects, such as holes, slubs, oil stains, etc., automatic defect detection is seen to be quite interesting. Using the provided fabric samples, this study introduces a novel method for fault and defect identification. The five-step process for textile defect detection begins with collecting picture samples from the industry-standard TILDA dataset. Grayscale transformation is a preprocessing technique that is used to enhance the picture quality and eliminate undesired noise. As a last step in feature extraction, SVM takes the gray-level co-occurrence matrix (GLCM) into account. The testing phase, however, involves validating these two classifiers using the test data and calculating their sensitivity, specificity, and accuracy.

BackgroundEarly detection of marginal bone loss is vital for treatment planning and prognosis of teeth and implant. This study was conducted to assess diagnostic accuracy of CBCT compared to intra-oral (IO) radiography for detection, classification, and measurement of peri-implant bone defects in an animal model.MethodsFifty-four mandible blocks with implants were harvested from nine male health adult beagle dogs with acquisition of IO, CBCT and micro-CT images from all samples. Peri-implant bone defects from 16 samples were diagnosed using micro-CT and classified into 3 defect categories: dehiscence (n = 5), infrabony defect (n = 3) and crater-like defect (n = 8). Following training and calibration of the observers, they asked to detect location (mesial, distal, buccal, lingual) and shape of the defect (dehiscence, horizontal defect, vertical defect, carter-like defect) utilizing both IO and CBCT images. Both observers assessed defect depth and width on IO, CBCT and micro-CT images at each side of peri-implant bone defect via CT-analyzer software. Data were analyzed using SPSS software and a p value of < 0.05 was considered as statistically significant.ResultsOverall, there was a high diagnostic accuracy for detection of bone defects with CBCT images (sensitivity: 100%/100%), while IO images showed a reduction in accuracy (sensitivity: 69%/63%). Similarly, diagnostic accuracy for defect classification was significantly higher for CBCT, whereas IO images were unable to correctly identify vestibular dehiscence, with incorrect assessment of half of the infrabony defects. For accuracy of measuring defect depth and width, a higher correlation was observed between CBCT and gold standard micro-CT (r = 0.91, 95% CI 0.86–0.94), whereas a lower correlation was seen for IO images (r = 0.82, 95% CI 0.67–0.91).ConclusionsThe diagnostic accuracy and reliability of CBCT was found to be superior to IO imaging for the detection, classification, and measurement of peri-implant bone defects. The application of CBCT adds substantial information related to the peri-implant bone defect diagnosis and decision-making which cannot be achieved with conventional IO imaging.

In this paper a system for web surface inspection is described. It has three parts: an image acquisition part, a defect detection part, and a defect classification part. The self-organizing maps (SOMs) are used both in defect detection and in defect classification which makes the system adaptable to different types of surfaces and defects. Our main focus is on defect classification where a generic content-based image retrieval (CBIR) system called PicSOM is utilized. The PicSOM uses tree-structured SOMs (TS-SOMs) and relevance feedback. It is trained with the feature sets of the defects in the database. For defect description, features from the MPEG-7 standard (the homogeneous texture, the color structure, and the edge histogram) are used and for the shape description our own shape feature set is applied. Results indicate that the system works with a high level of success.