Industrial Vision Systems and Defect Detection

Design of Economic Order Quantity on Polyester Yarn Raw Material Based on Artificial Neural Network Forecasting

2025-06-26

Muhammad Fathurrohman , Nunung Nurhasanah , Wawan Tripiawan | Deleted Journal

Polyester yarn is a key raw material in textile manufacturing due to its durability and affordability. PT ABC relies on external suppliers for polyester yarn, making inventory management crucial for … Polyester yarn is a key raw material in textile manufacturing due to its durability and affordability. PT ABC relies on external suppliers for polyester yarn, making inventory management crucial for production efficiency. However, the company's current ordering approach has led to occasional stock shortages, impacting operations. This study develops an inventory control model using the Economic Order Quantity (EOQ) method, incorporating safety stock and reorder point calculations to minimize stockouts and reduce inventory costs. Additionally, Artificial Neural Networks (ANN) are used to forecast demand for 2022, improving estimation accuracy. Based on historical demand data from 2019 to 2021, the EOQ method lowers inventory costs compared to the company’s approach, achieving efficiency gains of 19%, 12%, and 29%, saving IDR45,745,000, IDR23,735,000, and IDR98,020,000, for each respective year. The ANN model utilizing the TrainLM training function achieves the lowest Mean Squared Error (MSE) of 0.063528 and forecasts a total raw material requirement of 2,510,628 kg for 2022. The EOQ value for 2022 is set at 44,817 kg, with safety stock and reorder point levels of 8,438 kg and 29,360 kg, respectively.Keywords – Artificial Neural Network, Economic Order Quantity, Inventory, Reorder Point, Safety Stock.

AI – Powered PCOD Detection Platform

2025-06-25

Prasad Vadkar | INTERANTIONAL JOURNAL OF SCIENTIFIC RESEARCH IN ENGINEERING AND MANAGEMENT

Polycystic Ovary Syndrome (PCOS), also referred to as Polycystic Ovarian Disease (PCOD), is one of the most prevalent endocrine disorders affecting women of reproductive age worldwide. It is a leading … Polycystic Ovary Syndrome (PCOS), also referred to as Polycystic Ovarian Disease (PCOD), is one of the most prevalent endocrine disorders affecting women of reproductive age worldwide. It is a leading cause of anovulatory infertility and is characterized by hormonal imbalances that result in symptoms such as irregular menstrual cycles, excessive weight gain, acne, hair loss, and skin darkening. Despite its high prevalence, early-stage detection and accurate prediction of PCOS remain challenging due to limitations in existing diagnostic methods and treatment strategies. This research aims to address these challenges by developing an advanced, computer-aided detection system utilizing machine learning (ML) and deep learning (DL) techniques. The system leverages ovary ultrasound (USG) images— one of the most reliable diagnostic modalities for PCOS—and incorporates a Convolutional Neural Network (CNN) for robust feature extraction. To enhance classification performance, a stacking ensemble model is implemented using a combination of traditional machine learning classifiers as base learners and bagging or boosting techniques as meta-learners. The CNN architecture is further strengthened through transfer learning and modern feature selection techniques such as I-SQUARE and CHI-square. The study involves training and evaluating the proposed model on a dataset comprising 4000 ovary USG images, sourced from a publicly available PCOS dataset on Kaggle by Parson Kottarathil. Additionally, five ML classifiers— Random Forest, Support Vector Machine (SVM), Logistic Regression, Gaussian Naïve Bayes, and K-Nearest Neighbors—were evaluated on a subset of the dataset containing 41 clinical and physiological features, with the top 30 features selected for classification. Experimental results indicate that the Random Forest Classifier outperforms other models in terms of accuracy and reliability. The proposed hybrid system significantly improves detection accuracy while reducing execution time, making it a promising solution for aiding healthcare professionals in the early diagnosis and management of PCOS. This research lays the foundation for intelligent and scalable PCOS detection systems that integrate clinical data and medical imaging, thereby advancing personalized and timely healthcare delivery for women suffering from this condition. Keywords: Polycystic Ovary Syndrome (PCOS), Machine Learning, Deep Learning, Convolutional Neural Network (CNN), Medical Imaging, Ultrasound, Classification, Data Mining, Healthcare, Prediction System, Early Diagnosis

A Review of deep learning-based detection and segmentation of steel surface defects: advances in resolving intra-class differences and inter-class similarities

2025-06-25

Zibo Zhao | Advances in Engineering Innovation

Steel surface defect detection is a key part of the production process in the steel industry. The traditional manual inspection methods are inefficient and costly. With the rapid development of … Steel surface defect detection is a key part of the production process in the steel industry. The traditional manual inspection methods are inefficient and costly. With the rapid development of deep learning technology, automatic detection and segmentation of steel surface defects based on deep neural networks has received widespread attention and demonstrated good performance in several real-world scenarios. However, challenges remain due to the obvious inter-class similarity and intra-class variation problems in steel defect images, impacting accuracy and robustness. This paper systematically summarizes the representative researches on solving the inter-class similarity and intra-class variation problems in recent years, and focuses on analyzing the innovations of different methods in network structure design. In addition, this paper also discusses the shortcomings of the current research, and proposes a new idea of fusion of mainstream modeling method. By employing a subcomparison module to enhance feature similarity within classes and differentiate across classes, followed by pyramid feature fusion to optimize computational efficiency, this study aims to advance high-precision intelligent recognition of steel surface defects. This study reveals that the approach not only addresses existing challenges but also provides a foundation for future advancements in steel defect detection technologies.

Deep Learning for Robust Defect Detection in Industrial X-Ray Images with Minimal Real-Data Dependency

2025-06-24

Bishwajit Mohan Gosswami , Rajib Kumar Chanda , Tobias Schön +1 more | e-Journal of Nondestructive Testing

Defect detection in aluminum casting components via X-ray imaging is essential for quality assurance but is hindered by challenges like low contrast and diverse defect geometries. Traditional automated defect recognition … Defect detection in aluminum casting components via X-ray imaging is essential for quality assurance but is hindered by challenges like low contrast and diverse defect geometries. Traditional automated defect recognition (ADR) methods often lack robustness across varying datasets. This study introduces a convolutional neural network (CNN) framework that combines simulated and real X-ray images to address these limitations. A simulation pipeline generates realistic 2D radiographs by embedding segmented 3D defect models, obtained from industrial CT scans, into arbitrary 3D objects, followed by 2D projection. This framework incorporates a carefully curated real-image dataset with annotations to overcome the gap between synthetic and real data. The model is built on a U-Net architecture optimized for single-channel grayscale images, using bilinear up-sampling to reduce overfitting while improving generalization. A tiling-based approach enables compatibility with arbitrarily shaped images, enhancing adaptability across datasets. Additionally, transfer learning minimizes the need for extensive real-world data and allows efficient retraining for new scenarios. Despite annotation inaccuracies affecting conventional metrics like precision and recall, qualitative evaluation highlights the model's effectiveness in detecting subtle defects. Its scalability and ability to infer from diverse X-ray images position it as a transformative tool for automated defect detection, reducing reliance on extensive real data and improving industrial quality control.

Flange Surface Defect Detection Based on the MROC-YOLO Model

2025-06-24

Guo Yuan , Yiqiu Fan , Hongquan She +1 more | Engineering Research Express

Abstract Aiming to address the issues of low accuracy and high leakage rate in multi-scale defect detection on flange surfaces in industrial scenarios, this paper proposes the MROC-YOLO algorithm, which … Abstract Aiming to address the issues of low accuracy and high leakage rate in multi-scale defect detection on flange surfaces in industrial scenarios, this paper proposes the MROC-YOLO algorithm, which improves upon YOLOv8m. A heterogeneous backbone network, termed MRNet, is constructed by integrating MobileNetV3 and ResNet to strengthen multi-scale feature extraction. Traditional convolutions are replaced with the ODConv dynamic convolution module, improving target area sensing and reducing missed detections. Furthermore, the CSCF module, which combines the SCConv module with two C2f layers, is introduced to minimize redundant information interference and enhance the detection accuracy of small target defects, enabling more efficient and precise detection. To tackle the scarcity of open flange defect data, the FL-DET dataset is constructed, encompassing four defect types: pits, patches, pitted surfaces, and scratches. Experiments show that the average detection accuracy of the model reaches 87.5%, which is 11.0%, 4.0%, and 13.9% higher than that of the original YOLOv8m in terms of precision, recall, and [email protected] indexes, respectively. Additionally, the computation amount is reduced to 20.0 GFLOPs, with a detection speed of 53.7 frames per second. These results demonstrate that the MROC-YOLO algorithm can effectively improve detection accuracy, reduce the missed detection rate, and is more suitable for real-time industrial detection.

Morphology-prior-focused detection method for high-performance localization of fabric defects

2025-06-24

Kaixin Xu , Ning Dai , Xudong Hu +2 more | Textile Research Journal

With the growing demand for efficient and high-precision quality control in textiles, the limitations of traditional fabric defect detection methods have become increasingly apparent, particularly in handling complex backgrounds, diverse … With the growing demand for efficient and high-precision quality control in textiles, the limitations of traditional fabric defect detection methods have become increasingly apparent, particularly in handling complex backgrounds, diverse defect types, and small-object detection. To address these challenges, this study proposes a fabric defect detection method based on morphological prior features. By deeply integrating morphological prior information into the feature extraction and fusion modules, the robustness and accuracy of defect detection are improved significantly. Specifically, the divergent path feature enhancement module enhances the detection of small-object defects through fine-grained feature extraction. The oriented space and multiscale feature extraction module combines multiscale and directional feature extraction techniques to improve the recognition of defects with extreme aspect ratios. Furthermore, the efficient multipath feature fusion network achieves comprehensive capture of defect features by integrating shallow and deep features. In addition, the proposed fabric mosaic data augmentation strategy dynamically adjusts the cropping offset points, effectively preserving the pixel and feature integrity of target defects under conditions of data scarcity and imbalance. Experimental results demonstrate that the proposed model achieves a precision of 75.5%, a recall of 75.8%, and an F1 score of 75.6%. The [email protected] is improved by 9.8% compared with the baseline model. The proposed approach strikes a favorable balance between detection accuracy, computational complexity, and inference speed, showcasing excellent generalization capability and practical application potential.

An improved design for ferrous metal detection and removal with edge-aware adaptive algorithm in nonwoven fabric production lines

2025-06-24

P. Wang , Chengzu Li , Deming Liu +4 more | Textile Research Journal

With the advancement of manufacturing toward intelligence and digitalization, automation technologies have become integral across various industries. However, the nonwoven fabric industry continues to face challenges such as high labor … With the advancement of manufacturing toward intelligence and digitalization, automation technologies have become integral across various industries. However, the nonwoven fabric industry continues to face challenges such as high labor costs and low automation levels, creating an urgent need for technological innovation to enhance production efficiency and product quality. This study introduces an intelligent ferrous metal detection and removal automation system, designed specifically for industrial deployment in nonwoven fabric production lines. It integrates an optimized coil arrangement, an edge-aware adaptive cutting optimization algorithm, and a pneumatic–hydraulic driven removal device, ensuring high detection precision, minimal material waste, and seamless integration into continuous production workflows. The system employs a triangular coil arrangement to optimize magnetic field distribution, significantly improving detection sensitivity and accuracy. The edge-aware adaptive cutting algorithm precisely locates ferrous metal impurities while minimizing damage to material edges, thereby preventing twisting or deformation of the fabric during transport that could lead to machine jamming. In addition, a pneumatic–hydraulic driven removal device is designed to efficiently excise impurity areas with high precision. Experimental results demonstrate the system’s capability to detect and remove iron particles as small as 1.2 mm in diameter, substantially improving automation levels and production efficiency in nonwoven fabric manufacturing.

Fabric defect detection based on improved YOLOv8s and machine vision

2025-06-24

Can Hao , Ying Wang , Chengjun Cui +2 more | Textile Research Journal

As a key part of the fabric quality control for the textile industry, it is important to detect fabric defects quickly, accurately, and efficiently. To address the missed detection of … As a key part of the fabric quality control for the textile industry, it is important to detect fabric defects quickly, accurately, and efficiently. To address the missed detection of the defects with tiny, extreme aspect ratios, and low contrast in fabric images, an improved YOLOv8s-HCG algorithm is proposed in this paper. First, the histogram specification algorithm is used to enhance the defect feature expression of low-contrast fabric images at the input side. Second, the content-aware reassembly of features (CARAFE) operator is used instead of the nearest-neighbor interpolation operator for up-sampling in the YOLOv8s. The CARAFE operator aggregates contextual information in a large receptive field to reassemble abundant detailed features, and reassembles features by using targets in the feature map with the adaptive reassembled kernel. Finally, the global attention mechanism module is added into the YOLOv8s neck network to construct the interdependence relationship between fabric defect image channels and spatial dimensions to capture important features. The algorithm was validated on a self-made fabric dataset and two public fabric datasets. The comparison experimental results show that the detection performance of the proposed algorithm in this paper is better than the other algorithms for defects with extreme aspect ratios, tiny, and low contrast. This research is of great significance to the fabric defect detection industry.

Debris Pattern Recognition Based on Visual Sensor and Image Stitching Technology

2025-06-24

Wenhao Zheng , Jih-Gau Juang | Sensors and Materials

Transformer-Based Vision-CNC Fusion for Real-Time Error Compensation and Adaptive Optimization in Intelligent Machining Systems

2025-06-24

Z. Li | International Journal of Knowledge Management

It is difficult for current intelligent machining systems to accurately compensate for machining errors caused by thermal deformation of the spindle and fixture offset. This paper introduces a transformer-based vision-computerized … It is difficult for current intelligent machining systems to accurately compensate for machining errors caused by thermal deformation of the spindle and fixture offset. This paper introduces a transformer-based vision-computerized numerical control fusion optimization model, combines computer vision technology with the computerized numerical control system, and drives error adaptive compensation through visual monitoring to improve machining accuracy and system stability. The experimental results show that the compensation effect of the model under different working conditions is significant, the error correction ability in the processing of various workpieces is strong, and the average precision is 87.64%. In ultra-high-speed mode, path optimization reduces the processing time to 180 s and the tool wear to 10.5 μm. The research in this paper provides an optimization method for intelligent machining, which has broad application prospects.

Few-Shot Learning for Industrial Defect Detection Using Meta-Learning Techniques

2025-06-24

Xin Nie , Chee Fen Yu | International Journal of Scientific Research and Management (IJSRM)

Industrial defect detection plays a pivotal role in maintaining quality and safety across manufacturing processes. Traditional deep learning methods for visual inspection and defect classification rely heavily on large volumes … Industrial defect detection plays a pivotal role in maintaining quality and safety across manufacturing processes. Traditional deep learning methods for visual inspection and defect classification rely heavily on large volumes of annotated data, which are often costly and difficult to obtain in real-world industrial settings. This data scarcity poses a significant barrier to deploying robust and generalizable computer vision models for rare or evolving defect types. To address this challenge, we explore the use of few-shot learning (FSL), a paradigm that enables models to generalize to new classes with only a handful of labeled examples. Building upon this foundation, we integrate meta-learning strategies specifically model-agnostic algorithms and metric-based learners—that are trained to quickly adapt to new tasks with minimal supervision. To further enhance feature discrimination under limited data conditions, we incorporate contrastive learning, which encourages the model to learn meaningful representations by maximizing inter-class differences and minimizing intra-class variations through self-supervised instance discrimination. This study presents a hybrid framework combining contrastive pretraining with meta-learning to achieve superior performance in few-shot defect detection tasks. Experiments conducted on benchmark industrial datasets such as MVTec AD and DAGM demonstrate that our approach outperforms conventional few-shot baselines in both accuracy and generalization. The inclusion of contrastive learning boosts feature separability and improves recognition performance in low-shot settings. Our findings indicate that the proposed method is a viable and scalable solution for deploying intelligent inspection systems in real-world manufacturing environments, especially where annotated data is limited or difficult to collect.

From Color to Spectrum—The Next Leap in Display Metrology

2025-06-24

| News Digital Object Group

Classical machine learning and artificial neural network (ANN) to predict rejection in weaving industry

2025-06-24

Toufique Ahmed | Journal of Electrical Systems and Information Technology

Abstract Weaving is a crucial technology in textile production. Rejection is an inherent aspect of the industrial output. The textile industry experiences significant wastage due to wrong assumptions about rejection. … Abstract Weaving is a crucial technology in textile production. Rejection is an inherent aspect of the industrial output. The textile industry experiences significant wastage due to wrong assumptions about rejection. This study found that fabric allowance can be predicted from required gray fabrics by using logarithmic function. Similarly, required gray fabrics can be calculated from a linear equation with 99% goodness of fit. Crimp percentage and warp beam length (yards) are most informative for fabric production and rejection, derived from mutual information. The first six Principle Components Analysis (PCA) components account for 94% of the information, underscoring crucial features like required gray fabrics, fabric allowances, yarn density, and yarn fineness. Besides, the PC1–PC2 Biplot represents the required gray fabric, required finished fabric, and required warp beam length, which have the highest impact on the first principal components (PC1). Then, 14 classical machine learning techniques were applied to the datasets. Among these, random forest, decision tree, and LightGBM demonstrated the highest accuracy. The optimal hyperparameters for these best-performing algorithms were also selected using RandomizedSearchCV. Interestingly, traditional machine learning models achieved more than 95% accuracy without any data preprocessing. In contrast, artificial neural networks (ANN) require data preprocessing to achieve high accuracy rates. Additionally, adjusting hidden layers adjustment is crucial. A seven-layer ANN model with one hot encoded (OHE) and scaled with a min–max scaler demonstrates an accuracy exceeding 96%.

Hierarchy-based fuzzy segmentation and marker learning layer: theory and algorithms

2025-06-24

Gabriel Barbosa da Fonseca , Romain Negrel , Benjamin Perret +2 more | Journal of Mathematical Imaging and Vision

Improving Implantation Accuracy with Visual Direction-Indicating Guide (VDING)

2025-06-24

Yufeng Liang | International Journal of Oral and Maxillofacial Surgery

Combining the RoboCT Technology with ISAR using ADR for In- and At-Line Inspection of Cast Parts

2025-06-24

Frank Sukowski , Thomas F. Stocker , Markus Eberhorn +2 more | e-Journal of Nondestructive Testing

The still young RoboCT technology is based on two cooperating robots, each carrying one of the imaging X-ray components, the radiation source and detector. Their freedom of movement and flexibility … The still young RoboCT technology is based on two cooperating robots, each carrying one of the imaging X-ray components, the radiation source and detector. Their freedom of movement and flexibility make it possible to examine large and complex-shaped objects even in otherwise hard-to-reach areas using 2D radioscopic imaging and even 3D computed tomography (CT). Over the past few years, RoboCT has been developed as a laboratory testing system to product maturity. However, production integrated inspection poses further requirements for a testing system. In addition to robust and secure teach-in of the components to be tested, a high degree of automation is generally necessary for conducting test procedures, especially for the evaluation and assessment of image data, due to cycle time requirements. One of the most advanced software systems for automated defect detection (ADR) is ISAR, which has been used for over 20 years for 2D testing of wheels and structural cast components. A further development of RoboCT for integrated production quality testing is the seamless integration of ISAR into the RoboCT system solution, which even allows the execution of escalating test procedures, i.e., an automated execution of 3D CT scans when there is an unclear result at the corresponding position in the 2D transmission image that needs to be verified.

Quantifying the Impact of XCT Data Compression on Defect Shapes

2025-06-24

Thomas Lang , Christoph Heinzl | e-Journal of Nondestructive Testing

X-ray computed tomography represents an essential tool in non-destructive analysis of industrial components in a wide variety of applications, ranging from fast inline inspection, to imaging of very large structures … X-ray computed tomography represents an essential tool in non-destructive analysis of industrial components in a wide variety of applications, ranging from fast inline inspection, to imaging of very large structures such as cars, and micro-scale characterization of novel material candidates. Due to the high demand regarding detailed analyses, both hardware and software components are improving continuously, yielding larger quantities of three-dimensional datasets. Data compression techniques aim at reducing storage requirements, but yet there is a risk of losing valuable information if a compression is too strong. In this work, we thus investigate a three-dimensional lossy compression technique tailor-made for XCT datasets and evaluate the deterioration of defect representations due to compression. For this purpose, a virtual XCT dataset is generated by modeling defects of randomized shapes and using X-ray simulation to generate realistic XCT data, whereas the pipeline will be applied on real XCT data as well in the full paper. While varying the maximum admissible compression error, an image processing pipeline will segment the defects and evaluate a selection of shape factors. When increasing the compression factor, the quantitative results indicate a continuous drop in shape factors, i.e., a moderate loss in shape is encountered at high compression rates.

Effective SSD Based Multiple Object Detection in Videos

2025-06-24

Jehu Sheran | International Journal for Research in Applied Science and Engineering Technology

This paper aims to analyze the objects in any format of video or filmography. The web application is just simple like uploading any format of video, it processes the video … This paper aims to analyze the objects in any format of video or filmography. The web application is just simple like uploading any format of video, it processes the video and gives the finest progressive output with a bounding box. For processing the video for object detection, the SSD [algorithm] is used because the SSD [algorithm] has additional accuracy than the YOLO [algorithm]. And with SSD no other video format or other object detection processes are done, only for image classification SSD [algorithm] is used. In this project, introduced the framework of the video object detection process using the SSD [algorithm]. This technique works for surveillance video and any format of the video.

YOLO-LitePV: a lightweight detection algorithm for photovoltaic panel defects

2025-06-23

Xinru Li , Huaishu Hou , Yinyin Yu +2 more | Nondestructive Testing And Evaluation

Advancing photovoltaic cells defect detection in electroluminescence images through exploring multiple object detectors

2025-06-23

M. Waqar Akram , Jianbo Bai , Chen Xuan +3 more | Solar Energy Materials and Solar Cells

Augmenting energy efficiency in automotive paint ovens: a review of future prospects and potential for lean, six sigma, AI, and IoT integration

2025-06-23

Mohammad M. Almomani , Yazan O. Mayyas , Osama H. Alomari +3 more | Management of Environmental Quality An International Journal

Purpose Automobile manufacturing is an energy-conscious business, specifically paint ovens. With energy costs continuously increasing and environmental issues, more work must be conducted to make the paint ovens more energy-efficient. … Purpose Automobile manufacturing is an energy-conscious business, specifically paint ovens. With energy costs continuously increasing and environmental issues, more work must be conducted to make the paint ovens more energy-efficient. This project identified trends and avenues for adopting a Green Lean Six Sigma Energy Management System (GLSS-EnMS), Artificial Intelligence (AI), and Internet of Things (IoT) for improved energy efficiency in Electrostatic Powder paint ovens. The project aimed to transform the automotive industry by exploring trends and possibilities for implementing GLSS-EnMS, AI, and IoT technologies to improve the energy efficiency of the automotive paint oven process. Design/methodology/approach A literature review was conducted to evaluate the latest trends and problems of automotive paint oven energy efficiency technologies. Top car makers and oven paint producers were studied to find their approach and suggest energy efficiency improvement methods. The project considered the feasibility and application of integrating GLSS-EnMS, artificial intelligence, and Internet of Things technologies in the automobile paint oven processes. A framework is proposed to enhance paint quality, energy use, productivity, and system efficiency as a whole. A model was also developed to adopt GLSS-EnMS, AI, and IoT technologies to realize the optimal energy efficiency of car paint ovens. Findings The results and suggestions of this project will help automotive producers implement cost-saving and environmentally friendly energy management practices, ultimately making the industry more environmentally sustainable and economical. The GLSS-EnMS, AI, and IoT integration technologies into automotive paint oven processes will not save energy; they can also track and shut down the onset of issues from the very start, provide high availability and minimize maintenance expenses to join the new age of predictive maintenance. It can bring value by making smart grid integration and demand response programs within the power system more reliable. Originality/value This research offered an overview of trends and prospects for integrating the Green, Lean, and Six Sigma Energy Management System model with IoT and Deep Learning-based Predictive Energy Modeling (DL-PEM). It offers in-practice guidelines towards ISO 50001:2018 conformance and real-time data analysis, like automated energy-saving measures that form a true solution towards sustainability.

Automated Shrinkage and Gas Porosity Detection Using YOLO Model in Additive-Manufactured Aluminum Alloy Parts

2025-06-23

Laura Arias-Martínez , Francisco Jáñez-Martino , Eduardo Fidalgo +4 more | Integrating materials and manufacturing innovation

Abstract Developing intelligent and automatic systems to detect defects in additive manufacturing parts is still a great challenge. In this work, the binder jetting technology has been used to manufacture … Abstract Developing intelligent and automatic systems to detect defects in additive manufacturing parts is still a great challenge. In this work, the binder jetting technology has been used to manufacture molds for aluminum casting, and the defects of the parts obtained in these molds have been analyzed and compared with the parts obtained in molds manufactured with the traditional sand-casting method. The main defects obtained in both casting methods have been due to porosity, which is one of the most critical defects affecting the mechanical behavior of the parts. We propose a methodology to develop an automatic system for detecting these defects in casting parts. We also presented the Porosity Detection Corpus , a novel and publicly available dataset containing 204 pictures taken after cross-sectioning the manufactured parts, 102 for each casting technique. Then, we manually annotated the bounding boxes that include gas and shrinkage pores using two different labeling: only pores and their type. Finally, we trained and evaluated a pre-trained You Only Look Once (YOLO) model on the Porosity Detection Corpus, considering different thresholds in terms of recall. For detecting unique pores, we recommended using 25% of threshold, with a recall of 0.599. For classifying the type of pore, gas or shrinkage, we recommended a threshold of 25% with a mAP of 0.377.

A Method of Trash Balloon Classification Using Object Detection Models

2025-06-23

Seungjo Beom , Dongsu Kang | KIISE Transactions on Computing Practices

Quantifying Printing Quality for Printed Electrodes via Deep Learning and Spatial Association: Empowering Process Optimization

2025-06-22

Juhuhn Kim , Younsu Jung , Sajjan Parajuli +4 more | Advanced Intelligent Systems

Process optimization is a critical element in manufacturing, especially in the field of flexible and printed electronic (FPE) devices, particularly by a roll‐to‐roll (R2R) gravure printing. This technology is essential … Process optimization is a critical element in manufacturing, especially in the field of flexible and printed electronic (FPE) devices, particularly by a roll‐to‐roll (R2R) gravure printing. This technology is essential for the production of next‐generation electronic devices due to its high throughput rates and cost‐effective fabrication. However, the assessment of R2R gravure printing outcomes still relies on human visual inspection, introducing subjectivity susceptible to human fatigue. To address this challenge, a novel approach is proposed to quantify printing quality in real‐time using deep learning and spatial association. This not only reduces the need for human labor in evaluating the performance of the R2R gravure printing system but also ensures objective quantifiability. Quantification facilitates optimal process settings through response surface methodology, as printing quality can be expressed numerically. The article demonstrates the application of deep learning for quantitatively analyzing printed patterns in real‐time using the gate layer in the FPE devices as a reference sample, showing its effectiveness in optimizing the R2R gravure printing process.

Defect Detection and Error Source Tracing in Laser Marking of Silicon Wafers with Machine Learning

2025-06-22

H.M. Wang , Teng-To Yu , Wen-Fei Peng | Applied Sciences

Laser marking on wafers can introduce various defects such as inconsistent mark quality; under- or over-etching, and misalignment. Excessive laser power and inadequate cooling can cause burning or warping. These … Laser marking on wafers can introduce various defects such as inconsistent mark quality; under- or over-etching, and misalignment. Excessive laser power and inadequate cooling can cause burning or warping. These defects were inspected using machine vision, confocal microscopy, optical and scanning electron microscopy, acoustic/ultrasonic methods, and inline monitoring and coaxial vision. Machine learning has been successfully applied to improve the classification accuracy, and we propose a random forest algorithm with a training database to not only detect the defect but also trace its cause. Four causes have been identified as follows: unstable laser power, a dirty laser head, platform shaking, and voltage fluctuation of the electrical power. The object-matching technique ensures that a visible image can be utilized without a precise location. All inspected images were compared to the standard (qualified) product image pixel-by-pixel, and then the 2D matrix pattern for each type of defect was gathered. There were 10 photos for each type of defect included in the training to build the model with various labels, and the synthetic testing images altered by the defect cause model for laser marking defect inspection had accuracies of 97.0% and 91.6% in sorting the error cause, respectively

RST-YOLOv8: An Improved Chip Surface Defect Detection Model Based on YOLOv8

2025-06-21

Wenjie Tang , Yang‐Jun Deng , Xu Luo | Sensors

Surface defect detection in chips is crucial for ensuring product quality and reliability. This paper addresses the challenge of low identification accuracy in chip surface defect detection, which arises from … Surface defect detection in chips is crucial for ensuring product quality and reliability. This paper addresses the challenge of low identification accuracy in chip surface defect detection, which arises from the similarity of defect characteristics, small sizes, and significant scale differences. We propose an enhanced chip surface defect detection algorithm based on an improved version of YOLOv8, termed RST-YOLOv8. This study introduces the C2f_RVB module, which incorporates RepViTBlock technology. This integration effectively optimizes feature representation capabilities while significantly reducing the model’s parameter count. By enhancing the expressive power of deep features, we achieve a marked improvement in the identification accuracy of small defect targets. Additionally, we employ the SimAM attention mechanism, enabling the model to learn three-dimensional channel information, thereby strengthening its perception of defect characteristics. To address the issues of missed detections and false detections of small targets in chip surface defect detection, we designed a task-aligned dynamic detection head (TADDH) to facilitate interaction between the localization and classification detection heads. This design improves the accuracy of small target detection. Experimental evaluations on the PCB_DATASET indicate that our model improved the [email protected] by 10.3%. Furthermore, significant progress was achieved in experiments on the chip surface defect dataset, where [email protected] increased by 5.4%. Simultaneously, the model demonstrated significant advantages in terms of computational complexity, as both the number of parameters and GFLOPs were effectively controlled. This showcases the model’s balance between high precision and a lightweight design. The experimental results show that the RST-YOLOv8 model has a significant advantage in detection accuracy for chip surface defects compared to other models. It not only enhances detection accuracy but also achieves an optimal balance between computational resource consumption and real-time performance, providing an ideal technical pathway for chip surface defect detection tasks.

Intelligent real-time error correction in additive manufacturing via context-aware deep learning

2025-06-21

Prateek Mehta , Vivek Kamat , Daniel Ewing +1 more | Progress in Additive Manufacturing

An Improved Lightweight Model for Defect Detection on Paths in Images

2025-06-21

Zhaoning Cui , Yuejia Xu , Xinyi Jin +1 more | Applied Sciences

To address the challenges of detecting multi-scale road defects and the lack of lightweight designs in conventional detection models, we propose ACD-YOLOv8, an enhanced model based on YOLOv8s. Our model … To address the challenges of detecting multi-scale road defects and the lack of lightweight designs in conventional detection models, we propose ACD-YOLOv8, an enhanced model based on YOLOv8s. Our model enhances baseline architecture by integrating three key components: a lightweight Cross-Scale Feature Fusion Module (CCFM), an ADown sampling operation, and a Dynamic Head (DyHead). Experimental results on the RDD2022 dataset demonstrate the superiority of our approach. Compared to the baseline YOLOv8s, ACD-YOLOv8 achieves a 0.9% increase in [email protected] and a 1.6% increase in the more stringent [email protected]:0.95 metric. Simultaneously, the model’s parameter count is reduced by 3.72 million (a 33.3% reduction) and its size is reduced by 7.4 MB. This work provides a practical and scalable solution for deploying high-accuracy defect detection on resource-constrained mobile platforms, offering significant potential to enhance traffic safety and maintenance efficiency.

Multidimensional Comprehensive Evaluation Method for Sonar Detection Efficiency Based on Dynamic Spatiotemporal Interactions

2025-06-21

Shizhe Wang , Weiyi Chen , Zongji Li +2 more | Journal of Marine Science and Engineering

The detection efficiency evaluation of sonars is crucial for optimizing task planning and resource scheduling. The existing static evaluation methods based on single indicators face significant challenges. First, static modeling … The detection efficiency evaluation of sonars is crucial for optimizing task planning and resource scheduling. The existing static evaluation methods based on single indicators face significant challenges. First, static modeling has difficulty coping with complex scenes where the relative situation changes in real time in the task process. Second, a single evaluation dimension cannot characterize the data distribution characteristics of efficiency indicators. In this paper, we propose a multidimensional detection efficiency evaluation method for sonar search paths based on dynamic spatiotemporal interactions. We develop a dynamic multidimensional evaluation framework. It consists of three parts, namely, spatiotemporal discrete modeling, situational dynamic deduction, and probability-based statistical analysis. This framework can achieve dynamic quantitative expression of the sonar detection efficiency. Specifically, by accurately characterizing the spatiotemporal interaction process between the sonars and targets, we overcome the bottleneck in entire-path detection efficiency evaluation. We introduce a Markov chain model to guide the Monte Carlo sampling; it helps to specify the uncertain situations by constructing a high-fidelity target motion trajectory database. To simulate the actual sensor working state, we add observation error to the sensor, which significantly improves the authenticity of the target’s trajectories. For each discrete time point, the minimum mean square error is used to estimate the sonar detection probability and cumulative detection probability. Based on the above models, we construct the multidimensional sonar detection efficiency evaluation indicator system by implementing a confidence analysis, effective detection rate calculation, and a data volatility quantification analysis. We conducted relevant simulation studies by setting the source level parameter of the target base on the sonar equation. In the simulation, we took two actual sonar search paths as examples and conducted an efficiency evaluation based on multidimensional evaluation indicators, and compared the evaluation results corresponding to the two paths. The simulation results show that in the passive and active working modes of sonar, for the detection probability, the box length of path 2 is reduced by 0∼0.2 and 0∼0.5, respectively, compared to path 1 during the time period from T = 11 to T = 15. For the cumulative detection probability, during the time period from T = 15 to T = 20, the box length of path 2 decreased by 0∼0.1 and 0∼0.2, respectively, compared to path 1, and the variance decreased by 0∼0.02 and 0∼0.03, respectively, compared to path 1. The numerical simulation results show that the data distribution corresponding to path 2 is more concentrated and stable, and its search ability is better than path 1, which reflects the advantages of the proposed multidimensional evaluation method.

An Intelligent Optimization System Using Neural Networks and Soft Computing for the FMM Etching Process

2025-06-20

Wen‐Chin Chen , An-Xuan Ngo , Jianfeng Zhong | Mathematics

The rapid rise of flexible AMOLED displays has prompted manufacturers to advance technologies to meet growing global demand. However, high costs and quality inconsistencies hinder industry competitiveness and sustainability. This … The rapid rise of flexible AMOLED displays has prompted manufacturers to advance technologies to meet growing global demand. However, high costs and quality inconsistencies hinder industry competitiveness and sustainability. This study addresses these challenges by developing an intelligent optimization system for the fine metal mask (FMM) etching process, a critical step in producing high-resolution AMOLED panels. The system integrates advanced optimization techniques, including the Taguchi method, analysis of variance (ANOVA), back-propagation neural network (BPNN), and a hybrid particle swarm optimization–genetic algorithm (PSO-GA) approach to identify optimal process parameters. Experimental results demonstrate a marked improvement in product yield and process stability while reducing manufacturing costs. By ensuring consistent quality and efficiency, this system overcomes limitations of traditional process control; strengthens the AMOLED industry’s global competitiveness; and provides a scalable, sustainable solution for smart manufacturing in next-generation display technologies.

HSC-YOLO: steel surface defect detection model based on improved YOLOv10n

2025-06-20

Chunhua Zhao , Q. Wang , Julian Yi Hong Tan +4 more | Measurement Science and Technology

Abstract During the production process of steel, the control of surface quality is crucial to the performance of the final molded product, so it is necessary to detect defects on … Abstract During the production process of steel, the control of surface quality is crucial to the performance of the final molded product, so it is necessary to detect defects on its surface during the production process. Aiming at the problems of low detection accuracy and insufficient feature extraction and fusion capability in steel surface defect detection, a lightweight and multi-scale feature fusion model HSC-YOLO based on the improved YOLOv10n is proposed. Firstly, the backbone feature extraction network is reset using an improved lightweight network structure based on high performance GPU network (HGNetv2) to reduce the model size. Secondly, the multilevel feature fusion module semantics and detail infusion (SDI) is used instead of the two Concat modules in neck to enhance the semantic and detail information in the image. Finally, an iterative attentional feature fusion (iAFF) mechanism is introduced and combined with cross stage partial bottleneck with 2 convolutions (C2f) to solve the problems that occur when features are fused at different scales, especially the feature fusion problem with inconsistent semantics and scale. Test results on the datasets NEU-DET and GC10-DET show that the mean average precision (mAP) of HSC-YOLO improves by 3.9% and 2.1% over the mAP of YOLOv10n, and the detection speed improves by 46.2% and 43.1%, which provides the best detection results compared to other models.

Integrated Intelligent System for Automatic Classification and Counting in a Miniature Packaging Conveyor Using Machine Learning and Deep Learning

2025-06-20

Erwin Sitompul , Muhammad Ghiffaari Ilham Ramadhan , Iksan Bukhori | Journal of Intelligent & Fuzzy Systems

Material handling is one of the most important components in manufacturing industries. Conveyor belts belong to one of the material handling equipment categories that move objects and connect distant sections … Material handling is one of the most important components in manufacturing industries. Conveyor belts belong to one of the material handling equipment categories that move objects and connect distant sections of a manufacturing process. This paper proposes an integrated approach of deep-learning object recognition in a packaging conveyor for real-time automatic object counting, shape detection, and color description. The deep neural network model was built in TensorFlow by using the Single-Shot Detection (SSD) method and Feature Pyramid Network (FPN) extraction. The model was improved by implementing k-means clustering and k-nearest neighbors algorithms in the object detection and color description. The novelty of the research is the implementation of deep learning with three integrated features in a manufacturing-related process. A prototype of a packaging conveyor was constructed to function as a miniature manufacturing process. The proposed system's accuracy was evaluated in terms of object counting, shape detection, and color description. Test objects varied in 2 different shapes (cube and box) and three different colors (blue, orange, and green). The average accuracy of the proposed system is 93.7% in object counting, 97.2% in shape detection, and 77.5% in color description, with an overall average accuracy of 90.0%. The color description was found to be sensitive to illumination level. Future improvements for the color descriptor's accuracy include creating a specific color data set and covering data with various illumination levels.

Time-Varying Change Tracking of Fingernail Biometrics Based on MATLAB and Origin Software

2025-06-20

Rohail Hassan , A. Mustafa | Indian Journal of Science and Technology

Objectives: This work explores the 90-day temporal variation of fingernail biometric features and compares the performance of MATLAB and Origin software to analyze these changes. The study is primarily set … Objectives: This work explores the 90-day temporal variation of fingernail biometric features and compares the performance of MATLAB and Origin software to analyze these changes. The study is primarily set to determine the temporal stability of fingernail biometrics and the efficacy of computational tools in monitoring and analyzing these changes. Method: A case study design was employed, capturing fingernail photographs at three time points: Day 1, Day 8, and Day 90. Feature extraction and classification were performed using a support vector machine (SVM) in MATLAB, while contour mapping and histogram analysis were conducted using Origin software. The accuracy of classification, standard deviation of the pixel intensity, and contour complexity were measured to quantify biometric changes. Findings: The experiment found the deterioration of classification performance with time due to physiological change. The performance of the MATLAB model reduced from 98.5% on Day 1 to 92.1% on Day 8 and 84.3% on Day 90, impacting biometric recognition. Origin contour mapping identified higher structural complexity with 18.4% and 35.7% rise in Day 8 contour and Day 90 contour density, respectively. Histogram analysis produced a 50.2% increase in standard deviation on the 90th day, indicating variation in biometric parameters. MATLAB was able to pick these up but was reset, and Origin was used for demonstration purposes but not for predictive modeling. Novelty: This study addresses time-varying changes in fingernail biometrics against the time-invariance hypothesis. It implies vulnerabilities in traditional systems and encourages hybrid machine learning and statistical visualization-based approaches for improved accuracy. Keywords: Biometric Identification, Fingernail Characteristics, Machine Learning, Biometric Changes, MATLAB, Origin Software

Wavelet-Guided Multi-Scale Inpainting Framework Using AOT Blocks and GANs

2025-06-20

| International journal of intelligent engineering and systems

A High-Accuracy PCB Defect Detection Algorithm Based on Improved YOLOv12

2025-06-20

Zhi Chen , Bingxiang Liu | Symmetry

To address the common issues of high small object miss rates, frequent false positives, and poor real-time performance in PCB defect detection, this paper proposes a multi-scale fusion algorithm based … To address the common issues of high small object miss rates, frequent false positives, and poor real-time performance in PCB defect detection, this paper proposes a multi-scale fusion algorithm based on the YOLOv12 framework. This algorithm integrates the Global Attention Mechanism (GAM) into the redesigned A2C2f module to enhance feature response strength of complex objects in symmetric regions through global context modeling, replacing conventional convolutions with hybrid weighted downsampling (HWD) modules that preserve copper foil textures in PCB images via hierarchical weight allocation. A bidirectional feature pyramid network (BiFPN) is constructed to reduce bounding box regression errors for micro-defects by fusing shallow localization and deep semantic features, employing a parallel perception attention (PPA) detection head combining dense anchor distribution and context-aware mechanisms to accurately identify tiny defects in high-density areas, and optimizing bounding box regression using a normalized Wasserstein distance (NWD) loss function to enhance overall detection accuracy. The experimental results on the public PCB dataset with symmetrically transformed samples demonstrate 85.3% recall rate and 90.4% mAP@50, with AP values for subtle defects like short circuit and spurious copper reaching 96.2% and 90.8%, respectively. Compared to the YOLOv12n, it shows an 8.7% enhancement in recall, a 5.8% increase in mAP@50, and gains of 16.7% and 11.5% in AP for the short circuit and spurious copper categories. Moreover, with an FPS of 72.8, it outperforms YOLOv5s, YOLOv8s, and YOLOv11n by 12.5%, 22.8%, and 5.7%, respectively, in speed. The improved algorithm meets the requirements for high-precision and real-time detection of multi-category PCB defects and provides an efficient solution for automated PCB quality inspection scenarios.

Design and Implementation of Automatic Photovoltaic Panel Cleaning Using IOT

2025-06-19

M. Perarasi , Priscilla Whitin , R. Pavaiyarkarasi +3 more | CRC Press eBooks

Deep integration of conditional gan, attention mechanism, and image clustering for automated color separation and correction in textile screen printing

2025-06-19

Chao-Lung Yang , Yulius Harjoseputro , Chi-Hao Chien +1 more | Applied Intelligence

Casting defect-detection method based on multi-image feature of fusion artificial intelligence model

2025-06-19

Deyan Yang , Hongquan Jiang , Yang He +4 more | Applied Intelligence

High-precision Defect Detection of Glass for Thin-film Transistor Liquid Crystal Display Using YOLO Algorithms

2025-06-19

Rongri Tan , Chih‐Cheng Chen , Kai-An Kuo +7 more | Sensors and Materials

A cloud-edge collaborative deep network for signal compression and reconstruction in aerospace testing

2025-06-19

Youlong Lv , Bo Zhao , Hui Cheng +2 more | Engineering Research Express

Abstract To address the real-time processing requirements of massive multi-source signals in aerospace product integrated testing, this paper proposes a cloud-edge collaborative signal compression and reconstruction method based on a … Abstract To address the real-time processing requirements of massive multi-source signals in aerospace product integrated testing, this paper proposes a cloud-edge collaborative signal compression and reconstruction method based on a deep compressed sensing network. Targeting the transmission bottlenecks in cloud-edge architectures and the fragmentation of temporal signal dependencies, a dual-stage optimization method is developed: 1) At the edge side, a dual-branch convolutional compression network is designed to achieve adaptive compression of multi-form signals through global feature observation and local attention enhancement. 2) On the cloud side, a bidirectional LSTM (BiLSTM) combined with a progressive stacking structure is employed to establish a cross-temporal signal correlation reconstruction mechanism. The proposed method is evaluated on both public and real-world datasets. Experimental results demonstrate superior performance over traditional compressed sensing and deep learning methods, achieving lower reconstruction errors while maintaining high compression rates, thereby effectively balancing the trade-off between compression efficiency and reconstruction fidelity.

Integrating Manufacturing Intelligence, Computer Vision, and Process Observation for Yield Improvement and Failure Prediction in Electronics Manufacturing

2025-06-19

Vinit Vithalrai Shenvi , Ashutosh Sharma | Deleted Journal

Electronics manufacturing processes are complex and prone to yield loss and latent failures due to subtle process deviations and quality escapes. This paper presents a holistic approach to improving first-pass … Electronics manufacturing processes are complex and prone to yield loss and latent failures due to subtle process deviations and quality escapes. This paper presents a holistic approach to improving first-pass yield and predicting failures by integrating a~Manufacturing Intelligence for Reliability and Automated Insights (MIRAI) data platform with computer vision-based monitoring of~Standard Operating Procedure (SOP) adherence. The proposed system combines self-serve data analytics workflows for yield and field failure analysis with real-time process observation using deep learning vision models. Manufacturing data from production tests, reliability screenings, and field returns are aggregated and analyzed to identify key signals correlating with yield drops and field fallouts. Simultaneously, a PROSPECT tool employs AI cameras at assembly stations to record operator actions and detect deviations from standard procedures. A machine learning failure prediction model is then trained on the enriched dataset (including vision-detected deviations) to proactively flag high-risk units in real time.

A lightweight detection model for rice grain with dense bonding distribution based on YOLOv5s

2025-06-19

Dongquan Chen , Wenbin Sun , Kang Xu +3 more | Computers and Electronics in Agriculture

DScanNet: Packaging Defect Detection Algorithm Based on Selective State Space Models

2025-06-19

Luo Yi-rong , Yanping Du , Zhaohua Wang +3 more | Algorithms

With the rapid development of e-commerce and the logistics industry, the importance of logistics packaging defect detection as a key link in product quality control is becoming increasingly prominent. However, … With the rapid development of e-commerce and the logistics industry, the importance of logistics packaging defect detection as a key link in product quality control is becoming increasingly prominent. However, existing target detection models often face the problems of difficulty in improving detection accuracy and high model complexity when dealing with small-scale targets in logistics packaging. For this reason, an improved target detection model, DScanNet, is proposed in this paper. To address the problem that the model’s detailed feature extraction for small target defects is not sufficient and thus leads to low detection accuracy, the MEFE module, the local feature extraction module (LFEM Block), and the PCR module of the multi-scale convolution and feature enhancement strategy are proposed to enhance the model’s capability of capturing defective features and focusing on specific features, and to improve the detection accuracy. To address the problem of excessive model complexity, a Mamba module incorporating a channel attention mechanism is proposed to optimize the model via its linear complexity. Through experiments on its own dataset, BIGC-LP, DScanNet achieves a high accuracy of 96.8% on the defect detection task compared with the current mainstream detection algorithms, while the number of model parameters and the computational volume are effectively controlled.

Instance segmentation based on global-local attention and local Chan-Vese model

2025-06-18

Chengcheng Wang , Zhize Wu , Ming Sun +4 more | Signal Image and Video Processing

Characterization and Prediction of Stitch-Type Textile Sensors Using Deep Learning Model

2025-06-18

Jiseon Kim , Jooyong Kim | Fibers and Polymers

Surface defect detection on rolling mill rolls: a transfer learning approach with improved YOLOv5

2025-06-18

HU Yu-long , Peng Wen , Shiyi Chen +4 more | Deleted Journal

Abstract The roller is a crucial tool in the hot-rolled strip steel manufacturing process. Rapid and accurate detection of roll surface defects is essential for enhancing product surface quality, ensuring … Abstract The roller is a crucial tool in the hot-rolled strip steel manufacturing process. Rapid and accurate detection of roll surface defects is essential for enhancing product surface quality, ensuring dimensional precision, and reducing scrap rates. Currently, industrial roll inspection primarily relies on manual visual assessment, which is prone to subjectivity, low accuracy, and inefficiency. To overcome these limitations, this study introduces a defect detection method based on transfer learning and an enhanced YOLOv5 model.Given the lack of publicly available datasets for roll surface defects, a representative dataset was first constructed using defect images collected from a steel production line. To further optimize the detection model, the NEU-DET dataset—containing strip surface defect images—was employed to pre-train the improved YOLOv5 model, refining its parameters. The pre-trained model was then adapted for roll defect detection using transfer learning, where the learned parameters from strip surface defects served as initial weights for training on the roll defect dataset. Experimental results demonstrate that the proposed TR-CNF-YOLOv5 model, integrating transfer learning with an improved YOLOv5 architecture, outperforms existing models. Specifically, it achieves an mAP improvement of approximately 9.8% over the original YOLOv5 and 7.1% over YOLOv7 on the roll surface defect dataset.

Industrial Vision Systems and Defect Detection

Description

Keywords