Small Size Defects Research Articles

Radiation Resistivity of Ti-5331 Alloy with Different Microstructures

Titanium alloys have promising potential for applications in marine nuclear power systems owing to their exceptional mechanical and corrosion properties. Nevertheless, their radiation resistivity is a determining factor for their extensive use as structural materials in nuclear energy systems. In this study, the radiation resistivity of Ti-5Al-3V-3Zr-Cr (Ti-5331) alloys with three different microstructures was examined using a combination of characterization techniques including positron annihilation Doppler broadening spectroscopy (DBS), Elastic Recoil Detection Analysis (ERDA), Transmission Electron Microscope (TEM), Small-Angle X-ray scattering (SAXS) and nanoindentation. The results reveal that the equiaxed structure, bimodal structure, and Widmanstatten structure of Ti-5331 alloy obtained through different annealing processes exhibit differences in the number of interfaces and the content of the β phase. The α/β interfaces can significantly enhance the radiation resistance of titanium alloys by inhibiting the diffusion of helium atoms after helium ion irradiation. Notably, the alloy with a bimodal structure exhibited the best overall radiation resistivity, showing small defect size, low hardening effect, and low swelling rate of 0.003%. Moreover, the size of helium bubbles of the bimodal structure is half of that in the equiaxed structure and Widmanstatten structure. Thus, the bimodal structure of the Ti-5331 alloy possesses superior radiation resistivity.

A Candy Defect Detection Method Based on StyleGAN2 and Improved YOLOv7 for Imbalanced Data.

Quality management in the candy industry is a vital part of food quality management. Defective candies significantly affect subsequent packaging and consumption, impacting the efficiency of candy manufacturers and the consumer experience. However, challenges exist in candy defect detection on food production lines due to the small size of the targets and defects, as well as the difficulty of batch sampling defects from automated production lines. A high-precision candy defect detection method based on deep learning is proposed in this paper. Initially, pseudo-defective candy images are generated based on Style Generative Adversarial Network-v2 (StyleGAN2), thereby enhancing the authenticity of these synthetic defect images. Following the separation of the background based on the color characteristics of the defective candies on the conveyor belt, a GAN is utilized for negative sample data enhancement. This effectively reduces the impact of data imbalance between complete and defective candies on the model's detection performance. Secondly, considering the challenges brought by the small size and random shape of candy defects to target detection, the efficient target detection method YOLOv7 is improved. The Spatial Pyramid Pooling Fast Cross Stage Partial Connection (SPPFCSPC) module, the C3C2 module, and the global attention mechanism are introduced to enhance feature extraction precision. The improved model achieves a 3.0% increase in recognition accuracy and a 3.7% increase in recall rate while supporting real-time recognition scenery. This method not only enhances the efficiency of food quality management but also promotes the application of computer vision and deep learning in industrial production.

A lightweight defect detection algorithm for escalator steps

In this paper, we propose an efficient target detection algorithm, ASF-Sim-YOLO, to address issues encountered in escalator step defect detection, such as an excessive number of parameters in the detection network model, poor adaptability, and difficulties in real-time processing of video streams. Firstly, to address the characteristics of escalator step defects, we designed the ASF-Sim-P2 structure to improve the detection accuracy of small targets, such as step defects. Additionally, we incorporated the SimAM (Similarity-based Attention Mechanism) by combining SimAM with SPPF (Spatial Pyramid Pooling-Fast) to enhance the model’s ability to capture key information by assigning importance weights to each pixel. Furthermore, to address the challenge posed by the small size of step defects, we replaced the traditional CIoU (Complete-Intersection-over-Union) loss function with NWD (Normalized Wasserstein Distance), which alleviated the problem of defect missing. Finally, to meet the deployment requirements of mobile devices, we performed channel pruning on the model. The experimental results showed that the improved ASF-Sim-YOLO model achieved an average accuracy (mAP50) of 96.8% on the test data set, which was a 22.1% improvement in accuracy compared to the baseline model. Meanwhile, the computational complexity (in GFLOPS) of the model was reduced to a quarter of that of the baseline model, while the frame rate (FPS) was improved to 575.1. Compared with YOLOv3-tiny, YOLOv5s, YOLOv8s, Faster-RCNN, TOOD, RTMDET and other deep learning-based target recognition algorithms, ASF-Sim-YOLO has better detection accuracy and real-time processing capability. These results demonstrate that ASF-Sim-YOLO effectively balances lightweight design and performance improvement, making it highly suitable for real-time detection of step defects, which can meet the demands of escalator inspection operations.

A Novel YOLOv5_ES based on lightweight small object detection head for PCB surface defect detection

In the manufacturing process of printed circuit boards (PCBs), surface defects have a significant negative impact on product quality. Considering that traditional object detection algorithms have low accuracy in handling PCB images with complex backgrounds, various types, and small-sized defects, this paper proposes a PCB defect detection algorithm based on a novel YOLOv5 multi-scale attention mechanism(EMA) spatial pyramid dilated Convolution (SPD-Conv) (YOLOv5_ES) network improved YOLOv5s framework. Firstly, the detection head is optimized by removing medium and large detection layers, fully leveraging the capability of the small detection head to identify minor target defects. This approach not only improves model accuracy but also achieves lightweighting. Secondly, in order to further reduce the number of parameters and computational costs, the SPD-Conv is introduced to improve the feature extraction capability by reducing information loss. Thirdly, a EMA module is introduced to fuse context information of different scales, enhancing the model’s generalization ability. Compared to the YOLOv5s model, there is a 3.1% improvement in mean average precision (mAP0.5), a 55.8% reduction in model parameters, and a 4.8% reduction in giga floating-point operations per second (GFLOPs). These results demonstrate a significant improvement in both accuracy and model parameter efficiency.

A customised ConvNeXt-SCC network: integrating improved principal component analysis with ConvNeXt to enhance tire crown defect detection

ABSTRACT Defects such as bubbles and rubber shortages occasionally occur during vehicle tire production. Tiny defects in the crown gradually expand under load, ultimately causing tire failure. However, machine-vision-based crown defect detection faces challenges like limited datasets, varying groove patterns, low colour contrast, and small defect sizes. Therefore, this article integrates principal component analysis (PCA) and deep learning techniques for the visual detection of tire crown defects. A dataset of tire crown defects suitable for deep learning was constructed using an improved-PCA method and composite sampling. We propose the ConvNeXt-Shallow Conv CBAM (ConvNeXt-SCC) model, which includes an improved convolutional block attention module (CBAM) in a shallow stacked structure. And we propose an efficient and accurate method called Improved Selective Search (ISS) to detect and locate tire crown defective regions. The ISS pre-selected box generation method, enhanced with a priori knowledge filtering module, was used alongside a classification model to achieve online inference detection of large-size tire crown images. The experimental results show that the performance indices of the proposed ConvNeXt-Shallow Conv CBAM (ConvNeXt-SCC) model significantly improve compared to the original ConvNeXt-T. The defect detection and localisation method based on ISS meets the online inspection requirements of tire production.

A three-stage framework for accurate detection of high-speed train body paint film defects

Detecting paint film defects on high-speed train (HST) body is a crucial step towards zero-defect manufacturing (ZDM). The effectiveness of defect detection in industrial environments, however, is significantly impacted by factors such as high reflection noise, weak defect features, small defect size, and various defect scales, resulting in unsatisfactory detection accuracy. To address these challenges, this paper proposes a technical framework for the accurate detection of paint film defects on HST body, encompassing three stages. A reflection detection removal method that takes into account the reflection transition region as well as the preservation of defect features is at first used to reduce reflection noise interference. Then a lightness-weighted adaptive image enhancement algorithm based on the fractional order differentiation is utilized to enhance the features of paint film defects while avoiding the introduction of reflection noise. Finally, an improved YOLOv5 algorithm, incorporating attention mechanism feature extraction and multi-scale feature fusion, is employed to detect and classify paint film defects. Experimental results demonstrate that the proposed framework effectively reduces the reflection in paint film images, and enhances the contrast between paint film defects and the background, with a mean average precision (mAP) for defect detection reaching up to 90.13%. This work serves as a valuable reference for the automated detection of defects arising in the painting process of HST body.

A lightweight parallel attention residual network for tile defect recognition

In modern industrial production, permanent magnet motors are an indispensable part of industrial manufacturing. The quality of the magnetic tiles directly affects the working performance of the permanent magnet motors, making the detection of defects on the surface of magnetic tiles critically important. However, due to the small size of defects on the tile image and the reflectivity of the defective surface, the details of image characteristics are not prominently acquired.These problems bring a lot of difficulties for the recognition of magnetic tile defects. In this paper, a magnetic tile defect detection method is proposed for the probAlems of unclear image features and small defects. First, the image is processed using linear variation to enhance the image detail features. Then, by introducing the inverted bottleneck block structure in MobileNetV2, the Attention Parallel Residual Convolution Block (APR) is proposed, and the Lightweight Parallel Attention Residual Network (LPAR-Net) is built. In APR Block, 7 × 7 convolution is introduced so that the model can extract spatial features from a larger range, and weighted fusion of input images by residual structure. In addition, in this paper, CBAM is improved, split into two parts and inserted into APR Block. Finally, the mainstream image classification models and the LPAR-Net proposed in this paper are used for comparison, respectively. The experimental results show that the method achieves 93.63% accuracy on the adopted dataset, which is better than the existing mainstream image classification network models DenseNet, MobileNet, ConvNext and so on. In addition, this paper introduces a strip steel surface defect dataset and compares it with the above image classification model, which verifies that the detection method proposed in this paper still has strong recognition capability.

Influence of dimensional deviation and thickness non-uniformity on the small-size specimen tensile results of CLF-1 steel

Tensile testing of small-size specimens has been widely used to evaluate the influence of neutron irradiation on the mechanical properties for structural materials of fusion reactors. In recent years, researchers have been trying to standardize small-size specimen tensile test method. In the present paper, we have investigated the influence of dimensional deviation and thickness non-uniformity on the tensile results of small-size specimen by both experiments and finite element simulations, based on one of the candidate structural materials of fusion reactor, i.e. CLF-1 steel. It was found that under the conditions of small grain size, small metallurgical defect size and suitable specimen preparation, the tensile results of SS-J3 specimen could be close to the results of large-size specimen. Slightly higher yield and ultimate strengths for small specimens may be caused by the surface treatment of the wheel grinding. The repeatability of ultimate strength was better than that of yield strength, and the repeatability of elongation was worse than that of strength. For tensile tests on eight SS-J3 specimens, the maximum difference in total elongation was about 3 %. With ±0.1 mm deviation of thickness and parallel width, the maximum variation in total elongation was about 1.8 %. The ductility properties were sensitive to specimen's thickness non-uniformity, the measured uniform and total elongations would obviously decrease even with 0.01 mm thickness non-uniformity.

Insulator Defect Detection Based on YOLOv5s-KE

To tackle the issue of low detection accuracy in insulator images caused by intricate backgrounds and small defect sizes, as well as the requirement for real-time detection on embedded and mobile devices, this research introduces the YOLOv5s-KE model. Integrating multiple strategies, YOLOv5s-KE aims to boost detection accuracy significantly. Initially, an enhanced anchor generation method utilizing the K-means++ algorithm is proposed to generate more appropriate anchor boxes for insulator defects. Moreover, an attention mechanism is integrated into both the backbone and neck networks to enhance the model’s capacity to focus on defect features and resist interference. To improve the detection of small defects, the EIoU loss function is implemented in place of the original CIoU loss function. In order to meet the real-time detection needs on embedded and mobile devices, the model is further refined through the integration of Ghost convolution for lightweight feature extraction and a linear transformation to reduce the computational burden of standard convolution. A channel pruning strategy is deployed to optimize the sparsely trained network, diminishing redundancy, and improving model generalization. Additionally, the CARAFE operator replaces the original upsampling operator to minimize model parameters and elevate detection speed. Experimental outcomes demonstrate that YOLOv5s-KE achieves a detection accuracy of 92.3% on the Chinese transmission line insulator dataset, marking a 5.2% enhancement over the original YOLOv5s. The streamlined version of YOLOv5s-KE achieves a detection speed of 94.3 frames per second, indicating an improvement of 30.1 frames per second compared to the original model. Model parameters are condensed to 9.6 M, resulting in a detection accuracy of 91.1%. This study underscores the precision and efficiency of the proposed approach, suggesting that the advanced strategies explored introduce novel possibilities for insulator defect detection.

An insight into annealing mechanism of graphitized structures after irradiation

Graphite components are constantly subjected to a combined actions of annealing and irradiation due to high temperatures and thermal spike caused by irradiation when reactors are operating, resulting in complex microstructures along with matching changes in the material properties and dimensions. This study investigates the evolution process of initial irradiation defects at different annealing temperatures, which is difficult to captured by experiment. The findings indicate that 923K reactor-temperature annealing can also quickly restore isolated self-interstitial atoms and small-sized point defect clusters to intralayer locations on the nanosecond scale. However, multi-interlayer penetration damages and localized point defect aggregation from high-energy irradiation can lead to the disappearance of layered structural information, which allows these damage structures to develop into interlayer dislocations during annealing process. These interlayer dislocations further exacerbate the interlayer expansion of graphite crystal and may elucidate the mechanism behind volume expansion in nuclear graphite within reactors. Fully amorphized regions can also regain a layered structure approximately when guided by residual layered structures at 2000K, while chaotic regions or disordered layered structures form in the absence of guidance. These chaotic regions significantly exacerbate the volume expansion of graphite model, which may be related to the rapid volume increase observed in nuclear graphite components at the end of reactor life. The study provides insights into the transformation of initial irradiation defects into different types of defects under different temperatures and damage states, which serve as a critical foundation for assessing the evolution of irradiation defects in nuclear graphite for reactor applications and annealing studies.

A practical path planning method for optimal repair paths between multiple small-size defects

PurposeThis study aims to solve the problem of repair path planning between multiple small-size defects in the field of additive manufacturing (AM) repair by using Python-based ant colony algorithm (ACO). The optimal parameter combination scheme is obtained by discussing the influencing factors of parameters in the ACO.Design/methodology/approachThe effects of the information heuristic factor α, the expected heuristic factor ß and the pheromone volatile factor ρ on the simulation results were investigated by designing a three-factor and three-level orthogonal experiment. The fast convergence of ACO in finding the optimal solution of multiple small-size defect repair path problem is proved by comparing the simulation results with those of genetic algorithm (GA) on the same data set.FindingsThe ACO can effectively solve the repair path planning problem between multiple small-size defects by optimizing the parameters. In the case of 50 defect locations, the simulation results of the ACO with optimized parameters are 159.8 iterations and 3,688 average path lengths, while the GA has 4,027.2 average path lengths under the same data set and the same number of iterations, and by comparison, it is proved that the ACO can find the optimal solution quickly in the small-size defects repair path planning problem, which greatly improves the efficiency of defect repair.Originality/valueThe parameter-optimized ACO can be quickly applied to the planning problem of repair paths between multiple small-size defects in the field of AM repair, which can better improve the defect repair efficiency and reduce the waste of resources.

REDef-DETR: real-time and efficient DETR for industrial surface defect detection

Abstract Industrial surface defect detection is an important part of industrial production, which aims to identify and detecting various defects on the surface of product to ensure quality and meet customer requirements. With the development of deep learning and image processing technologies, the surface defect detection methods based on computer vision has become the mainstream method. However, the prevalent convolutional neural network-based defect detection methods also have many problems. For example, these methods rely on post-processing of Non-Maximum Suppression and have poor detection ability for small targets, which affects the speed and accuracy of surface defect detection in industrial scenarios. Therefore, we propose a novel DEtection TRansformer-based surface defect detection method. Firstly, we propose a Multi-scale Contextual Information Dilated module and fuse it into the backbone. The module is mainly composed of large kernel convolutions, which aims to expand the receptive field of the model, thus reducing the leakage rate of the model. Moreover, we design an efficient encoder which mainly contains two important modules, namely feature enhancement based on cascaded group attention module and efficient feature fusion module based on content-aware. The former module effectively enhances the high-level semantic information extracted by the backbone, thus enabling the model to better interpret features, and it can improve the problem of high computational cost of transformer encoder, thus increasing the detection speed. The latter module performs multi-scale feature fusion across the feature information of various scales, thus improving the detection accuracy of the model for small-size defects. Experimental results show that the proposed method achieves 80.6%mAP and 80.3FPS on NEU-DET, and 98.0%mAP and 79.4FPS on PCB-DET. Our proposed method exhibits excellent detection performance and achieves real-time and efficient surface defect detection capability to meet the needs of industrial surface defect detection.

Tree internal defects detection method based on ResNet improved subspace optimization algorithm

The erosion behavior of trunk borers leads to the destruction of trunk structure and the formation of internal defects, which significantly impacts the ecological and economic value of trees. Traditional non-destructive testing (NDT) methods are costly and have low resolution, whereas electromagnetic NDT methods are more suitable for high-resolution detection and imaging. However, solving the highly nonlinear electromagnetic inverse scattering problems (ISPs) for small-sized defects with high contrast is challenging. Therefore, this paper proposes an improved subspace optimization algorithm based on a ResNet network called SOM-ResNet. SOM-ResNet incorporates physical principles into deep learning networks by simulating the iterative process of induced current and contrast, thereby enhancing its ability to accurately detect small objects with high contrast. Experimental results demonstrate that SOM-ResNet outperforms single inversion algorithms in detecting complex scatterers with small to medium-sized targets, validating its excellent performance.

Posterior Tibial Artery Perforator Based Propeller Flap for Reconstruction of 143 Skin and Soft Tissue Defect Around Ankle Joint

Background: Soft tissue coverage of distal leg and ankle region represents a challenge and such defect usually requires a free flap. However, this needs facility of microsurgery, may lead to donor site morbidity and is time consuming. With the introduction of perforator flap, management of small and medium size defects of distal leg and ankle region is easier, less time consuming, and with minimal donor site morbidity. When local perforator flap is designed as propeller and rotated to 180-degree, donor site can be closed primarily and increases the distance crossed by flap, thus increasing versatility. Objectives: To observe the outcome of soft tissue coverage of defects around ankle joint by posterior tibial artery perforator based propeller flaps. Methods: This was a prospective type of observational study carried out on 20 patients with skin and soft tissue defects in and around ankle joint for a period of 14 months in the Department of Burn and Plastic Surgery, Dhaka Medical College and Hospital. Results: Among 20 cases 17 were male and 3 females; age ranged from 11 to 60 years. Post operative outcome of flap were excellent in 11 cases, good in 6 cases, poor in 3 cases as per defined criteria. Uneventful outcome was observed in 14 cases and minor complications were observed in 5 cases; and total flap loss occurred in 1 case which required revision surgery. Conclusions: Perforator flap provides adequate soft tissue coverage for small & medium size wound around ankle. It is an easy, reliable & effective procedure with reasonable outcome. J Dhaka Med Coll. 2022; 31(2) : 243-250

Crystallization management of CsPbI2Br perovskites by PbAc2-incorporated twice spin-coating process for efficient and stable CsPbI2Br perovskite solar cells

CsPbI2Br perovskite solar cell has been extensively studied due to its exceptional thermal stability and relatively stable perovskite phase structure. However, the presence of bromine leads to a rapid crystallization rate of CsPbI2Br films, resulting in small grain size and high defect density. Additionally, CsPbI2Br demonstrates poor light absorption due to its wide bandgap. Therefore, it is crucial to control the crystallization rate and increase the film thickness to reduce defect density, enhance light absorption, and improve photovoltaic performance. In this study, we utilized a PbAc2-incorporated twice spin-coating (PTS) process to address these issues. Initially, PbAc2 was added to the CsPbI2Br precursor solution to form a CsPbI2Br film, which was then coated with the CsPbI2Br precursor solution to produce the PTS film. Ac− can delay the perovskite crystallization, leading to the formation of thicker and denser CsPbI2Br films. Moreover, lone-pair electrons of the oxygen atom provided by Ac− formed coordination bonds with under-coordinated Pb2+ ions to fill halogen ion vacancies, thereby reducing the defect density. Ultimately, the PTS CsPbI2Br device achieved a peak power conversion efficiency (PCE) of 16.19% and maintained 96.7% of its initial PCE over 1500 h at room temperature under 25% relative humidity without any encapsulation.

Context-Enhanced Network with Spatial-Aware Graph for Smartphone Screen Defect Detection.

Interactive devices such as touch screens have gained widespread usage in daily life; this has directed the attention of researchers to the quality of screen glass. Consequently, defect detection in screen glass is essential for improving the quality of smartphone screens. In recent years, defect detection methods based on deep learning have played a crucial role in improving detection accuracy and robustness. However, challenges have arisen in achieving high-performance detection due to the small size, irregular shapes and low contrast of defects. To address these challenges, this paper proposes CE-SGNet, a Context-Enhanced Network with a Spatial-aware Graph, for smartphone screen defect detection. It consists of two novel components: the Adaptive Receptive Field Attention Module (ARFAM) and the Spatial-aware Graph Reasoning Module (SGRM). The ARFAM enhances defect features by adaptively extracting contextual information to capture the most relevant contextual region of defect features. The SGRM constructs a region-to-region graph and encodes region features with spatial relationships. The connections among defect regions are enhanced during the propagation process through a graph attention network. By enriching the feature representations of defect regions, the CE-SGNet can accurately identify and locate defects of various shapes and scales. Experimental results demonstrate that the CE-SGNet achieves outstanding performance on two public datasets.

Disparities in Healthcare Utilization: An Analysis of Disease Specific and Patient Level Factors in a Congenital Diaphragmatic Hernia Clinic

ObjectivesOur study examines if the disease severity profile of our Congenital Diaphragmatic Hernia (CDH) patient cohort adherent to long-term follow-up differs from patients lost to follow-up after discharge and examines factors associated with health care utilization. MethodsRetrospective review identified CDH survivors born 2005–2019 with index repair at our institution. Primary outcome was long-term follow-up status: “active” or “inactive” according to clinic guidelines. Markers of CDH disease severity including CDH defect classification, oxygen use, tube feeds at discharge, and sociodemographic factors were examined as exposures. ResultsOf the 222 included patients, median age [IQR] was 10.2 years [6.7–14.3], 61% male, and 57 (26%) were insured by Medicaid. Sixty-three percent (139/222) of patients were adherent to follow-up. Seventy-six percent of patients discharged on tube feeds had active follow-up compared to 55% of patients who were not, with similar findings for oxygen at discharge (76% vs. 55%). Kaplan–Meier analysis showed patients with smaller defect size had earlier attrition compared to patients with larger defect size. Other race (Hispanic, Asian, Middle Eastern) patients had 2.87 higher odds of attrition compared to white patients (95% CI 1.18–7.0). Medicaid patients had 2.64 higher odds of attrition compared to private insurance (95% CI 1.23–5.66). ConclusionLoss to follow-up was associated with race and insurance type. Disease severity was similar between the active and inactive clinic cohorts. Long-term CDH clinic publications should examine attrition to ensure reported outcomes reflect the discharged population. This study identified important factors to inform targeted interventions for follow-up adherence. Level of EvidenceLevel III.

Effects of different preparation protocols of blood serum eyedrops on corneal healing after alkali burn: Clinical evaluation and α-SMA expression

Background: In cases of alkali burn, blood serum eyedrops may supply protein and growth factors. A standard preparation protocol is necessary for consistency in epitheliotropic capacity. Objective: We aim to evaluate how post-alkali-burn administration of blood serum eyedrops from different preparation protocols results in different outcomes on rabbits. Methods: This randomized experimental study used three treatments: placebo, non-diluted serum, and 25% concentration serum. We analyzed the corneal haziness, neovascularization degree, histopathological examination of neovascularization scores, and alpha-smooth muscle actin (α-SMA) expression. Results: Twelve rabbits (24 eyes) were split evenly into three treatment groups. On day 14, the group receiving the 25% concentration serum had the least corneal haziness (p = 0.005). No adverse effect was observed on the treated eyes. Histopathologically, no significant difference in neovascularization scores and α-SMA expression was observed. Conclusions: Treatment using the 25% concentration serum resulted in less corneal haziness, smaller defect size, and greater healing rate. There was also no significant difference in histopathological outcomes among the three treatment groups.

Deep learning cigarette defect detection method based on saliency feature guidance

AbstractCigarette defect detection is important in industrial production. Existing methods extract features for defect detection manually or using deep learning. However, due to the small size of cigarette defects, these methods are unable to effectively extract discriminative features, limiting detection performance. Hence, a deep learning‐based method called significant feature‐guided cigarette defect detection (SFGCD) is proposed, which combines saliency feature extraction methods with deep learning to enhance feature representation and improve detection. First, edge saliency features are extracted using the proposed target gradient saliency feature extraction (GSFE) strategy. Then, a dense multi‐level feature fusion network is designed to combine the original features with the saliency features obtained from the target gradient saliency feature extraction strategy. This network enriches feature representation and improves detection by fusing original and saliency features at different levels and scales. Experimental results demonstrate that the proposed method achieves a higher accuracy of 0.02 mean average precision (MAP) value and a detection speed of 5 frames per second (FPS) on the authors' own labeled cigarette defect dataset compared to existing state‐of‐the‐art methods.

Multi-scale surface defect detection method for bottled products based on variable receptive fields and Gather–Distribute feature fusion mechanism

The inspection of visual quality represents a crucial step in the production process of bottled products. Numerous machine vision methodologies have demonstrated proficient identification of significant defects on bottle surfaces in well-controlled imaging environments. However, the actual production scenario introduces a myriad of surface defect types on bottled products, exhibiting diverse shapes, with the majority of defect instances characterized by relatively small individual areas. Faced with such diverse and small-sized defects, traditional convolutional neural networks (CNNs) encounter limitations due to the utilization of fixed-size convolutional kernels. This choice results in a constrained receptive field, hindering the capture of sufficiently effective contextual information. Additionally, pooling operations within traditional CNNs lead to a substantial reduction in feature map dimensions, causing excessive blurring or outright neglect of small-sized defects. Consequently, this detrimentally impacts the accuracy of surface defect detection and recognition. This study proposes a novel multi-scale defect detection model incorporating a variable receptive field and Gather–Distribute feature fusion mechanism to overcome limitations of traditional CNNs. Nine different scale defect images of bottled products, including worn, wrinkles, joint markings and etc, were collected and enhanced to construct a surface defect detection dataset for the actual production of bottled products. Enhancements to convolutional layers and the C2f module improve feature extraction for different defect shapes and sizes. Integration of the Gather–Distribute feature fusion module (GDFFM) reduces feature information loss and enhances shallow-layer feature utilization. An efficient detection head (E-Detect) with ”parameter sharing” is proposed to reduce computational complexity and improve detection speed without experiencing significant accuracy loss. Experimental results demonstrate that the model’s superiority over advanced defect detection algorithms across various categories. Notably, it achieves accuracies of 89.9%, 55.6%, 89.4%, and 75% for challenging defects like small worn, subtle wrinkles, inclined external cover, and misaligned labels, with improvements of 3.2%, 7%, 2%, and 11.8% over the baseline model, respectively. The model’s mean Average Precision (mAP) also increases by 2.7%.