Defects In Region Research Articles

NFCF: Industrial Surface Anomaly Detection with Normalizing Flow Cross-Fitting Network

The effective fully-supervised defect detection methods in industry are done by training many defect labels. These methods face application challenges because of insufficient defect samples. Moreover, the subtlety of defects, the similarity between defect-free and defect regions, and the interference factors carried by defect-free images also pose problems. To solve these difficulties above, we propose a Normalizing Flow Cross-Fitting network (NFCF) to use only defect-free images as priori knowledge for training. Specifically, the incremental broadening module in the network focuses on information expansion for minor defects, and the interactive filtering module completes the weight filtering through same-scale mutual reasoning. The processed information is fitted to a defect-free distribution by the normalizing flow module and the distribution is used as the basis for the determination. The NFCF does experiments on four types of datasets, and it achieves an average of 96.3% and 95.88% on two metrics, AUC-Image and AUC-Pixel. Experimental results show that the network can distinguish defect images using only defect-free training. Moreover, visualization experiments show that it can also complete localization segmentation for minor defects. Based on semi-supervision with achieved accuracy, NFCF alleviates the labeling dependence problem of defects and significantly increases the application value.

Memory linked knowledge domain transfer few-shot learning for thermography nondestructive evaluation system

Infrared NDT techniques are suitable for detecting near-surface and sub-surface defects in composite materials. However, significant detection challenges remain while data volume is quantitatively limited for training weak contrast between defective and non-defective regions. In this paper, we propose an end-to-end memory linked knowledge domain with transfer few-shot learning segmentation network with a wider general model that solves the problems of insufficient data support. The proposed method is shown to attain high detection performance for multiple types of specimens by using cross domain adaption. By improving the support branch, a foreground-background joint guidance linked with memory bank is proposed. This provides better guidance information for the query branch. In addition, the intra-class and inter-class loss metric distance between defective and non-defective features is increased for better segmentation performance. In order to enable interpretability of the model, robust experiments have been conducted to detect the inner debond on multiple carbon fiber reinforced polymer (CFRP) composites. A comparative analysis is presented and compared with the state-of-the-art machine learning algorithms.

Highly tough and conductive hydrogel based on defect-patched reduction graphene oxide for high-performance self-powered flexible sensing micro-system

Tough and conductive reduction graphene oxide (rGO)-based hydrogels have broad application prospects in various flexible wearable electronics. However, an often ignored phenomenon was that numerous defective regions were inevitably formed within the basal plane of rGO during synthesis process, deteriorating the physical characteristics as well as overall performances of wearable devices. Herein, a facile and effective defect patching engineering was proposed to heal rGO by carbonized metal–organic frameworks (CMOF). The fixed CMOF at defect regions acted as a bridge to allow electrons to pass quickly and enhance tolerance of rGO to loaded mechanical energy. The resultant optimized hydrogel containing defect-patched rGO flakes not only possesses satisfied mechanical properties (e.g., tensile strength of 195 kPa) and high electrical conductivity (2.42 S/m), but also exhibits reinforced electric output performances when assembled into triboelectric nanogenerator, supercapacitors and sensor devices. More importantly, the sensor with high sensitivity (gauge factor (GF) of 14.68, short response time of 40 ms) was capable of effective distinguishing complex human activities and accurate detecting the temperature fluctuation of skin within narrow range. As proof-of-concept, the as-synthesized all-in-one flexible intelligent micro-system, composed of power supply, energy storage and sensing units, had proved the great compatibility and feasibility of wearable electronics in detecting and recognizing human expressions, motion and physiological signals. Our findings proposed a reliable defect patching strategy for large-scale production of defect-free 2D carbonaceous materials, boosting the development of high-performance self-powered sensing micro-system in wearable devices.

Pull & Push: Leveraging Differential Knowledge Distillation for Efficient Unsupervised Anomaly Detection and Localization

Recently, much attention has been paid to segmenting subtle unknown defect regions by knowledge distillation in an unsupervised setting. Most previous studies concentrated on guiding the student network to learn the same representations on the normality, neglecting the different behaviors of the abnormality. This leads to a high probability of false detection of subtle defects. To address such an issue, we propose to push representations on abnormal areas of the teacher and student network as far as possible while pulling representations on normal areas as close as possible. Based on this idea, we design an efficient teacher-student model for anomaly detection and localization, which maximizes pixel-wise discrepancies for anomalous regions approximated by data augmentation and simultaneously minimizes discrepancies for pixel-wise normal regions between these two networks. The explicit differential knowledge distillation enlarges the margin between normal representations and abnormal ones in favour of discriminating them. Then, the appropriate small student network is not only efficient, but more importantly, helps inhibit the generalization ability of anomalous patterns when learning normal patterns, facilitating the precise decision boundary. The experimental results on the MVTec AD, Fashion-MNIST, and CIFAR-10 datasets demonstrate that our proposed method achieves better performance than current state-of-the-art (SOTA) approaches. Especially, For the MVTec AD dataset with high resolution images, we achieve 98.1 AUROC% and 93.6 AUPRO% in anomaly localization, outperforming knowledge distillation based SOTA methods by 1.1 AUROC% and 1.5 AUPRO% with a lightweight model.

Applying SDR with CNN to Identify Weld Defect: A New Processing Method

X-ray weld images are investigated to achieve nondestructive identification of defects in oil and gas pipeline welds. The content mainly includes image filtering and enhancement, weld defect suspected defect region (SDR) segmentation, and defect identification. (1) The x-ray image quality is poor, as shown by a lot of noise and low contrast. Therefore, the mean filtering method is used to filter out the noise, and a nonlinear enhancement using sin function is proposed to improve the contrast between the weld and the background. (2) In the image segmentation, a segmentation method of SDR is proposed. Region of interest (ROI) is first extracted, and then SDRs are segmented using clustering. (3) In defect identification, deep learning networks are used for SDRs. A 6-stage, 10-layer convolutional neural network (CNN) network structure is designed, and the convolutional kernels are set to be 5×5 and 3×3. In the experiments, the proposed method is able to automate the processing of x-ray weld images and improve the processing efficiency. The identification accuracy rate is 98.9%.

Watermark Detection in CMOS Image Sensors Using Cosine-Convolutional Semantic Networks

This article proposes a deep learning approach for automatic segmentation of low-contrast watermark defects on Complementary Metal Oxide Semiconductor (CMOS) Image Sensor (CIS) surfaces. The task of surface defect detection has been tackled by machine vision or deep learning methods. Traditional machine vision methods require expert domain knowledge for feature extraction, and they are computationally intensive. Deep learning methods based on Convolutional Neural Network (CNN) are extremely fast, but can be affected by illumination variations and perform poorly for low-contrast images. In this study, a U-Net type of encoder-decoder architecture is proposed. The Convolutional operation is substituted with the proposed Cosine-Convolutional operation in the encoder layers of the U-Net model with the objective of minimizing the effects of uneven illumination and highlighting the defect region. To further reduce the computation complexity and improve defect segmentation, an attention-like mechanism is used in the semantic encoder-decoder network. Pixel-wise multiplication is applied in the skip connections between encoder and decoder layers of the U-Net to replace the classical concatenation, thus reducing the parameter count of the model. The proposed methods are computationally fast, with an evaluation time of less than 10-ms for an image of 384×256 pixels on a single low-end GPU. Experimental results show that the proposed Cosine-Convolutional encoder-decoder architectures can effectively and accurately be applied for defect detection and segmentation of watermarks in low-contrast CIS images.

Size-controllable Rh2P nanoparticles on reduced graphene oxide toward highly hydrogen production

Abundant of defective regions and oxygen-containing groups on graphene oxide (GO) could promote the nucleation and growth of metallic nanoparticles. In this work, the composite catalyst is synthesized by pyrolysis of the mixture powder of GO and rhodium organometallic reactant at 900 °C, Ar atmosphere and low pressure, in which dispersive Rh2P nanoparticles are uniformly attached on reduced graphene oxide (Rh2P/RGO). The sizes of Rh2P nanoparticles are controllable by adjusting the mass proportion of reactants. For the optimized size of Rh2P nanoparticles at 6.0 nm, only 11.7 and 10.3 mV of the overpotentials are required at the current density of 10 mA cm−2 in 0.5 M H2SO4 and 1 M KOH electrolytes, respectively, exhibiting superior HER activity over the commercial Pt/C. Theoretical calculations indicate that the hydrogen adsorption free energy of Rh2P/RGO composites is close to zero, and Bede charge analysis confirms the accelerated electron transport between the RGO and Rh2P. Therefore, the introduction of RGO substrate could not only promote a uniform distribution and tunable size of Rh2P NPs, but also be advantageous to the electron transport from RGO to Rh2P NPs, and a favorable H2 evolution during the HER process. This work provides a facile strategy to control the size of metal phosphide nanoparticles on RGO for the design of advanced HER electrocatalysts.

Perspective on pulse compression favorable infrared imaging techniques for non-destructive testing and evaluation of solids

Abstract Inspection reliability of sub-surface defects is imperative for safer functionality of critical components/materials/structures used in a wide variety of applications in various industries. The need for reliable, fast, remote, safe inspection and evaluation methods for detection of hidden defects increases in parallel with the demand for more sustainable solutions which helps in inherent modifications in design and manufacturing specifications. During in-service operation, the hidden defects are typically originated from various loading conditions leading to catastrophic failure. This perspective explores the best possible reliable, fast, remote, safe, and easy to implement for field inspection and evaluation experimental method and the associated post-processing approach using InfraRed Imaging (IRI) for Thermal Non-Destructive Testing and Evaluation (TNDT&E) of Glass Fibre Reinforced Polymer (GFRP) materials having delamination defects. This perspective further explores the state-of-the-art infrared imaging modalities used in TNDT&E by highlighting their advantages and limitations in terms of the detection sensitivity and depth resolvability to detect the subsurface defects located at interior of material. Most of the proposed experimental and post-processing techniques presented in the literature for TNDT&E, explore the observed spatial thermal contrast over the defective region of sample to provide the depth resolvability from the reconstructed thermograms even though it is obtained from the temporal data processing on the captured image sequence. This perspective provides an insight on the state-of-the-art research in the field of thermal/infrared non-destructive testing and evaluation and associated post-processing approaches to visualize the hidden subsurface defects not only resolved by spatial thermal gradients but also simultaneously provide temporal thermal gradients to locate defective regions.

High-quality InGaAs films bonded on Si substrate with a thin polycrystalline Si intermediate layer

We report the fabrication of high-quality InGaAs films on Si by wafer bonding at low temperatures (300 ℃). A thin polycrystalline Si (poly-Si) interlayer is introduced between InGaAs and Si to block the lattice mismatch and absorb the bubble by-products produced at the bonded interface using its disordered crystal orientation and gain boundaries. To achieve a high bonding strength, argon plasma treatment was employed on InGaAs and poly-Si surfaces before the pre-bonding process. By using post-wet etching, a defect-rich region near the surface of exfoliated InGaAs film that was initially caused by the thermal mismatch between InGaAs and InP during epitaxy growth can be removed. As a result, a Si-based InGaAs film with a 1.05 MPa bonding strength and bubble-free bonding interface was produced. The defective region is 1.2 μm-thick. After etching the defective area, a high-quality, 1.8 μm-thick Si-based InGaAs film was obtained with an XRD FWHM of 94 arcsec. The results reveal that this method is prospective for the fabrication of high-quality III-V materials on Si, proving that a prospective approach for the fabrication of low-noise InGaAs/Si PIN photodiodes and InGaAs/Si avalanche photodiodes with ultralow dark currents.

Thermal integrity profiling of cast-in-situ piles in sand using fiber-optic distributed temperature sensing

Defects in cast-in-situ piles have an adverse impact on load transfer at the pile‒soil interface and pile bearing capacity. In recent years, thermal integrity profiling (TIP) has been developed to measure temperature profiles of cast-in-situ piles, enabling the detection of structural defects or anomalies at the early stage of construction. However, using this integrity testing method to evaluate potential defects in cast-in-situ piles requires a comprehensive understanding of the mechanism of hydration heat transfer from piles to surrounding soils. In this study, small-scale model tests were conducted in laboratory to investigate the performance of TIP in detecting pile integrity. Fiber-optic distributed temperature sensing (DTS) technology was used to monitor detailed temperature variations along model piles in sand. Additionally, sensors were installed in sand to measure water content and matric suction. An interpretation method against available DTS-based thermal profiles was proposed to reveal the potential defective regions. It shows that the temperature difference between normal and defective piles is more obvious in wet sand. In addition, there is a critical zone of water migration in sand due to the water absorption behavior of cement and temperature transfer-induced water migration in the early-age concrete setting. These findings could provide important insight into the improvement of the TIP testing method for field applications.

Global and local defect detection for 3D printout surface based on geometric shape comparison

A 3D printer produces parts from CAD models. The printout must be checked for defects. We present a surface defect detection algorithm that constructs a 3D mesh of the printout by multi-view scanning and compares it to the part CAD model. It computes and displays the distance from each vertex of the printout mesh to the closest point on the part model. These distances comprise the global deviation of the printout from the model. When the error is within a threshold, the algorithm compares the curvature at each printout vertex to that at the closest part point. Qualitative differences in curvature indicate local defects, such as blobs, cracks, surface roughness, and faded embossing and engraving. The results are grouped into local defect regions that are displayed. We demonstrate the algorithm on metal printing and on PLA 3D printing.

Texture and semantic convolutional auto‐encoder for anomaly detection and segmentation

AbstractAnomaly detection is a challenging task, especially detecting and segmenting tiny defect regions in images without anomaly priors. Although deep encoder‐decoder‐based convolutional neural networks have achieved good anomaly detection results, existing methods operate uniformly on all extracted image features without considering disentangling these features. To fully explore the texture and semantic information of images, A novel unsupervised anomaly detection method is proposed. Specifically, discriminative features are extracted from images by using a deep pre‐trained network, where shallow and deep features are aggregated into texture and semantic modules, respectively. Then, a feature fusion module is developed to interactively enable feature information in two different modules. The texture and semantic segmentation results are obtained by comparing the texture features and semantic features before and after reconstruction, respectively. Finally, an anomaly segmentation module is designed to generate anomaly detection results by integrating the results of the texture and semantic modules by setting a threshold. Experimental results on benchmark datasets for anomaly detection demonstrate that our proposed method can efficiently and effectively detect anomalies, outperforming some state‐of‐the‐art methods by 2.7% and 0.6% in classification and segmentation.

Research on anti-interference detection of 3D-printed ceramics surface defects based on deep learning

The inherent brittleness of 3D-printed ceramic parts makes the surface more prone to produce defects. In the 3D-printing environment, the defect regions in the surface are easily covered by interference factors. This brings great difficulties to surface defect detection. Based on this, this paper proposes an anti-interference detection method for surface defects of ceramic parts based on deep learning. This anti-interference detection method is mainly divided into three stages: interference factors identification, interference factors repair, defect detection. Interfering factors in the surface are located and identified through the built Multimodal Feature Layer Fusion-PSP network (MFLF-PSP net) model. MFLF-PSP net's mAP for interference factor identification is up to 95.67%. Then, based on the results of the interference identification, the proposed Parallel Spatial-Channel Attention Mechanism (PSCAM)-RFR net model is used to perform pixel filling and repair in the regions where the interference factor is located. This solves the difficult problem of defect detection caused by interference factors. On this basis, the constructed Inception-SSD network model is used to perform effective defect detection on ceramic surface images. The mAP of the model for the detection of crack defects and bulge defects is 96.34% and 94.92%, respectively. Through the close cooperation of the above three stages, the problem of defect detection caused by interference factors, such as misjudgment and defect region segmentation, has been solved. It provides technical support for automatic detection of ceramic surface quality during 3D-printing.

Papillary vitreous detachment as a possible accomplice in non-arteritic anterior ischaemic optic neuropathy

AimTo evaluate the role of papillary vitreous detachment in the pathogenesis of non-arteritic anterior ischaemic optic neuropathy (NAION) by comparing the features of vitreopapillary interface between NAION patients and normal...

Effect of low-intensity pulsed ultrasound on osteogenic differentiation of human induced membrane-derived cells in Masquelet technique

IntroductionThe Masquelet technique is a relatively new method for large bone defect treatment. In this technique, grafted bone tissue is used, and after the cement is removed, the induced membrane (IM; that form around the cement spacers placed in the bone defect region) is thought to play an important role in promoting bone formation. On the other hand, low-intensity pulsed ultrasound (LIPUS) is known to promote fracture healing and angiogenesis through mechanical stimulation. This study aimed to investigate the in vitro effects of LIPUS on the osteogenic differentiation of human induced membrane-derived cells (IMCs). MethodsSeven patients who had been treated using the Masquelet technique were enrolled. The IM was harvested during the second stage of the technique. IMCs were isolated, cultured in growth medium, and then divided into two groups: (1) control group, IMCs cultured in osteogenic medium without LIPUS, and (2) LIPUS group, IMCs cultured in osteogenic medium with LIPUS treatment. Adherent cells from the IM samples were harvested after the first passage and evaluated for cell surface protein expression using immunostaining. A cell proliferation assay was used to count the number of IMCs using a hemocytometer. Osteogenic differentiation capability was assessed using an alkaline phosphatase (ALP) activity assay, Alizarin Red S staining, and real-time reverse transcription-polymerase chain reaction. ResultsCell surface antigen profiling revealed that the IMCs contained cells positive for the mesenchymal stem cell–related markers CD73, CD90, and CD105. No significant difference in cell numbers was found between the control and LIPUS groups. The ALP activity of IMCs in the LIPUS group was significantly higher than that in the control group on days 7 and 14. Alizarin red S staining intensity was significantly higher in the LIPUS group than in the control group on day 21. Runx2 and VEGF expression was significantly upregulated on days 7 and 14, respectively, compared with levels in the control group. ConclusionWe demonstrated the significant effect of LIPUS on the osteogenic differentiation of human IMCs. This study indicates that LIPUS can be used as an additional tool for the enhancement of the healing process of the Masquelet technique.

Helium focused ion beam induced subsurface damage on Si and SiC substrates: experiments and generative deep neural network modeling via position-dependent input

Deep neural networks (DNNs) are reshaping many fields due to their end-to-end ability to learn directly from data, leading to outstanding performance, especially in image processing. However, their training requires large image datasets while image preparation can be expensive/time-consuming. If the underlying behavior can be formulated as spatial-dependent, we propose that every pixel from every image can be considered as a different data source, thus enabling a truly large pixel-based dataset with millions of elements (‘position-dependent input’), even when obtained from a few images. The approach is applied to helium focused ion beam nanofabrication, where the cross-section of the helium-damaged region essentially resembles that of a lightbulb, while close inspection reveals the presence of (i) an outer defective region in direct contact with the bulk substrate and (ii) an inner amorphous region filled with helium bubbles of gradually increasing size. Interestingly, the amorphous phase may swell upwards to form a protruding mesa, depending on the beam energy and dose, a feature that has resisted modeling so far. Through dedicated experiments on both Si and SiC substrates, and careful image analysis (segmentation), we describe the transformations that take place in the defective and amorphous regions with increasing energy and dose, and use the outlined position-dependent input together with a simple DNN of 4 hidden layers and 16 neurons per layer to describe all damage features realistically, inherently demonstrating generative behavior. This is the first time that any model predicts swelling satisfactorily. Generalization is surprisingly smooth and accurate.

Study of the Radiation Damage Kinetics in NbTiVZr High-Entropy Alloys Irradiated by Heavy Ions

The purpose of this paper was to study the kinetics of accumulation of radiation damage in the near-surface layer of a high-entropy NbTiVZr alloy upon irradiation with heavy Kr15+ ions and fluences of 1010–1015 ion/cm2. According to the data of X-ray diffraction analysis, it was found that irradiation with heavy ions lead to structural changes associated with the accumulation of deformation distortions and stresses. In this case, the nature of deformation distortions was associated with tensile distortions of the crystal lattice and swelling of the near-surface layer. An analysis of the strength properties of the irradiated samples showed that the accumulation of deformation distortions in the near-surface layer lead to a decrease in the resistance to cracking and destruction of the near-surface layer. In this case, the main changes occurred at fluences above 1013 ion/cm2, which are characterized by an excess of the threshold for overlapping defective regions that appear along the trajectory of incident ions. Tribological tests showed that an increase in defective inclusions in the structure of the near-surface layer lead to an increase in friction and a decrease in crack resistance.

A hybrid model- and deep learning-based framework for functional lung image synthesis from multi-inflation CT and hyperpolarized gas MRI.

Hyperpolarized gas MRI is a functional lung imaging modality capable of visualizing regional lung ventilation with exceptional detail within a single breath. However, this modality requires specialized equipment and exogenous contrast, which limits widespread clinical adoption. CT ventilation imaging employs various metrics to model regional ventilation from non-contrast CT scans acquired at multiple inflation levels and has demonstrated moderate spatial correlation with hyperpolarized gas MRI. Recently, deep learning (DL)-based methods, utilizing convolutional neural networks (CNNs), have been leveraged for image synthesis applications. Hybrid approaches integrating computational modeling and data-driven methods have been utilized in cases where datasets are limited with the added benefit of maintaining physiological plausibility. To develop and evaluate a multi-channel DL-based method that combines modeling and data-driven approaches to synthesize hyperpolarized gas MRI lung ventilation scans from multi-inflation, non-contrast CT and quantitatively compare these synthetic ventilation scans to conventional CT ventilation modeling. In this study, we propose a hybrid DL configuration that integrates model- and data-driven methods to synthesize hyperpolarized gas MRI lung ventilation scans from a combination of non-contrast, multi-inflation CT and CT ventilation modeling. We used a diverse dataset comprising paired inspiratory and expiratory CT and helium-3 hyperpolarized gas MRI for 47 participants with a range of pulmonary pathologies. We performed six-fold cross-validation on the dataset and evaluated the spatial correlation between the synthetic ventilation and real hyperpolarized gas MRI scans; the proposed hybrid framework was compared to conventional CT ventilation modeling and other non-hybrid DL configurations. Synthetic ventilation scans were evaluated using voxel-wise evaluation metrics such as Spearman's correlation and mean square error (MSE), in addition to clinical biomarkers of lung function such as the ventilated lung percentage (VLP). Furthermore, regional localization of ventilated and defect lung regions was assessed via the Dice similarity coefficient (DSC). We showed that the proposed hybrid framework is capable of accurately replicating ventilation defects seen in the real hyperpolarized gas MRI scans, achieving a voxel-wise Spearman's correlation of 0.57±0.17 and an MSE of 0.017±0.01. The hybrid framework significantly outperformed CT ventilation modeling alone and all other DL configurations using Spearman's correlation. The proposed framework was capable of generating clinically relevant metrics such as the VLP without manual intervention, resulting in a Bland-Altman bias of 3.04%, significantly outperforming CT ventilation modeling. Relative to CT ventilation modeling, the hybrid framework yielded significantly more accurate delineations of ventilated and defect lung regions, achieving a DSC of 0.95 and 0.48 for ventilated and defect regions, respectively. The ability to generate realistic synthetic ventilation scans from CT has implications for several clinical applications, including functional lung avoidance radiotherapy and treatment response mapping. CT is an integral part of almost every clinical lung imaging workflow and hence is readily available for most patients; therefore, synthetic ventilation from non-contrast CT can provide patients with wider access to ventilation imaging worldwide.

Amorphized Defective Fullerene with a Single‐Atom Platinum for Room‐Temperature Hydrogen Storage

AbstractThe search for hydrogen storage materials allowing the storage of hydrogen in its molecular or atomic form at room temperature to meet the multistage targets such as the US Department of Energy (DOE) ultimate gravimetric and volumetric capacities of 6.5 wt% and 50 kg m‐3 is of global importance. Here, it is reported that an amorphized defective fullerene (C60‐x) offers a promising solution to this challenge. C60‐x immobilized with single‐atom platinum has ≈14‐fold higher surface area accessible for CH bonds compared to a crystalline C60, and its micro/meso pores give a ≈20‐fold larger volume for fast hydrogen transport. Indeed, hydrogen storage via spillover on C60‐x through pressure swing at room temperature is experimentally demonstrated to enable high reversible gravimetric (6.8 wt%) and volumetric (64.9 kg m‐3) capacities, hitherto the highest reversible capacities close to DOE targets at room temperature. Also, the density functional theory calculations show that a key to efficient hydrogen storage is the preservation of a curved sp2‐type local carbon geometry for spillover, which holds H radicals loosely for fast hydrogen migration. Moreover, H‐atom diffusion on the intact region of C60‐x is faster than that on the defect region. Furthermore, excellent capacity retention is achieved over repeated hydrogen adsorption/desorption cycles.

Fabric defect detection based on low-rank decomposition with factor group-sparse regularizer

Recently, many low-rank-based methods in terms of detecting defects in fabric images have been proposed. However, there are two disadvantages of these methods. First, current low-rank based methods use the nuclear norm as the surrogate of rank, which causes inefficient optimization process and sub-optimal performance. Second, low-rank defective regions cannot be detected by low-rank based models. Thus, we propose a factor group-sparse regularized low-rank decomposition model (FGSRLRD) to solve these problems. This method takes the factor group-sparse regularizer as the surrogate of rank, which is more efficient as singular value decomposition (SVD) is not applied in the optimization process. Better performance is achieved as the factor group-sparse regularizer is a more accurate approximation of the rank. In addition, the weight matrix generated by the lightweight autoencoder is incorporated into the object function of FGSRLRD to guide locating defective regions. Besides, as low-rank defective regions cannot be segmented by low-rank models, this method constructs a fusion image of the prior image and the sparse image to highlight the defective regions. The performance of the proposed method is evaluated on two standard datasets, and the results indicate that the suggested method outperforms the existing state-of-the-art methods in locating the defective regions on fabric images.