Fabric Defect Image Research Articles

Fabric Defect Detection Using Transfer Learning

Transfer learning in fabric defect detection involves utilizing pre-trained deep learning models on a large dataset, typically from a different domain, and fine-tuning them on a smaller dataset that is specific to fabric defects. By leveraging transfer learning, the limitations of limited annotated data for fabric defect detection can be overcome by utilizing the knowledge gained from a more extensive and diverse dataset. The pre-trained model's learned features are adjusted to recognize specific fabric defect patterns, resulting in more accurate and efficient defect detection. This approach reduces the reliance on a massive labelled dataset for training, which is particularly beneficial in industrial applications where obtaining a vast amount of annotated fabric defect images may be challenging. Ultimately, transfer learning enhances the model's ability to generalize and detect fabric defects with higher precision, thereby contributing to improved quality control in textile manufacturing processes.

Fabric defect image generation method based on the dual-stage W-net generative adversarial network

Due to the intricate and diverse nature of textile defects, detecting them poses an exceptionally challenging task. In comparison with conventional defect detection methods, deep learning-based defect detection methods generally exhibit superior precision. However, utilizing deep learning for defect detection requires a substantial volume of training data, which can be particularly challenging to accumulate for textile flaws. To augment the fabric defect dataset and enhance fabric defect detection accuracy, we propose a fabric defect image generation method based on Pix2Pix generative adversarial network. This approach devises a novel dual-stage W-net generative adversarial network. By increasing the network depth, this model can effectively extract intricate textile image features, thereby enhancing its ability to expand information sharing capacity. The dual-stage W-net generative adversarial network allows generating desired defects on defect-free textile images. We conduct quality assessment of the generated fabric defect images resulting in peak signal-to-noise ratio and structural similarity values exceeding 30 and 0.930, respectively, and a learned perceptual image patch similarity value no greater than 0.085, demonstrating the effectiveness of fabric defect data augmentation. The effectiveness of dual-stage W-net generative adversarial network is established through multiple comparative experiments evaluating the generated images. By examining the detection performance before and after data augmentation, the results demonstrate that mean average precision improves by 6.13% and 14.57% on YOLO V5 and faster recurrent convolutional neural networks detection models, respectively.

Fabric defect classification using prototypical network of few-shot learning algorithm

Computer-vision technology plays a vital role in automated fabric defect classification. In this paper, a novel prototypical network is presented for improving the fabric defect classification performance, especially in the case of an imbalanced distribution over the number of class samples. A traditional neural network (such as a convolutional neural network) usually inputs a batch of samples each time in training until the entire training dataset is covered, and thus it is not robust to cope with imbalanced data. The proposed network follows an N-way K-shot paradigm to split the training set into a support set and query set, and thereby forces the number of samples within each class to be uniformly distributed. The support set is used to learn the common knowledge of each class, whereas the query set is utilized to fine-tune the model parameters gained from the respective support set. The prototype of each class in the support set is computed as the representation of the class. For samples in the query set, the loss function is designed to match them with the corresponding prototypes as accurately as possible. In addition, the class activation mapping is used to visualize and interpret the discriminative regions of interest most relevant to specific defect classes. The classification performance of the proposed method is tested against five existing models on a labeled dataset of fabric defect images collected by a commercial inspection system. The proposed method achieves the highest classification accuracy (96.04%) over seven defect categories among the six tested methods.

Automatic Fabric Defect Detection Based on an Improved YOLOv5

Fabric defect detection is particularly remarkable because of the large textile production demand in China. Traditional manual detection method is inefficient, time-consuming, laborious, and costly. A deep learning technique is proposed in this work to perform automatic fabric defect detection by improving a YOLOv5 object detection algorithm. A teacher-student architecture is used to handle the shortage of fabric defect images. Specifically, a deep teacher network could precisely recognize fabric defects. After information distillation, a shallow student network could do the same thing in real-time with minimal performance degeneration. Moreover, multitask learning is introduced by simultaneously detecting ubiquitous and specific defects. Focal loss function and central constraints are introduced to improve the recognition performance. Evaluations are performed on the publicly available Tianchi AI and TILDA databases. Results indicate that the proposed method performs well compared with other methods and has excellent defect detection ability in the collected textile images.

Multi-Class Object Learning with Application to Fabric Defects Detection

Deep convolutional neural networks (CNNs) have shown great success in single-class fabric image detection. However, real-world fabric defect images generally contain several types of defects in one image. Accurately recognizing and classifying multi-class fabric defect images is still an unsolved issue due to the complexity of intersected defects, as well as difficulty in distinguishing small-size defects. To address these challenges, this study develops a methodology based on the deep learning feature pyramid networks (FPN) approach to detect multi-class fabric defects. To evaluate the proposed detection model, we built a unique multi-class fabric defects database (DHU-MO1000), where multi-class defect images are generated by industrial monitors from a textile factory. We used the dataset as the benchmark for multi-class defects detection training and testing the FPN. Furthermore, we conducted extensive experimental validations for various design choices. The experimental results show that the model outperformed existing multi-class object detection methods.

Fabric defect detection via low-rank decomposition with gradient information and structured graph algorithm

The low rank decomposition model shows the potential of fabric defect detection, in which a matrix is decomposed into a sparse matrix representing the defect free region (background) and identifying the defect area (foreground). However, there are still two shortcomings. Firstly, when the texture of defect image has high gradient feature, the sparse matrix obtained by the existing model still retains a large number of edges of undetected regions. Secondly, due to the imprecision of prior information, most models will misjudge the defect free points around the defect block when dealing with the small defect area or the defect area containing multiple cycles. In order to solve these problems, this paper proposes a fabric defect detection method based on low rank decomposition of gradient information and structured graph algorithm: 1) structured graphics algorithm, according to the characteristics of fabric defect image, fabric defect image is divided into defect free block with local feature and defect damage period. 2) In the merging process, an adaptive threshold is set according to the number of cycles contained in the current block to encourage intra lattice merging and prevent the merging of defective blocks and surrounding non defective blocks. 3) The defect prior information calculated from the segmentation results is used to guide matrix decomposition to weaken the defect free region and highlight the defect area under the sparse term. We evaluated our model on a standard database and compared it with the four latest methods. The total TPR and fpr of this method are 87.3% and 1.21% respectively on box, star and point databases, which is the best performance.

Automatic Fabric Defect Detection Using Cascaded Mixed Feature Pyramid with Guided Localization.

Generic object detection algorithms for natural images have been proven to have excellent performance. In this paper, fabric defect detection on optical image datasets is systematically studied. In contrast to generic datasets, defect images are multi-scale, noise-filled, and blurred. Back-light intensity would also be sensitive for visual perception. Large-scale fabric defect datasets are collected, selected, and employed to fulfill the requirements of detection in industrial practice in order to address these imbalanced issues. An improved two-stage defect detector is constructed for achieving better generalization. Stacked feature pyramid networks are set up to aggregate cross-scale defect patterns on interpolating mixed depth-wise block in stage one. By sharing feature maps, center-ness and shape branches merges cascaded modules with deformable convolution to filter and refine the proposed guided anchors. After balanced sampling, the proposals are down-sampled by position-sensitive pooling for region of interest, in order to characterize interactions among fabric defect images in stage two. The experiments show that the end-to-end architecture improves the occluded defect performance of region-based object detectors as compared with the current detectors.

A visual long-short-term memory based integrated CNN model for fabric defect image classification

Fabric defect classification is traditionally achieved by human visual examination, which is inefficient and labor-intensive. Therefore, using intelligent and automated methods to solve this problem has become a hot research topic. With the increasing diversity of fabric defects, it is urgent to design effective methods to classify defects with a higher accuracy, which can contribute to ensuring the fabric products' quality. Considering that the fabric defect is not obvious against the texture background and many kinds of them are too confusing to distinguish, a visual long-short-term memory (VLSTM) based integrated CNN model is proposed in this paper. Inspired by the human visual perception and visual memory mechanism, three categories of features are extracted, which are the visual perception (VP) information extracted by stacked convolutional auto-encoders (SCAE), the visual short-term memory (VSTM) information characterized by a shallow convolutional neural network (CNN), and the visual long-term memory (VLTM) information characterized by non-local neural networks. Experimental results on three fabric defect datasets have shown that the proposed model provides competitive results to the current state-of-the-art methods on fabric defect classification.

Unsupervised segmentation and elm for fabric defect image classification

In order to solve the problem of low accuracy and efficiency for surface defects in common woven fabrics, a novel fabric defect classification method is proposed based on unsupervised segmentation and ELM. The classification method is divided into four steps including defect segmentation, feature extraction, ELM classifier training, and Bayesian probability fusion. Firstly, an unsupervised segmentation is presented for the Grayscale fabric defect image after preprocessing. Secondly, geometric and texture features were extracted by using the segmented image and the undivided Grayscale image. Then, features and labels in fabric defect images are considered as training sets to train the ELM classifier. Finally, the input fabric defect image is classified by the trained ELM classifier and the Bayesian probability fusion method. Experimental results show that the proposed method can classify the fabric defect image with high accuracy and efficiency that can better meet the requirements for practical applications.

The design of optimal real Gabor filters and their applications in fabric defect detection

Fabric defect detection has been recognised as one of the key challenges for automatic production, and Gabor filters are one of the most useful tools in detecting fabric defects. The half‐peak tangent method is applied in real Gabor filter design so that the filters can cover the frequency of defects as much as possible. Meanwhile, the half‐peak‐magnitude contours of the neighbouring filters are tangential. On this basis, two optimal orientations are selected by applying direction masks, and the optimal scale at each optimal orientation is determined according to the signal‐to‐noise ratio. In this way, two optimal real Gabor filters are obtained. A new algorithm based on the two optimal filters is proposed for fabric defect detection. A series of experiments are carried out for 46 fabric defect images combined with 46 corresponding reference fabric images, in order to verify the effectiveness of the new algorithm. The experimental results obtained show that the new algorithm can accurately detect defects in grey fabric defect images as well as in colour images. For the 46 fabric defect images, the detection rate is 95.66%, indicating that the new algorithm performs well. In addition, comparison of the new algorithm with other algorithms in the literature demonstrates that the new algorithm is more effective in the detection of several fabric defect images.

Fabric defect image segmentation based on the visual attention mechanism of the wavelet domain

In order to explore the accurate image segmentation of fabric defects, we will introduce the visual attention mechanism of the wavelet domain to the dynamic detection of fabric defects. First of all, feature maps are formed by extracting simple features from a collection image. Secondly, feature maps by multi-layer wavelet decomposition are decomposed into a lot of feature sub-maps of the wavelet domain. On this basis, the center-surround operator among feature sub-maps of the wavelet domain is adopted to build the feature difference sub-maps, which are fused into feature saliency maps through fuse strategy. Finally, the defect interest areas are segmented based on the maximum between-cluster variance method in saliency maps, and the fabric defects through the region growing method are detected in the defect interest areas. Comparing with the wavelet transform algorithm, experimental results show that the proposed method is able to segment the defect information completely, and it has a strong ability to resist noise interference, which can improve the accuracy of defect detection.

Segmentation of Fabric Defect Images Based on PCNN Model and Symmetric Tsallis Cross Entropy

Segmentation of defect images is an important step in the automatic fabric defect detection. In order to extract fabric defects effectively, a segmentation method of fabric defect images based on pulse coupled neural network (PCNN) model and symmetric Tsallis cross entropy is proposed. The image is segmented by PCNN according to the gray strength difference between fabric defect area and non-defect area. To guarantee that the grayscale inside the object and background is uniform after segmentation, symmetric Tsallis cross entropy is used as the image segmentation criterion to select the optimal threshold and iteration number. A large number of experimental results show that, compared with the related segmentation methods such as Otsu method, PCNN method, the method based on PCNN and cross entropy, the segmentation effect of the proposed method is the best. The texture of non-defect area is removed more completely, and the defect area is segmented more accurately.

Image analysis measurement of cottonseed coat fragments in 100% cotton woven fabric

Our research presented an objective and reliable evaluation method for measuring of fabric cottonseed coat fragment (CSCF) using a scanner with higher accuracy. After the surface image of woven fabric was firstly transformed into the frequency domain using discrete Fourier transform, the image of CSCF parts was then enhanced with Gabor filter to highlight CSCF. Segmentation and localization of CSCF areas parts were achieved by inverse discrete Fourier transform in the reconstructed image. The number and area of CSCF were extracted by threshold at last, and three parameters including recognition ratio, error recognition ratio and omission ratio were used to evaluate the accuracy of the measurements. The results can be used as descriptors of remain of cottonseed coat. The experimental results demonstrated that the fabric CSCF evaluation system developed in this study represented the CSCF properties and also showed high accuracy in the degree of CSCF. By such an inspection system, the CSCF recognition ratio of the fabric CSCF evaluation system defect images comes up to 91 %.

Fabric Defect Automated Detection Technology

To solve the problem of automated defect detection for textile fabrics, this paper proposed a method for fabric defect detection which is based on computer vision. After the operations in many aspects of basis image processing, such as gray-scale, denoising, contour detection and morphological, and it can transmit the fabric defect image information to host-computer with the USB interface in time. In order to acquire high processing speed, the captured images from each camera are sent into one dedicate computer for distributed and parallel image processing. The experimental result confirms that the proposed method can is feasible in rapid defect detection, and has a broad market prospect.

New approach of fabric defects detection based on saliency region feature

Concerning the diversity of fabric defect type and low efficiency of traditional artificial detection methods,in order to detect the fabric defect more effectively,a new approach,SGE,based on saliency region feature for fabric defect detection was studied.In this approach,the original image was divided into two parts,one was used to extracted the saliency region feature of fabric defect by improved Frequency-tuned Saliency Region(FSR) method roughly,another was used to employ the Gabor filter with the amplitude as an output characteristic,and to extract the saliency region feature of fabric defect by Pixel Saliency Region(PSR) method accurately,then by using maximum entropy to segment the saliency region respectively and to fuse the sub-images.The result was got by calculating perimeter and area of the contours to remove the isolated points.The experiment selected four types of typical fabric defect images and OpenCV library was used.The experimental result shows that the algorithm,without prior learning,meets the real-time requirement.

Research of Postprocessing Algorithm after Fabric Defect Image Segmentation Based on Downsampling

This paper research the algorithm after image segmentation based on using downsampling to preprocess the image and Otsu to segment the image, because defect region is weakened in the process of preprocessing to remove the texture, which lead to the defect region in the binary image can not reflect the original defect information. In view of this shortcoming, this paper put forward a new method to link the different connected regions which belong to a defect region and the new method solve the problem of weakening the defect region in the process of preprocessing, and then the integral information of defect region can be obtained. At last, the experiments’ results prove the validity of the arithmetic in this paper.

Fabric Defect Detection Method Based on Optimal Gabor Filter Bank

A method of fabric defect detection online based on Gabor filter bank is proposed. First, optimal Gabor filter bank is designed to match texture features of normal fabric image and built adopting the half-peak tangent method. The filter bank’s parameters are obtained by using texture features. Then, fabric defect image is treated using this filter bank, and filtered sub-images are obtained which characterize the fabric defect in different orientations and frequencies. Finally, these sub-images are fused to reconstruct a binary defect image. And the binary defect image is output to finish the detection. Experiment results show that the Gabor filter bank based on texture features can reduce the number of filters effectively and detect most of the fabric defects.

Combined fabric defects detection approach and quadtree decomposition

In this article, various approaches are investigated and synthesized for real-time inspection of diversified fabric texture. Improved Gabor filter bank was proposed to detect fabric defects by using its property of steerable directions and scales. Normal fabric textured image was processed with optimal filter and then gauss smooth. Two threshold value of thenormal fabric image will be determined, which are considered as standard of image parameter for segmenting defect areas. For some intense background texture of fabric defect images, background analysis approach is proposed to detect the flaws. The third detection method is multi-scale wavelet, and typical results of the test are provided. Threeapproaches synthesized into a cascade detection system to develop better detection results. Inspection results that conducted on real fabric defects for various approaches reveal that combined detection scheme meet good expectation. Further, preliminary graphical user interface is designed so as to facilitate the operation. After integrate graphical user interface with procedure, parameters material can be acquired, which is vital to applying the inspection system on industrial computer in future.

Defect Detection Study of Fabric Based on Gabor Filter Algorithm

This thesis exploits a multichannel Gabor filters detection algorithm. Analysis filtering images from different orientations and scales, then fuses the multichannel data. Finally, a threshold iterative algorithm and mathematical morphology post-processing is used to achieve the fabric defect detection. The experiment selects five types of fabric defect image. Experimental results suggest that this algorithm can effectively identify blob-shaped, linear and planar defect and has well real-time character.

Fabric defect image de-noising based on contourlet transform and nonlinear diffusion

Fabric defect image de-noising based on contourlet transform and nonlinear diffusion