With the rise of labor costs and advancement automation in textile industry, fabric defect detection has become a hot research field recent years. We proposed learning-based framework for automatic defects. Firstly, we use fixed-size square slider to crop original image certain step regularity. Then an improved histogram equalization is used enhance each cropped image. Furthermore, Inception-V1 model employed predict existence defects local area. Finally, apply LeNet-5 model, which plays...

Automatic video segmentation is the first and necessary step for organizing a long file into several smaller units. The smallest basic unit shot. Relevant shots are typically grouped high-level called scene. Each scene part of story. Browsing these scenes unfolds entire story film, enabling users to locate their desired segments quickly efficiently. Existing definitions rather broad, making it difficult compare performance existing techniques develop better one. This paper introduces...

Purpose The problem of fabric defect detection is a particularly challenging task, as the defects occupy only small portion image pixels and it difficult to collect sufficient samples for training deep learning-based models. purpose this work present novel self-supervised learning method address problem. Design/methodology/approach In order solve lack samples, based on fabric-specific degree texture regularity, we propose an anomaly generation create synthetic by destroying normal texture....

In this paper, we present a new fabric defect detection algorithm based on learning an adaptive dictionary. Such dictionary can efficiently represent columns of normal images using linear combination its elements. Benefiting from the fact that defects appear to be small in size, learned directly testing image itself instead reference, allowing more flexibility adapt varying textures. When modeling test dictionary, involving anomalies are likely have larger reconstruction errors than ones....

In this work, a new method based on local patch approximation is presented to address automated defect segmentation textile fabrics. The proposed adopts unsupervised scheme without the need of reference images or any other prior information. Image approximated by dictionary learned from testing sample in least squares sense. With clue differentiation error, abnormal map (each pixel’s anomalous likelihood) can be computed patch-level difference. 2D maximum entropy with neighbourhood...

The computational complexity grows exponentially for multi-level thresholding (MT) with the increase of number thresholds. Taking Kapur’s entropy as optimized objective function, paper puts forward modified quick artificial bee colony algorithm (MQABC), which employs a new distance strategy neighborhood searches. experimental results show that MQABC can search out optimal thresholds efficiently, precisely, and speedily, are very close to examined by exhaustive In comparison EMO...

For the problem of hairiness information missed in existing measurement method, goal this work is to accurately measure length long yarn and obtain path over every point whole hairiness. To achieve goal, images were captured by video microscope (MOTIC) thinned obtained a series image processing. The different baseline step value choose segment method segmentation, lengths obtained, results show that 0.5mm (baseline)and 3 pixels (step value) closest real length. And then, more accurate...

Fabric density measurement plays a key role in the analysis of fabric structural parameters. Existing automatic methods lack varieties adaptability and present poor performance practical application. In order to solve these problems, we use convolutional neural networks (CNNs) locate warps wefts for woven measurement. First, portable wireless device capture high-resolution images set up new dataset with labeled yarns location. Based on this dataset, propose an effective multi-scale network...

Inspired by the image de-noising techniques using learned dictionaries and sparse representation, we present a fabric defect detection scheme via dictionary reconstruction. Fabric defects can be regarded as local anomalies against relatively homogeneous texture background. Following from flexibility of normal samples efficiently represented linear combination few elements dictionary. When modeling new with dictionary, tuned to input data containing structural features, abnormal or defective...

To obtain a stable fabric texture representation result and improve the computation speed, novel method based on dictionary learning is presented. The learned by alternating least-squares using discrete cosine transform (DCT) as initiation dictionary. test effectiveness of dictionary, we comprehensively investigated 42 diverse fault-free woven samples, three fabrics with defects. After preprocessing procedure, samples were characterized experiments 37 different densities demonstrated that...

Deep learning-based defect inspection has gained popularity in recent years. The dataset requirements for the supervised method are currently high, but types of defects numerous and difficult to gather. This work proposes a local image reconstruction-based unsupervised fabric segmentation address this problem. Cyclic structures make up normal portion image, whereas anomalous minor comparison. As result, will be recreated as texture utilizing information from its surrounding areas, preserved...

To locate a video clip in large collections is very important for retrieval applications, especially digital rights management. In this paper, we present novel technique automatic identification of video. This new algorithm based on dynamic programming that fully uses the temporal dimension to measure similarity between two sequences. A normalized chromaticity histogram used as feature which illumination-invariant. Dynamic applied shot-level find optimal nonlinear mapping Two distance...

Abstract In the process of designing and analyzing yarn‐dyed fabric, yarn color pattern has an important effect on appearance fabric. An automatic recognition method for fabric is proposed in this study. The uses images obtained from a high‐resolution digital camera image acquisition system. local statistical texture features are used segmentation. classification problem then formulated research framework multiregion fuzzy segmentation, which can be added auxiliary variables solved...

In textile and garment industries, misarranged warp yarns of yarn-dyed fabrics disorganize the layout lead to poor product quality. This series studies aims develop a computer vision-based system for automatic detection color in terms high efficiency good accuracy. Four main parts are included this studies: yarn segmentation, fabric image stitching, regional proofing. paper proposes continuous segmentation method detect automatically, which is foundation developed system. The proposed...

In recent years, robotic salespersons have been rapidly deployed in various shops especially Japan. However, it is still unclear what kind of action effective for sales, and are often ignored by customers. this paper, we analyzed the multimodal conversations between a salesperson visitors conducting field experiment. We aim to investigate actions robot that lead interactive conversations. The results experiment showed order conduct an conversation, necessary give short speech easy answer....

Computer color matching can improve production efficiency and reduce costs in spun. However, practice the computer success rate for pre-colored fiber blends has not been good, leading to customers being unsatisfied with accuracy of results. Aiming accuracy, a hybrid least squares grid search method proposed spectrophotometric blend based on improved Kubelka–Munk (K-M) double-constant theory. Two-primary, three-primary, four-primary, five-primary cotton were prepared as standard samples...

Abstract Nowadays, community detection is one of the important fields for understanding network topology and has many applications in information diffusion, interaction mining migration behaviour analysis. Therefore, social networks can help to understand user characteristics. There are methods, which often designed single-layer networks. However, real-world use several types relationships establish connections between users, each different Hence, be modelled as multiplex In general, an...