Kaixin Xu’s research while affiliated with Zhejiang Sci-Tech University and other places

In industrial applications where device capacity, computational performance, and thermal management are limited, we propose the YOLOvT-Light model for fabric defect detection. This model incorporates the convolutional block attention module (CBAM)-EfficientNet backbone network, balancing detection speed and precision while significantly reducing complexity and maintaining high precision. GhostConv replaces standard convolution in the neck section, effectively reducing parameters and computational cost through simple linear transformations. Additionally, the integration of Faster Block and C2f modules retains local feature fusion capabilities while further decreasing parameters and computation. Experimental results using the DAGM2007 dataset demonstrate that YOLOvT-Light significantly reduces weight size (9.50 MB), computation performance (13.9 Gflops), and parameter count (6.11 M) compared with the baseline model, while improving inference speed (223 fps), without sacrificing precision. This lightweight architecture ensures the feasibility of deploying the model on resource-constrained devices, making it suitable for real-time, cost-effective, and safe defect detection in textile manufacturing environments. This study provides a reliable solution for developing efficient, lightweight detection models applicable to real-world industrial settings.

The diversity of weaving equipment has led to inconsistencies in communication protocols, impeding data collection and interoperability between devices, and ultimately reducing production efficiency. Additionally, fabric defects significantly impact product quality, while current visual inspection technologies are primarily reactive and traditional quality prediction methods often exhibit considerable errors. This study leverages the standardization and interoperability features of open platform communications unified architecture technology to facilitate data acquisition within the weaving department, establishing a reliable Internet of Things framework that supports subsequent fabric quality prediction, and optimizing the back propagation neural network through the K-means clustering algorithm and particle swarm optimization to predict the type and number of fabric defects. A comparative analysis with traditional BP and PSO-BP prediction models was conducted, ultimately verifying the feasibility of using OPC UA to transmit weaving data for fabric quality prediction. The research results demonstrate that using OPC UA technology enables the unified transmission of weaving equipment data, addressing the issue of heterogeneity in weaving department equipment. The K-means-PSO-BP model can effectively predict defects such as double weft, hundred feet, and broken warp with minimal error, achieving a root mean square error of less than 0.15.

In the field of fabric defect detection, the development of algorithms has been hindered by issues such as poor quality and limited quantity of open-source datasets. Traditional data augmentation methods offer limited improvements in model performance, while generative data augmentation methods are plagued by difficulties in training generative models, susceptibility to artifacts, and the need for re-labeling. To address these challenges, this paper proposes a blind super-resolution algorithm for fabric defect data augmentation. The model is based on Real-ESRGAN and has been optimized specifically for the resolution degradation module to better adapt to the resolution degradation process in fabric images. Subsequently, a novel loss function named Local Blur Discrimination Loss is designed to address the local blur phenomenon and suppress the generation of fabric artifacts during the super-resolution process. Finally, both subjective evaluations of super-resolution effects and objective comparisons of data augmentation performance were conducted during the experimental phase. The subjective assessments demonstrate that the proposed method outperforms the baseline model. Additionally, in terms of objective performance, augmenting the DAGM2007 dataset using the proposed model, the detection model's accuracy (P) increased by 7.4%, recall (R) increased by 1.0%, and the mean average precision (mAP) increased by 2.5%, surpassing commonly used traditional vision-based data augmentation algorithms.

Facing the shift in the weaving industry from mass production to a more diversified, small-batch production model, traditional production scheduling systems are no longer capable of meeting the demand for rapid market response. To address this issue, this paper first analyzes the production process of a weaving workshop, identifying key scheduling challenges such as order allocation, equipment selection, and operation sequencing. Based on this analysis, a flexible job shop multi-objective scheduling model tailored for weaving workshops is developed. To handle the multiple constraints and optimization goals inherent in the model, an improved NSGA-II algorithm is proposed. This algorithm combines artificial bee colony (ABC) algorithm for population initialization with simulated annealing (SA) for population filtering. Simulation examples and case studies from actual workshops demonstrate that the improved NSGA-II algorithm outperforms other algorithms in solving the scheduling problem for weaving workshops. The proposed multi-objective scheduling model and its improved algorithm provide accurate and efficient optimization solutions for workshop scheduling.

The automated replacement of empty tubes in the yarn bank is a critical step in the process of automatic winding machines with yarn banks, as the real-time detection of depleted yarn on spools and accurate positioning of empty tubes directly impact the production efficiency of winding machines. Addressing the shortcomings of traditional methods, such as poor adaptability and low sensitivity in optical and visual tube detection, and aiming to reduce the computational and detection time costs introduced by neural networks, this paper proposes a lightweight yarn spool detection model based on YOLOv8. The model utilizes Darknet-53 as the backbone network, and due to the dense spatial distribution of yarn spool targets, it incorporates large selective kernel units to enhance the recognition and positioning of dense targets. To address the issue of excessive focus on local features by convolutional neural networks, a bi-level routing attention mechanism is introduced to capture long-distance dependencies dynamically. Furthermore, to balance accuracy and detection speed, a FasterNeck is constructed as the neck network, replacing the original convolutional blocks with Ghost convolutions and integrating with FasterNet. This design minimizes the sacrifice of detection accuracy while achieving a significant improvement in inference speed. Lastly, the model employs weighted IoU with a dynamic focusing mechanism as the bounding box loss function. Experimental results on a custom yarn spool dataset demonstrate a notable improvement over the baseline model, with a high-confidence mAP of 94.2% and a compact weight size of only 4.9 MB. The detection speed reaches 223FPS, meeting the requirements for industrial deployment and real-time detection.

The rapid development of the logistics industry has driven innovations in parcel sorting technology, among which the swift and precise positioning and classification of parcels have become key to enhancing the performance of logistics systems. This study aims to address the limitations of traditional light curtain positioning methods in logistics sorting workshops amidst high-speed upgrades. This paper proposes a high-speed classification and location algorithm for logistics parcels utilizing a monocular camera. The algorithm combines traditional visual processing methods with an enhanced version of the lightweight YOLOv5 object detection algorithm, achieving high-speed, high-precision parcel positioning. Through the adjustment of the network structure and the incorporation of new feature extraction modules and ECIOU loss functions, the model’s robustness and detection accuracy have been significantly improved. Experimental results demonstrate that this model exhibits outstanding performance on a customized logistics parcel dataset, notably enhancing the model's parameter efficiency and computational speed, thereby offering an effective solution for industrial applications in high-speed logistics supply.

The rapid development of the logistics industry poses significant challenges to the sorting work within this sector. The fast and precise identification of moving express parcels holds immense significance for the performance of logistics sorting systems. This paper proposes a motion express parcel positioning algorithm that combines traditional vision and AI-based vision. In the traditional vision aspect, we employ a brightness-based traditional visual parcel detection algorithm. In the AI vision aspect, we introduce a Convolutional Block Attention Module (CBAM) and Focal-EIoU to enhance YOLOv5, improving the model’s recall rate and robustness. Additionally, we adopt an Optimal Transport Assignment (OTA) label assignment strategy to provide a training dataset based on global optimality for the model training phase. Our experimental results demonstrate that our modified AI model surpasses traditional algorithms in both parcel recognition accuracy and inference speed. The combined approach of traditional vision and AI vision in the motion express parcel positioning algorithm proves applicable for practical logistics sorting systems.

The piezoelectric needle selector is a crucial component of computerized dobby weft knitting machines. With the continuous development of weft knitting machine technology, enhancing the accuracy of piezoelectric needle selector control is essential. Accurate determination of whether the blades are in the correct position would significantly improve the precision of piezoelectric needle selector control. In this study, piezoelectric ceramic sensors were used to collect impact vibration signals when the blades struck the damper baffle. Key hardware circuits were designed for this purpose. A self-learning algorithm was employed to capture the highest point and the time it takes to reach the highest point in the impact vibration signal. A fault detection algorithm was used to implement closed-loop fault detection for piezoelectric needle selectors. Experimental results and practical applications have demonstrated that this research effectively addresses the accurate determination of whether the piezoelectric needle selector blades are in the correct position. It has reduced the defect rate of fabric production in weft knitting, thereby improving the overall efficiency and profitability of businesses.

With the diversification of spinning order varieties and process parameters, the conventional method of determining production plans through trial spinning no longer satisfies the processing requirements of enterprises. Currently, deficiencies exist in predicting spinning quality relying on manual experience and traditional methods. The back propagation (BP) neural network within the realm of deep learning theory faces challenges in handling time series data, while the long short-term memory (LSTM) neural network, despite its intricate mechanism, exhibits an overall lower predictive accuracy. Consequently, a more precise predictive methodology is imperative to assist production personnel in efficiently ascertaining cotton-blending schemes and processing parameters, thereby elevating the production efficiency of the enterprise. In response to this challenge, we propose an attention-GRU-based cotton yarn quality prediction model. By employing the attention mechanism, the model is directed towards the input features most significantly impacting yarn quality. Real-world performance indicators of raw cotton and process parameters are utilized to predict yarn tensile strength. A comparative analysis is conducted against prediction results of BP, LSTM, and gated recurrent unit (GRU) neural networks that do not incorporate the attention mechanism. The outcomes reveal that the GRU model enhanced with the attention mechanism demonstrates reductions of 56.3%, 38.5%, and 36.4% in root mean square error (RMSE), along with 0.367%, 0.158%, and 0.190% in mean absolute percentage error (MAPE), respectively. The model attains a coefficient of determination R-squared of 0.954, indicating a high degree of fitness. This study underscores the potential of the proposed attention-GRU model in refining cotton yarn quality prediction and its consequential implications for process optimization and enhanced production efficiency within textile enterprises.