Production management system and method for oxford fabric

Abstract

The invention relates to the technical field of intelligent management, and more specifically discloses a production management system and method for Oxford fabric, and the method comprises the steps: collecting a slurry supply rate, a slurry yarn tension value and monitoring video data, extracting feature vectors, and fusing the feature vectors to obtain an associated feature vector of a sizing process, thereby achieving the production management of the Oxford fabric. Based on the feature vectors, it can be judged that the sizing rate of the pulp silk at the current time point should be increased or decreased, so that the strength and wear resistance of the oxford fabric can be improved, and the anti-cutting effect and durability of products such as tents are improved.

Description

Technical Field

[0001] This application relates to the field of intelligent management technology, and more specifically, to a production management system and method for Oxford cloth. Background Technology

[0002] A tent is a shelter erected on the ground to protect against wind, rain, and sun, providing temporary accommodation. Tents are generally composed of fabric and a frame. The fabric used in most tents is Oxford cloth, also known as oxford cloth, a versatile and widely used material. The main types of Oxford cloth on the market include full-stretch Oxford cloth, checkered Oxford cloth, nylon Oxford cloth, and weft-striped Oxford cloth. Oxford cloth is an important raw material for making bags, defensive and flood control supplies, and other products requiring high strength and abrasion resistance.

[0003] Currently, existing Oxford cloth, when used in tents in harsh outdoor environments, is frequently rubbed against sharp branches or small pebbles, leading to cuts and rendering it ineffective at wind and insect protection. Its cut resistance is poor, and it lacks durability. Therefore, improving the processing technology of Oxford cloth to enhance its strength and abrasion resistance has been a ongoing research topic in the fabric processing industry. Specifically, the most critical aspects of Oxford cloth production preparation are the control of warp tension, sizing rate, and warp stiffness. Sizing rate refers to the rate at which sizing agent is applied to the fabric during the sizing process. If the sizing rate is not accurately controlled, uneven sizing may occur, resulting in inconsistent fabric density and stiffness during weaving, leading to wasted warp yarns.

[0004] Therefore, a production management system and method for Oxford cloth is desired. Summary of the Invention

[0005] To address the aforementioned technical problems, this application is proposed. Embodiments of this application provide a production management system and method for Oxford cloth. This system collects sizing supply rate, sizing yarn tension value, and monitoring video data, extracts feature vectors, and then fuses these feature vectors to obtain a correlation feature vector for the sizing process. Based on these feature vectors, it can be determined whether the sizing rate at the current time point should be increased or decreased. This can improve the strength and abrasion resistance of Oxford cloth, and enhance the cut resistance and durability of products such as tents.

[0006] Accordingly, according to one aspect of this application, a production management system for Oxford cloth is provided, comprising:

[0007] The Oxford cloth sizing process data module is used to collect the sizing rate at multiple predetermined time points in the Oxford cloth sizing process through a flow sensor, monitor the sizing yarn tension value at multiple predetermined time points in the Oxford cloth sizing process in real time through a tension sensor, and collect monitoring video of the Oxford cloth sizing process using a video acquisition device.

[0008] The feature vector extraction module is used to extract the sizing process change feature vector, sizing supply rate feature vector, and sizing tension feature vector from the monitoring video of the Oxford cloth sizing process, the sizing supply rate at the multiple predetermined time points, and the sizing tension value at the multiple predetermined time points, respectively.

[0009] The feature fusion module is used to construct a sizing process correlation feature vector among the sizing process change feature vector, the sizing supply rate feature vector, and the sizing fiber tension feature vector, and optimize it to obtain an optimized sizing process correlation feature vector.

[0010] The sizing rate variation module is used to classify the feature vector associated with the optimized sizing process through a classifier to obtain a classification result, which is used to indicate whether the sizing rate of the sizing fiber at the current time point should be increased or decreased.

[0011] According to another aspect of this application, a production management method for Oxford cloth is provided, comprising:

[0012] The sizing process of Oxford cloth is monitored by collecting the sizing rate at multiple predetermined time points using a flow sensor, the sizing tension value of the sizing yarn at multiple predetermined time points using a tension sensor, and the monitoring video of the sizing process of Oxford cloth is collected using a video acquisition device.

[0013] The sizing process variation feature vector, sizing supply rate feature vector, and sizing tension feature vector are extracted from the monitoring video of the Oxford cloth sizing process, the sizing supply rate at the multiple predetermined time points, and the sizing tension value at the multiple predetermined time points, respectively.

[0014] Construct a sizing process correlation feature vector among the sizing process change feature vector, the sizing supply rate feature vector, and the sizing fiber tension feature vector, and optimize it to obtain an optimized sizing process correlation feature vector;

[0015] The feature vector associated with the optimized sizing process is passed through a classifier to obtain a classification result, which is used to indicate whether the sizing rate of the filaments at the current time point should be increased or decreased.

[0016] Compared with the prior art, the production management system and method for Oxford cloth provided in this application collects sizing supply rate, sizing tension value and monitoring video data, extracts feature vectors, and then fuses these feature vectors to obtain the associated feature vectors of the sizing process. Based on these feature vectors, it can be determined whether the sizing rate of the sizing yarn at the current time point should be increased or decreased, which can improve the strength and abrasion resistance of Oxford cloth and improve the cut resistance and durability of products such as tents. Attached Figure Description

[0017] The above and other objects, features, and advantages of this application will become more apparent from the more detailed description of the embodiments of this application in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of this application and form part of the specification. They are used together with the embodiments of this application to explain this application and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same components or steps.

[0018] Figure 1 This is a block diagram of a production management system for Oxford cloth according to an embodiment of this application.

[0019] Figure 2 This is a block diagram of a feature vector extraction module in a production management system for Oxford cloth according to an embodiment of this application.

[0020] Figure 3 This is a block diagram of a video monitoring feature unit in a production management system for Oxford cloth according to an embodiment of this application.

[0021] Figure 4 This is a block diagram of a slurry supply rate feature unit in a production management system for Oxford cloth according to an embodiment of this application.

[0022] Figure 5 This is a schematic diagram of the architecture of a production management system for Oxford cloth according to an embodiment of this application.

[0023] Figure 6 This is a flowchart of a production management method for Oxford cloth according to an embodiment of this application. Detailed Implementation

[0024] Various exemplary embodiments, features, and aspects of this application will now be described in detail with reference to the accompanying drawings. The same reference numerals in the drawings denote elements that have the same or similar functions. Although various aspects of the embodiments are shown in the drawings, they are not necessarily drawn to scale unless specifically indicated otherwise.

[0025] The term “exemplary” as used herein means “serving as an example, embodiment, or illustration.” Any embodiment illustrated herein as “exemplary” is not necessarily to be construed as superior to or better than other embodiments.

[0026] Furthermore, to better illustrate this application, numerous specific details are provided in the following detailed description. Those skilled in the art should understand that this application can be implemented without certain specific details. In some instances, methods, means, components, and circuits well-known to those skilled in the art have not been described in detail in order to highlight the main points of this application.

[0027] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0028] Figure 1 The illustration shows a block diagram of a production management system for Oxford cloth according to an embodiment of this application. Figure 1 As shown, the production management system 100 for Oxford cloth according to an embodiment of this application includes: an Oxford cloth sizing process data module 110, used to collect the sizing material supply rate at multiple predetermined time points in the Oxford cloth sizing process via a flow sensor, monitor the sizing yarn tension value at multiple predetermined time points in real time via a tension sensor, and collect monitoring video of the Oxford cloth sizing process using a video acquisition device; a feature vector extraction module 120, used to extract sizing process change feature vectors, sizing material supply rate feature vectors, and sizing yarn tension feature vectors from the collected monitoring video of the Oxford cloth sizing process, the sizing material supply rate at the multiple predetermined time points, and the sizing yarn tension value at the multiple predetermined time points, respectively; a feature fusion module 130, used to construct a sizing process correlation feature vector between the sizing process change feature vector, the sizing material supply rate feature vector, and the sizing yarn tension feature vector, and optimize it to obtain an optimized sizing process correlation feature vector; and a sizing yarn sizing rate change module 140, used to pass the optimized sizing process correlation feature vector through a classifier to obtain a classification result, the result being used to indicate whether the sizing yarn sizing rate at the current time point should increase or decrease.

[0029] In this embodiment, the Oxford cloth sizing process data module 110 is used to collect the sizing material supply rate at multiple predetermined time points during the Oxford cloth sizing process via a flow sensor, monitor the sizing yarn tension value at multiple predetermined time points during the Oxford cloth sizing process in real time via a tension sensor, and collect monitoring video of the Oxford cloth sizing process using a video acquisition device. It should be understood that the sizing material supply rate refers to the rate at which sizing material is applied to the fabric during the Oxford cloth sizing process. Collecting sizing material supply rate data via a flow sensor allows for real-time monitoring of the sizing material supply, ensuring its stability and accuracy. This helps control the amount of sizing material used, avoiding excessive or insufficient sizing material supply, thereby improving production efficiency and finished product quality. The sizing yarn tension value refers to the magnitude of tension experienced by the sizing yarn during the Oxford cloth sizing process. Real-time monitoring of changes in sizing yarn tension value via a tension sensor allows for timely detection of tension anomalies, such as excessively high or low tension, enabling timely adjustment measures to maintain stable sizing yarn operation. Stable sizing yarn tension helps ensure the uniformity and quality of the Oxford cloth. Collecting monitoring video of the Oxford cloth sizing process using a video acquisition device provides comprehensive visual information. Through video surveillance, production managers can observe the sizing process in real time, including sizing supply, sizing tension, and the entire sizing process. This helps to promptly identify any anomalies or problems and allows for rapid intervention and adjustments. The surveillance video can also be used for playback and analysis to help improve production processes and procedures, thereby increasing production efficiency and product quality. In this way, by collecting data on sizing supply rate, sizing tension, and surveillance video, the Oxford cloth production management system can achieve real-time monitoring and control of the production process to improve production efficiency, optimize product quality, and promptly identify and resolve potential problems.

[0030] In this embodiment, the feature vector extraction module 120 is used to extract feature vectors representing changes in the sizing process, the sizing rate, and the sizing tension from the monitoring video of the Oxford cloth sizing process, the sizing supply rate at multiple predetermined time points, and the sizing tension values ​​at multiple predetermined time points. It should be understood that by extracting feature vectors from the monitoring video, key changes and features in the sizing process can be captured. These features may include the uniformity of the sizing, the tension distribution of the sizing fibers, and the degree of fabric stretching. Extracting these feature vectors can help analyze the changing trends and abnormalities in the sizing process, thereby enabling timely adjustments and improvements. The sizing supply rate feature vector reflects key information in the sizing supply process. By extracting the sizing supply rate feature vectors at multiple predetermined time points, the stability, changing trends, and efficiency of sizing supply can be understood. These feature vectors can be used to monitor and control the sizing supply system, ensuring the accuracy and stability of the sizing supply and avoiding excessive or insufficient sizing supply. The sizing tension feature vector reflects the tension experienced by the sizing fibers during the sizing process. By extracting the characteristic vectors of the fiber tension at multiple predetermined time points, the changing trend and stability of the fiber tension can be understood. These characteristic vectors can be used to monitor and control the fiber tension, ensuring stable fiber operation and avoiding the impact of excessively high or low tension on product quality.

[0031] In other words, by extracting feature vectors, complex monitoring videos, slurry supply rates, and filament tension data can be transformed into simpler, quantifiable representations, facilitating subsequent data analysis, model building, and decision-making. These feature vectors can be used to build models, perform data mining, and conduct machine learning, thereby optimizing and controlling the production process.

[0032] Accordingly, in one embodiment of this application, Figure 2 The diagram illustrates a block diagram of a feature vector extraction module in a production management system for Oxford cloth according to an embodiment of this application. Figure 2 As shown, in the production management system 100 for Oxford cloth described above, the feature vector extraction module 120 includes: a monitoring video feature unit 121, used to extract multiple key frames of the sizing process from the monitoring video of the Oxford cloth sizing process, and then obtain a sizing process change feature vector through a neural network model; a sizing supply rate feature unit 122, used to arrange the sizing supply rates at multiple predetermined time points into a sizing supply rate input vector according to the time dimension, and then obtain a sizing supply rate feature vector through feature extraction; and a sizing tension value unit 123, used to obtain a sizing tension feature vector by passing the sizing tension values ​​at multiple predetermined time points through a sizing tension extractor based on a time encoder.

[0033] Accordingly, in a specific example of this application, the monitoring video feature unit 121 is used to extract multiple key frames of the sizing process from the monitoring video of the Oxford cloth sizing process, and then obtain a sizing process change feature vector through a neural network model. It should be understood that the monitoring video can provide rich information, including color, texture, deformation, etc., allowing for a more comprehensive understanding of the changes and characteristics during the sizing process. Extracting multiple key frames of the sizing process monitoring can capture key states and changes at different time points, thus providing a more comprehensive understanding of the dynamic changes in the sizing process. Through the neural network model, the change feature vector of the sizing process can be automatically extracted from the monitoring video, avoiding the tedious process of manual feature extraction. The neural network model can learn and extract features from images and transform them into high-dimensional feature vector representations, which are more suitable for subsequent data analysis and processing. After obtaining the sizing process change feature vector, more advanced data analysis and pattern recognition can be performed. These feature vectors can be used to build models, perform data mining and machine learning, thereby achieving intelligent analysis and optimization of the sizing process. Based on these feature vectors, anomaly detection, quality assessment, process optimization, etc., can be performed, providing support and reference for production management decisions.

[0034] Specifically, firstly, monitoring video data of the Oxford cloth sizing process is collected. This video data can be acquired through cameras or other visual sensors and recorded for subsequent processing and analysis. Several keyframes are selected from the acquired monitoring videos. Keyframes are frames in the video that contain important or representative information and can be used to represent the state and changes of the entire sizing process. The selection of keyframes can be based on criteria such as time interval, motion changes, and image quality. Image preprocessing is performed on the selected keyframes. This includes image denoising, image enhancement, and image cropping to improve the accuracy and robustness of subsequent feature extraction. A neural network model is used to extract features from the preprocessed keyframes. The neural network model can be a convolutional neural network (CNN) or other model suitable for image processing. By inputting the keyframes into the neural network model, the feature vectors of the sizing process changes can be obtained. The features output by the neural network model are represented as vectors. This can be done by using the outputs of intermediate layers or fully connected layers of the neural network as feature vectors, converting the high-dimensional representation of the image into a low-dimensional feature vector. The obtained feature vectors of the sizing process changes are then analyzed and processed. Statistical analysis, machine learning, and other methods can be used to perform clustering, dimensionality reduction, and anomaly detection on feature vectors to gain deeper insights and understanding.

[0035] In this way, through the above steps, multiple key frames of the sizing process can be extracted from the collected monitoring video, and feature vectors of sizing process changes can be obtained through a neural network model. These feature vectors can be used for subsequent data analysis, model building, and decision-making, thereby achieving optimization and control of the Oxford cloth production process.

[0036] Accordingly, in one embodiment of this application, Figure 3 The diagram illustrates a block diagram of a monitoring video feature unit in a production management system for Oxford cloth according to an embodiment of this application. Figure 2 As shown, in the above-mentioned production management system 100 for Oxford cloth, the monitoring video feature unit 121 includes: a monitoring keyframe subunit 1211, used to extract multiple sizing process monitoring keyframes from the monitoring video of the Oxford cloth sizing process; a deep-shallow fusion subunit 1212, used to pass the multiple sizing process monitoring keyframes through a first convolutional neural network model containing a deep-shallow fusion module to obtain multiple sizing process monitoring feature matrices; and a convolutional coding subunit 1213, used to aggregate the multiple sizing process monitoring feature matrices along the time dimension into a three-dimensional sizing process feature tensor and then use a second convolutional neural network model with a three-dimensional convolutional kernel to obtain a sizing process change feature vector.

[0037] Furthermore, the monitoring keyframe subunit 1211 is used to extract multiple monitoring keyframes of the sizing process from the monitoring video of the Oxford cloth sizing process. It should be understood that the sizing process is dynamic, and the state and appearance of the fabric change over time. By extracting multiple keyframes, the key states of the sizing process at different points in time can be captured, thus providing a more comprehensive understanding of the changes throughout the process. Monitoring videos are typically continuous image sequences containing a large number of frames. However, not every frame is necessary for the analysis and understanding of the sizing process. Extracting keyframes reduces the amount of data, allowing only the most representative and information-rich frames to be selected for subsequent processing and analysis, thereby reducing computational and storage requirements. Processing the entire monitoring video sequence may require significant computational resources and time. By extracting keyframes, computation can be concentrated on a few frames, reducing the time and resources required for computation and improving processing efficiency. Keyframes typically contain important information about the sizing process, such as the distribution of the coating liquid, changes in fabric texture, and coating uniformity. By extracting keyframes, this key information can be highlighted, making it easier for manual or automated analysis algorithms to identify and understand.

[0038] Furthermore, the deep-shallow fusion subunit 1212 is used to pass the multiple sizing process monitoring keyframes through a first convolutional neural network model containing a deep-shallow fusion module to obtain multiple sizing process monitoring feature matrices. It should be understood that the sizing process monitoring keyframes contain rich visual information, such as texture, color, and shape, which is crucial for analyzing and understanding the sizing process. Convolutional Neural Networks (CNNs) are powerful deep learning models that can effectively extract features from images. By inputting the keyframes into the first convolutional neural network model, its convolutional and pooling layers can be used to automatically learn and extract visual features such as texture, color, and shape during the sizing process. Each keyframe generates a corresponding feature matrix after passing through the first convolutional neural network model. Since the keyframes represent different states and changes during the sizing process, the feature matrix of each keyframe can reflect information from different stages or features. By extracting multiple feature matrices, multiple aspects and details of the sizing process can be comprehensively considered, thus providing a more comprehensive description and analysis of the sizing process.

[0039] The shallow-deep fusion module is a technique that combines shallow and deep features. It introduces shallow features, such as pixel-level information, into a convolutional neural network and fuses them with deep features. This fusion provides richer feature representations while preserving the correlation between low-level and high-level features. By using the shallow-deep fusion module, feature extraction capabilities can be enhanced, improving the accuracy of understanding and analyzing the sizing process.

[0040] Accordingly, firstly, each keyframe of the sizing process monitoring is used as input for feature extraction through a first convolutional neural network model. This model contains convolutional layers, pooling layers, and other layers that learn and extract visual features from the keyframes. A shallow-deep fusion module is introduced into this model to combine shallow and deep features. Deep features capture more abstract and high-level information, while shallow features contain lower-level pixel-level information. The shallow-deep fusion module merges the shallow and deep features to generate a comprehensive feature representation. For each sizing process monitoring keyframe, a corresponding sizing process monitoring feature matrix is ​​obtained through this model.

[0041] Specifically, the shallow-deep fusion subunit includes: a shallow feature extraction subunit for extracting a shallow feature map from the i-th layer of the first convolutional neural network model, where j is greater than or equal to 1 and less than or equal to 6; a deep feature extraction subunit for extracting a deep feature map from the j-th layer of the first convolutional neural network model, where the ratio between the j-th layer and the i-th layer is greater than or equal to 5 and less than or equal to 10; a feature map fusion subunit for fusing the shallow feature map and the deep feature map using the shallow-deep feature fusion module of the first convolutional neural network model to obtain a fused feature map; and a feature map dimensionality reduction subunit for performing global pooling along the channel dimension on the fused feature map to obtain the multiple sizing process monitoring feature matrices.

[0042] In other words, to address the aforementioned technical issues, a first convolutional neural network model incorporating a deep-shallow fusion module can leverage the feature extraction capabilities of convolutional neural networks to extract visual features from multiple keyframes monitoring the sizing process. The introduction of the deep-shallow fusion module enhances feature extraction capabilities and generates multiple sizing process monitoring feature matrices, which can be used for further data analysis and processing.

[0043] Furthermore, the convolutional coding subunit 1213 is used to aggregate the multiple sizing process monitoring feature matrices along the time dimension into a three-dimensional sizing process feature tensor, and then use a second convolutional neural network model with a three-dimensional convolutional kernel to obtain a sizing process change feature vector. It should be understood that the sizing process is a process that changes over time. The sizing process monitoring feature matrix at different time points may contain different information, such as the distribution of sizing material, changes in texture, etc. By aggregating the feature matrix along the time dimension, these temporal changes can be taken into account, thereby better describing the temporal characteristics of the sizing process. The three-dimensional convolutional neural network is suitable for processing data with a time dimension. It can effectively learn spatiotemporal features and capture the dynamic changes of the sizing process in a time series. By using a three-dimensional convolutional kernel, convolution operations can be performed in the time dimension to extract the features of the sizing process changes. By using a three-dimensional convolutional neural network model, the change features of the sizing process can be extracted from the three-dimensional sizing process feature tensor. These feature vectors can represent the evolution of the sizing process over time, including sizing material transport, sizing material distribution uniformity, changes in texture, etc. These feature vectors can be used for further analysis, classification, anomaly detection, and other tasks.

[0044] Specifically, firstly, multiple sizing process monitoring feature matrices are arranged chronologically to form a time series. Then, using this time series as input, the multiple feature matrices are aggregated along the time dimension to form a three-dimensional sizing process feature tensor. This tensor's dimensions include width, height, and time. Next, a second convolutional neural network model is used, employing a three-dimensional convolutional kernel to process the sizing process feature tensor. The three-dimensional convolutional kernel can perform convolution operations along the time dimension, thereby capturing the temporal variation features of the sizing process. Through convolution operations and other layer processing on the three-dimensional sizing process feature tensor, variation feature vectors of the sizing process are extracted. These feature vectors can represent key information about the sizing process, such as the distribution of the sizing agent and changes in texture.

[0045] In this way, by aggregating multiple sizing process monitoring feature matrices along the time dimension into a three-dimensional sizing process feature tensor, and using a second convolutional neural network model with three-dimensional convolutional kernels, the changing feature vectors of the sizing process can be obtained. These feature vectors can provide information about the dynamic changes in the sizing process, providing more comprehensive data analysis and decision support for the Oxford cloth production management system.

[0046] Accordingly, the convolutional coding subunit includes: a second-level coding subunit, used to perform three-dimensional convolutional coding on the three-dimensional sizing process feature tensor using the second convolutional neural network model to obtain a sizing process change feature map; and a second-level dimensionality reduction subunit, used to perform global mean pooling on each feature matrix along the channel dimension of the sizing process change feature map to obtain the sizing process change feature vector. The second-level coding subunit is used to: use the second convolutional neural network model with the three-dimensional convolutional kernel to perform the following on the input data during the forward propagation of the layer: perform three-dimensional convolution processing on the input data based on the three-dimensional convolutional kernel to obtain a convolutional feature map; perform mean pooling processing on the convolutional feature map based on the local feature matrix to obtain a pooled feature map; and perform nonlinear activation on the pooled feature map to obtain an activation feature map; wherein the output of the last layer of the second convolutional neural network model is the sizing process change feature map, and the input of the first layer of the second convolutional neural network model is the three-dimensional sizing process feature tensor.

[0047] Accordingly, in a specific example of this application, the sizing supply rate feature unit 122 is used to arrange the sizing supply rates at multiple predetermined time points into a sizing supply rate input vector according to the time dimension, and then extract features to obtain a sizing supply rate feature vector. It should be understood that the sizing supply rate is an important parameter in the Oxford cloth production process. The sizing supply rate may differ at different time points, such as the stability and uniformity of the sizing supply. By arranging the sizing supply rates according to the time dimension, these temporal variations can be taken into account, better describing the temporal characteristics of the sizing supply rate. Through feature extraction methods, relevant features of the sizing supply rate can be extracted from the sizing supply rate input vector. These features may include average supply rate, peak supply rate, and the trend of supply rate changes. Extracting these features helps in the quantitative analysis and monitoring of the sizing supply rate. Through feature extraction, a feature vector of the sizing supply rate can be obtained. This feature vector can represent the dynamic changes in the sizing supply rate, including the stability of the sizing supply and the degree of fluctuation in the supply rate. These feature vectors can be used for further analysis, modeling, prediction, and other tasks.

[0048] Specifically, in one embodiment of this application, Figure 4 The illustration shows a block diagram of a slurry supply rate feature unit in a production management system for Oxford cloth according to an embodiment of this application. Figure 2 As shown, in the production management system 100 for Oxford cloth described above, the slurry supply rate feature unit 122 includes: an arrangement vector subunit 1221, used to arrange the slurry supply rates at the multiple predetermined time points into a slurry supply rate input vector according to the time dimension; and a multi-scale extraction subunit 1222, used to extract the slurry supply rate input vector through a multi-scale neighborhood feature extraction module to obtain a slurry supply rate feature vector.

[0049] Furthermore, the arrangement vector subunit 1221 is used to arrange the slurry supply rates at multiple predetermined time points into a slurry supply rate input vector according to the time dimension. It should be understood that the slurry supply rate is a parameter that changes over time, and the supply rate at different time points may differ. By arranging the slurry supply rates according to the time dimension, the temporal sequence information can be preserved, better describing the temporal changes in the supply rate. Arranging the slurry supply rates into an input vector according to the time dimension integrates the slurry supply rates at multiple time points into a single vector. This vector representation is concise and compact, facilitating subsequent data processing and analysis. After arranging the slurry supply rates at multiple time points into an input vector, various feature extraction methods and models can be easily applied for further analysis. For example, convolutional neural networks, recurrent neural networks, and other models can be used to extract features of the slurry supply rate, or used as input to a model for prediction and decision-making.

[0050] Furthermore, the multi-scale extraction subunit 1222 is used to extract the slurry supply rate input vector through a multi-scale neighborhood feature extraction module to obtain a slurry supply rate feature vector. It should be understood that the slurry supply rate may exhibit characteristics at different scales, such as rapid fluctuations over short periods and trend changes over long periods. By using the multi-scale neighborhood feature extraction module, features at different scales can be analyzed and extracted. This allows for a more comprehensive understanding of the slurry supply rate's variation patterns. The multi-scale neighborhood feature extraction module can capture features of the slurry supply rate at different scales. For example, features such as the average supply rate, maximum supply rate, and rate of change of the supply rate within a local neighborhood can be extracted. These features provide more details and descriptions about the slurry supply rate, aiding in further analysis and decision-making. By combining the features extracted at different scales into a feature vector, a comprehensive feature representation of the slurry supply rate can be obtained. This feature vector can contain feature information at different scales, reflecting multiple aspects of the slurry supply rate. Such a feature vector can be used for further data analysis, modeling, and prediction.

[0051] Specifically, the multi-scale extraction subunit 1222 includes: a first-scale encoding secondary subunit, used to perform one-dimensional convolutional encoding on the slurry supply rate input vector using the first convolutional layer of the multi-scale neighborhood feature extraction module with a one-dimensional convolutional kernel of the first scale to obtain a first-scale slurry supply rate feature vector, wherein the first convolutional layer has a first one-dimensional convolutional kernel of the first length; a second-scale encoding secondary subunit, used to perform one-dimensional convolutional encoding on the slurry supply rate input vector using the second convolutional layer of the multi-scale neighborhood feature extraction module with a one-dimensional convolutional kernel of the second scale to obtain a second-scale slurry supply rate feature vector, wherein the second convolutional layer has a second one-dimensional convolutional kernel of the second length, and the first length is different from the second length; and a multi-scale cascade secondary subunit, used to cascade the first-scale slurry supply rate feature vector and the second-scale slurry supply rate feature vector to obtain the slurry supply rate feature vector.

[0052] The first scale encoding secondary subunit is used to: use the first convolutional layer of the multi-scale neighborhood feature extraction module to perform one-dimensional convolutional encoding on the slurry supply rate input vector with the following first convolution formula to obtain the first scale slurry supply rate feature vector;

[0053] The first convolution formula is:

[0054]

[0055] Where, a is the width of the first convolution kernel in the x-direction, F(a) is the parameter vector of the first convolution kernel, G(xa) is the local vector matrix operated with the convolution kernel function, w is the size of the first convolution kernel, X represents the slurry supply rate input vector, and Cov1(X) represents the first-scale slurry supply rate feature vector; the second-scale encoding sub-unit is used to: use the second convolution layer of the multi-scale neighborhood feature extraction module to perform one-dimensional convolution encoding on the slurry supply rate input vector with the following second convolution formula to obtain the second-scale slurry supply rate feature vector;

[0056] The second convolution formula is:

[0057]

[0058] Where b is the width of the second convolution kernel in the x direction, F(b) is the parameter vector of the second convolution kernel, G(xb) is the local vector matrix operated with the convolution kernel function, m is the size of the second convolution kernel, X represents the slurry supply rate input vector, and Cov2(X) represents the second-scale slurry supply rate feature vector.

[0059] Accordingly, in a specific example of this application, the paddle tension value unit 123 is used to obtain a paddle tension feature vector by passing the paddle tension values ​​at multiple predetermined time points through a paddle tension extractor based on a time-series encoder. It should be understood that a time-series encoder is a model capable of processing time-series data, capturing patterns and dependencies in the time dimension. By using a paddle tension extractor based on a time-series encoder, the model's capabilities can be fully utilized to extract important features from the time series of paddle tension. Paddle tension is a parameter that varies over time, and tension values ​​at different time points may contain different information. Through the paddle tension extractor based on a time-series encoder, key paddle tension features, such as average tension, maximum tension, and volatility, can be extracted. These features provide important descriptions and insights into paddle tension variations. Integrating the extracted paddle tension features into a feature vector allows for convenient representation and processing. This feature vector can serve as input for subsequent analysis and modeling, such as for classification, regression, or other machine learning tasks. Simultaneously, the feature vector typically has a low dimensionality, helping to reduce data complexity and computational overhead.

[0060] Specifically, the propeller tension value unit first arranges the propeller tension values ​​at the multiple predetermined time points into a propeller tension input vector according to the time dimension. Then, the fully connected layer of the timing encoder is used to fully encode the propeller tension input vector using the following formula to extract the high-dimensional hidden features of the propeller tension values ​​at each position in the rotational speed input vector, wherein the formula is: Where X is the fiber tension input vector, Y is the output vector, W is the weight matrix, and B is the bias vector. This represents matrix multiplication. Finally, the one-dimensional convolutional layer of the propeller tension extractor based on the temporal encoder is used to perform one-dimensional convolutional encoding on the propeller tension input vector using the following formula to extract high-dimensional implicit correlation features between the propeller tension values ​​at each position in the propeller tension input vector; wherein, the formula is:

[0061]

[0062] Where a is the width of the convolution kernel in the x direction, F is the convolution kernel parameter vector, G is the local vector matrix of the operation with the convolution kernel function, w is the size of the convolution kernel, X represents the rotational speed input vector, and Cov(X) represents the one-dimensional convolution encoding of the propeller tension input vector.

[0063] In this embodiment, the feature fusion module 130 is used to construct a sizing process correlation feature vector among the sizing process change feature vector, the sizing supply rate feature vector, and the filament tension feature vector, and optimize it to obtain an optimized sizing process correlation feature vector. It should be understood that the sizing process involves changes in multiple aspects, including the sizing supply rate and filament tension. Fusing these different features together can more comprehensively describe the characteristics of the sizing process. By fusing the sizing process change feature vector, the sizing supply rate feature vector, and the filament tension feature vector, their information can be complementary, providing a richer and more accurate feature representation. Different aspects of the sizing process are often related. For example, changes in the sizing supply rate may affect the filament tension, and changes in filament tension may reflect changes in the state of the sizing process. By fusing these related features, the correlation between them can be captured, providing a more informative feature representation. Fusing multiple features can improve modeling performance. Different features may provide useful information in different aspects; by fusing them together, information loss can be reduced, and a better feature representation can be provided. This helps improve the performance of subsequent modeling tasks, such as classification, regression, or other machine learning tasks.

[0064] Specifically, in the technical solution of this application, feature locality is considered to exist in the feature vector associated with the sizing process due to the following reasons. First, the acquisition of monitoring video is affected by factors such as ambient lighting and shooting angle, which may cause the features extracted from different keyframes to exhibit differences due to the influence of the local environment. Second, the first convolutional neural network model of the deep-shallow fusion module may be affected by factors such as local texture, color, and shape when extracting keyframe features, further introducing feature locality. In addition, the second convolutional neural network model with three-dimensional convolutional kernels may also be affected by the above factors when processing the three-dimensional sizing process feature tensor, resulting in local feature prominence. Feature locality in the feature vector associated with the sizing process will bring a series of problems. First, this locality may cause features in specific areas to be overemphasized, while features in other areas are ignored or weakened, thus distorting the feature vector and failing to accurately reflect the features of the overall sizing process. Second, due to feature locality, the model's generalization ability decreases when processing new data, because the model relies too much on local features and ignores the changing patterns of the overall features, thus failing to make accurate predictions or classifications for new situations. Finally, feature locality can also reduce the resilience of feature vectors to noise or anomalies. Locality weakens the ability of feature vectors to handle changes and disturbances in the overall data, thus reducing the robustness of the model. Therefore, topological information aggregation between feature nodes is performed on the feature vectors associated with the sizing process to improve the model's focus on global features, reduce feature vector locality, and thereby improve classification accuracy and generalization ability.

[0065] Specifically, the topological information between feature nodes of the sizing process-related feature vector is aggregated using the following optimization formula to obtain an optimized sizing process-related feature vector; wherein, the optimization formula is:

[0066]

[0067] Among them, f i This represents the eigenvalue at the i-th position of the feature vector associated with the sizing process, where softmax represents the normalization exponential function, f. i ' represents the feature value at the i-th position of the feature vector associated with the optimized sizing process.

[0068] In other words, due to the feature locality of the sizing process-related feature vectors—meaning each feature node can only reflect its own information and cannot effectively integrate information from other feature nodes—the representational power of the feature matrix is ​​insufficient, failing to accurately depict the intrinsic structure of the data. To address this issue, this application proposes a method for optimizing sizing process-related feature vectors based on topological information aggregation. This method utilizes the topological relationships between feature nodes to aggregate topological information between them, thereby enhancing and reducing the dimensionality of the feature matrix. In this way, robustness is introduced by minimizing the loss around semantic information, improving the clustering performance where local features are equivalent to the overall features. Even with noise or outliers in the data, the stability and discriminability of the feature matrix are guaranteed. Thus, during the iteration process, the divergence in dependence on the desired features caused by parameter adjustments is reduced, improving the feature representational power of the sizing process-related feature vectors and thereby increasing the accuracy and efficiency of the classification task.

[0069] In this embodiment, the sizing rate variation module 140 is used to pass the optimized sizing process associated feature vector through a classifier to obtain a classification result, which indicates whether the sizing rate at the current time point should be increased or decreased. It should be understood that the goal of the sizing process is to achieve the optimal sizing rate while ensuring spinning quality. Since multiple factors influence the sizing process, such as the sizing material supply rate and sizing tension, changes in these factors may cause the sizing rate to deviate from the optimal state. By inputting the sizing process associated feature vector into the classifier, it is possible to determine whether the sizing rate should be increased or decreased based on the characteristics at the current time point, thereby adjusting the sizing rate in real time to bring it closer to the optimal state. Classifying the sizing process through a classifier can help optimize the control strategy of the sizing process. The classification result can guide the operator or automatic control system to take corresponding control measures, such as adjusting the sizing material supply rate or adjusting tension control, to achieve better sizing effect and spinning quality. Classifying the sizing process associated feature vector through a classifier enables an automated decision-making process. By training the classifier and using historical data for classification, an adaptive model can be established to make decisions based on the characteristics at the current time point. This can reduce the operator's workload and improve the efficiency and consistency of the sizing process.

[0070] Specifically, in one embodiment of this application, the sizing rate variation module 140 includes: using the fully connected layer of the classifier to perform fully connected encoding on the optimized sizing process associated feature vector to obtain a fully connected encoded feature vector; inputting the fully connected encoded feature vector into the Softmax classification function of the classifier to obtain probability values ​​of the optimized sizing process associated feature vector belonging to various classification labels, wherein the classification labels include those indicating that the sizing rate of the sizing process should increase at the current time point and those indicating that the sizing rate of the sizing process should decrease at the current time point; and determining the classification label corresponding to the largest probability value as the classification result.

[0071] This application also provides a system architecture diagram, as shown in the embodiment below. Figure 5 As shown. Figure 5 This is a schematic diagram of the architecture of a production management system for Oxford cloth according to an embodiment of this application. In this system architecture, firstly, the sizing rate at multiple predetermined time points during the sizing process of Oxford cloth is collected by a flow sensor; the tension value of the sizing yarn at multiple predetermined time points during the sizing process of Oxford cloth is monitored in real time by a tension sensor; and monitoring video of the sizing process of Oxford cloth is collected using a video acquisition device. Then, multiple sizing process monitoring keyframes are extracted from the collected monitoring video of the sizing process of Oxford cloth. Next, the multiple sizing process monitoring keyframes are respectively processed through a first convolutional neural network model including a deep and shallow fusion module to obtain multiple sizing process monitoring feature matrices. Then, the multiple sizing process monitoring feature matrices are aggregated along the time dimension into a three-dimensional sizing process feature tensor, and then processed through a second convolutional neural network model with a three-dimensional convolutional kernel to obtain a sizing process change feature vector. Finally, the sizing rate at the multiple predetermined time points is arranged along the time dimension into a sizing rate input vector, and then processed through a multi-scale neighborhood feature extraction module to obtain a sizing rate feature vector. Next, the fiber tension values ​​at the multiple predetermined time points are processed by a fiber tension extractor based on a time encoder to obtain fiber tension feature vectors. Then, a fiber tension process correlation feature vector is constructed, which includes the fiber tension feature vector, the fiber tension feature vector, and the fiber tension feature vector. This vector is then optimized to obtain an optimized fiber tension process correlation feature vector. Finally, the optimized fiber tension process correlation feature vector is processed by a classifier to obtain a classification result, which indicates whether the fiber sizing rate should increase or decrease at the current time point.

[0072] In summary, the production management system and method for Oxford cloth described in the embodiments of this application collect sizing supply rate, sizing tension value and monitoring video data, extract feature vectors, and then fuse these feature vectors to obtain the associated feature vectors of the sizing process. Based on these feature vectors, it can be determined whether the sizing rate of the sizing yarn at the current time point should be increased or decreased, which can improve the strength and abrasion resistance of Oxford cloth and improve the cut resistance and durability of products such as tents.

[0073] As described above, the production management system 100 for Oxford cloth according to the embodiments of this application can be implemented in various terminal devices, such as servers for the production management system of Oxford cloth. In one example, the production management system 100 for Oxford cloth can be integrated into the terminal device as a software module and / or a hardware module. For example, the production management system 100 for Oxford cloth can be a software module in the operating system of the terminal device, or it can be an application developed for the terminal device; of course, the production management system 100 for Oxford cloth can also be one of many hardware modules of the terminal device.

[0074] Alternatively, in another example, the production management system 100 for Oxford cloth and the terminal device can also be separate devices, and the production management system 100 for Oxford cloth can be connected to the terminal device via wired and / or wireless networks, and transmit interactive information in accordance with an agreed data format.

[0075] Figure 6 This is a flowchart illustrating a production management method for Oxford cloth according to an embodiment of this application. Figure 6 As shown, the production management method for Oxford cloth according to an embodiment of this application includes: S110, collecting the sizing rate at multiple predetermined time points during the sizing process of Oxford cloth using a flow sensor, monitoring the sizing tension value at multiple predetermined time points during the sizing process of Oxford cloth using a tension sensor, and collecting monitoring video of the sizing process of Oxford cloth using a video acquisition device; S120, extracting sizing process change feature vectors, sizing rate feature vectors, and sizing tension feature vectors from the collected monitoring video of the sizing process of Oxford cloth, the sizing rate at the multiple predetermined time points, and the sizing tension value at the multiple predetermined time points, respectively; S130, constructing a sizing process correlation feature vector among the sizing process change feature vector, the sizing rate feature vector, and the sizing tension feature vector, and optimizing it to obtain an optimized sizing process correlation feature vector; S140, passing the optimized sizing process correlation feature vector through a classifier to obtain a classification result, the result being used to indicate whether the sizing rate at the current time point should increase or decrease.

[0076] Here, those skilled in the art will understand that the specific operations of each step in the above-described method for production management of Oxford cloth have been referenced above. Figures 1 to 5 The description of the production management system used for Oxford cloth is detailed therein, and therefore, its repeated description will be omitted.

[0077] In the several embodiments provided by this invention, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0078] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0079] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0080] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0081] Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be embraced within the invention. No appended diagram markings in the claims should be construed as limiting the scope of the claims.

[0082] Furthermore, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices recited in a system claim may also be implemented by a single unit or device through software or hardware. The term "second class" is used to indicate names and does not indicate any specific order.

[0083] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit of the technical solutions of the present invention.

Claims

1. A production management system for Oxford cloth, characterized in that, include: The Oxford cloth sizing process data module is used to collect the sizing rate at multiple predetermined time points in the Oxford cloth sizing process through a flow sensor, monitor the sizing yarn tension value at multiple predetermined time points in the Oxford cloth sizing process in real time through a tension sensor, and collect monitoring video of the Oxford cloth sizing process using a video acquisition device. The feature vector extraction module is used to extract the sizing process change feature vector, sizing supply rate feature vector, and sizing tension feature vector from the monitoring video of the Oxford cloth sizing process, the sizing supply rate at the multiple predetermined time points, and the sizing tension value at the multiple predetermined time points, respectively. The feature fusion module is used to construct a sizing process correlation feature vector among the sizing process change feature vector, the sizing supply rate feature vector, and the sizing fiber tension feature vector, and optimize it to obtain an optimized sizing process correlation feature vector. The sizing rate variation module is used to classify the feature vector associated with the optimized sizing process through a classifier to obtain a classification result, which is used to indicate whether the sizing rate of the sizing fiber at the current time point should be increased or decreased.

2. The production management system for Oxford cloth according to claim 1, characterized in that, The feature vector extraction module includes: The monitoring video feature unit is used to extract multiple key frames of the sizing process from the monitoring video of the Oxford cloth sizing process, and then obtain the sizing process change feature vector through a neural network model. The slurry supply rate feature unit is used to arrange the slurry supply rates at the multiple predetermined time points into a slurry supply rate input vector according to the time dimension, and then extract features to obtain a slurry supply rate feature vector. The fiber tension value unit is used to obtain the fiber tension feature vector by passing the fiber tension values ​​at the multiple predetermined time points through a fiber tension extractor based on a time encoder.

3. The production management system for Oxford cloth according to claim 2, characterized in that, The surveillance video feature unit includes: The monitoring keyframe subunit is used to extract multiple monitoring keyframes of the sizing process from the monitoring video of the Oxford cloth sizing process. The deep-shallow fusion subunit is used to pass the multiple sizing process monitoring key frames through a first convolutional neural network model containing a deep-shallow fusion module to obtain multiple sizing process monitoring feature matrices. The convolutional coding subunit is used to aggregate the multiple sizing process monitoring feature matrices along the time dimension into a three-dimensional sizing process feature tensor, and then use a second convolutional neural network model with a three-dimensional convolutional kernel to obtain the sizing process change feature vector.

4. The production management system for Oxford cloth according to claim 3, characterized in that, The deep-shallow fusion subunit includes: The shallow feature extraction subunit is used to extract shallow feature maps from the i-th layer of the first convolutional neural network model, where j is greater than or equal to 1 and less than or equal to 6. A deep feature extraction subunit is used to extract a deep feature map from the j-th layer of the first convolutional neural network model, wherein the ratio between the j-th layer and the i-th layer is greater than or equal to 5 and less than or equal to 10. The feature map fusion secondary subunit is used to fuse the shallow feature map and the deep feature map using the shallow feature fusion module of the first convolutional neural network model to obtain a fused feature map; The feature map dimensionality reduction second-level sub-unit is used to perform global pooling along the channel dimension on the fused feature map to obtain the multiple sizing process monitoring feature matrices.

5. The production management system for Oxford cloth according to claim 4, characterized in that, The convolutional coding subunit includes: The encoding second-level subunit is used to perform three-dimensional convolutional encoding on the three-dimensional sizing process feature tensor using the second convolutional neural network model to obtain a sizing process change feature map; The dimension reduction second-level sub-unit is used to perform global mean pooling on each feature matrix along the channel dimension of the sizing process change feature map to obtain the sizing process change feature vector.

6. The production management system for Oxford cloth according to claim 5, characterized in that, The slurry supply rate characteristic unit includes: The arrangement vector sub-unit is used to arrange the slurry supply rates at the multiple predetermined time points into a slurry supply rate input vector according to the time dimension; The multi-scale extraction subunit is used to extract the slurry supply rate input vector through the multi-scale neighborhood feature extraction module to obtain the slurry supply rate feature vector.

7. The production management system for Oxford cloth according to claim 6, characterized in that, The multi-scale extraction subunit includes: The first scale encoding secondary subunit is used to perform one-dimensional convolutional encoding on the slurry supply rate input vector using the first convolutional layer of the multi-scale neighborhood feature extraction module with a one-dimensional convolutional kernel of the first scale to obtain the first scale slurry supply rate feature vector, wherein the first convolutional layer has a first one-dimensional convolutional kernel of the first length. The second-scale encoding subunit is used to perform one-dimensional convolutional encoding on the slurry supply rate input vector using the second convolutional layer of the multi-scale neighborhood feature extraction module with a one-dimensional convolutional kernel having a second scale to obtain a second-scale slurry supply rate feature vector, wherein the second convolutional layer has a second one-dimensional convolutional kernel of a second length, and the first length is different from the second length. A multi-scale cascaded two-level sub-unit is used to cascade the first-scale slurry supply rate feature vector and the second-scale slurry supply rate feature vector to obtain the slurry supply rate feature vector.

8. The production management system for Oxford cloth according to claim 7, characterized in that, The fusion feature module includes: A fusion unit is used to fuse the sizing process change feature vector, the sizing supply rate feature vector, and the filament tension feature vector to obtain a sizing process associated feature vector. The optimization unit is used to aggregate the topological information between feature nodes of the sizing process-related feature vector to obtain an optimized sizing process-related feature vector.

9. The production management system for Oxford cloth according to claim 8, characterized in that, The optimization unit is used to: aggregate the topological information between feature nodes of the sizing process-related feature vector using the following optimization formula to obtain an optimized sizing process-related feature vector; The optimization formula is as follows: Among them, f i This represents the eigenvalue at the i-th position of the feature vector associated with the sizing process, where softmax represents the normalization exponential function, f. i ' represents the feature value at the i-th position of the feature vector associated with the optimized sizing process.

10. A production management method for Oxford cloth, characterized in that, include: The sizing process of Oxford cloth is monitored by collecting the sizing rate at multiple predetermined time points using a flow sensor, the sizing tension value of the sizing yarn at multiple predetermined time points using a tension sensor, and the monitoring video of the sizing process of Oxford cloth is collected using a video acquisition device. The sizing process variation feature vector, sizing supply rate feature vector, and sizing tension feature vector are extracted from the monitoring video of the Oxford cloth sizing process, the sizing supply rate at the multiple predetermined time points, and the sizing tension value at the multiple predetermined time points, respectively. Construct a sizing process correlation feature vector among the sizing process change feature vector, the sizing supply rate feature vector, and the sizing fiber tension feature vector, and optimize it to obtain an optimized sizing process correlation feature vector; The feature vector associated with the optimized sizing process is passed through a classifier to obtain a classification result, which is used to indicate whether the sizing rate of the filaments at the current time point should be increased or decreased.

Images