Intelligent woolen sweater production monitoring method and system based on Internet of Things

Abstract

The invention relates to the technical field of image processing, and discloses a woolen sweater intelligent production monitoring method and system based on the Internet of Things. The method comprises the following steps: acquiring a wool fiber three-dimensional point cloud through multispectral imaging and a space stereo matching algorithm, and analyzing the twist and crimpness of yarns through a gradient convolutional network to obtain a quality feature set; a high-speed acquisition system is adopted to obtain a knitting needle motion sequence, and a fabric structure map is obtained through a line enhancement algorithm. And constructing knit fault prediction based on the spatial-temporal characteristic network and the memory model, and obtaining a density uniformity index. And inputting the quality feature set and the density index into a fusion network, and performing multi-layer attention processing to obtain a quality evaluation model. And constructing an early warning system by using the depth map network, and generating a process optimization scheme. According to the method, comprehensive perception, dynamic early warning and intelligent optimization of the production process are realized, and the problems of incomplete data acquisition, inflexible parameter control, untimely quality early warning and the like in the prior art are solved.

Description

Technical Field

[0001] The present application relates to the field of image processing technology, and in particular to an intelligent production monitoring method and system for wool sweaters based on the Internet of Things. Background Art

[0002] With the rapid development of the textile industry, the level of intelligence and automation in the wool sweater production process continues to improve. Existing methods for monitoring wool sweater production primarily include manual visual inspection, single-sensor monitoring, and fixed parameter control. These methods use cameras to capture the fabric surface during production and detect defects using image processing algorithms; tension sensors monitor yarn tension changes; and temperature and humidity sensors monitor the production environment. At the same time, some companies are beginning to experiment with integrating machine vision and deep learning technologies to automatically identify and warn of quality issues during wool sweater production.

[0003] However, existing technologies have the following shortcomings: First, traditional single-sensor monitoring methods make it difficult to obtain complete production process data, resulting in insufficient accuracy and real-time performance of quality control; second, fixed parameter control methods lack the ability to adapt to dynamic changes in the production process and cannot respond to quality fluctuations in a timely manner; third, existing image processing algorithms are mainly aimed at static fabric defect detection, which makes it difficult to predict and prevent the spread of quality problems; finally, multi-source heterogeneous data in the production process have not yet been effectively integrated and utilized, which restricts the improvement of the level of intelligent control. Summary of the Invention

[0004] The present application provides an intelligent production monitoring method and system for wool sweaters based on the Internet of Things, which is used to achieve comprehensive perception, dynamic early warning and intelligent optimization of the production process, and solve the problems existing in the prior art such as incomplete data acquisition, inflexible parameter control, and untimely quality warning.

[0005] In the first aspect, the present application provides an intelligent production monitoring method for wool sweaters based on the Internet of Things, and the intelligent production monitoring method for wool sweaters based on the Internet of Things includes: acquiring images of wool yarn fiber bundles through a multispectral imaging system, and processing them through a spatial stereo matching algorithm to obtain a three-dimensional feature point cloud of wool fibers; based on the three-dimensional feature point cloud of wool fibers, analyzing the yarn twist and fiber curl through a gradient convolution network to obtain a wool yarn quality feature set; for the knitted fabric forming process, acquiring the knitting needle motion sequence through a high-speed acquisition system, and processing it through a texture enhancement algorithm to obtain a fabric structure map; based on the fabric structure map, constructing a weaving defect prediction model through a spatiotemporal feature network and a memory model to obtain a fabric density uniformity index; inputting the wool yarn quality feature set and the fabric density uniformity index into a feature fusion network, and processing them through a multi-layer attention mechanism to obtain a garment quality assessment model; based on the output parameters of the garment quality assessment model, constructing a wool sweater quality control early warning system through a deep graph network to obtain a production process optimization plan.

[0006] In a second aspect, the present application provides an IoT-based intelligent production monitoring system for wool sweaters, the IoT-based intelligent production monitoring system for wool sweaters comprising: The acquisition module is used to collect images of wool yarn fiber bundles through a multispectral imaging system and obtain a three-dimensional feature point cloud of wool fibers through spatial stereo matching algorithm processing; An analysis module is used to analyze the yarn twist and fiber curl based on the three-dimensional feature point cloud of the wool fiber through a gradient convolution network to obtain a wool yarn quality feature set; The processing module is used to obtain the knitting needle motion sequence through a high-speed acquisition system during the knitted fabric forming process, and then process it through a texture enhancement algorithm to obtain a fabric structure map; A prediction module is used to construct a weaving defect prediction model based on the fabric structure map through a spatiotemporal feature network and a memory model to obtain a fabric density uniformity index; A fusion module is used to input the wool yarn quality feature set and the fabric density uniformity index into a feature fusion network, and process them through a multi-layer attention mechanism to obtain a garment quality assessment model; The optimization module is used to build a wool sweater quality control early warning system through a deep graph network based on the output parameters of the garment quality assessment model to obtain a production process optimization plan.

[0007] The third aspect of the present invention provides a computer device, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor calls the instructions in the memory so that the computer device executes the above-mentioned IoT-based intelligent production monitoring method for wool sweaters.

[0008] A fourth aspect of the present invention provides a computer-readable storage medium, which stores instructions. When the computer-readable storage medium is run on a computer, it enables the computer to execute the above-mentioned smart production monitoring method of wool sweaters based on the Internet of Things.

[0009] In the technical solution provided by this application, the wool yarn fiber bundle is imaged by a multispectral imaging system, and a three-dimensional feature point cloud of the fiber is constructed by combining a spatial stereo matching algorithm, thereby achieving accurate characterization of the fiber morphology. The gradient convolutional network analyzes the yarn twist and fiber curl, extracts the wool yarn quality feature set, and provides a reliable data basis for quality assessment. The high-speed acquisition system obtains the needle motion sequence in real time, and generates a fabric structure map through a texture enhancement algorithm, thereby achieving dynamic monitoring of the weaving process. The spatiotemporal feature network and memory model construct a weaving defect prediction model, accurately outputs the fabric density uniformity index, and improves the accuracy of quality warning. The feature fusion network adopts a multi-layer attention mechanism to process multi-source data, establishes a garment quality assessment model, and achieves a comprehensive assessment of product quality. The wool sweater quality control warning system constructed by the deep graph network can generate production process optimization plans in a timely manner, thereby improving the intelligence level of the production process. The entire solution utilizes a multispectral imaging algorithm to provide highly accurate raw material feature data, a spatial stereo matching algorithm to achieve 3D fiber reconstruction, a gradient convolutional network to ensure the accuracy of quality feature extraction, a spatiotemporal feature network to enhance defect prediction, a multi-layer attention mechanism to optimize feature fusion, and a deep graph network to improve the reliability of the early warning system. This significantly improves the monitoring accuracy and intelligence level of the wool sweater production process. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following is a brief introduction to the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0011] Figure 1 This is a schematic diagram of an embodiment of a method for intelligent production monitoring of wool sweaters based on the Internet of Things in an embodiment of the present application; Figure 2 Schematic diagram of the process of analyzing yarn twist and fiber curl by using a gradient convolutional network in an embodiment of the present application; Figure 3 This is a schematic diagram of an embodiment of an intelligent production monitoring system for wool sweaters based on the Internet of Things in an embodiment of the present application; Figure 4 It is a schematic block diagram of the structure of a computer device in an embodiment of the present invention. DETAILED DESCRIPTION

[0012] The embodiments of the present application provide a method and system for intelligent production monitoring of wool sweaters based on the Internet of Things. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0013] For ease of understanding, the specific process of the embodiment of the present application is described below. Figure 1 In the embodiment of the present application, an embodiment of the smart production monitoring method of wool sweaters based on the Internet of Things includes: Step S101: Capture images of wool yarn fiber bundles using a multispectral imaging system, and process them using a spatial stereo matching algorithm to obtain a three-dimensional feature point cloud of wool fibers; Step S102: Analyze the yarn twist and fiber curl based on the wool fiber three-dimensional feature point cloud using a gradient convolutional network to obtain a wool yarn quality feature set; Step S103: for the knitted fabric forming process, the knitting needle motion sequence is acquired by a high-speed acquisition system, and processed by a texture enhancement algorithm to obtain a fabric structure map; Step S104: constructing a weaving defect prediction model based on the fabric structure map through a spatiotemporal feature network and a memory model to obtain a fabric density uniformity index; Step S105: Input the wool yarn quality feature set and the fabric density uniformity index into the feature fusion network, and process them through a multi-layer attention mechanism to obtain a garment quality assessment model; Step S106: Based on the output parameters of the garment quality assessment model, a wool sweater quality control early warning system is constructed through a deep graph network to obtain a production process optimization plan.

[0014] It is understandable that the execution subject of this application can be the smart production monitoring system for wool sweaters based on the Internet of Things, or it can be a terminal or a server, and the specific implementation is not limited here. The embodiment of this application is described by taking the server as the execution subject as an example.

[0015] Specifically, a multispectral imaging system is used to acquire image data of wool yarn fiber bundles. The multispectral imaging system includes a visible light camera, a near-infrared camera, and an ultraviolet camera, capturing image data in the wavelength ranges of 400-700 nm, 700-1000 nm, and 200-400 nm, respectively. A spatial stereo matching algorithm processes the acquired multispectral images using structured light encoding technology. The structured light encoder projects a specific fringe pattern onto the fiber surface and calculates depth information based on the fringe deformation. A stereo matching network calculates the disparity of depth information from different perspectives to generate multi-viewpoint point cloud data. This data is further integrated with geometric feature parameters and optical property data to form a three-dimensional feature point cloud of the wool fiber. Based on the acquired 3D feature point cloud of the wool fiber, a gradient convolutional network performs spatial morphological analysis on the point cloud data to extract fiber bundle geometric morphology data. Feature decomposition is used to generate a yarn surface texture feature map, which is then used to quantitatively analyze yarn twist and fiber curl. Yarn twist analysis generates a twist distribution curve, while fiber curl analysis generates a curl parameter matrix. Feature mapping integrates the twist distribution curve and the curl parameter matrix into a yarn structure characteristic map. The wool yarn quality evaluation system comprehensively processes the map and outputs a wool yarn quality feature set.

[0016] During the fabric forming process, a high-speed camera system dynamically captures the reciprocating motion of knitting needles, generating a time-series image set. Optical flow analysis extracts knitting needle trajectory data and generates a needle displacement matrix. Spatial positioning analysis determines the knitting needle node map, and a grain enhancement network extracts knitting grain features to generate knitted fabric texture data. Density analysis generates a fabric density distribution map, and feature integration and mapping output a fabric structure map. During defect prediction, a spatiotemporal feature analysis unit processes the fabric structure map and extracts a time-series data set for the fabric structure. Sequence feature analysis classifies fabric defects and generates a defect type distribution matrix. Time domain transformation and fabric deformation feature analysis generate a defect propagation trend map. Spatial correlation analysis extracts fabric deformation patterns and outputs a fabric stress distribution field. A density calibration algorithm aggregates local features to generate fabric structure uniformity data. The defect assessment system comprehensively analyzes and outputs a fabric density uniformity index.

[0017] During the feature fusion phase, the wool yarn quality feature set undergoes multi-scale feature decomposition to obtain a sequence of yarn physical properties. Fabric density and uniformity indicators are processed through time-series sampling to generate a fabric structural deformation field. Feature correspondence analysis generates a sweater quality map, and hierarchical attention calculation generates garment-piece structure correlation data. Deep feature decomposition outputs a sweater process parameter matrix, which is used to construct a garment quality assessment model to quantitatively evaluate appearance quality, structural stability, and comfort. During the early warning system construction process, multi-dimensional data mapping processes the output parameters of the garment quality assessment model to generate a sweater quality status matrix. A deep graph feature network analyzes regional connectivity to identify quality defect propagation chains. Time-series feature decomposition, combined with process parameter constraints, outputs a process anomaly correlation map. Warning level classification and risk propagation analysis generate a quality warning decision tree. Multi-objective optimization determines a process adjustment sequence, resulting in a production process optimization plan that includes needle parameter optimization indicators, fabric structure optimization indicators, and garment quality control indicators.

[0018] For example, fiber image data collected by a multispectral imaging system is processed through spatial stereo matching to create an accurate three-dimensional feature point cloud model. Gradient convolutional network analysis found that when yarn twist varies within the range of 120-150 turns / meter, the fiber curl remains stable at 65%-75%. A high-speed camera system captures needle movement at a sampling rate of 1000 frames per second, and after processing with a texture enhancement algorithm, the detailed features of the fabric structure are clearly displayed. A spatiotemporal feature network combined with a memory model predicts weaving defects, and the resulting fabric density uniformity index reflects the consistency of the fabric structure. A feature fusion network and a multi-layer attention mechanism integrate various indicators into a garment quality assessment model, from which a deep graph network generates process optimization recommendations, ensuring that the production process always remains within the optimal process parameter range.

[0019] In a specific embodiment, the process of executing step S101 may specifically include the following steps: (1) Using a multispectral camera to image the wool yarn fiber bundle in different wavelength bands, we can obtain visible light image data, near infrared image data, and ultraviolet image data; (2) Inputting visible light image data, near-infrared image data, and ultraviolet image data into a structured light encoder to obtain a fiber surface depth coding map; (3) Based on the fiber surface depth coding map, disparity calculation is performed through the stereo matching network to obtain multi-view point cloud data; (4) Input the multi-view point cloud data into the feature extraction network, obtain the geometric feature parameters through the deep learning method, and obtain the optical property data of the fiber surface; (5) Based on the optical property data and geometric characteristic parameters of the fiber surface, the three-dimensional feature point cloud of the wool fiber is obtained through multi-scale feature fusion processing; (6) Based on the three-dimensional feature point cloud of wool fiber, the spatial position calibration is performed through the binocular camera calibration system to complete the accurate reconstruction of the point cloud data.

[0020] Specifically, the multispectral imaging system collects multi-band data on wool yarn fiber bundles. The system includes three cameras: a visible light camera captures images in the 400-700nm band, primarily to obtain fiber surface morphology; a near-infrared camera captures images in the 700-1000nm band, reflecting the fiber's internal structure; and a UV camera captures images in the 200-400nm band, used to detect surface microscopic defects. After preprocessing, the image data for each band forms a multidimensional feature matrix M: in, Represents the image intensity value of the nth band, is the corresponding weight coefficient, Represents three-dimensional spatial coordinates. This matrix integrates image information from the three bands, providing the foundational data for subsequent structured light encoding. In practical applications, when processing a bundle of 2mm diameter wool fibers, the system captures an image with a resolution of 4096 × 3072 pixels, with a sampling interval of 0.5 microns for each band.

[0021] The structured light encoder receives multi-dimensional feature matrix data and calculates depth information by projecting a specific stripe pattern. The encoding process uses a phase offset encoding method, and the calculation formula for the encoding intensity E is: in, represents the background light intensity, is the modulation amplitude, is the phase value, =Phase offset. By analyzing fringe deformation, the system generates a depth-encoded map of the fiber surface. At the production site, the encoder projects fringe spacing of 0.1 mm, with a phase offset of 8 steps, and a complete encoding cycle of 20 ms.

[0022] The subsequent stereo matching network processes the depth coding map and uses an improved semi-global matching algorithm to calculate the disparity value. This algorithm introduces an adaptive cost aggregation strategy, and the disparity cost The calculation formula is: in, 、 and Represents color similarity, gradient similarity and Census transformation similarity respectively, 、 、 is the weight coefficient, and p represents the pixel coordinates (x, y) in the reference image, that is, the pixel position where the matching point needs to be found. d represents the disparity value, in pixels, which represents the horizontal offset of the corresponding points in the left and right images. By minimizing the disparity cost, multi-view point cloud data is obtained. In practical applications, when processing wool fiber bundles, the matching accuracy reaches 0.05mm, and the point cloud density is 1000 points / mm². The feature extraction network adopts a multi-layer neural network structure, including a point feature learning layer, a feature aggregation layer, and a global feature extraction layer. The network inputs multi-view point cloud data and extracts geometric feature parameters, including fiber diameter, curvature, roughness, etc., through deep learning methods. At the same time, combined with spectral information, optical property data of the fiber surface is generated, including features such as reflectivity and scattering coefficient.

[0023] The multi-scale feature fusion stage adopts a pyramid structure to analyze and integrate feature information at different spatial scales. The processing process starts from the microscale (1 micron) and gradually expands to the macroscale (1 mm) to generate a three-dimensional feature point cloud. The fused point cloud data contains the morphological characteristics and optical properties of the fiber. The binocular camera calibration system calibrates the spatial position of the three-dimensional feature point cloud. The calibration process uses a 9×9 calibration plate and establishes the correspondence between the camera parameters and the world coordinate system by analyzing the position of the calibration points. The spatial accuracy of the calibrated point cloud data reaches 0.01mm, recording the three-dimensional morphological characteristics of the wool fiber.

[0024] In a specific embodiment, the process of executing step S102 may specifically include the following steps: (1) Perform spatial morphological analysis on the three-dimensional feature point cloud of wool fibers to obtain fiber bundle geometric morphology data; (2) The fiber bundle geometric morphology data is input into the gradient convolutional network, and the yarn surface texture feature map is obtained through feature decomposition; (3) Based on the yarn surface texture feature map, the yarn twist is quantitatively analyzed to obtain the twist distribution curve; (4) Based on the yarn surface texture feature map, the spatial variation of fiber curl is analyzed to obtain the curl parameter matrix; (5) Feature mapping is performed between the twist distribution curve and the curl parameter matrix to obtain the yarn structure characteristic map; (6) The yarn structure characteristic map is comprehensively processed through the wool yarn quality evaluation system to obtain the wool yarn quality feature set.

[0025] Specifically, if Figure 2 As shown, it is a schematic diagram of the process of analyzing yarn twist and fiber curl by gradient convolution network in an embodiment of the present application, wherein spatial morphological features of point cloud data are extracted and the geometric morphological parameters G(x, y, z) of the fiber bundle are calculated: in represents the shape characteristic function, represents the curvature characteristic function, and is the weight coefficient, and (x, y, z) is the spatial coordinate. This formula comprehensively calculates characteristics such as the fiber bundle's diameter distribution, cross-sectional shape, and surface roughness, outputting fiber bundle geometry data. In practice, when processing a bundle of 30-tex wool yarn, the system's sampling point spacing is 0.1 mm, resulting in approximately 1000 feature points per millimeter of length.

[0026] The gradient convolution network receives geometric data and extracts the surface texture features of the yarn through multi-layer convolution operations. The network adopts an improved residual structure, and the calculation formula of the texture feature T is: in represents feature points, is the convolution operation, is the convolution kernel, is the weight coefficient, is the residual function, is the residual coefficient. Through this calculation, the system generates a texture feature map containing information about the yarn surface microstructure. On the production line, the system processes 10 meters of yarn per second and outputs texture feature data in real time.

[0027] Based on the texture feature map, the system performs twist and curl analysis respectively. The twist analysis uses a spiral structure recognition algorithm to calculate the fiber spiral angle θ to obtain the twist value: in is the yarn radius (mm), is the twist length (mm). The system records a twist value every 1 mm along the yarn axis to form a twist distribution curve. At the same time, the fiber curl is spatially analyzed, the curl coefficient in each direction is calculated, and the curl parameter matrix is generated. When processing Merino wool with a fineness of 21.5 micron, the twist value detected by the system fluctuates within the range of 600-800 T / m, and the curl is distributed between 70-80%. The feature mapping link merges the twist distribution curve with the curl parameter matrix to construct a yarn structure characteristic map. The map contains multi-dimensional information such as the geometric characteristics, physical properties and surface characteristics of the yarn, providing a data basis for quality evaluation. The wool yarn quality evaluation system calculates quality indicators such as strength, elasticity and feel based on these data, and outputs a wool yarn quality feature set.

[0028] In a specific embodiment, the process of executing step S103 may specifically include the following steps: (1) The reciprocating motion of the knitting needle is dynamically captured by a high-speed camera system to obtain a set of knitting needle time-series images; (2) Perform optical flow analysis on the knitting needle time series image set, extract the knitting needle motion trajectory data, and obtain the needle displacement movement matrix; (3) The knitting needle intersection points are spatially located according to the needle displacement matrix to obtain the knitting needle knitting node graph; (4) Input the knitting needle node graph into the texture enhancement network to extract the knitting texture features and obtain the knitted fabric texture data; (5) Process the knitted fabric texture data using a density analysis algorithm to obtain a fabric density distribution map; (6) Perform feature integration and mapping on the fabric density distribution map to obtain the fabric structure map.

[0029] Specifically, a high-speed camera system captures the knitting needle motion in real time. Using an industrial camera with a frame rate of 2000 fps, the reciprocating motion of the needles is continuously sampled and timestamped. The motion state V of each needle can be expressed as: in is the knitting needle position coordinate (mm), is the angular velocity (rad / s), is the acceleration (mm / s²). In actual production, 480 knitting needles on a size 12 knitting machine are synchronously captured to form a set of knitting needle time-series images. The knitting needles have a reciprocating motion cycle of 0.3 seconds, during which the system collects 600 frames of image data.

[0030] In the optical flow analysis stage, the improved Lucas-Kanade algorithm is used to process the time series images and calculate the needle motion trajectory. Needle displacement matrix The calculation formula is: in is the displacement weight coefficient, is the displacement between adjacent frames (mm), with i and j representing the needle's transverse and longitudinal indices, respectively. This matrix records the complete trajectory of the needle's motion. During loom operation, each needle's reciprocating amplitude is 15 mm, and the motion of each needle is tracked in real time.

[0031] Based on the needle displacement matrix, the system performs positioning analysis on the needle intersection points. The spatial geometric constraint method is used to identify the knitting nodes and calculate the node feature vector H: in is the node space coordinate (mm), is the intersection angle (deg), is the knitting tension (N). The system integrates these feature data into a knitting node graph for characterizing the fabric structure. When producing a wool sweater with a density of 12 needles per inch, 144 knitting nodes are identified per square centimeter, accurately recording the fabric structure. The texture enhancement network receives the knitting node graph and extracts the knitting texture features through multi-layer convolution operations. At the same time, the density analysis algorithm calculates the local density distribution of the fabric and generates a density distribution map that reflects the uniformity of the fabric structure. Through feature integration and mapping, all data are comprehensively processed and a fabric structure map is output.

[0032] In a specific embodiment, the process of executing step S104 may specifically include the following steps: (1) Input the fabric structure map into the spatiotemporal feature analysis unit to extract the fabric surface change characteristics and obtain the fabric structure time series data set; (2) Based on the fabric structure time series data set, fabric defects are classified through sequence feature analysis to obtain the fabric defect type distribution matrix; (3) Performing time domain transformation on the defect type distribution matrix and combining it with the fabric deformation characteristics to obtain a defect propagation trend diagram; (4) Based on the defect propagation trend diagram, the fabric deformation law is extracted through spatial correlation analysis to obtain the fabric stress distribution field; (5) Based on the stress distribution field of the fabric, the density calibration algorithm is used to perform local feature aggregation to obtain the fabric structural uniformity data; (6) The fabric structure uniformity data is comprehensively analyzed through the weaving defect evaluation system to obtain the fabric density uniformity index.

[0033] Specifically, the spatiotemporal feature analysis stage performs image segmentation on the fabric structure atlas, demarcating fabric surface feature regions. A sliding window approach is used to capture local variations in the fabric surface, including texture orientation, density distribution, and deformation. Temporal feature analysis combines consecutive frames of data into a fabric structure temporal dataset, recording the dynamic changes in the fabric structure over time. During the wool sweater production process, spatiotemporal feature analysis processes 120 frames of image data per second to record changes in fabric surface features. Sequential feature analysis processes the fabric structure temporal dataset and identifies fabric defects through morphological feature extraction. The analysis process includes edge detection, region segmentation, and feature matching. Identified defects are classified by type, forming a defect type distribution matrix. This matrix records the spatial distribution and severity of each defect. Common defects in wool sweater fabrics include broken needles, skipped stitches, and loop loss, each with its own unique characteristic pattern.

[0034] The time domain conversion process transforms the defect type distribution matrix from the spatial domain to the time domain. Combined with the deformation characteristics of the fabric during the production process, a defect propagation trend chart is constructed. This trend chart reflects the temporal expansion and impact of defects, providing a basis for predicting defect development. In practical applications, defect propagation trends are analyzed to predict the impact of defects on finished product quality. The spatial correlation analysis stage further processes the defect propagation trend chart to extract fabric deformation patterns. This analysis includes stress distribution, strain field calculation, and deformation trend prediction, generating a fabric stress distribution field. The stress distribution field reflects the stress state and deformation trends of each fabric region, providing data support for subsequent quality control. During the wool sweater weaving process, changes in fabric stress distribution are monitored in real time to promptly identify areas of abnormal stress concentration.

[0035] The density calibration algorithm aggregates local features of fabric stress distribution field data. Through regional density calculation, feature point extraction, and density uniformity assessment, fabric structural uniformity data is generated. This data contains information such as fabric density distribution, structural stability, and local deformation. At the production site, the density calibration algorithm processes fabric structural data in real time to assess the fabric's uniformity level. The fabric defect assessment system comprehensively analyzes fabric structural uniformity data and calculates a fabric density uniformity index. The assessment process considers multiple aspects of the fabric, including appearance quality, structural stability, and performance, providing a basis for production process adjustments. In actual application, the fabric defect assessment system conducts comprehensive testing on wool sweater fabrics and outputs a comprehensive score reflecting fabric quality.

[0036] Taking the production of a 12-needle-per-inch wool sweater fabric as an example, the spatiotemporal feature analysis unit dynamically monitors the fabric surface, capturing 120 frames of image per second to record changes in the fabric structure. Sequential feature analysis identifies defects such as broken and skipped stitches and generates a defect distribution matrix. Time-domain transformation analysis reveals that a broken needle defect at a specific location spreads to eight surrounding stitches within 5 seconds, and stress analysis reveals increased stress concentration in that area. The density calibration algorithm calculates a decrease in density uniformity in the defective area. Based on this information, the defect assessment system outputs a fabric quality score and recommends adjustments to the operating parameters of the relevant knitting needles.

[0037] In a specific embodiment, the process of executing step S105 may specifically include the following steps: (1) The wool yarn quality feature set is decomposed into multi-scale features, and the yarn physical property sequence is obtained through dynamic registration; (2) The fabric density uniformity index is sampled in time series, and the fabric structure deformation field is obtained by gradient feature extraction; (3) Based on the correspondence between the yarn physical property sequence and the fabric structure deformation field generation characteristics, the sweater quality mapping map is obtained; (4) Perform hierarchical attention calculation on the sweater quality map to obtain garment piece structure association data; (5) Decompose the garment piece structure association data through deep feature decomposition to obtain the sweater process parameter matrix; (6) A garment quality evaluation model is constructed based on the wool sweater process parameter matrix, where the garment quality evaluation model is used to quantitatively characterize the appearance quality, structural stability and comfort indicators of the wool sweater.

[0038] Specifically, the feature decomposition uses the wavelet transform method to separate the yarn features at different scales. The decomposition expression R is defined as: in is the scale parameter, is the time parameter, is the characteristic component, is the wavelet basis function. This calculation captures the physical properties of the yarn at different scales, including strength, elastic modulus, and surface friction coefficient. In actual production, when analyzing 30tex wool yarn, the decomposition scale is gradually adjusted from microscopic (0.1mm) to macroscopic (10mm).

[0039] After the fabric density uniformity index is sampled in time series, the local deformation characteristics of the fabric are calculated using the gradient feature extraction algorithm. The calculation formula is: in represents the basic deformation mode, is the corresponding weight coefficient, and (x, y) is the spatial coordinate. This formula calculates the deformation state of each area of the fabric, generating structural deformation field data. During the production process, each square centimeter is divided into 100 sampling points to monitor the fabric deformation state in real time.

[0040] The feature correspondence analysis uses a deep neural network structure to establish the mapping relationship between yarn properties and fabric deformation. The quality mapping function L is expressed as: in is the yarn property parameter, is the fabric deformation parameter, is the weight coefficient, is a nonlinear mapping function, and φ is an activation function. This calculation generates a sweater quality map, reflecting the correlation between raw material characteristics and finished product quality. A hierarchical attention mechanism processes the quality map and calculates the strength of correlation between different features. The analysis process consists of two stages: local feature extraction and global feature integration, outputting garment structure correlation data. Deep feature decomposition further extracts key process parameters, forming a sweater process parameter matrix. This matrix includes key parameters such as knitting tension, loop length, and loop depth.

[0041] Based on the process parameter matrix, the garment quality assessment model uses a multi-layer perceptron architecture to perform feature mapping, outputting multi-dimensional evaluation metrics including appearance quality, structural stability, and comfort. In wool sweater production, this model quantitatively assesses fabric flatness, dimensional stability, and tactile comfort, providing data support for quality control.

[0042] Taking the production of a 100% Merino wool sweater as an example, a multi-scale analysis of the yarn was performed, extracting physical properties such as a surface friction coefficient of 0.2-0.3 and an elastic modulus of 4000-5000 cN / tex. Time-series sampling of the fabric density uniformity index revealed that local deformation fluctuated within a range of 0.5 mm. Feature mapping analysis showed that when the yarn surface friction coefficient increased by 0.05, the local deformation of the fabric increased by 0.2 mm. Hierarchical attention calculation identified deformation-sensitive areas, and deep feature decomposition determined that the optimal weaving tension should be within the range of 25-30 cN. Based on these parameters, the quality assessment model calculated the fabric's flatness score, dimensional stability index, and tactile comfort level, providing a basis for optimizing the production process.

[0043] In a specific embodiment, the process of executing step S106 may specifically include the following steps: (1) Feature analysis is performed based on the output parameters of the garment quality assessment model, and the sweater quality status matrix is obtained through multidimensional data mapping; (2) The sweater quality status matrix is input into the deep graph feature network, and the quality defect propagation chain is obtained through regional connectivity analysis; (3) Decompose the time series characteristics of the quality defect propagation chain and obtain the process anomaly correlation map by combining the process parameter constraints; (4) Classify the warning levels based on the process anomaly correlation map and obtain the quality warning decision tree through the risk propagation model; (5) Input the quality warning decision tree into the process parameter optimization unit and obtain the process adjustment sequence through multi-objective optimization; (6) Construct a production process optimization plan based on the process adjustment sequence. The production process optimization plan includes knitting needle parameter optimization indicators, fabric structure optimization indicators and garment quality control indicators.

[0044] Specifically, appearance quality parameters (smoothness, color uniformity), structural stability parameters (anti-pilling, shrinkage resistance), and comfort parameters (breathability, softness) are extracted. Multidimensional data mapping converts these parameters into a standardized scoring space to construct a sweater quality status matrix. Each element in this matrix corresponds to a specific quality indicator, and correlation analysis establishes mapping relationships between different quality indicators. A deep graph feature network processes the quality status matrix data and uses graph convolution to analyze the correlations between different quality indicators. Regional connectivity analysis is performed by constructing a feature graph, with quality indicators as nodes and correlations between quality parameters as edges. This analysis identifies the propagation paths of quality defects and forms a quality defect propagation chain. This chain records the location, spread, and impact of defects, reflecting the evolution of quality issues during the production process.

[0045] Time series feature decomposition dynamically analyzes the quality defect propagation chain and generates a process anomaly correlation map based on process parameter constraints such as needle parameters, tension control, and temperature and humidity. This correlation map displays the corresponding relationship between quality defects and process parameters, reflecting the degree of impact of different process parameters on product quality. The map includes information such as process parameter trends, abnormal fluctuation points, and key control nodes. Warning level classification categorizes quality issues into different warning levels based on the degree of anomaly in the process anomaly correlation map. The risk propagation model analyzes the development trends of various quality issues, predicts the scope of problem spread and impact, and constructs a quality warning decision tree. The decision tree includes treatment plans for different warning levels and clarifies the priority and range of process parameter adjustments.

[0046] The process parameter optimization unit receives the output of the quality warning decision tree and calculates the optimal process parameter combination using a multi-objective optimization algorithm. The optimization process comprehensively considers multiple objectives, including product quality improvement, production efficiency, and resource consumption, and outputs a process adjustment sequence. The adjustment sequence includes specific parameter adjustment steps, adjustment timing, and adjustment range, providing specific guidance for production process optimization. The production process optimization plan integrates the content of the process adjustment sequence to form an optimization indicator system. Needle parameter optimization indicators include parameters such as needle movement speed, acceleration, and working angle; fabric structure optimization indicators include parameters such as loop density, structure, and loop depth; and garment quality control indicators include requirements such as dimensional tolerance, appearance quality, and wearability. These indicators form a process optimization plan to guide quality control during the production process.

[0047] Taking the production of a high-end wool sweater as an example, multidimensional data mapping identified pilling issues at the collar and dimensional deviations at the cuffs. Deep graph feature network analysis revealed a correlation between these two defects, both related to deviations in the needle working angle. Time series feature decomposition revealed that every 1-degree deviation in the needle angle resulted in a 0.1-mm change in the loop length, which in turn affected the fabric density. The early warning system classified this issue as a Level 2 alert, and the risk propagation model predicted that if not addressed promptly, the problem would spread to the entire cuff area. Process parameter optimization calculated that the needle angle needed to be adjusted by 0.8 degrees and provided a combination of process parameters for adjusting the machine: reducing the needle speed by 50 rpm and increasing the mid-section lateral tension. The optimization plan also included recommendations for fabric structure adjustments: increasing the amount of anti-pilling finish and adjusting the density ratio of the cuff ribbing.

[0048] The above describes the smart production monitoring method of wool sweaters based on the Internet of Things in the embodiment of the present application. The following describes the smart production monitoring system of wool sweaters based on the Internet of Things in the embodiment of the present application. Figure 3 In the embodiment of the present application, an embodiment of the smart sweater production monitoring system based on the Internet of Things includes: The acquisition module is used to collect images of wool yarn fiber bundles through a multispectral imaging system and obtain a three-dimensional feature point cloud of wool fibers through spatial stereo matching algorithm processing; The analysis module is used to analyze the yarn twist and fiber curl based on the three-dimensional feature point cloud of wool fibers through a gradient convolutional network to obtain the wool yarn quality feature set; The processing module is used to obtain the knitting needle motion sequence through a high-speed acquisition system during the knitted fabric forming process, and then process it through a texture enhancement algorithm to obtain a fabric structure map; The prediction module is used to construct a weaving defect prediction model based on the fabric structure map through the spatiotemporal feature network and memory model to obtain the fabric density uniformity index; A fusion module is used to input the wool yarn quality feature set and the fabric density uniformity index into a feature fusion network, and process them through a multi-layer attention mechanism to obtain a garment quality assessment model; The optimization module is used to build a wool sweater quality control early warning system through a deep graph network based on the output parameters of the garment quality assessment model to obtain a production process optimization plan.

[0049] Through the collaborative efforts of these components, a multispectral imaging system captures images of wool yarn fiber bundles, and a spatial stereo matching algorithm is used to construct a three-dimensional fiber feature point cloud, achieving precise characterization of fiber morphology. A gradient convolutional network analyzes yarn twist and fiber curl to extract a set of wool yarn quality features, providing a reliable data foundation for quality assessment. A high-speed acquisition system captures needle motion sequences in real time and generates a fabric structure map using a texture enhancement algorithm, enabling dynamic monitoring of the weaving process. A spatiotemporal feature network and memory model construct a weaving defect prediction model, accurately outputting fabric density and uniformity indicators and improving the accuracy of quality warnings. A feature fusion network utilizes a multi-layer attention mechanism to process multi-source data and establish a garment quality assessment model, enabling comprehensive evaluation of product quality. A wool sweater quality control warning system, constructed using a deep graph network, can promptly generate production process optimization plans, enhancing the intelligence of the production process. The entire solution utilizes a multispectral imaging algorithm to provide highly accurate raw material feature data, a spatial stereo matching algorithm to achieve 3D fiber reconstruction, a gradient convolutional network to ensure the accuracy of quality feature extraction, a spatiotemporal feature network to enhance defect prediction, a multi-layer attention mechanism to optimize feature fusion, and a deep graph network to improve the reliability of the early warning system. This significantly improves the monitoring accuracy and intelligence level of the wool sweater production process.

[0050] Reference Figure 4 In an embodiment of the present invention, a computer device is also provided. The computer device may be a server, and its internal structure may be as follows: Figure 4 As shown. The computer device includes a processor, memory, display screen, input device, network interface and database connected via a system bus. The processor of the computer design is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, the above method is implemented.

[0051] Those skilled in the art will understand that Figure 4 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present invention and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0052] An embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, which implements the above-described method when executed by a processor. It is understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0053] Those skilled in the art will appreciate that all or part of the processes in the above-described method embodiments can be implemented by instructing the relevant hardware using a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the above-described method embodiments. Any reference to memory, storage, database, or other media provided herein and used in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-speed SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct RAMbus dynamic RAM (DRDRAM), and RAMbus dynamic RAM.

[0054] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0055] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0056] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application is described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for intelligent production monitoring of wool sweaters based on the Internet of Things, characterized in that: The smart production monitoring method for wool sweaters based on the Internet of Things includes: The wool yarn fiber bundle is imaged using a multispectral imaging system and processed using a spatial stereo matching algorithm to obtain a three-dimensional feature point cloud of the wool fiber. According to the wool fiber three-dimensional feature point cloud, the yarn twist and fiber curl are analyzed by gradient convolution network to obtain the wool yarn quality feature set; In the knitted fabric forming process, the needle motion sequence is acquired through a high-speed acquisition system and processed by a texture enhancement algorithm to obtain the fabric structure map; According to the fabric structure map, a fabric defect prediction model is constructed through a spatiotemporal feature network and a memory model to obtain a fabric density uniformity index; Inputting the wool yarn quality feature set and the fabric density uniformity index into a feature fusion network, and processing them through a multi-layer attention mechanism to obtain a garment quality assessment model; Based on the output parameters of the garment quality assessment model, a wool sweater quality control early warning system was constructed through a deep graph network to obtain a production process optimization plan.

2. The method for intelligent production monitoring of wool sweaters based on the Internet of Things according to claim 1 is characterized in that: The method of collecting images of wool yarn fiber bundles by a multispectral imaging system and processing them by a spatial stereo matching algorithm to obtain a three-dimensional feature point cloud of wool fibers includes: The wool yarn fiber bundle is imaged in different wavelength bands using a multispectral camera to obtain visible light image data, near-infrared image data, and ultraviolet image data; Inputting the visible light image data, near infrared image data and ultraviolet image data into a structured light encoder to obtain a fiber surface depth coding map; According to the fiber surface depth coding map, disparity calculation is performed through a stereo matching network to obtain multi-view point cloud data; Inputting the multi-view point cloud data into a feature extraction network, obtaining geometric feature parameters through a deep learning method, and obtaining fiber surface optical property data; According to the fiber surface optical property data and the geometric characteristic parameters, a three-dimensional feature point cloud of the wool fiber is obtained through multi-scale feature fusion processing; Based on the three-dimensional feature point cloud of the wool fiber, spatial position calibration is performed through a binocular camera calibration system to complete the accurate reconstruction of the point cloud data.

3. The method for intelligent production monitoring of wool sweaters based on the Internet of Things according to claim 1, characterized in that: According to the wool fiber three-dimensional feature point cloud, the yarn twist and fiber curl are analyzed by a gradient convolution network to obtain a wool yarn quality feature set, including: Perform spatial morphological analysis on the three-dimensional feature point cloud of wool fibers to obtain fiber bundle geometric morphology data; Inputting the fiber bundle geometric morphology data into a gradient convolutional network, and obtaining a yarn surface texture feature map through feature decomposition; Quantitatively analyzing the yarn twist according to the yarn surface texture characteristic diagram to obtain a twist distribution curve; According to the yarn surface texture characteristic map, a spatial variation analysis of the fiber curl is performed to obtain a curl parameter matrix; Performing feature mapping on the twist distribution curve and the curl parameter matrix to obtain a yarn structure characteristic map; The yarn structure characteristic map is comprehensively processed by a wool yarn quality evaluation system to obtain a wool yarn quality feature set.

4. The method for intelligent production monitoring of wool sweaters based on the Internet of Things according to claim 1, characterized in that: For the knitted fabric forming process, the knitting needle motion sequence is acquired through a high-speed acquisition system and processed by a texture enhancement algorithm to obtain a fabric structure map, including: The reciprocating motion of the knitting needles is dynamically captured by a high-speed camera system to obtain a set of knitting needle time-series images; Performing optical flow analysis on the knitting needle time series image set, extracting knitting needle motion trajectory data, and obtaining a needle displacement movement matrix; Positioning the knitting needle intersections in space according to the needle displacement movement matrix to obtain a knitting needle knitting node graph; Inputting the knitting needle knitting node graph into a texture enhancement network to extract knitting texture features and obtain knitted fabric texture data; Processing the knitted fabric texture data using a density analysis algorithm to obtain a fabric density distribution map; The fabric density distribution map is subjected to feature integration and mapping to obtain a fabric structure map.

5. The method for intelligent production monitoring of wool sweaters based on the Internet of Things according to claim 1, characterized in that: The method of constructing a weaving defect prediction model based on the fabric structure map by using a spatiotemporal feature network and a memory model to obtain a fabric density uniformity index includes: Inputting the fabric structure map into a spatiotemporal feature analysis unit to extract fabric surface variation features and obtain a fabric structure time series data set; According to the fabric structure time series data set, fabric defects are classified by sequence feature analysis to obtain a fabric defect type distribution matrix; Performing time domain conversion on the weaving defect type distribution matrix and combining it with fabric deformation characteristics to obtain a weaving defect propagation trend diagram; Based on the defect propagation trend diagram, the fabric deformation law is extracted through spatial correlation analysis to obtain the fabric stress distribution field; Based on the fabric stress distribution field, local feature aggregation is performed using a density calibration algorithm to obtain fabric structural uniformity data; The fabric structure uniformity data is comprehensively analyzed by a fabric defect evaluation system to obtain a fabric density uniformity index.

6. The method for intelligent production monitoring of wool sweaters based on the Internet of Things according to claim 1, characterized in that: The wool yarn quality feature set and the fabric density uniformity index are input into a feature fusion network and processed through a multi-layer attention mechanism to obtain a garment quality assessment model, including: Performing multi-scale feature decomposition on the wool yarn quality feature set and obtaining a yarn physical property sequence through dynamic registration; Performing time series sampling on the fabric density uniformity index and obtaining the fabric structure deformation field by gradient feature extraction; Obtaining a sweater quality mapping atlas based on a correspondence between the yarn physical property sequence and the fabric structure deformation field generation feature; Performing hierarchical attention calculation on the sweater quality map to obtain garment piece structure association data; Decomposing the garment piece structure association data through deep feature decomposition to obtain a sweater process parameter matrix; A garment quality evaluation model is constructed based on the wool sweater process parameter matrix, wherein the garment quality evaluation model is used to quantitatively characterize the appearance quality, structural stability and comfort index of the wool sweater.

7. The method for intelligent production monitoring of wool sweaters based on the Internet of Things according to claim 1, characterized in that: The output parameters of the garment quality assessment model are used to construct a wool sweater quality control early warning system through a deep graph network to obtain a production process optimization plan, including: Performing feature analysis based on the output parameters of the garment quality assessment model and obtaining a sweater quality status matrix through multi-dimensional data mapping; The sweater quality status matrix is input into the deep graph feature network, and the quality defect propagation chain is obtained through regional connectivity analysis; Performing time series feature decomposition on the quality defect propagation chain and combining it with process parameter constraints to obtain a process anomaly correlation map; Based on the process anomaly correlation map, the warning level is divided and the quality warning decision tree is obtained through the risk propagation model; Inputting the quality warning decision tree into the process parameter optimization unit, and obtaining the process adjustment sequence through multi-objective optimization; A production process optimization plan is constructed according to the process adjustment sequence, and the production process optimization plan includes knitting needle parameter optimization indicators, fabric structure optimization indicators and garment quality control indicators.

8. An intelligent production monitoring system for wool sweaters based on the Internet of Things, used to implement the intelligent production monitoring method for wool sweaters based on the Internet of Things as claimed in any one of claims 1 to 7, characterized in that: The IoT-based smart sweater production monitoring system includes: The acquisition module is used to collect images of wool yarn fiber bundles through a multispectral imaging system and obtain a three-dimensional feature point cloud of wool fibers through spatial stereo matching algorithm processing; An analysis module is used to analyze the yarn twist and fiber curl based on the three-dimensional feature point cloud of the wool fiber through a gradient convolution network to obtain a wool yarn quality feature set; The processing module is used to obtain the knitting needle motion sequence through a high-speed acquisition system during the knitted fabric forming process, and then process it through a texture enhancement algorithm to obtain a fabric structure map; A prediction module is used to construct a weaving defect prediction model based on the fabric structure map through a spatiotemporal feature network and a memory model to obtain a fabric density uniformity index; A fusion module is used to input the wool yarn quality feature set and the fabric density uniformity index into a feature fusion network, and process them through a multi-layer attention mechanism to obtain a garment quality assessment model; The optimization module is used to build a wool sweater quality control early warning system through a deep graph network based on the output parameters of the garment quality assessment model to obtain a production process optimization plan.

9. A computer device, characterized in that: It includes a memory and a processor, the memory stores a computer program that can be run on the processor, and is characterized in that when the processor executes the computer program, it implements the smart production monitoring method of wool sweaters based on the Internet of Things as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the processor is enabled to execute the method for intelligent production monitoring of wool sweaters based on the Internet of Things according to any one of claims 1 to 7.

Images