An automated yarn production control method and control system

Abstract

The application relates to the technical field of yarn production, in particular to an automatic yarn production control method and a control system, which comprises the following steps: collecting a multi-view polarization image sequence of yarn in a production process online to obtain optical characteristic differences; extracting a yarn main body polarization reflection signal through polarization state separation and eliminating fluff scattering noise; completing multi-view registration and fusing into a yarn panoramic image through feature point matching; processing a hair point cloud based on z coordinate grouping according to density clustering; combining contour information based on multi-view contour transformation to construct a yarn three-dimensional form model; calculating yarn diameter and surface uniformity in real time according to the three-dimensional model; comparing quality indexes with a qualified threshold range, and generating a control instruction by an analysis unit when the quality indexes exceed the qualified threshold range, adjusting production equipment parameters through an execution unit until the indexes return to the qualified threshold range. The application can effectively overcome the measurement error caused by yarn fluff scattering and improve production efficiency.

Description

Technical Field

[0001] This invention relates to the field of yarn production, specifically to an automated yarn production control method and control system. Background Technology

[0002] With the rapid development of the textile industry towards intelligence and automation, automated yarn production control methods based on machine vision have become a key technology for improving product quality and production efficiency. This method uses high-definition industrial cameras to acquire yarn images online and employs image processing algorithms to analyze key indicators in real time, thereby achieving real-time monitoring and closed-loop feedback control of the production process. Currently, this technology is widely used in processes such as yarn clearing and winding, and has become an important component of modern digital spinning workshops.

[0003] However, in practical production applications, the accuracy and reliability of existing control methods face serious challenges. The core problem lies in the inherent interference to imaging quality caused by the complex optical and physical properties of the yarn material itself. Specifically, yarns with surfaces rich in fluff, such as wool, have dense fiber fluff that triggers strong, irregular light scattering, resulting in blurred yarn edges and diffused contours in the acquired images. This severely interferes with the accurate measurement and judgment of the yarn's diameter and surface uniformity.

[0004] In summary, existing automated control methods exhibit significant perceptual deficiencies when faced with image source noise introduced by the light scattering of yarn fibers and its unique luster. Traditional solutions often focus on optimizing backend algorithms but fail to effectively address this fundamental interference at the level of imaging mechanisms or frontend feature decoupling. Consequently, in high-value-added, complex yarn production scenarios, the stability and accuracy of quality control systems struggle to meet the requirements of high-end manufacturing. Therefore, a novel automated yarn production control method capable of overcoming or distinguishing such optical interference is urgently needed. Summary of the Invention

[0005] (1) Technical problems to be solved

[0006] The purpose of this invention is to solve the technical problem that irregular light scattering from the surface hairs of yarn causes blurred imaging, which interferes with the accurate measurement of the yarn body diameter and surface uniformity.

[0007] (2) Technical solution

[0008] To achieve the above objectives, in one aspect, the present invention provides an automated yarn production control method, the method comprising:

[0009] A front-end imaging system is built, which works in conjunction with a dual-plane mirror assembly and a polarization control module to acquire multi-view polarization image sequences of yarn during the production process online, so that the yarn body and the fibers exhibit distinguishable optical feature differences due to different responses, and the optical feature differences are acquired simultaneously.

[0010] The polarization reflection signal of the yarn body is extracted by polarization state separation to remove the source noise generated by non-polarization scattering of the fibers; then, multi-view image registration is completed by feature point matching and fused to form a panoramic image of the yarn.

[0011] Based on the multi-view polarization image sequence and optical feature differences, the polarization reflection signal of the yarn body is extracted, the source noise generated by non-polarization scattering of the fluff is removed, and the fluff point cloud is processed in groups according to the z coordinate; based on the principle of multi-view contour transformation, the contour information of each view image is merged to construct a three-dimensional morphological model of the yarn.

[0012] Based on the three-dimensional morphology model of the yarn, quality indicators are calculated in real time. The quality indicators include yarn diameter and surface uniformity. The yarn diameter is calculated by fitting a circle to the cross-sectional profile, and the surface uniformity is characterized by the coefficient of variation of the diameters of adjacent cross sections. The qualified threshold range of each indicator is set.

[0013] The quality indicators are compared with the acceptable threshold range. When an indicator exceeds the acceptable threshold range, the analysis unit generates a targeted control instruction. The execution unit adjusts the operating parameters of the production equipment, including the yarn clearer cutting threshold, the winding machine speed, and the fiber feed rate, until the indicator returns to the acceptable threshold range to ensure the stability of yarn production quality.

[0014] Furthermore, the polarization state separation is achieved based on the difference in polarization characteristics between the yarn body and the fluff. The reflected signal of the yarn body has fixed polarization characteristics, while the scattering noise of the fluff is in an irregular polarization state. By distinguishing the difference in fixed polarization characteristics, the reflected signal of the yarn body is filtered, thereby achieving effective separation between the body signal and the scattering noise of the fluff.

[0015] Furthermore, the method for completing multi-view image registration and fusing to form a yarn panoramic image through feature point matching includes:

[0016] Key feature points of yarn images from each viewpoint of a multi-view polarization image sequence are extracted. These key feature points are determined based on the edge contour, structural inflection points, and continuous morphological features of the yarn body.

[0017] Select a yarn image from one perspective as the reference image, and perform consistency matching between the key feature points of the other perspective images and the corresponding feature points of the reference image to establish the spatial relationship between the perspective images.

[0018] Based on the aforementioned spatial correlation, yarn images from all perspectives are calibrated to the same spatial coordinate system to achieve multi-view image matching.

[0019] Information is integrated from the matched images at each viewpoint, retaining the effective morphological information of the yarn that is not obscured from each viewpoint, and eliminating duplicate and redundant data, ultimately forming a panoramic image of the yarn that can fully present all morphological features of the yarn in the circumferential direction.

[0020] Furthermore, the method for constructing a three-dimensional yarn morphology model by merging contour information from images at various perspectives based on the principle of multi-view contour transformation includes:

[0021] The contour information of the yarn body in the registered panoramic image of yarn from each viewpoint is extracted. The contour information includes the edge contour, cross-sectional shape and continuous extension features of the yarn. Non-main contours corresponding to fuzz and residual noise are removed.

[0022] Based on the principle of multi-view contour transformation, and combined with the spatial coordinate system relationship of each view image, the yarn contour of each view is mapped to a unified three-dimensional space, and the correspondence of each contour in the spatial depth direction is established.

[0023] The mapped multi-view contours are superimposed and merged to retain the effective structural information of the yarn body in each contour, eliminate redundant data in the overlapping area of ​​the contours, and fill in the contour details not covered under a single view to form a complete set of yarn space contours.

[0024] The yarn spatial contour set is converted into three-dimensional point cloud data. By integrating the spatial position information of the point cloud, a three-dimensional morphological model of the yarn body is constructed, which can accurately restore the three-dimensional structure, shape and extension state of the yarn body.

[0025] Furthermore, the method for calculating the yarn diameter includes:

[0026] Multiple cross sections are uniformly selected from the three-dimensional morphological model of the yarn along the yarn extension direction. The main outline of the yarn is extracted from each cross section. Residual fuzz and non-main interference parts in the outline are removed. Circle fitting is performed based on the outline edge points of each cross section. The diameter of the fitted circle is used as the yarn diameter of the corresponding cross section to ensure that the diameter calculation result fits the actual cross-sectional shape of the yarn.

[0027] Furthermore, the surface uniformity is characterized by the coefficient of variation of adjacent cross-sectional diameters, specifically including:

[0028] Multiple adjacent cross sections are continuously selected from the three-dimensional morphological model of the yarn along the yarn extension direction. The yarn diameter data of each cross section is extracted to form a diameter sequence. The dispersion of this diameter sequence is calculated. The uniformity of the yarn surface is quantitatively characterized by the coefficient of variation of the diameter of adjacent cross sections. This coefficient of variation intuitively reflects the diameter fluctuation of the yarn along the length direction, ensuring the objectivity and accuracy of the uniformity evaluation.

[0029] On the other hand, the present invention provides an automated yarn production control system, the system comprising:

[0030] Image acquisition unit: It has multi-view acquisition and polarization control functions, including a dual-plane mirror assembly and a polarization control module, which is used to acquire multi-view polarization image sequences of yarn online and simultaneously output the multi-view polarization image sequences and optical feature differences;

[0031] Data processing unit: used to perform polarization state separation, feature point matching, image registration and fusion, denoising and thinning, contour transformation and 3D modeling processing on multi-view polarized images and optical feature differences, and calculate yarn diameter and surface uniformity based on the constructed yarn 3D morphology model;

[0032] Analysis unit: Pre-stores qualified threshold ranges for yarn diameter and surface uniformity, receives quality indicators output by the data processing unit and compares them with the thresholds, and generates targeted control instructions when the indicators exceed the thresholds;

[0033] Execution unit: responds to control commands to adjust the operating parameters of the production equipment, including the yarn clearer cutting threshold, the winding machine speed and the fiber feeding amount, to realize closed-loop feedback control of the production process.

[0034] Furthermore, the dual-plane mirror assembly in the image acquisition unit is fixedly set to expand the imaging angle through reflection, thereby enabling multi-view image acquisition in conjunction with an industrial camera; the polarization control module includes a polarization light source and an electrically controlled polarization filter, which work together to enhance the optical characteristic differences between the yarn body and the fluff, suppress the interference of irregular light scattering from the fluff, and ensure the effectiveness of the multi-view polarization image sequence and optical characteristic differences.

[0035] Furthermore, the data processing unit includes an industrial computer and an embedded processing module. The embedded processing module integrates relevant processing logic for image separation, registration, fusion, modeling, and index calculation. It can quickly process multi-view polarized image sequences, complete the construction of a three-dimensional yarn morphology model and the calculation of quality indicators, and ensure the real-time performance of data processing to meet the needs of online production control.

[0036] Furthermore, the analysis unit and the execution unit establish a connection through an industrial communication protocol. The analysis unit can modify and save the qualified threshold range online, and the generated control commands can accurately match the corresponding operating parameter adjustment scheme according to the type and degree of the quality index exceeding the standard. The execution unit includes a yarn clearer, a winding machine, and a fiber feed regulator, which can adjust the operating parameters in real time according to the control commands to ensure that the yarn quality index quickly returns to the qualified threshold range.

[0037] (3) Beneficial effects

[0038] Compared with the prior art, the beneficial effects of the present invention are:

[0039] An imaging system integrating a dual-plane mirror assembly and a polarization control module effectively suppressed irregular light scattering noise from yarn surface fibers at the data acquisition source. By utilizing the difference in polarization characteristics, the optical feature contrast between the yarn body and the fibers was enhanced, providing image data with a higher signal-to-noise ratio for subsequent processing. Attached Figure Description

[0040] Figure 1 This is a flowchart of the automated yarn production control method according to Embodiment 1 of the present invention.

[0041] Figure 2 This is a system module diagram of the automated yarn production control system of Embodiment 2 of the present invention. Detailed Implementation

[0042] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0043] Before providing examples, it is necessary to describe the application scenarios of this invention. Wool yarns, air-jet spun yarns, and similar products have surfaces rich in dense fiber fluff. These fluff fibers typically have a diameter between 1 and 5 μm and a length between 1 and 10 mm. When illumination from an industrial camera shines on the yarn surface, the fluff fibers induce strong, random light scattering (belonging to the Mie scattering category). Due to the random spatial distribution and irregular shape of the fluff fibers, the scattered light forms large areas of grayscale dispersion in the image, resulting in blurred edges and unclear outlines of the yarn body, severely interfering with the accurate identification of the yarn body's diameter and surface uniformity.

[0044] According to optical scattering theory, the intensity of scattered light from yarn fibers is closely related to the fiber density, diameter, refractive index, and wavelength of the illuminating light. Experimental data shows that when the fiber density on the yarn surface exceeds 50 fibers / mm, the scattered light reduces the gray-scale gradient at the edge of the yarn body by more than 30%, resulting in a contour offset of the yarn body in a two-dimensional image of 0.05~0.1mm, far exceeding the ±0.01mm measurement accuracy threshold required for high-value-added yarns.

[0045] Example 1: As Figure 1 As shown, this embodiment provides an automated yarn production control method, the method comprising:

[0046] Step 1: Multi-view polarization co-imaging acquisition:

[0047] A front-end imaging system is built, which works in conjunction with a dual-plane mirror assembly and a polarization control module to acquire multi-view polarization image sequences of yarn during the production process online; the multi-view polarization image sequences and optical feature differences are output synchronously.

[0048] The dual-plane mirror assembly is designed with a 72° angle between the two mirrors. Through optical reflection, the industrial camera simultaneously captures one real image of the yarn and four evenly distributed virtual images in a single image, achieving 360° full-view coverage. This angle is chosen based on the geometric symmetry of a regular pentagon. The four virtual and real images are evenly distributed in the image, with adjacent viewing angles spaced 72° apart, ensuring no imaging blind spots. The dual-plane mirror assembly is fixed inside a darkroom. The mirrors are made of high-reflectivity optical glass (reflectivity ≥95%), and the surface is treated with anti-glare to avoid secondary interference from reflected light.

[0049] The working mechanism of the polarization control module: The polarization control module includes a light source with switchable polarization direction and a synchronously operating electrically controlled polarization filter. The yarn body is composed of dense fiber bundles, and its reflected light has fixed polarization characteristics (the polarization direction is consistent with the fiber arrangement direction); while the fluff is a loosely distributed single fiber, and its scattered light is in a state of irregular polarization. By switching the polarization direction of the light source (0°, 90°, 135°, 180°) and controlling the synchronous matching of the electrically controlled polarization filter, the optical feature difference between the yarn body and the fluff in the multi-view polarization image sequence is enhanced, increasing the gray value difference between the two by more than 40%, laying the foundation for subsequent signal separation.

[0050] Image acquisition parameter optimization: A high-definition industrial camera with a resolution of ≥1920×1200 pixels is selected, and the frame rate is set to 30fps to ensure that the online detection requirements of the spinning process (yarn running speed 5~30m / min) are met; the exposure time is set to 10~20μs to avoid image blurring caused by yarn movement; a white polarized LED light source with a wavelength of 400~760nm is used, and the illumination uniformity is ≥90% to ensure consistent illumination on the yarn surface and avoid local reflection interference.

[0051] Step 2: Polarization state separation and multi-view image fusion.

[0052] The polarization reflection signal of the yarn body is extracted by polarization state separation to remove the source noise generated by non-polarization scattering of the fibers; then, multi-view image registration is completed by feature point matching and fused to form a panoramic image of the yarn.

[0053] Polarization separation: Based on the difference in polarization characteristics between the yarn body and the fluff, signal separation is performed on multi-view polarized image sequences. First, the grayscale value change rate of the same pixel under different polarization directions is calculated. The grayscale value change rate of the yarn body is ≤10% (fixed polarization characteristics), while the grayscale value change rate of fluff scattering noise is ≥30% (irregular polarization). Second, a screening threshold is set according to the grayscale value change rate, retaining pixels whose change rate matches the characteristics of the main body and removing noise pixels with excessive change rates, thus achieving effective separation of the main body signal and noise. This process requires no complex algorithms and can be completed simply by comparing grayscale values, ensuring real-time processing.

[0054] Feature point extraction: Key feature points of yarn images from each perspective are extracted. The key feature points are determined based on the edge contour of the yarn body (such as the upper and lower edges of the yarn strip), structural inflection points (such as the diameter abrupt change), and continuous morphological features (such as the straight extension segment of the yarn strip). The coordinates of the feature points are extracted using an edge detection algorithm (such as the Canny operator). The number of feature points extracted from each perspective is ≥50 to ensure matching reliability.

[0055] Reference image selection: Select the yarn image corresponding to the real image as the reference image. This image has the closest imaging distance and the highest clarity, and can be used as the registration reference.

[0056] Feature point matching: The feature points of the remaining 4 virtual images are matched with the feature points of the reference image. Similarity criteria (such as Euclidean distance ≤ 2 pixels) are used to filter matching point pairs and eliminate mismatched points. The matching accuracy is ≥ 95%.

[0057] Coordinate system calibration: Based on the matching point pairs, establish the spatial relationship between each view image and the reference image. Through coordinate transformation, calibrate all view images to the same spatial coordinate system to ensure that the position of the yarn body is accurately corresponding under each view, with a registration error of ≤1 pixel.

[0058] Panoramic image fusion: Information is integrated from the registered images from various perspectives. A strategy of optimizing and deduplicating is adopted to retain the effective morphological information of the yarn that is not obscured in each perspective (such as the clear left outline of the yarn in a certain perspective) and eliminate duplicate and redundant data (such as the overlapping main body area in multiple perspectives). Finally, a panoramic image of the yarn that can completely present all the morphological features of the yarn in the circumferential direction is formed, eliminating the imaging blind spots in a single perspective.

[0059] Step 3: Image optimization and yarn 3D morphology modeling.

[0060] Based on the multi-view polarization image sequence and optical feature differences, the polarization reflection signal of the yarn body is extracted, the source noise generated by non-polarization scattering of the fluff is removed, and the fluff point cloud is processed in groups according to the z coordinate; based on the principle of multi-view contour transformation, the contour information of each view image is merged to construct a three-dimensional morphological model of the yarn.

[0061] Image denoising and thinning include:

[0062] Statistical filtering denoising: Statistical filtering is used to remove noise from yarn edges. A 3×3 pixel neighborhood window is selected, and the average grayscale value and standard deviation of the pixels within the window are calculated. Pixels deviating from the average value by more than twice the standard deviation are identified as noise and removed. This method can effectively remove a small amount of residual fuzz noise after polarization separation while preserving the details of the main subject's outline. The signal-to-noise ratio of the denoised image is ≥35dB.

[0063] Density-based clustering: The feather point cloud is grouped according to the z-coordinate (yarn extension direction), with each group corresponding to the point cloud data of the same cross-section. A density-based clustering method is used, setting a neighborhood range and a minimum group size to group point clouds with similar densities together, further separating the feather point cloud from the main point cloud. The minimum group size is set to 1 to ensure that the point cloud of a single isolated feather is not missed, while avoiding noise from being mixed into the feather groupings.

[0064] Yarn 3D morphology modeling, including:

[0065] Contour information extraction: The contour information of the main body of the yarn in the registered panoramic image of the yarn from each viewpoint is extracted. The contour information includes the edge contour (boundary in the xy plane), cross-sectional shape (cross-sectional contour in the xz plane), and continuous extension features (length direction shape in the yz plane). Non-main body contours corresponding to fuzz and residual noise are removed by threshold filtering to ensure that the purity of the contour information is ≥99%.

[0066] Contour Space Mapping: Based on the principle of multi-view contour transformation and combined with the spatial coordinate system relationship of images from different viewpoints, the yarn contour of each viewpoint is mapped to a unified three-dimensional space. According to the reflection path of the two plane mirrors and the camera imaging parameters, the depth coordinates of each contour in three-dimensional space are calculated, and the correspondence of each contour in the spatial depth direction is established. The mapping error is ≤0.005mm.

[0067] Contour overlay and merging: The mapped multi-view contours are overlaid and merged to retain the effective structural information of the yarn body in each contour (such as the clear left contour of the cross section under one view and the clear right contour of the cross section under another view), eliminate redundant data in the overlapping area of ​​the contours, and complete the contour details not covered under a single view (such as the contour of the back of the yarn) to form a complete set of yarn spatial contours.

[0068] 3D Model Construction: The set of yarn spatial contours is converted into 3D point cloud data, and the spatial coordinates (x, y, z) of each point cloud are accurately calculated through contour mapping results. By integrating the spatial position information of the point clouds, a surface reconstruction algorithm is used to construct a 3D morphological model of the yarn that can accurately restore the 3D structure, shape, and extension state of the yarn body. The 3D dimensional error of the model is ≤0.01mm.

[0069] Step 4: Real-time extraction of core quality indicators.

[0070] Based on the three-dimensional morphology model of the yarn, quality indicators are calculated in real time. The quality indicators include yarn diameter and surface uniformity. The yarn diameter is calculated by fitting a circle to the cross-sectional profile, and the surface uniformity is characterized by the coefficient of variation of the diameters of adjacent cross sections. The qualified threshold range of each indicator is set.

[0071] The core of this step is to achieve accurate and quantitative extraction of quality indicators, and the specific process is as follows:

[0072] Yarn diameter calculation includes:

[0073] Cross-section selection: Multiple cross-sections are uniformly selected from the three-dimensional morphology model of the yarn along the yarn extension direction (z-axis direction). The selection interval is set according to the yarn running speed and detection accuracy requirements, usually 0.1mm, to ensure coverage of key areas along the entire length of the yarn.

[0074] Contour purification: Extract the main contour of the yarn in each cross section, and remove residual fuzz and non-main interference parts in the contour through morphological processing to ensure the integrity and accuracy of the contour.

[0075] Circular Fitting Calculation: A circular fitting process is performed based on the contour edge points of each cross-section. The least squares method is used to solve for the center coordinates and radius of the fitted circle, and the diameter of the fitted circle is taken as the yarn diameter of the corresponding cross-section. This method can effectively compensate for measurement errors caused by slight ellipticization of the yarn cross-section, with a diameter calculation accuracy of ≤ ±0.008 mm.

[0076] Surface uniformity calculation includes:

[0077] Diameter sequence acquisition: Multiple adjacent cross-sections are continuously selected from the three-dimensional morphological model of the yarn along the yarn extension direction, and the yarn diameter data of each cross-section is extracted to form a diameter sequence. The sequence length is set according to the detection requirements, usually 100 consecutive cross-sections, to ensure that the diameter fluctuation in the yarn length direction can be reflected;

[0078] Coefficient of variation calculation: The dispersion of the diameter sequence is calculated, and the uniformity of the yarn surface is quantitatively characterized by the coefficient of variation of the diameters of adjacent cross-sections. The coefficient of variation is characterized by the ratio of the standard deviation to the mean, where the standard deviation reflects the fluctuation range of the diameter, and the mean reflects the average diameter of the yarn;

[0079] Threshold setting: Set the acceptable threshold range according to the yarn type and production requirements. For high value-added yarns, the acceptable threshold for yarn diameter is usually set to ±0.01mm (relative to the design diameter), and the acceptable threshold for surface uniformity is set to a coefficient of variation ≤5%. For ordinary yarns, the threshold range can be appropriately relaxed to ensure the flexibility and adaptability of the threshold setting.

[0080] Step 5: Execution of closed-loop feedback control.

[0081] The quality indicators are compared with the acceptable threshold range. When an indicator exceeds the acceptable threshold range, the analysis unit generates a targeted control instruction. The execution unit adjusts the operating parameters of the production equipment, including the yarn clearer cutting threshold, the winding machine speed, and the fiber feed rate, until the indicator returns to the acceptable threshold range to ensure the stability of yarn production quality.

[0082] The core of this step is to achieve precise closed-loop control of the production process, and the specific logic is as follows:

[0083] Indicator comparison and anomaly judgment: The extracted yarn diameter and surface uniformity indicators are compared with the preset qualified threshold range in real time. When the indicators exceed the threshold range, it is judged as a production anomaly, and the anomaly type (such as diameter too large, uniformity exceeding the standard), anomaly degree (such as diameter deviation of 0.02mm, coefficient of variation of 6%) and anomaly location (such as specific section in the yarn length direction) are recorded.

[0084] Control command generation: The analysis unit generates targeted control commands based on the type and severity of the anomaly, employing multi-parameter linkage optimization logic.

[0085] When the yarn diameter deviation is ≥ ±0.02mm, adjust the winding machine speed. The speed adjustment range is positively correlated with the diameter deviation. Generally, for every 0.01mm diameter deviation, the speed is adjusted by ±3% to ±5%. The yarn diameter is corrected by changing the winding tension.

[0086] When the coefficient of variation of surface uniformity is ≥6%, adjust the fiber feed rate. The adjustment range of the feed rate is ±2%~±8%, and improve the surface uniformity of the yarn by stabilizing the fiber supply.

[0087] When diameter deviation and uniformity exceed the standard at the same time, the fiber feed rate should be adjusted first, and then the winding machine speed should be fine-tuned according to the adjustment effect to ensure the rationality and effectiveness of the control logic.

[0088] Control command execution and feedback: After receiving control commands, the execution units (yarn clearer, winding machine, fiber feed regulator) adjust operating parameters in real time, with an adjustment response time ≤200ms. Simultaneously, the anti-interference imaging system continuously acquires yarn images, and the data processing unit calculates quality indicators in real time, forming a closed-loop feedback until the indicators return to the acceptable threshold range. If the indicators still fail to meet the standards after three consecutive adjustments, the system issues an audible and visual alarm signal and records the abnormal data for subsequent troubleshooting.

[0089] Furthermore, the polarization state separation is achieved based on the difference in polarization characteristics between the yarn body and the fluff. The reflected signal of the yarn body has fixed polarization characteristics, while the scattering noise of the fluff is in an irregular polarization state. By distinguishing the difference in fixed polarization characteristics, the reflected signal of the yarn body is filtered, thereby achieving effective separation between the body signal and the scattering noise of the fluff.

[0090] Furthermore, the method for completing multi-view image registration and fusing to form a yarn panoramic image through feature point matching includes:

[0091] Key feature points of yarn images from each viewpoint of a multi-view polarization image sequence are extracted. These key feature points are determined based on the edge contour, structural inflection points, and continuous morphological features of the yarn body.

[0092] Select a yarn image from one perspective as the reference image, and perform consistency matching between the key feature points of the other perspective images and the corresponding feature points of the reference image to establish the spatial relationship between the perspective images.

[0093] Based on the aforementioned spatial correlation, yarn images from all perspectives are calibrated to the same spatial coordinate system to achieve multi-view image matching.

[0094] Information is integrated from the matched images at each viewpoint, retaining the effective morphological information of the yarn that is not obscured from each viewpoint, and eliminating duplicate and redundant data, ultimately forming a panoramic image of the yarn that can fully present all morphological features of the yarn in the circumferential direction.

[0095] Furthermore, the method for constructing a three-dimensional yarn morphology model by merging contour information from images at various perspectives based on the principle of multi-view contour transformation includes:

[0096] The contour information of the yarn body in the registered panoramic image of yarn from each viewpoint is extracted. The contour information includes the edge contour, cross-sectional shape and continuous extension features of the yarn. Non-main contours corresponding to fuzz and residual noise are removed.

[0097] Based on the principle of multi-view contour transformation, and combined with the spatial coordinate system relationship of each view image, the yarn contour of each view is mapped to a unified three-dimensional space, and the correspondence of each contour in the spatial depth direction is established.

[0098] The mapped multi-view contours are superimposed and merged to retain the effective structural information of the yarn body in each contour, eliminate redundant data in the overlapping area of ​​the contours, and fill in the contour details not covered under a single view to form a complete set of yarn space contours.

[0099] The yarn spatial contour set is converted into three-dimensional point cloud data. By integrating the spatial position information of the point cloud, a three-dimensional morphological model of the yarn body is constructed, which can accurately restore the three-dimensional structure, shape and extension state of the yarn body.

[0100] Furthermore, the method for calculating the yarn diameter includes:

[0101] Multiple cross sections are uniformly selected from the three-dimensional morphological model of the yarn along the yarn extension direction. The main outline of the yarn is extracted from each cross section. Residual fuzz and non-main interference parts in the outline are removed. Circle fitting is performed based on the outline edge points of each cross section. The diameter of the fitted circle is used as the yarn diameter of the corresponding cross section to ensure that the diameter calculation result fits the actual cross-sectional shape of the yarn.

[0102] Furthermore, the surface uniformity is characterized by the coefficient of variation of diameters of adjacent cross sections, specifically including:

[0103] Multiple adjacent cross sections are continuously selected from the three-dimensional morphological model of the yarn along the yarn extension direction. The yarn diameter data of each cross section is extracted to form a diameter sequence. The dispersion of this diameter sequence is calculated. The uniformity of the yarn surface is quantitatively characterized by the coefficient of variation of the diameter of adjacent cross sections. This coefficient of variation intuitively reflects the diameter fluctuation of the yarn along the length direction, ensuring the objectivity and accuracy of the uniformity evaluation.

[0104] Example 2: To implement the above method, the present invention provides an automated yarn production control system. The system includes an image acquisition unit, a data processing unit, an analysis unit, and an execution unit. Each unit works in concert to achieve full closed-loop control of the yarn production process.

[0105] System Overall Architecture: The system adopts a distributed control combined with centralized management architecture. Image acquisition and execution units are deployed on the production site, while data processing and analysis units are deployed in the control center. All units interact via industrial Ethernet, with a communication latency of ≤50ms, ensuring real-time control. The system supports multi-device collaborative operation and can simultaneously control multiple spinning production lines, exhibiting excellent scalability.

[0106] The specific structure and function of each unit:

[0107] Image acquisition unit: It has multi-view acquisition and polarization adjustment functions and is the core hardware for realizing anti-interference imaging. It includes a dual-plane mirror assembly, a polarization adjustment module, an industrial camera and a dark box.

[0108] Dual-plane mirror assembly: Fixedly installed inside the darkroom, the two mirrors are at a 72° angle. Made of high-reflectivity optical glass with a mirror flatness ≤0.001mm, it forms four evenly distributed virtual images of the yarn through reflection. These, combined with the real image, achieve 360° full-view coverage, eliminating blind spots in single-view imaging. The assembly is equipped with an adjustable bracket, allowing for fine-tuning of the mirror position according to the yarn thickness (18~36tex) to ensure optimal imaging.

[0109] Polarization control module: This module includes a polarization light source and an electrically controlled polarization filter, which work in tandem through a synchronous control module. The polarization light source uses a white LED array with a spectral range of 400~760nm, illumination uniformity ≥90%, and can switch between four polarization directions: 0°, 90°, 135°, and 180°, with a switching response time ≤10ms. The electrically controlled polarization filter is installed at the front of the industrial camera lens and can synchronously follow the light source to switch polarization directions, with a transmittance ≥85%, ensuring effective transmission of polarization signals.

[0110] Industrial Camera: A high-definition CMOS industrial camera is selected, with a resolution ≥1920×1200 pixels, a frame rate ≥30fps, a pixel size ≤3.45μm, and a dynamic range ≥60dB, ensuring the clarity and detail of the acquired images. The camera is equipped with a C-mount lens with an adjustable focal length within the range of 12~25mm to adapt to the imaging needs of yarns of different thicknesses.

[0111] Dark box: Made of light-shielding material with an internal light-absorbing coating to avoid interference from ambient light; the dark box is equipped with a pull-out slide rail for easy placement of yarn samples and equipment maintenance; it is also equipped with a temperature sensor to monitor the imaging environment temperature in real time (operating temperature range: 0~50℃) to ensure stable operation of the equipment.

[0112] Data processing unit: Used to perform polarization state separation, feature point matching, image registration and fusion, denoising and thinning, contour transformation and 3D modeling on multi-view polarized images and optical feature differences. It calculates yarn diameter and surface uniformity based on the constructed yarn 3D morphology model and is the core of the system's data processing.

[0113] Hardware components include an industrial computer and an embedded processing module. The industrial computer is equipped with an Intel Core i7-12700H CPU (clock speed ≥ 2.7GHz), 32GB DDR5 memory, and a 1TB SSD hard drive, providing powerful data processing capabilities. The embedded processing module uses an FPGA chip (model: Xilinx Zynq UltraScale+), integrating image separation, registration, fusion, modeling, and index calculation logic. It can process multi-view image data in parallel with a data processing latency of ≤100ms.

[0114] Software Functions: The software integrates self-developed image processing software, supporting functions such as image acquisition, preprocessing, 3D modeling, and index calculation. It features a user-friendly interface and can display yarn images, 3D morphology models, and quality index data in real time. The software also supports data storage and export functions, allowing users to save original images, processing results, and quality index data for easy subsequent analysis and traceability.

[0115] Analysis Unit: It pre-stores the acceptable threshold ranges for yarn diameter and surface uniformity, receives the quality indicators output by the data processing unit and compares them with the thresholds. When the indicators exceed the thresholds, it generates targeted control instructions and is the core of the system's control decision.

[0116] Hardware components include a PLC controller (model: Siemens S7-1500) and a threshold configuration module. The PLC controller has high-speed computing capabilities, with an instruction execution time of ≤0.1μs, and can process multiple channels of quality indicator data simultaneously. The threshold configuration module is equipped with an industrial touch screen (size ≥10 inches, resolution ≥1280×800), which supports users to modify and save the qualified threshold range online, making operation convenient.

[0117] Software Functions: Integrated control decision algorithm, which can accurately match the corresponding operating parameter adjustment scheme according to the type (diameter deviation, uniformity deviation) and degree (deviation size, coefficient of variation value) of the quality indicators; Supports abnormal alarm function, which can issue audible and visual alarm signals (alarm volume ≥80dB, alarm light brightness ≥500cd / m²) when the indicators continuously exceed the threshold, and record abnormal information (including abnormal time, abnormal type, abnormal data); Supports communication with the MES system in the production workshop, and can upload production quality data and equipment operating status to realize multi-process collaborative control.

[0118] Execution unit: responds to control commands and adjusts the operating parameters of production equipment. It is the actuator that realizes closed-loop control, including yarn clearer, winding machine and fiber feed regulator.

[0119] Yarn clearer: The electronic yarn clearer (model: Uster Quantum 3) can adjust the cutting threshold according to the control command. The cutting response time is ≤10ms and the cutting accuracy is ≤0.01mm. It can accurately remove excessive hairs and impurities in the yarn.

[0120] Winding machine: Equipped with a variable frequency speed control system, the speed adjustment range is 500~1500r / min, the speed adjustment accuracy is ≤±1r / min, the speed can be adjusted in real time according to the control command, and the yarn diameter can be corrected by changing the winding tension;

[0121] Fiber feed regulator: The feed roller is servo controlled, with a feed rate adjustment range of 5~20g / min and an adjustment accuracy of ≤±0.1g / min. It can stabilize the fiber supply according to the control command and improve the uniformity of the yarn surface.

[0122] Communication interface: Each actuator is equipped with an industrial Ethernet interface that supports the Modbus TCP protocol, enabling real-time data interaction with the analysis unit and ensuring rapid response and accurate execution of control commands.

[0123] Furthermore, the dual-plane mirror assembly in the image acquisition unit is fixedly set to expand the imaging angle through reflection, thereby enabling multi-view image acquisition in conjunction with an industrial camera; the polarization control module includes a polarization light source and an electrically controlled polarization filter, which work together to enhance the optical characteristic differences between the yarn body and the pile, and suppress the interference of irregular light scattering from the pile.

[0124] Furthermore, the data processing unit includes an industrial computer and an embedded processing module. The embedded processing module integrates relevant processing logic for image separation, registration, fusion, modeling, and index calculation. It can quickly process multi-view polarized image sequences, complete the construction of a three-dimensional yarn morphology model and the calculation of quality indicators, and ensure the real-time performance of data processing to meet the needs of online production control.

[0125] Furthermore, the analysis unit and the execution unit establish a connection through an industrial communication protocol. The analysis unit can modify and save the qualified threshold range online, and the generated control commands can accurately match the corresponding operating parameter adjustment scheme according to the type and degree of the quality index exceeding the standard. The execution unit includes a yarn clearer, a winding machine, and a fiber feed regulator, which can adjust the operating parameters in real time according to the control commands to ensure that the yarn quality index quickly returns to the qualified threshold range.

[0126] It should be noted that the specific methods of the system in the above embodiments have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0127] Finally, it should be noted that although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An automated yarn production control method, characterized in that, The method includes: A front-end imaging system is built, which works in conjunction with a dual-plane mirror assembly and a polarization control module to acquire multi-view polarization image sequences of yarn during the production process online, so that the yarn body and the fibers exhibit distinguishable optical feature differences due to different responses, and the optical feature differences are acquired simultaneously. Based on the multi-view polarization image sequence and optical feature differences, the polarization reflection signal of the yarn body is extracted by polarization state separation to remove the source noise generated by non-polarization scattering of the fibers; then, multi-view image registration is completed by feature point matching and fused to form a panoramic image of the yarn. Based on the multi-view polarization image sequence and optical feature differences, the polarization reflection signal of the yarn body is extracted by polarization state separation, the source noise generated by non-polarization scattering of the fluff is eliminated, and the fluff point cloud is processed in groups according to the z coordinate; based on the principle of multi-view contour transformation, the contour information of each view image is merged to construct a three-dimensional morphological model of the yarn. Based on the three-dimensional morphology model of the yarn, quality indicators are calculated in real time. The quality indicators include yarn diameter and surface uniformity. The yarn diameter is calculated by fitting a circle to the cross-sectional profile, and the surface uniformity is characterized by the coefficient of variation of the diameters of adjacent cross sections. The qualified threshold range of each indicator is set. The quality indicators are compared with the acceptable threshold range. When an indicator exceeds the acceptable threshold range, a targeted control instruction is generated. The operating parameters of the production equipment, including the yarn clearer cutting threshold, the winding machine speed, and the fiber feed rate, are adjusted by the execution unit until the indicator returns to the acceptable threshold range, so as to ensure the stability of yarn production quality.

2. The method according to claim 1, characterized in that, The polarization separation is achieved based on the difference in polarization characteristics between the yarn body and the fluff. The reflected signal of the yarn body has fixed polarization characteristics, while the scattering noise of the fluff is in an irregular polarization state. By distinguishing the difference in fixed polarization characteristics, the reflected signal of the yarn body is filtered, thereby achieving effective separation between the body signal and the scattering noise of the fluff.

3. The method according to claim 1, characterized in that, The method for registering and fusing multi-view images through feature point matching to form a panoramic image of yarn includes: Key feature points of yarn images from each viewpoint of a multi-view polarization image sequence are extracted. These key feature points are determined based on the edge contour, structural inflection points, and continuous morphological features of the yarn body. Select a yarn image from one perspective as the reference image, and perform consistency matching between the key feature points of the other perspective images and the corresponding feature points of the reference image to establish the spatial relationship between the perspective images. Based on the aforementioned spatial correlation, yarn images from all perspectives are calibrated to the same spatial coordinate system to achieve multi-view image matching. Information is integrated from the matched images at each viewpoint, retaining the effective morphological information of the yarn that is not obscured from each viewpoint, and eliminating duplicate and redundant data, ultimately forming a panoramic image of the yarn that can fully present all morphological features of the yarn in the circumferential direction.

4. The method according to claim 1, characterized in that, The method for constructing a three-dimensional yarn morphology model by merging contour information from images at various perspectives based on the principle of multi-view contour transformation includes: The contour information of the yarn body in the registered panoramic image of yarn from each viewpoint is extracted. The contour information includes the edge contour, cross-sectional shape and continuous extension features of the yarn. Non-main contours corresponding to fuzz and residual noise are removed. Based on the principle of multi-view contour transformation, and combined with the spatial coordinate system relationship of each view image, the yarn contour of each view is mapped to a unified three-dimensional space, and the correspondence of each contour in the spatial depth direction is established. The mapped multi-view contours are superimposed and merged to retain the effective structural information of the yarn body in each contour, eliminate redundant data in the overlapping area of ​​the contours, and fill in the contour details not covered under a single view to form a complete set of yarn space contours. The yarn spatial contour set is converted into three-dimensional point cloud data. By integrating the spatial position information of the point cloud, a three-dimensional morphological model of the yarn body is constructed, which can accurately restore the three-dimensional structure, shape and extension state of the yarn body.

5. The method according to claim 1, characterized in that, The method for calculating the yarn diameter includes: Multiple cross sections are uniformly selected from the three-dimensional morphological model of the yarn along the yarn extension direction. The main outline of the yarn is extracted from each cross section. Residual fuzz and non-main interference parts in the outline are removed. Circle fitting is performed based on the outline edge points of each cross section. The diameter of the fitted circle is used as the yarn diameter of the corresponding cross section to ensure that the diameter calculation result fits the actual cross-sectional shape of the yarn.

6. The method according to claim 1, characterized in that, The surface uniformity is characterized by the coefficient of variation of adjacent cross-sectional diameters, specifically including: Multiple adjacent cross sections are continuously selected from the three-dimensional morphological model of the yarn along the yarn extension direction. The yarn diameter data of each cross section is extracted to form a diameter sequence. The dispersion of this diameter sequence is calculated. The uniformity of the yarn surface is quantitatively characterized by the coefficient of variation of the diameter of adjacent cross sections. This coefficient of variation intuitively reflects the diameter fluctuation of the yarn along the length direction, ensuring the objectivity and accuracy of the uniformity evaluation.

7. An automated yarn production control system, characterized in that, The system includes: Image acquisition unit: It has multi-view acquisition and polarization control functions, including a dual-plane mirror assembly and a polarization control module, which is used to acquire multi-view polarization image sequences of yarn online and simultaneously output the multi-view polarization image sequences and optical feature differences; Data processing unit: used to perform polarization state separation, feature point matching, image registration and fusion, denoising and thinning, contour transformation and 3D modeling processing on multi-view polarized images and optical feature differences, and calculate yarn diameter and surface uniformity based on the constructed yarn 3D morphology model; Analysis unit: Pre-stores qualified threshold ranges for yarn diameter and surface uniformity, receives quality indicators output by the data processing unit and compares them with the thresholds, and generates targeted control instructions when the indicators exceed the thresholds; Execution unit: responds to control commands to adjust the operating parameters of the production equipment, including the yarn clearer cutting threshold, the winding machine speed and the fiber feeding amount, to realize closed-loop feedback control of the production process.

8. The system according to claim 7, characterized in that, The dual-plane mirror assembly in the image acquisition unit is fixedly installed and expands the imaging angle through reflection, working with an industrial camera to achieve multi-view image acquisition; the polarization control module includes a polarization light source and an electrically controlled polarization filter, which work together to enhance the optical characteristic differences between the yarn body and the fluff, suppress the interference of irregular light scattering from the fluff, and ensure the effectiveness of the multi-view polarization image sequence and optical characteristic differences.

9. The system according to claim 7, characterized in that, The data processing unit includes an industrial computer and an embedded processing module. The embedded processing module integrates relevant processing logic for image separation, registration, fusion, modeling, and index calculation. It can quickly process multi-view polarized image sequences, complete the construction of yarn three-dimensional morphology models and the calculation of quality indicators, and ensure the real-time performance of data processing to meet the needs of online production control.

10. The system according to claim 7, characterized in that, The analysis unit and the execution unit are connected via an industrial communication protocol. The analysis unit can modify and save the qualified threshold range online, and the generated control commands can accurately match the corresponding operating parameter adjustment scheme according to the type and degree of the quality index exceeding the standard. The execution unit includes a yarn clearer, a winding machine, and a fiber feed regulator, which can adjust the operating parameters in real time according to the control commands to ensure that the yarn quality index quickly returns to the qualified threshold range.

Images