Machine vision-based elasticizer yarn hairiness online detection system and method

Abstract

The invention discloses an elasticizer yarn hairiness online detection system and method based on machine vision, and relates to the technical field of textile automatic detection. The system comprises a pneumatic yarn guide cleaning unit, a dual-band chromatic aberration imaging unit and an image processing unit. The pneumatic yarn guide cleaning unit utilizes an airflow guide structure to generate high-speed laminar flow parallel to the optical window, and the high-speed laminar flow performs the functions of cleaning the window and restraining the yarn in the air cushion layer by utilizing fluid dynamic pressure at the same time. The dual-band chromatic aberration imaging unit enables the first band to be focused on the window surface and the second band to be focused on the air cushion layer by using the axial chromatic aberration characteristic. The image processing unit quantifies attachment features on the window surface by using the first waveband data, and performs pixel-level suppression on the second waveband data according to the attachment features. Through physical coupling of pneumatic constraint and spectrum layering, the problem of imaging detection caused by coupling of yarn shaking and stain interference in a high-speed oil mist environment of the elasticizer is solved.

Description

Technical Field

[0001] This invention relates to the field of automated textile inspection technology, and in particular to an online detection system and method for yarn hairiness in texturing machines based on machine vision. Background Technology

[0002] In the high-speed production scenario of chemical fiber texturing machines, online detection systems face a single yet complex imaging technology challenge: how to achieve non-contact, high signal-to-noise ratio microscopic imaging in a complex environment filled with high-concentration oil mist interference and where the target object is in a state of high-speed random micro-motion.

[0003] Specifically, in order to capture micron-level hair-like features, the optical system must have high resolution, which physically determines its extremely shallow depth of field. However, the yarn in the production environment inevitably experiences random positional drift beyond this depth of field during high-speed operation, causing the image to go out of focus; at the same time, spinning oils in the environment splash onto the optical interface, forming stains that exhibit optical features in the image that are extremely similar to those of the out-of-focus yarn or hair.

[0004] This spatiotemporal coupling of "positional uncertainty (jitter)" and "environmental noise (oil stains)" renders existing single methods ineffective: mechanical restraints cannot solve the interference of oil stains and damage the yarn, while conventional optical imaging cannot distinguish between stains and lint when the depth of field is insufficient. Therefore, there is an urgent need for an integrated detection solution that can simultaneously lock the spatial position of the target and optically separate the environmental background noise. Summary of the Invention

[0005] This invention provides an online detection system and method for yarn hairiness in texturing machines based on machine vision. It aims to solve the spatiotemporal mismatch between the high-resolution shallow depth-of-field characteristics required for microscopic imaging and the large-scale random shaking of yarn and environmental oil mist interference in the high-speed production environment of texturing machines, and to achieve non-contact stable imaging detection.

[0006] In view of the above problems, the present invention provides an online detection system for yarn hairiness in texturing machines based on machine vision, comprising: A pneumatic yarn guiding and cleaning unit includes a cavity with a through yarn channel and an optical window embedded on one side of the yarn channel; the cavity is provided with an airflow guiding structure for generating a high-speed laminar flow parallel to the inner surface of the optical window; the high-speed laminar flow is configured to clean the optical window and use hydrodynamic pressure to constrain the yarn passing through the yarn channel within an air cushion layer at a predetermined distance from the inner surface of the optical window; A dual-band chromatic aberration imaging unit, with its optical axis perpendicular to the optical window, includes a multi-band light source and a fixed-focus chromatic aberration lens. The fixed-focus chromatic aberration lens has a preset axial chromatic aberration amount, such that the focal plane of the first band light from the multi-band light source is located on the inner surface of the optical window, and the focal plane of the second band light from the multi-band light source is located at the center of the air cushion layer. The image processing unit, connected to the dual-band chromatic aberration imaging unit, is configured to acquire an original image containing first-band data and second-band data, quantify the features of the attachments on the surface of the optical window using the first-band data, and perform pixel-level suppression correction on the second-band data based on the attachment features to output a yarn feature image.

[0007] Furthermore, the airflow guiding structure is configured to have an airflow channel that is smoothly tangential to the inner surface of the optical window; the airflow channel is used to guide the high-speed laminar flow tangentially to the inner surface of the optical window, thereby forming the air cushion layer by utilizing the fluid adhesion effect.

[0008] Furthermore, the multi-band light source employs synchronous strobe control; the first band light is red light with a center wavelength of 620nm to 640nm, and the second band light is green light with a center wavelength of 520nm to 540nm; the preset distance is 50 micrometers to 150 micrometers.

[0009] Furthermore, the image processing unit is configured to perform the pixel-level suppression correction through the following steps: Calculate the gradient plot of the first band data; Construct a suppression coefficient matrix based on the gradient graph; The suppression coefficient matrix is ​​applied to the second band data.

[0010] Furthermore, the pixels in the suppression coefficient matrix inhibition coefficient Calculated by the following formula: in, It is the hyperbolic tangent function. To suppress sensitivity factors, For the first band data in The normalized gradient value at that point; The pixel values ​​of the yarn feature image Calculated by the following formula: in, This represents the original intensity value of the second band data.

[0011] Furthermore, the normalized gradient value Calculated by the following formula: in, The magnitude of the Sobel gradient. This is the gradient cutoff threshold.

[0012] Furthermore, the image processing unit is also configured to perform a self-test step: periodically calculate the average gradient value of the first band data, and trigger an alarm signal when the average gradient value exceeds a preset dirt threshold.

[0013] The present invention also provides a machine vision-based online detection method for yarn hairiness in texturing machines, which is executed using the above-mentioned system and includes: S1. Control the pneumatic yarn guiding and cleaning unit to generate a high-speed laminar flow that adheres to the optical window, clean the optical window and form the air cushion layer that carries the yarn; S2. Control the dual-band chromatic aberration imaging unit to perform exposure, wherein the first band is focused on the optical window surface for imaging, and the second band is focused on the air cushion layer for imaging. S3. Calculate the gradient features of the first band image to quantify the degree of interference from the attachments on the surface of the optical window; S4. Using the gradient features, generate suppression weights and perform masking on the second band image to generate a yarn image; S5. Binarize and extract the skeleton of the yarn image, and calculate the hairiness index.

[0014] The technical solution provided in this application has at least the following technical effects: This invention solves the imaging problem under the above-mentioned combined interference by constructing a pneumatic-optical coupled detection field. The single high-speed laminar flow generated by the pneumatic yarn guiding and cleaning unit physically completes the continuous cleaning of the optical interface and the flexible compression of the yarn's Z-axis degree of freedom simultaneously, forcibly locking the yarn within a micrometer-thick air cushion layer, thereby providing the necessary spatial determinism for high-magnification imaging.

[0015] Based on this physical foundation, the dual-band chromatic aberration imaging unit utilizes axial chromatic aberration characteristics to map the physically layered "clean yarn (cushion layer)" and "residual stains (viewing window surface)" into a dual-band image with optically separated focal planes. The image processing unit then uses this optical layering information and a nonlinear suppression algorithm to precisely remove stain noise at the data level.

[0016] This solution achieves deep synergy between physical positioning, optical layering, and algorithmic denoising, ensuring that the detection system can continuously acquire clear, interference-free intrinsic yarn images in harsh industrial environments without contacting or damaging the yarn. Attached Figure Description

[0017] Figure 1This is a schematic diagram of the overall structure of the machine vision-based online detection system for yarn hairiness in texturing machines provided in this embodiment of the invention; Figure 2 A flowchart of an online detection method for yarn hairiness in a texturing machine based on machine vision, provided in an embodiment of the present invention. Detailed Implementation

[0018] The above technical solutions will now be described in detail with reference to the accompanying drawings and specific embodiments to provide a better understanding of them. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. It should be understood that the present invention is not limited to the exemplary embodiments used only to explain the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention. Furthermore, it should be noted that, for ease of description, only the parts related to the present invention are shown in the drawings, not all of them. Example

[0019] like Figure 1 and Figure 2 As shown, this embodiment provides an online detection system for yarn hairiness in texturing machines based on machine vision. This online detection system includes a pneumatic yarn guiding and cleaning unit, a dual-band color difference imaging unit, and an image processing unit.

[0020] The pneumatic yarn guiding and cleaning unit includes a cavity with a through-yarn channel and an optical window embedded in one side of the yarn channel. An airflow guiding structure is installed within the cavity to generate a high-speed laminar flow parallel to the inner surface of the optical window. This high-speed laminar flow is configured to have a dual function: first, to clean the optical window; and second, to use hydrodynamic pressure to confine the yarn passing through the yarn channel within an air cushion layer at a predetermined distance from the inner surface of the optical window.

[0021] The optical axis of the dual-band chromatic aberration imaging unit is perpendicular to the optical window. The dual-band chromatic aberration imaging unit includes a multi-band light source and a fixed-focus chromatic aberration lens. The fixed-focus chromatic aberration lens has a preset axial chromatic aberration amount. This axial chromatic aberration amount ensures that the focal plane of the first band of light from the multi-band light source is located on the inner surface of the optical window, and that the focal plane of the second band of light from the multi-band light source is located at the center of the air cushion layer.

[0022] The image processing unit is connected to the dual-band chromatic aberration imaging unit. The image processing unit is configured to acquire a raw image containing first-band data and second-band data, quantify the features of the adhering material on the surface of the optical window using the first-band data, and perform pixel-level suppression correction on the second-band data based on the adhering material features to output a yarn feature image.

[0023] Based on the above system structure, this embodiment also implements a machine vision-based online detection method for yarn hairiness in a texturing machine. The image processing unit controls the pneumatic yarn guiding and cleaning unit and the dual-band color difference imaging unit to achieve online detection of yarn hairiness through physical field construction, light and color layer acquisition, and anti-interference algorithm flow.

[0024] The implementation of this system and method begins with the physical construction of the pneumatic yarn guiding and cleaning unit and the establishment of a steady-state fluid field. The main body of the pneumatic yarn guiding and cleaning unit is made of stainless steel or aluminum alloy, and its interior features a through-yarn channel with a diameter of 5 mm. A sapphire optical window is installed on one side wall of the yarn channel using an embedded sealing process. To ensure a smooth flow field, the inner surface of the optical window is geometrically flush with the inner wall of the yarn channel.

[0025] The precise assembly of the airflow guiding structure determines the morphology of the flow field. A slit nozzle is positioned at the inlet of the yarn channel, and its outlet width is configured to a specific value within the range of 0.2 mm to 0.5 mm. A curved guide vane connects the outlet of the slit nozzle to the inner surface of the optical window. The curved guide vane has a geometric profile of an arc surface with a radius of 20 mm. One end of the curved guide vane is tangential to the jet direction of the slit nozzle, while the other end is smoothly tangential to the inner surface of the optical window. This specific geometric connection ensures that the airflow is deflected along the curved guide vane and ultimately enters the optical window surface tangentially.

[0026] After the external high-pressure air source is turned on, clean compressed air with a stable pressure of 0.5 MPa enters the slit nozzle. The compressed air is accelerated through the slit nozzle and ejected with an initial velocity greater than 30 m / s. Due to the entrainment of surrounding fluid by the jet stream, a negative pressure is generated on one side of the curved guide plate, and the high-speed airflow flows closely along the curved guide plate. The high-speed airflow flowing closely along the curved guide plate eventually forms a tangential high-speed laminar flow along the inner surface of the optical window.

[0027] High-speed laminar flow continuously washes over the inner surface of the optical window, using fluid shear force to remove any oil mist particles that may be adhering to the surface. With the establishment of this high-speed laminar flow, the hydrodynamic environment within the yarn channel changes. As the yarn output from the texturing machine passes through the yarn channel, it is subjected to the hydrodynamic pressure generated by the high-speed laminar flow. This hydrodynamic pressure exerts a lateral force on the yarn, pointing towards the optical window. The gas viscosity creates a boundary layer near the inner surface of the optical window, generating viscous resistance.

[0028] When the lateral force generated by the hydrodynamic pressure balances the viscous resistance generated by the boundary layer, the yarn is constrained within an air cushion layer ranging from 50 to 150 micrometers from the inner surface of the optical window. Before the air cushion layer forms, the vertical jiggling amplitude of the yarn within the yarn channel is approximately 3 millimeters. After the air cushion layer forms, the vertical displacement of the yarn is physically limited to within 0.1 millimeters.

[0029] After the physical field is constructed, the conjugate calibration of the dual-band chromatic aberration imaging unit is a prerequisite for achieving high-precision layered detection. The fixed-focus chromatic aberration lens is mounted perpendicular to the optical axis of the optical window. By mechanically adjusting the working distance between the fixed-focus chromatic aberration lens and the optical window, a conjugate mapping relationship between light wavelength and spatial depth can be established. The calibration process utilizes microscopic dust or scratches on the surface of the optical window as a physical reference. When the image texture gradient value formed by the first band light with a center wavelength of 630 nm on the inner surface of the optical window reaches its maximum, the position of the fixed-focus chromatic aberration lens is locked.

[0030] At this point, the physical position of the fixed-focus chromatic aberration lens ensures that the focal plane of the first-band light strictly coincides with the inner surface of the optical window. Based on the axial chromatic aberration characteristics preset by the fixed-focus chromatic aberration lens, while the first-band light is focused on the inner surface of the optical window, the focal plane of the second-band light, with a center wavelength of 530 nm, automatically shifts to a spatial position 100 micrometers away from the inner surface of the optical window. This 100-micrometer spatial position geometrically coincides with the central region of the air cushion layer formed by the pneumatic yarn guiding and cleaning unit. This optical path configuration ensures that the yarn located within the air cushion layer can be clearly imaged by the second-band light, while the oil stains adhering to the surface of the optical window are clearly imaged by the first-band light.

[0031] Image data acquisition follows microsecond-level synchronous timing logic to adapt to the high-speed movement characteristics of the yarn. After the detection task is initiated, the image processing unit sends a trigger signal according to a preset sampling frequency. Upon receiving the trigger signal, the red and green LED components inside the multi-band light source are simultaneously activated and illuminated. The duration of a single illumination by the red and green LED components is controlled by hardware circuitry to 20 microseconds. For a yarn moving at 800 meters per second, the 20-microsecond exposure time limits the yarn's displacement to within 16 micrometers, a displacement less than the equivalent of one pixel in the imaging system, thus eliminating motion blur.

[0032] Within a 20-microsecond window of illumination from the multi-band light source, the photosensors inside the dual-band chromatic aberration imaging unit synchronously perform exposure operations. The photosensors convert the received photon energy into digital signals, outputting a single-frame raw image containing data components from the first and second bands. The raw image data is transmitted to the image processing unit in Bayer array format via a high-speed data interface.

[0033] After receiving the high-speed transmitted single-frame raw image, the image processing unit first performs data unpacking and channel separation operations. Since the raw image is arranged in a Bayer array format, it internally stores spectral energy information of different bands. The image processing unit uses a color channel separation algorithm to decompose the raw image into independent two-dimensional data matrices. Data corresponding to the red light spectral component is extracted and reassembled into a first-band data matrix, which physically represents the imaging information of the optical window surface. Data corresponding to the green light spectral component is extracted and reassembled into a second-band data matrix, which physically represents the imaging information of the yarn within the cushion layer. The first-band data matrix and the second-band data matrix maintain a one-to-one spatial correspondence in the pixel coordinate system.

[0034] After obtaining the first-band data matrix, the image processing unit performs gradient quantization calculations on it to identify and quantify the features of deposits on the optical window surface. This calculation uses a 3x3 Sobel convolution operator. The Sobel convolution operator slides pixel-by-pixel across the first-band data matrix, calculating the grayscale gradient components in both the horizontal and vertical directions for each pixel, and then takes the square root of the sum of the squares of the two components to obtain the gradient magnitude. The magnitude of the gradient magnitude directly reflects whether there are oil or dust edges in focus at the corresponding location.

[0035] To unify the gradient magnitudes under different lighting conditions to a standard dimensionless range and prevent numerical overflow in subsequent calculations, the image processing unit calculates the first band data in... Normalized gradient value at The normalized gradient value Calculated by the following formula: in, For the first band data in Sobel gradient magnitude at point, In this embodiment, the gradient cutoff threshold is used. The value is set to 255. The normalized gradient values ​​of all pixels in the first-band data matrix together form a normalized gradient map. In the normalized gradient map, the bright spots with values ​​close to 1 physically correspond precisely to the edges of the oil stains attached to the surface of the optical window and clearly imaged by the first-band light.

[0036] Based on the normalized gradient map, the image processing unit constructs a suppression coefficient matrix. The construction process follows non-linear mapping logic, and the image processing unit calculates the pixel values ​​in the suppression coefficient matrix. inhibition coefficient The inhibition coefficient Calculated by the following formula: in, It is the hyperbolic tangent function. To suppress sensitivity factors, For the first band data in The normalized gradient value at that point. Suppression sensitivity factor. This is configured as a constant during the algorithm initialization phase to adjust the response rate of the suppression coefficient matrix to gradient changes. A specific numerical example demonstrates the ability of this nonlinear mapping process to distinguish different regions. For instance, the suppression sensitivity factor is set... The value is 4. For the clean area of ​​the optical window surface, the texture changes captured by the image sensor are weak, and the normalized gradient value is... The measured value is 0.01. Substituting this value into the formula, the hyperbolic tangent function... The output value approaches 0, and the calculated suppression coefficient... Approaching 1.0. For the oily edge region on the optical window surface, the image sensor captured significant texture abrupt changes, and the normalized gradient value... The measured value is 0.6. Substituting this value into the formula, the independent variable of the hyperbolic tangent function becomes 2.4. The output value is approximately 0.98, and the calculated inhibition coefficient is... It is 0.02.

[0037] After obtaining the suppression coefficient matrix, the image processing unit performs pixel-level mask fusion to generate the final yarn feature image. The pixel values ​​of the yarn feature image... Calculated by the following formula: in, This represents the original intensity value of the second band data. Based on the results of the aforementioned numerical examples, in the clean area, the original brightness signal representing the yarn body in the second band data is multiplied by a coefficient of 1.0, and the optical characteristics of the yarn are completely preserved in the yarn feature image. In the oily edge area, the original brightness signal representing oily interference in the second band data is multiplied by a coefficient of 0.02, and the oily interference signal that was originally overlapping or confused on the yarn image is forcibly attenuated to a gray level close to zero. Through this pixel-level suppression correction, the system preserves the yarn morphology in the green light channel while eliminating the interference from physical attachments identified by the red light channel.

[0038] It is important to note that this invention utilizes the high-frequency information attenuation characteristic caused by defocus blur. Because the first band of a fixed-focus chromatic aberration lens is focused on the viewfinder surface, while the yarn is constrained by the air cushion layer at a distance of 50-150 micrometers from the viewfinder, this distance far exceeds the depth of field range of high-magnification lenses. Therefore, in the first band image, the yarn in the background is severely out of focus, with its edges exhibiting a smooth, blurred transition (low gradient feature); while the stains adhering to the viewfinder surface are on the focal plane, exhibiting sharp edges (high gradient feature). Based on this frequency domain difference caused by optical defocus, the algorithm can accurately distinguish between stains and yarn simply by normalizing the gradient values, achieving physical-level segmentation without the need for complex deep learning.

[0039] After the yarn feature image is generated, the image processing unit initiates the morphology extraction and quantization analysis program. Based on the grayscale histogram distribution characteristics of the yarn feature image, the Otsu method is used to automatically calculate the optimal segmentation threshold. Based on the optimal segmentation threshold, the pixels in the yarn feature image are divided into foreground objects and background noise, thus generating a black-and-white binary image.

[0040] The binary image is then processed by a morphological thinning algorithm, shrinking the original optical projection of the yarn, which had a certain width, into a connected skeleton trajectory with a single pixel width. The connected skeleton trajectory preserves the topological information of the yarn body and its surface protrusions. Based on the connected skeleton trajectory, the system executes geometric measurement logic. The image processing unit analyzes the node connections of the connected skeleton trajectory, and the longest continuous path is identified as the yarn backbone. Branch paths extending from the nodes of the yarn backbone are marked as candidate fuzz features. According to the spatial resolution parameters of the dual-band chromatic aberration imaging unit, the pixel length of the candidate fuzz features in the image coordinate system is converted to millimeters in physical space. Branch paths with a physical length greater than 2 millimeters are determined as valid fuzz defects. The system counts the number of valid fuzz defects within a single detection field of view and accumulates the physical lengths of all valid fuzz defects to calculate the cumulative total fuzz length per unit yarn length, which is defined as the fuzz index H value. The fuzz index H value serves as the final quality assessment criterion.

[0041] The image processing unit internally stores a quality grading threshold table based on textile industry standards. The calculated hairiness index H value is compared step-by-step with the values ​​in the quality grading threshold table. Based on the comparison results, the quality status of the currently tested yarn segment is classified as superior, first-class, or second-class. The corresponding quality grade code signal is sent to an external programmable logic controller (PLC) via an industrial fieldbus interface. The PLC, based on the received quality grade code signal, controls the texturing machine's winding drum switching action or drives the marking mechanism to mark the quality of specific spindles.

[0042] The long-term operational stability of the system relies on a periodic self-checking mechanism. An internal dirt level monitoring program runs within the image processing unit. The statistical period of this monitoring program is exemplarily set to every 1000 frames. At the end of each statistical period, the image processing unit iterates through all first-band data acquired within that period. The gradient magnitudes of all pixels in the first-band data are accumulated and divided by the total number of pixels to calculate the average gradient value, characterizing the overall density of the current surface deposits on the optical window. The numerical determination logic based on this average gradient value determines the system's maintenance status.

[0043] The image processing unit compares the calculated average gradient value with a preset dirt threshold. In one possible implementation, the preset dirt threshold is configured to a value of 50. When the calculated average gradient value is less than or equal to the preset dirt threshold, the system maintains normal detection. When the calculated average gradient value is greater than the preset dirt threshold, the image processing unit generates a maintenance alarm signal. The maintenance alarm signal is transmitted to the central control system of the texturing machine via a communication interface. In response to the maintenance alarm signal, the central control system of the texturing machine immediately stops the yarn suction action and outputs a prompt message requesting manual cleaning and maintenance of the optical window.

[0044] Those skilled in the art will understand that the specific numerical values ​​(such as yarn channel diameter, airflow velocity, wavelength, threshold parameters, etc.) involved in the above embodiments are merely examples used to illustrate the implementation principle of the present invention, and are not intended to limit the present invention. Any numerical adjustments, equivalent substitutions, or improvements made based on the technical solutions of the present invention without departing from the concept of the present invention should be included within the scope of protection of the present invention.

[0045] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A machine vision-based online detection system for yarn hairiness in texturing machines, characterized in that, include: A pneumatic yarn guiding and cleaning unit includes a cavity with a through yarn channel and an optical window embedded on one side of the yarn channel; the cavity is provided with an airflow guiding structure for generating a high-speed laminar flow parallel to the inner surface of the optical window; the high-speed laminar flow is configured to clean the optical window and use hydrodynamic pressure to constrain the yarn passing through the yarn channel within an air cushion layer at a predetermined distance from the inner surface of the optical window; A dual-band chromatic aberration imaging unit, with its optical axis perpendicular to the optical window, includes a multi-band light source and a fixed-focus chromatic aberration lens. The fixed-focus chromatic aberration lens has a preset axial chromatic aberration amount, such that the focal plane of the first band light from the multi-band light source is located on the inner surface of the optical window, and the focal plane of the second band light from the multi-band light source is located at the center of the air cushion layer. The image processing unit, connected to the dual-band chromatic aberration imaging unit, is configured to acquire an original image containing first-band data and second-band data, quantify the features of the attachments on the surface of the optical window using the first-band data, and perform pixel-level suppression correction on the second-band data based on the attachment features to output a yarn feature image.

2. The system according to claim 1, characterized in that, The airflow guiding structure is configured to have an airflow channel that is smoothly tangential to the inner surface of the optical window; the airflow channel is used to guide the high-speed laminar flow tangentially to the inner surface of the optical window, thereby forming the air cushion layer by utilizing the fluid adhesion effect.

3. The system according to claim 1, characterized in that, The multi-band light source employs synchronous strobe control; the first band light is red light with a center wavelength of 620nm to 640nm, and the second band light is green light with a center wavelength of 520nm to 540nm; the preset distance is 50 micrometers to 150 micrometers.

4. The system according to claim 1, characterized in that, The image processing unit is configured to perform the pixel-level suppression correction through the following steps: Calculate the gradient plot of the first band data; Construct a suppression coefficient matrix based on the gradient graph; The suppression coefficient matrix is ​​applied to the second band data.

5. The system according to claim 4, characterized in that, The pixels in the suppression coefficient matrix inhibition coefficient Calculated by the following formula: in, It is the hyperbolic tangent function. To suppress sensitivity factors, For the first band data in The normalized gradient value at that point; The pixel values ​​of the yarn feature image Calculated by the following formula: in, This refers to the original intensity value of the second band data; The normalized gradient value Calculated by the following formula: in, The magnitude of the Sobel gradient. This is the gradient cutoff threshold.

6. The system according to claim 1, characterized in that, The image processing unit is also configured to perform a self-test step: periodically calculate the average gradient value of the first band data, and trigger an alarm signal when the average gradient value exceeds a preset dirt threshold.

7. A machine vision-based online detection method for yarn hairiness in texturing machines, characterized in that, The system applied to any one of claims 1-6 comprises: S1. Control the pneumatic yarn guiding and cleaning unit to generate a high-speed laminar flow that adheres to the optical window, clean the optical window and form the air cushion layer that carries the yarn; S2. Control the dual-band chromatic aberration imaging unit to perform exposure, wherein the first band is focused on the optical window surface for imaging, and the second band is focused on the air cushion layer for imaging. S3. Calculate the gradient features of the first band image to quantify the degree of interference from the attachments on the surface of the optical window; S4. Using the gradient features, generate suppression weights and perform masking on the second band image to generate a yarn image; S5. Binarize and extract the skeleton of the yarn image, and calculate the hairiness index.

Images