Air-jet vortex spinning yarn image acquisition device and wrapping effect detection method

Abstract

The invention discloses an air-jet vortex spinning yarn image acquisition device and a wrapping effect detection method, and belongs to the technical field of air-jet vortex spinning yarn quality detection.The air-jet vortex spinning yarn image acquisition method comprises the steps that in the image acquisition stage, two light sources with the irradiation angles parallel to a core fiber and not parallel to a wrapping fiber are arranged to enhance light and shade contrast of a wrapping area and improve visibility of the wrapping area; on the basis, the wrapping-core yarn region segmentation neural network model and the detection frame generation head of the candidate wrapping region are used for accurately segmenting the wrapping region and the core yarn region of the preprocessed air-jet vortex spinning yarn image, so that the detection accuracy is improved. The wrapping coefficient, the short wrapping rate, the long core exposure rate and the wrapping angle of the air-jet vortex spinning yarn are calculated according to the segmentation result so as to evaluate the yarn wrapping effect, and the defect that a traditional method is insufficient in detection of a wrapping structure is overcome. An efficient and objective wrapping effect detection means is provided, the production process can be optimized in the textile industry, and the yarn quality and the production efficiency are improved.

Description

Technical Field

[0001] The invention relates to an image acquisition device for jet vortex spinning yarn and a wrapping effect detection method, belonging to the technical field of jet vortex spinning yarn quality detection. Background Art

[0002] At present, as a new type of spinning technology, jet vortex spinning technology has become a key development direction of the textile industry with its remarkable automation and intelligent characteristics. Jet vortex spinning technology uses compressed air to form a high-speed airflow, and twists the fibers to form a special structure in which the outer layer of fibers wraps the inner layer of fibers, which significantly improves the strength and wear resistance of the yarn and reduces the generation of hairiness. Since the wrapping structure of the yarn is crucial to the quality of the yarn, the rapid and accurate detection of the wrapping effect has become an important means to optimize the production process and improve the quality of the yarn. However, most of the existing strength, evenness, hairiness, and twist detection methods use the ring spinning yarn detection method, and the structure of the ring spinning yarn is essentially different from that of the jet vortex spinning yarn. Therefore, the existing yarn quality detection is not suitable for the jet vortex spinning yarn with a special structure, and there are disadvantages such as inconsistent detection standards, low detection efficiency, few detection samples, and large errors in the detection results.

[0003] The paper "Discussion on the Image-Based Test Method for Twist of Jet Vortex Spinning Yarn" by Sun Zhihao, Li Hao, etc. uses the image method to measure the twist angle based on the ratio of the inner and outer layers of the yarn cross section and the appearance image to test the twist of the jet vortex spun yarn. However, there are relatively few studies on the research methods of yarn wrapping structure and the evaluation indicators of wrapping effect. The paper "The Effect of Wrapping on the Breaking Strength of Jet Vortex Spinning Coarse Special Yarn" by Luo Caihong, etc. uses an ultra-depth microscope and PS software to manually divide the wrapped area and the unwrapped area, proposes a quantitative indicator wrapping coefficient, and verifies the relationship between the breaking strength of jet vortex spinning coarse special yarn and the wrapping coefficient. However, the equipment used is not suitable for first-line production testing, is not generalizable, and has low image acquisition efficiency. The manual annotation workload using PS software is large and the proposed quantitative wrapping effect indicators are relatively one-sided. Summary of the invention

[0004] In order to solve the above technical problems, the present invention provides an image acquisition device for jet vortex spinning yarn and a wrapping effect detection method, and the technical solution is as follows:

[0005] As one aspect of the present invention, there is provided a jet vortex spinning line image acquisition device, comprising: a base, on which a yarn running platform and an optical axis are fixedly arranged, on which a camera, a lens, a lens holder and a support rod are fixedly arranged in sequence axially from top to bottom on the optical axis, the camera, the lens, the lens holder and the support rod are all located directly above the yarn running platform, and a first light source holder and a second light source holder are fixedly arranged at both ends of the support rod, respectively, the first light source holder is fixedly connected to a first light source, the second light source holder is fixedly connected to a second light source, the first light source and the second light source are symmetrically arranged on both sides of the yarn running platform, and illuminate the yarn running platform at an angle parallel to the core fiber and not parallel to the wrapping fiber.

[0006] As another aspect of the present invention, a method for detecting the wrapping effect of jet vortex spinning yarn is provided, comprising:

[0007] S100, continuously collecting the jet vortex spinning yarn wrapping image I according to a preset frequency by the jet vortex spinning yarn image collection device according to claim 1 i ,i∈[1,n];

[0008] S200, wrapping image I for each jet vortex spinning yarn i Perform image preprocessing to remove noise and background and extract the main area of ​​the yarn B i ;

[0009] S300, using the wrapping-core yarn area segmentation neural network model, from each jet vortex spinning yarn wrapping image I i The yarn trunk area B i Identify the wrapping area R i k Core yarn area C i k ;

[0010] S400, wrapping area R according to all the jet vortex spinning yarn wrapping images i k Core yarn area C i k , calculate the wrapping coefficient R, short wrapping rate S, long exposed core rate L, and wrapping angle A of the jet vortex spinning yarn;

[0011] S500, judging the wrapping effect of the jet vortex spun yarn according to the wrapping coefficient R, the short wrapping rate S, the long exposed core rate L, and the wrapping angle A of the jet vortex spun yarn, and obtaining the jet vortex spun yarn wrapping effect detection result.

[0012] Further, step S300 includes:

[0013] S310, sequentially performing local convolution, nonlinear activation and pooling operations in the image feature extraction module on the yarn trunk region B iPerform feature extraction and separate the yarn trunk area B i Converted to a high-dimensional feature image F(I i );

[0014] S320, generating a high-dimensional feature image F(I i ) generates a preset number of candidate frames for each candidate enveloping area, and obtains a candidate frame set P i , and obtain the candidate box set P i The location information of the candidate box in the image includes the center point coordinates (x c ,y c ), width w, angle θ and step size λ;

[0015] S330, the high-dimensional feature image F(I i ) and the candidate box set P i The position information of each candidate box in is input into the regional feature extraction module to extract the candidate box set P i The features of the regions corresponding to each candidate box in the output are the aligned region feature maps F corresponding to each candidate box. fs (I i );

[0016] S340: Set the candidate box set P i Input to the region screening module, remove the candidate boxes whose overlapping parts corresponding to each candidate enveloping area exceed the set IoU threshold, and obtain the intermediate candidate box set P i1 , and select the intermediate candidate box with the largest confidence corresponding to each candidate enveloping area, and output the valid candidate box set P i2 ;

[0017] S350: Set the valid candidate box set P i2 and high-dimensional feature image F(I i ) Input region calibration module, optimize the center point coordinates, width, angle and step size of the candidate box through regression network, and output the calibrated valid candidate box set P c ;

[0018] S360: The calibrated valid candidate frame set P c Input to the pixel classification module to identify the regional feature map F through pixel-level classification fs (I i ) belongs to the entangled region, and the Softmax function is used to convert the category score of each pixel into a probability, output the entangled region segmentation mask, and select the most likely category as the prediction result to identify the entangled region R i k , the remaining area is the core yarn area C i k .

[0019] Furthermore, the angle θ of the candidate box is fitted by minimizing the error term, and the angle θ is calculated by the inverse tangent function, θ∈[-170°,170°];

[0020] The step size λ of the candidate box is calculated based on the width w of the candidate box, the height h of the candidate box, and the angle θ of the candidate box. The calculation method of λ is:

[0021]

[0022] λ=min(S w ,S h )·f(θ)

[0023] Among them, L(a θ ,b θ ) represents the loss function, x m Indicates the horizontal coordinate of the point in the candidate box, y m Indicates the ordinate of the point in the candidate box, a θ and b θ Both represent parameters in the linear function, S w Represents a high-dimensional feature image F(I i ) to the width of the candidate box, S h Represents a high-dimensional feature image F(I i ) to the height of the candidate box, and f(θ) represents a fixed adjustment factor set according to the rotation angle θ.

[0024] Furthermore, the calculation formula of the wrapping coefficient R is:

[0025]

[0026] Among them, Count represents the number of pixels in the statistical area. Represents all jet vortex spinning yarn wrapping images I i The wrapping area R i k The sum of the areas, Represents all jet vortex spinning yarn wrapping images I i The yarn trunk area B i The sum of the areas.

[0027] Furthermore, the calculation formula of the short wrapping rate S is:

[0028]

[0029] Among them, Count represents the number of pixels in the statistical area. Indicates the width of the wrapping area, that is, the area of ​​the wrapping area Specific yarn height h i , τ represents the short wrapping threshold, τ∈[0.1,0.6], Indicates width Less than τ times the yarn trunk width W i The wrapping area, that is, the short wrapping area Represents all jet vortex spinning yarn wrapping images I i Width in Less than τ times the yarn trunk width W i The wrapping area R i k The sum of the areas, Represents all jet vortex spinning yarn wrapping images I i The yarn trunk area B i The sum of the areas.

[0030] Furthermore, the calculation formula of the long exposed core rate L is:

[0031]

[0032] Among them, Count represents the number of pixels in the statistical area. Indicates the width of the core yarn area, that is, the area of ​​the core yarn area Specific yarn height h i The value of γ represents the long dew core threshold, γ∈[0.1,0.6], Indicates width More than γ times the yarn trunk area width W i The exposed core area, that is, the long exposed core area Represents all jet vortex spinning yarn wrapping images I i Width in The sum of the exposed core areas exceeding γ times the main yarn width, Represents all jet vortex spinning yarn wrapping images I i The yarn trunk area B i The sum of the areas.

[0033] Furthermore, the calculation formula of the wrapping angle A is:

[0034]

[0035] in, Indicates the width centerline of each wrapping area and the corresponding height centerline of the yarn trunk area The angle between them, that is, the wrapping angle of each wrapping area, A i The jet vortex spinning yarn wrapping image I i Wrapping angle.

[0036] Beneficial effects of the present invention:

[0037] (1) The jet vortex spinning yarn image acquisition device provided by the present invention is provided with a yarn-walking platform, and a first light source and a second light source are respectively provided on both sides above the yarn-walking platform. The first light source and the second light source illuminate the yarn-walking platform at an angle parallel to the core fiber and not parallel to the wrapping fiber, forming a light-dark contrast effect in which the core fiber is dark and the wrapping fiber is bright, highlighting the wrapping area and the core yarn area, greatly improving the contrast and resolution of the wrapping area and the core yarn area in the captured image, and providing a yarn image that is easy to achieve regional segmentation. In addition, when the yarn passes through, the yarn-walking platform prevents the yarn from twisting, maintains the working distance between the yarn and the lens, and ensures that the camera captures stable and clear images.

[0038] (2) The method for detecting the wrapping effect of jet vortex spinning yarn provided by the present invention accurately segments the wrapping area and the core yarn area of ​​the pre-processed jet vortex spinning yarn image through the wrapping-core yarn area segmentation neural network model and the detection frame generation head of the candidate wrapping area therein, and calculates the wrapping coefficient, short wrapping rate, long core exposure rate and wrapping angle of the jet vortex spinning yarn according to the segmentation results to comprehensively evaluate the yarn wrapping effect from multiple dimensions, solving the insufficiency of the traditional method in detecting the wrapping structure. It provides an efficient and objective wrapping effect detection method, which helps the textile industry optimize the production process and improve the yarn quality and production efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0040] Figure 1 It is a brief structural schematic diagram of the jet vortex spinning line image acquisition device provided in Example 1 of the present invention.

[0041] Figure 2 It is a schematic structural diagram of a wrapping-core yarn region segmentation neural network model provided in Example 3 of the present invention;

[0042] Figure 3 is the collected image of the jet vortex spinning yarn provided in Example 3 of the present invention;

[0043] Figure 4 is the image of the main region of the yarn provided by Example 3 of the present invention;

[0044] Figure 5It is a schematic diagram of the division of the wrapping area and the core yarn area provided in Example 3 of the present invention. DETAILED DESCRIPTION

[0045] In order to make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0046] Embodiment 1:

[0047] The embodiment of the present invention provides an image acquisition device for jet vortex spinning yarn, such as Figure 1 As shown, it includes a camera 11 electrically connected to and controlled by a controller, a first light source 9 and a second light source 10 controlled by a light source controller 13, the camera 11 is equipped with a lens 12 and is located directly above the yarn-running platform 8, the first light source 9 and the second light source 10 are respectively fixed on an optical axis 15 by a first light source bracket 16 and a second light source bracket 17, and the first light source 9 and the second light source 10 are respectively located above both sides of the yarn-running platform 8 to ensure that the light evenly covers the yarn area, the first light source 9 and the second light source 10 respectively illuminate the yarn-running platform 8 at an angle parallel to the core fiber and not parallel to the wrapping fiber, forming a dark core fiber and a dark wrapping fiber. The light and dark contrast effect of the fiber highlights the wrapped area and the core yarn area. The plane of the yarn platform 8 is parallel to the imaging plane of the camera 11. The yarn platform 8 is fixed on the base 14. The optical axis 15 is also fixed on the base 14. The optical axis 15 is fixed with a camera 11, a lens 12, a lens bracket 18 and a support rod in sequence from top to bottom in the axial direction. The first light source bracket 16 and the second light source bracket 17 are fixed at both ends of the support rod respectively. The first yarn tension device 4 and the second yarn tension device 5 are fixed on the base 14 and are located on both sides of the yarn platform 8 to provide necessary tension for the yarn to ensure that the yarn always remains straight.

[0048] In the specific implementation, the yarn 2 on the bobbin 1 is unwound by the yarn cone 3, and passes through the traction roller driven by the motor. The motor speed can be adjusted to control the yarn speed, and the yarn tension is adjusted by the first yarn tension device 4. After the yarn 2 enters the yarn platform 8, the first light source 9 and the second light source 10 are turned on, and the motor is started to control the camera 11 to continuously collect 12,000 clear area yarn images. The yarn 2 is finally wound and collected by the first winding shaft 6 and the second winding shaft 7. The above 12,000 yarn images are used for the subsequent jet vortex spinning yarn wrapping effect detection.

[0049] As a preferred solution, the device can add fine-tuning rotation mechanisms to the installation positions of the first light source 9 and the second light source 10, respectively, so as to perform fine angle adjustment on the angles of the light sources.

[0050] As a preferred solution, the lens bracket 18 is a coarse focusing screw of a microscope, and the fixed bracket is fixed on the optical axis, and the working distance between the camera and the yarn-feeding platform can be adjusted within a certain range.

[0051] The jet vortex spinning yarn image acquisition device provided by the embodiment of the present invention is provided with a yarn-walking platform, and a first light source and a second light source are respectively provided on both sides above the yarn-walking platform. The first light source and the second light source illuminate the yarn-walking platform at an angle parallel to the core fiber and not parallel to the wrapping fiber, forming a light-dark contrast effect of dark core fiber and bright wrapping fiber, highlighting the wrapping area and the core yarn area, greatly improving the contrast and resolution of the wrapping area and the core yarn area in the captured image, and providing a yarn image that is easy to achieve regional segmentation. In addition, when the yarn passes through, the yarn-walking platform prevents the yarn from twisting, maintains the working distance between the yarn and the lens, and ensures that the camera captures stable and clear images.

[0052] Embodiment 2:

[0053] The better the wrapping effect of the yarn, the tighter the wrapping fibers are wrapped around the core yarn, the tighter the yarn forming structure, the better the breaking strength and wear resistance of the yarn, and the better the yarn spinning quality. However, too tight wrapping can easily cause uneven yarn, and too loose wrapping will lead to poor yarn breaking strength and poor yarn uniformity. The more evenly the wrapping area is distributed on the core yarn, the smaller the uneven strength and wear resistance of the yarn, the better the strength and wear resistance, and the better the yarn spinning quality. The more evenly the number of wrapping fibers in the wrapping area is distributed, the better the yarn wrapping effect, and the better the yarn spinning quality. Therefore, an embodiment of the present invention provides a method for detecting the wrapping effect of jet vortex spinning yarn, which quantifies the yarn wrapping effect by detecting the wrapping coefficient, short wrapping rate, long core exposure rate and wrapping angle of the yarn, and specifically includes the following steps:

[0054] S100, continuously collecting the jet vortex spinning yarn wrapping image I according to a preset frequency by the jet vortex spinning yarn image collection device. i ,i∈[1,n], such as Figure 3 As shown;

[0055] S200, wrapping image I for each jet vortex spinning yarn i Perform image preprocessing to remove noise and background and extract the main area of ​​the yarn B i ,like Figure 4 As shown;

[0056] S300, using the wrapping-core yarn area segmentation neural network model, from each jet vortex spinning yarn wrapping image I i The yarn trunk area B i Identify each image I i Each wrapping area R i k Core yarn area C i k ,like Figure 5 As shown;

[0057] S400, wrapping area R according to all the jet vortex spinning yarn wrapping images i k Core yarn area C i k , calculate the wrapping coefficient R, short wrapping rate S, long exposed core rate L, and wrapping angle A of the jet vortex spinning yarn;

[0058] S500, judging the wrapping effect of the jet vortex spun yarn according to the wrapping coefficient R, the short wrapping rate S, the long exposed core rate L, and the wrapping angle A of the jet vortex spun yarn, and obtaining the jet vortex spun yarn wrapping effect detection result.

[0059] Specifically, the wrapping coefficient is used to characterize the amount of yarn wrapped around the fiber. The more wrapped fibers, the better the fiber wrapping and the higher the yarn breaking strength. However, too much fiber wrapping may cause problems such as excessive hardness or poor elasticity of the yarn. Therefore, a wrapping coefficient that is too high or too low will affect the wrapping effect of the yarn.

[0060] Most of the wrapped fibers are gathered on the surface of the core fiber, but occasionally there will be a wrapped area with a small number of fibers. A small amount of wrapping will also affect the wear resistance and strength of the yarn. Therefore, the short wrapping rate is introduced to evaluate the wrapping quality of the spinning yarn. The higher the short wrapping rate, the worse the wear resistance of the spinning yarn, and the higher the wear resistance unevenness rate, the worse the wrapping effect.

[0061] The exposed core ratio refers to the proportion of core fibers exposed on the yarn surface. A larger exposed core ratio will inevitably cause a loss in yarn strength and wear resistance, and the wrapping effect will be worse.

[0062] The wrapping angle is of great significance for the calculation of yarn twist. The lower the measured wrapping angle unevenness, the more stable the airflow inside the hollow spindle during the spinning process, and the better the wrapping effect.

[0063] In the specific implementation, taking 30S polyester-viscose mixed jet vortex spun yarn as an example, if the wrapping coefficient R is within the preset wrapping coefficient threshold range (0.4-0.6), the short wrapping rate S is within the short wrapping rate threshold range (0-0.1), the long exposed core rate L is within the long exposed core rate threshold range (0.4-0.6), and the wrapping angle A is within the wrapping angle threshold range (30°-70°), it is determined that the wrapping effect of the jet vortex spun yarn is excellent; otherwise, it is determined that the wrapping effect of the jet vortex spun yarn is poor.

[0064] The method for detecting the wrapping effect of jet vortex spinning yarn provided by the embodiment of the present invention accurately segments the wrapping area and the core yarn area of ​​the pre-processed jet vortex spinning yarn image through the wrapping-core yarn area segmentation neural network model and the detection frame generation head of the candidate wrapping area therein, and calculates the wrapping coefficient, short wrapping rate, long core exposure rate and wrapping angle of the jet vortex spinning yarn according to the segmentation result to comprehensively evaluate the yarn wrapping effect from multiple dimensions, solving the problem of the traditional method's insufficient detection of the wrapping structure. It provides an efficient and objective wrapping effect detection method, which helps the textile industry optimize the production process and improve the yarn quality and production efficiency.

[0065] Embodiment 3:

[0066] The embodiment of the present invention provides a preferred solution of a method for detecting the wrapping effect of jet vortex spinning yarn, comprising:

[0067] S100, continuously collecting the jet vortex spinning yarn wrapping image I according to a preset frequency by the jet vortex spinning yarn image collection device. i ,i∈[1,n];

[0068] Specifically, first turn on the light source, start the motor, and control the camera to continuously capture n images, n>5, each image is defined as I i ,i∈[1,n], such as Figure 3 shown.

[0069] S200, wrapping image I for each jet vortex spinning yarn i Perform image preprocessing to remove noise and background and detect the main area of ​​the yarn B i ,like Figure 4 As shown;

[0070] Specifically, the image preprocessing method provided by the embodiment of the present invention mainly includes:

[0071] S210, wrapping image I of jet vortex spinning yarn i Filter the noise. The method of filtering the noise here may be median filtering, Gaussian filtering, etc.;

[0072] S220, jet vortex spinning yarn wrapping image I after noise filtering i Find the maximum connected area in , where the maximum connected area refers to the minimum area including the yarn trunk and all remaining hairiness;

[0073] S230, perform binarization processing, morphological operation and threshold segmentation on the largest connected area in sequence to obtain the yarn trunk area B i .

[0074] S300, using the wrapping-core yarn area segmentation neural network model, from each jet vortex spinning yarn wrapping image I i The yarn trunk area B i Identify the wrapping area R i k Core yarn area C i k ;

[0075] Specifically, Figure 2 As shown, step S300 includes:

[0076] S310, extracting features from the main region of the yarn by local convolution, nonlinear activation and pooling operations in the image feature extraction module, and converting the main region of the yarn into i Converted to a high-dimensional feature image F(I i ), providing abstract semantic information of the image for subsequent tasks;

[0077] S320, generating a high-dimensional feature image F(I i ) generates a preset number of candidate frames for each candidate enveloping area, and obtains a candidate frame set P i , candidate box P i include Represents a high-dimensional feature image F(I i ) corresponds to the g-th candidate box of the j-th candidate wrapping area, where the candidate box set P i The position information of each candidate box in includes the center point coordinates (x c ,y c ), width w, angle θ and step length λ, the center of the candidate box and the yarn trunk area B i Overlap as much as possible to accurately capture the wrapping area. The height of the candidate frame is the same as the height of the yarn trunk h. i The shape of the wrapped area is similar, which is convenient for matching the shape of the wrapped area and forming a more accurate candidate area. The angle θ of the candidate box is fitted by minimizing the error term, and the angle θ is calculated by the inverse tangent function, θ∈[-170°,170°]. The step length λ of the candidate box is calculated based on the width w of the candidate box, the height h of the candidate box and the angle θ of the candidate box.

[0078] Specifically, the calculation method of λ is shown in formula (1) and formula (2):

[0079]

[0080] λ=min(S w ,S h )·f(θ) (2)

[0081] Among them, L(a θ ,bθ ) represents the loss function, x m Indicates the horizontal coordinate of the point in the candidate box, y m Indicates the ordinate of the point in the candidate box, a θ and b θ Both represent parameters in the linear function, S w Represents a high-dimensional feature image F(I i ) to the width of the candidate box, S h Represents a high-dimensional feature image F(I i ) is the ratio of the height of the candidate box to the height of the candidate box, and f(θ) represents a fixed adjustment factor set according to the rotation angle θ;

[0082] By minimizing the loss function L(a θ ,b θ ) can find the best a θ and b θ ;

[0083] S330, the high-dimensional feature image F(I i ) and the candidate box set P i The position information of each candidate box in is input into the regional feature extraction module to extract the candidate box set P i The features of the regions corresponding to each candidate box in the output are the aligned region feature maps F corresponding to each candidate box. fs (I i ) for subsequent classification, regression and segmentation tasks;

[0084] Here, the features of the region corresponding to the candidate box refer to multi-scale fusion features such as semantic features, graphic features, and position features of the region corresponding to the candidate box;

[0085] S340: Set the candidate box set P i Input to the region screening module, remove the candidate boxes whose overlapping parts corresponding to each candidate enveloping area exceed the set IoU threshold, and obtain the intermediate candidate box set P i1 , and select the intermediate candidate box with the largest confidence corresponding to each candidate enveloping area, and output the valid candidate box set P i2 ;

[0086] Specifically, the expression of the area screening module is shown in the following formula (3) and formula (4):

[0087] P i1 =NMS(P i ,IoU threshold ) (3)

[0088] P i2 = argmax{score(P i1 )} (4)

[0089] Among them, score(P i1 ) represents the confidence score of the candidate box, IoU threshold represents the IoU threshold;

[0090] It should be understood that after the region screening module, each candidate wrapping region corresponds to only one valid candidate box;

[0091] S350: Set the valid candidate box set P i2 and high-dimensional feature image F(I i ) Input region calibration module, optimize the center point coordinates, width, step size and angle of the candidate box through regression network, fine-tune the candidate box, and output the calibrated valid candidate box set P c , which contains the exact center point position (x′) of each candidate box. c , y′ c ), width w′ j , step length λ j and angle θ′ j ;

[0092] Specifically, the expression of the regional calibration module is shown in the following formula (5):

[0093]

[0094] The calibrated valid candidate box set P c The expression of is as follows:

[0095] P c = {P i 1′ ,P i 2′ ,P i 3′ …P i j′} (6)

[0096] Among them, dx represents the offset of the horizontal coordinate of the center point of the valid candidate box, dy represents the offset of the vertical coordinate of the center point of the valid candidate box, and d w Indicates the width adjustment of the valid candidate box, d θ Indicates the angle adjustment amount of the valid candidate box;

[0097] S360: The calibrated valid candidate frame set P c Input to the pixel classification module to identify the regional feature map F through pixel-level classification fs (I i) belongs to the entangled region, and the Softmax function is used to convert the category score of each pixel into a probability, output the entangled region segmentation mask, and select the most likely category as the prediction result to identify the entangled region R i k , the remaining area is the core yarn area C i k ,like Figure 5 As shown;

[0098] Specifically, as shown in formula (7) and formula (8):

[0099]

[0100] y p = argmaxP pixel (p,c) (8)

[0101] Among them, f p (F fs (I i ), c) represents the regional feature map F fs (I i ) The category score obtained after the extracted features are processed by the classification network, ∑ c′ exp(f p (F(I i ), c′)) represents the normalization factor calculated by summing up the indices of the non-normalized scores of all categories, P pixel (p,c) represents the probability that pixel p belongs to category c, y p represents the final category label of pixel p.

[0102] S400, wrapping area R according to all the jet vortex spinning yarn wrapping images i k Core yarn area C i k , calculate the wrapping coefficient E, short wrapping rate S, long exposed core rate L, and wrapping angle A of the jet vortex spinning yarn;

[0103] Specifically, S400 includes:

[0104] S410, by calculating all the jet vortex spinning yarn wrapping images I i The wrapping area R i k The sum of the areas accounts for all the jet vortex spinning yarn wrapping images I i The yarn trunk area B i The ratio of the sum of the areas is used to obtain the wrapping coefficient E, as shown in formula (9):

[0105]

[0106] Among them, Count represents the number of pixels in the statistical area;

[0107] S420, by calculating all the jet vortex spinning yarn wrapping images I i Width in Less than τ times the yarn trunk width W i The wrapping area R i k The sum of the areas accounts for all the jet vortex spinning yarn wrapping images I i The yarn trunk area B i The ratio of the sum of the areas is obtained to obtain the short wrapping ratio S, as shown in formulas (10) and (11):

[0108]

[0109] in, Indicates the width of the wrapping area, that is, the area of ​​the wrapping area Specific yarn height h i , τ represents the short wrapping threshold, τ∈[0.1,0.6], Indicates width Less than τ times the yarn trunk width W i The wrapping area, that is, the short wrapping area

[0110] S 430, by calculating all the jet vortex spinning yarn wrapping images I i Width in More than γ times the yarn trunk width W i The sum of the exposed core areas accounts for all the wrapped images of the jet vortex spinning yarn I i The yarn trunk area B i The ratio of the core length L is obtained, as shown in formulas (12) and (13):

[0111]

[0112] Among them, Count represents the number of pixels in the statistical area. Indicates the width of the core yarn area, that is, the area of ​​the core yarn area Specific yarn height h i The value of γ represents the long dew core threshold, γ∈[0.1,0.6], Indicates width More than γ times the yarn trunk area width W i The exposed core area, that is, the long exposed core area

[0113] S440, obtaining the wrapping angle A by calculating the angle between the width center line of the wrapping area and the height center line of the yarn trunk area, as shown in formulas (14) to (16):

[0114]

[0115] in, Indicates the width centerline of the wrapping area, Indicates the height centerline of the yarn trunk area, Indicates the width centerline of each wrapping area and the corresponding height centerline of the yarn trunk area The angle between them, that is, the wrapping angle of each wrapping area, A i Image showing the wrapping of jet vortex spinning yarn i I wrapping angle;

[0116] Wrapping area R i k The pixel coordinates near the center line of the medium width are (x j ,y j ), by minimizing the error term to fit the center line of the wrapping area, to calculate the wrapping area R i k Width Centerline Slope and yarn trunk area B i The height centerline slope To calculate the wrapping angle;

[0117] Specifically, the wrapping area R i k Width Centerline Slope The calculation method of is shown in formulas (17) and (18):

[0118]

[0119] in, Represents the intercept of the fitted width centerline function, which is set to 0 here;

[0120] Yarn trunk area B i The height centerline slope The calculation method of is shown in formulas (19) and (20):

[0121]

[0122] in, Represents the intercept of the fitted height centerline function, which is set to 0 here.

[0123] S500, judging the wrapping effect of the jet vortex spun yarn according to the wrapping coefficient R, the short wrapping rate S, the long exposed core rate L, and the wrapping angle A of the jet vortex spun yarn, and obtaining the jet vortex spun yarn wrapping effect detection result.

[0124] In the embodiment of the present invention, two light sources with illumination angles parallel to the core fiber and not parallel to the wrapping fiber are set in the image acquisition stage of the jet vortex spinning yarn. The light source is used to enhance the light-dark contrast of the wrapping area and improve the visibility of the wrapping area. On this basis, the wrapping-core yarn area segmentation neural network model and the detection frame generation head of the candidate wrapping area are used to accurately segment the wrapping area and the core yarn area of ​​the pre-processed jet vortex spinning yarn image. The innovation of the detection frame generation head of the candidate wrapping area makes the positioning of the wrapping area more accurate and improves the accuracy of subsequent detection. The wrapping coefficient, short wrapping rate, long exposed core rate and wrapping angle of the jet vortex spinning yarn are calculated according to the segmentation results to comprehensively evaluate the yarn wrapping effect from multiple dimensions, solving the problem of insufficient detection of the wrapping structure by the traditional method. An efficient and objective wrapping effect detection method is provided, which helps the textile industry to optimize the production process and improve the yarn quality and production efficiency.

[0125] Some steps in the embodiments of the present invention may be implemented using software, and the corresponding software program may be stored in a readable storage medium, such as a CD or a hard disk.

[0126] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. Air jet vortex spinning yarn image acquisition device, characterized in that: include: A base, on which a yarn-feeding platform and an optical axis are fixedly arranged, and a camera, a lens, a lens holder and a support rod are fixedly arranged in sequence axially from top to bottom on the optical axis, the camera, the lens, the lens holder and the support rod are all located directly above the yarn-feeding platform, and a first light source holder and a second light source holder are fixedly arranged at both ends of the support rod, the first light source holder is fixedly connected to a first light source, and the second light source holder is fixedly connected to a second light source, the first light source and the second light source are symmetrically arranged on both sides of the yarn-feeding platform, and illuminate the yarn-feeding platform at an angle parallel to the core fiber and not parallel to the wrapping fiber.

2. A method for detecting the wrapping effect of jet vortex spinning yarn, characterized in that: include: S100, continuously collecting the jet vortex spinning yarn wrapping image I according to a preset frequency by the jet vortex spinning yarn image collection device according to claim 1. i , i∈[1,n]; S200, wrapping image I for each jet vortex spinning yarn i Perform image preprocessing to remove noise and background and extract the main area of ​​the yarn B i ; S300, using the wrapping-core yarn area segmentation neural network model, from each jet vortex spinning yarn wrapping image I i The yarn trunk area B i Identify the wrapping area R i k Core yarn area C i k ; S400, wrapping area R according to all the jet vortex spinning yarn wrapping images i k Core yarn area C i k , calculate the wrapping coefficient R, short wrapping rate S, long exposed core rate L, and wrapping angle A of the jet vortex spinning yarn; S500, judging the wrapping effect of the jet vortex spun yarn according to the wrapping coefficient R, the short wrapping rate S, the long exposed core rate L, and the wrapping angle A of the jet vortex spun yarn, and obtaining the jet vortex spun yarn wrapping effect detection result.

3. The method according to claim 2, characterized in that The step S300 includes: S310, sequentially performing local convolution, nonlinear activation and pooling operations in the image feature extraction module on the yarn trunk region B i Perform feature extraction and separate the yarn trunk area B i Converted into a high-dimensional feature image F(I i ); S320, generating a high-dimensional feature image F(I i ) generates a preset number of candidate frames for each candidate enveloping area, and obtains a candidate frame set P i , and obtain the candidate box set P i The location information of the candidate box in the image includes the center point coordinates (x c ,y c ), width w, angle θ and step size λ; S330, the high-dimensional feature image F(I i ) and the candidate box set P i The position information of each candidate box in is input into the regional feature extraction module to extract the candidate box set P i The features of the regions corresponding to each candidate box in the output are the aligned region feature maps F corresponding to each candidate box. fs (I i ); S340: Set the candidate box set P i Input to the region screening module, remove the candidate boxes whose overlapping parts corresponding to each candidate enveloping area exceed the set IoU threshold, and obtain the intermediate candidate box set P i1 , and select the intermediate candidate box with the largest confidence corresponding to each candidate enveloping area, and output the valid candidate box set P i2 ; S350: Set the valid candidate box set P i2 and high-dimensional feature image F(I i ) Input region calibration module, optimize the center point coordinates, width, angle and step size of the candidate box through regression network, and output the calibrated valid candidate box set P c ; S360: The calibrated valid candidate frame set P c Input to the pixel classification module to identify the regional feature map F through pixel-level classification fs (I i ) belongs to the entangled region, and the Softmax function is used to convert the category score of each pixel into a probability, output the entangled region segmentation mask, and select the most likely category as the prediction result to identify the entangled region R i k , the remaining area is the core yarn area C i k .

4. The method according to claim 3, characterized in that The angle θ of the candidate box is fitted by minimizing the error term, and the angle θ is calculated by the inverse tangent function, θ∈[-170°,170°]; The step size λ of the candidate box is calculated based on the width w of the candidate box, the height h of the candidate box, and the angle θ of the candidate box. The calculation method of λ is: λ=min(S w ,S h )·f(θ) Among them, L(a θ ,b θ ) represents the loss function, x m Indicates the horizontal coordinate of the point in the candidate box, y m Indicates the ordinate of the point in the candidate box, a θ and b θ Both represent parameters in the linear function, S w Represents a high-dimensional feature image F(I i ) to the width of the candidate box, S h Represents a high-dimensional feature image F(I i ) to the height of the candidate box, and f(θ) represents a fixed adjustment factor set according to the rotation angle θ.

5. The method according to claim 4, characterized in that The calculation formula of the wrapping coefficient R is: Among them, Count represents the number of pixels in the statistical area. Represents all jet vortex spinning yarn wrapping images I i The wrapping area R i k The sum of the areas, Represents all jet vortex spinning yarn wrapping images I i The yarn trunk area B i The sum of the areas.

6. The method according to claim 4, characterized in that The calculation formula of the short wrapping rate S is: Among them, Count represents the number of pixels in the statistical area. Indicates the width of the wrapping area, that is, the area of ​​the wrapping area Specific yarn height h i , τ represents the short wrapping threshold, τ∈[0.1,0.6], Indicates width Less than τ times the yarn trunk width W i The wrapping area, that is, the short wrapping area Represents all jet vortex spinning yarn wrapping images I i Width in Less than τ times the yarn trunk width W i The wrapping area R i k The sum of the areas, Represents all jet vortex spinning yarn wrapping images I i The yarn trunk area B i The sum of the areas.

7. The method according to claim 4, characterized in that The calculation formula of the long exposed core rate L is: Among them, Count represents the number of pixels in the statistical area. Indicates the width of the core yarn area, that is, the area of ​​the core yarn area Specific yarn height h i The value of γ represents the long dew core threshold, γ∈[0.1,0.6], Indicates width More than γ times the yarn trunk area width W i The exposed core area, that is, the long exposed core area Represents all jet vortex spinning yarn wrapping images I i Width in The sum of the exposed core areas exceeding γ times the main yarn width, Represents all jet vortex spinning yarn wrapping images I i The yarn trunk area B i The sum of the areas.

8. The method according to claim 4, characterized in that The calculation formula of the wrapping angle A is: in, Indicates the width centerline of each wrapping area and the corresponding height centerline of the yarn trunk area The angle between them, that is, the wrapping angle of each wrapping area, A i The jet vortex spinning yarn wrapping image I i Wrap angle.

Images