Non-woven fabric surface defect detection method and system

Abstract

The invention discloses a non-woven fabric surface defect detection method and system, and relates to the field of image processing, a non-woven fabric surface image and a transmission image are synchronously collected through double cameras, a space mapping coordinate system is constructed through checkerboard calibration, accurate correlation of two image pixels is achieved, and the non-woven fabric surface defect detection method and system are obtained based on the coordinate system. Extracting shrinkage ring width and node diagonal length of a calendered diamond node from the surface image, and extracting aperture and frame bundle lamination number of a transparent window from the transmission image; by constructing a coupling vector, the size difference influence is eliminated; calculating an offset index based on the coupling vector to quantify the unbalance degree of the defect, comparing the defect proportion with a preset threshold value, judging a risk level and generating a production control instruction; according to the scheme, cross-modal coupling analysis from the surface of the non-woven fabric to the internal structure is realized, the defect detection precision is improved, the defect omission ratio is reduced, and automatic closed-loop control of production quality is supported.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of image processing, in particular to a non-woven fabric surface defect detection method and system. BACKGROUND

[0002] As a flexible material with the characteristics of air permeability, flexibility, acid and alkali resistance, non-woven fabric is widely used in medical and health, packaging and protection fields. Among them, the non-woven fabric formed by calendering process has regular diamond node structure on the surface, and has high tensile strength and air permeability. In the production process, the integrity of the calendering diamond node directly determines the product performance: the node shrinkage ring width reflects the uniformity of fiber distribution, the air window aperture affects the air permeability, and the number of edge frame bundle stacks is related to the structural strength. The coordinated matching of the three is the core to ensure the quality of non-woven fabric. However, non-woven fabric production is mostly carried out on high-speed continuous production line, and traditional detection methods cannot meet the real-time quality monitoring requirements under high-speed production. Conventional machine vision detection only analyzes the surface image alone, and it is difficult to establish the correlation between the surface structure and the internal pore, resulting in that the quality judgment depends on experience and the precision is limited.

[0003] The surface and internal pore of the calendering diamond node are the coordinated reflection of the calendering process parameters, and the imbalance of the two directly reflects the process abnormality. However, the existing detection technology does not construct the spatial mapping relationship between the surface image and the transmission image, and cannot accurately extract the structure data and internal pore data of the non-woven fabric surface. At the same time, the extraction of non-woven fabric image data by the existing detection technology is mostly single dimension, lacks correlation judgment, and leads to a high missing detection rate of more than 15% for the surface node structure exceeding the standard range and pore imbalance defects caused by uneven calendering force in high-speed production, which seriously affects the strict requirements of medical and health and other high-end fields on the quality of non-woven fabric. SUMMARY

[0004] In view of the deficiencies of the prior art, the present application provides a non-woven fabric surface defect detection method and system to solve the problem that the structure data and internal pore data of the non-woven fabric surface cannot be accurately extracted.

[0005] To achieve the above purpose, the present application realizes the following technical scheme: obtaining a non-woven fabric surface image and a non-woven fabric transmission image; extracting the shrinkage ring width and node diagonal length of the calendering diamond node according to the non-woven fabric surface image, the calendering diamond node being a diamond texture unit formed by calendering process of the non-woven fabric, including a central air window and a surrounding fiber shrinkage ring, and the diagonal direction of the calendering diamond node being consistent with the transmission direction of the calendering force; extracting the air window aperture and the number of edge frame bundle stacks of the air window according to the non-woven fabric transmission image, the air window being a pore structure formed at the center of the calendering diamond node due to the fiber being extruded to the edge in the calendering process, including the boundary of the pore itself and the fiber frame around it; a coupling vector is constructed based on a ratio of the shrink ring width and the node diagonal length to a ratio of the aperture of the openable window and the number of the frame beam layers; a difference value is calculated based on the coupling vector, and a sum of the difference value is obtained to obtain a deviation index, and a defect is determined according to a preset deviation threshold and the deviation index.

[0006] Preferably, the non-woven fabric surface image is collected by a first linear array camera above the running direction of the non-woven fabric, and the non-woven fabric transmission image is collected by a second linear array camera below the running direction of the non-woven fabric; A spatial mapping relationship between the non-woven fabric surface image and the non-woven fabric transmission image is established by a preset checkerboard calibration plate, and a spatial mapping coordinate system is constructed based on the spatial mapping relationship.

[0007] Preferably, the surface calibration image and the transmission calibration image of the checkerboard calibration plate are synchronously collected based on the first linear array camera and the second linear array camera; The pixel coordinates of the checkerboard corner points in the surface calibration image and the transmission calibration image are extracted by the Harris corner point detection algorithm, and a surface corner point set and a transmission corner point set are obtained respectively; The homography matrix is solved by the least square method, and the spatial mapping relationship between the transmission corner point set and the surface corner point set is constructed by multiplying the homography matrix and the transmission corner point set, and the spatial mapping relationship between the transmission corner point set and the surface corner point set is converted into a spatial mapping table; The pixels of the non-woven fabric surface image are mapped to the checkerboard corner points of the surface calibration image, and the upper left corner pixel of the non-woven fabric surface image is set as the original point of the initial spatial mapping coordinate system, the running direction of the non-woven fabric is set as the X axis, and the width direction is set as the Y axis; The pixels of the non-woven fabric transmission image are mapped to the checkerboard corner points of the transmission calibration image, and are converted to the initial spatial mapping coordinate system through the spatial mapping table, to obtain a spatial mapping coordinate system including the pixel coordinates of the non-woven fabric surface image and the pixel coordinates of the non-woven fabric transmission image corresponding to each other.

[0008] Preferably, the non-woven fabric surface image is subjected to gray projection along the 45° and 135° diagonal lines in the running direction and the width direction of the non-woven fabric respectively, and the brightness peak columns distributed at 45° and 135° are extracted; for each brightness peak column, the center line of the brightness peak column is determined as the diagonal line of the calendered rhombus node by Gaussian fitting the center coordinate distribution of the peak. According to the design interval D of the calendering rhombus node, each brightness peak column is divided into a local window with a length of 1.5D; in each local window, a local gray scale peak and a gray scale valley in the neighborhood of the local gray scale peak within a range of ±0.3D are detected, and a peak-valley gray scale difference is calculated; when the peak-valley gray scale difference is greater than or equal to a preset peak-valley difference threshold, the peak is determined as a center point of the corresponding calendering rhombus node, the coordinates of the center point in a spatial mapping coordinate system are recorded, and a 3*3 neighborhood gray scale average G of the center point is calculated, so as to obtain a center point set containing the coordinates and the gray scale average; The center point is set as a center, and is sequentially expanded outward according to a preset radial step length, so as to form K sampling circumferences, and the radial step length is a radius difference between adjacent sampling circumferences; the number N of pixels of each sampling circumference is counted, and the gray scale value corresponding to each pixel is recorded; when the Mth sampling circumference is reached, the number of pixels with a gray scale value less than or equal to G on the Mth sampling circumference is marked as N1; when N1 is greater than or equal to a proportion threshold multiplied by N, the Mth sampling circumference is determined as an inner boundary of the fiber shrinkage ring. The shrinkage ring width W of the calendering rhombus node is obtained by calculating the product of the radial step length and M, and the diagonal length L of the node is obtained by calculating the Euclidean distance between the opposite vertices on the diagonal.

[0009] Preferably, the center point coordinates of the calendering rhombus node in the non-woven fabric surface image are mapped to the non-woven fabric transmission image through the spatial mapping coordinate system, so as to obtain a coordinate set of the center point on the non-woven fabric transmission image; Delaunay triangulation is performed on the coordinate set, and the smallest rhombus quadrilateral formed by every four coordinates is taken as a search window. The search window is divided into a center area and an annular area, and the gray scale averages of the center area and the annular area are calculated respectively; when the gray scale average of the center area minus the gray scale average of the annular area is greater than a preset gray scale difference threshold, it is determined that the search window contains a through window. The center area edge is outwardly expanded pixel by pixel, and the gray scale gradients of adjacent pixels in each direction are calculated; when the gray scale gradients of N2 consecutive pairs of adjacent pixels in the same direction are less than or equal to a preset gradient threshold, the pixel expansion in the direction is stopped, and the coordinates of the first pixel in the N2 consecutive pairs of adjacent pixels are marked as boundary coordinates, and the last coordinate is marked as a frame coordinate. When the pixel expansion in all directions is stopped, a closed contour formed by all the marked boundary coordinates is taken as a through window boundary, the through window boundary is fitted through a minimum circumscribed rhombus algorithm, the short diagonal length of the minimum circumscribed rhombus obtained by fitting is calculated, and a through window aperture V is obtained. The formed area between the closed contour formed by all the frame coordinates and the open window boundary is constructed as an annular sampling band, the annular sampling band is divided into four areas, the fiber bundle texture period in different areas of the annular sampling band is extracted through a Gabor filter, the number of local maxima of the Gabor filter response fiber bundle texture period in each area is counted, and the average of the number of local maxima in the four areas is taken as the number of stack layers C of the frame bundle.

[0010] Preferably, the ratio of the shrink ring width corresponding to each center point in the center point set on the non-woven fabric surface image to the node diagonal length is calculated to obtain K1. The ratio of the aperture of the open window corresponding to each center point in the center point set on the non-woven fabric transmission image to the number of stack layers of the frame bundle is calculated to obtain K2. The center point coordinates corresponding to K1 and K2 and the values of K1 and K2 form a coupling vector F.

[0011] Preferably, the offset index is obtained by calculating the sum of the absolute value of the difference between K1 in the coupling vector and the reference value and the absolute value of the difference between K2 and the calibration constant K0; the reference value is the designed ratio of the shrink ring width to the node diagonal length. A preset offset threshold is set, if the offset index of the i-th coupling vector is greater than the offset threshold, the center point corresponding to the i-th coupling vector is marked as an unbalanced defect, otherwise it is marked as qualified. The number of unbalanced defects is counted, and risk control is performed according to the number of unbalanced defects.

[0012] Preferably, the number of center points marked as unbalanced defects in the center point set of the non-woven fabric surface image is counted, and the number of center points marked as unbalanced defects is divided by the total number of center points to obtain a defect ratio. A preset defect threshold is set, when the defect ratio is greater than the defect threshold, it is determined as a first-level risk, a control instruction is generated, and the time of determination and the coordinate set of the defect center point are recorded.

[0013] Preferably, a non-woven fabric surface defect detection system is provided for implementing any one of the above defect detection methods, comprising: An image acquisition module, the image acquisition module comprises a first subunit, a second subunit and a matching light source, the first subunit is a first linear array camera arranged above the non-woven fabric running direction, the second subunit is a second linear array camera arranged below the non-woven fabric running direction, and the matching light source is a backlight source installed above the non-woven fabric running direction; the first subunit and the second subunit simultaneously acquire non-woven fabric surface images and non-woven fabric transmission images through a double-camera synchronous triggering mechanism; The data extraction module is configured to construct a space mapping coordinate system for the collected non-woven fabric surface image and non-woven fabric transmission image, extract the shrink ring width and node diagonal length of the calendered rhombus node from the non-woven fabric surface image, and extract the aperture of the air-through window and the number of frame bundle stacks from the non-woven fabric transmission image; The defect judgment module is configured to construct a coupling vector according to the extracted shrink ring width, node diagonal length, aperture of the air-through window and number of frame bundle stacks, calculate an offset index according to the coupling vector, and perform defect judgment according to the offset index. The risk control module is configured to count the results of the defect judgment to obtain a defect proportion, determine the risk level of the non-woven fabric production line according to the defect proportion, and generate a control instruction.

[0014] Compared with the prior art, the following beneficial effects are achieved: The non-woven fabric surface defect detection method and system have the following beneficial effects compared with the prior art: The non-woven fabric surface defect detection method and system have the following beneficial effects compared with the prior art: The non-woven fabric surface defect detection method and system have the following beneficial effects compared with the prior art:

[0015] In summary, the non-woven fabric defect detection method realizes a closed loop of defect detection, and the non-woven fabric defect detection system upgrades the non-woven fabric defect detection from traditional qualitative judgment to quantitative, cross-modal and full-process automatic control, significantly improves the detection accuracy and production stability, and is especially suitable for quality monitoring in high-demand fields such as medical health. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 The figure is a flowchart of the method. Figure 2 A schematic diagram of a system framework of the present application; DETAILED DESCRIPTION

[0017] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of the present application.

[0018] Referring to Figure 1 The present application provides a non-woven fabric surface defect detection method. The method specifically comprises the following steps: Step one: a first linear array camera is disposed above along the running direction of the non-woven fabric and is matched with a coaxial backlight source to collect a non-woven fabric surface image, and a second linear array camera is disposed below along the running direction of the non-woven fabric to collect a non-woven fabric transmission image transmitted by the coaxial backlight source. Specifically, the first linear array camera and the second linear array camera are synchronously exposed by the same trigger signal to ensure that the non-woven fabric surface image and the non-woven fabric transmission image collected at the same time correspond to the same area of the non-woven fabric (spatial position matching), thereby avoiding image misplacement caused by high-speed running of the production line.

[0019] Step two: a preset checkerboard calibration plate (the distance between checkerboard corner points is known, and the accuracy is ±0.01 mm) is placed in the detection area of the non-woven fabric production line, a surface calibration image (surface reflection image of the calibration plate) is collected by the first linear array camera, and a transmission calibration image (transmission image of the calibration plate) is collected by the second linear array camera. Specifically, the checkerboard calibration plate used should be of semi-transparent material, and is removed from the detection area after the calibration images are collected. The pixel coordinates of the checkerboard corner points in the surface calibration image and the transmission calibration image are extracted by the Harris corner point detection algorithm to obtain a surface corner point set and a transmission corner point set, respectively. A homography matrix H (3×3 matrix) is solved by the least square method, and a spatial mapping relationship between the transmission corner point set and the surface corner point set is constructed by multiplying the homography matrix and the transmission corner point set, the spatial mapping relationship between the transmission corner point set and the surface corner point set is converted into a spatial mapping table (records the coordinate mapping relationship between each pixel of the surface image and the corresponding pixel of the transmission image), and one-to-one correspondence between the surface image pixel and the transmission image pixel is realized. The pixels of the non-woven fabric surface image are mapped to the checkerboard corner points of the surface calibration image, the top-left corner pixel of the non-woven fabric surface image is set as the origin of the coordinate system, the running direction of the non-woven fabric (along the length direction of the production line) is set as the X axis, and the width direction of the non-woven fabric (along the width direction of the production line) is set as the Y axis, so as to construct an initial spatial mapping coordinate system. The pixels of the non-woven fabric transmission image are mapped to the transmission calibration image chessboard corner points, and are converted to the initial space mapping coordinate system through a space mapping table, and finally a space mapping coordinate system is formed which contains one-to-one correspondence between the non-woven fabric surface image pixel coordinates and the non-woven fabric transmission image pixel coordinates; Specifically, by unifying the coordinate system, the pixel coordinates of the non-woven fabric surface image and the non-woven fabric transmission image are standardized, so that the subsequent extracted surface features (such as the shrink ring width, the node diagonal length) and the transmission features (such as the through-hole aperture, the number of edge frame beam stacks) can be associated in the same coordinate, providing a unified spatial reference for the correlation analysis of the non-woven fabric surface and the interior.

[0020] Step three: along the 45° and 135° diagonal lines of the running direction and the width direction of the non-woven fabric, respectively, the non-woven fabric surface image is subjected to gray scale projection (the pixel gray scale values of the image along the direction are accumulated), and two cross-distributed gray scale projection curves are obtained; from the curve, the continuous region with gray scale value significantly higher than the background is extracted, that is, the brightness peak column (corresponding to the diagonal line of the calendered rhombic node, because the fiber is dense and the reflection is strong at the diagonal line) distributed at 45° and 135°, for each brightness peak column, the center coordinate distribution of the Gaussian fitting peak is determined (the gray scale distribution of the peak conforms to the Gaussian distribution, and the Gaussian fitting reduces the interference of noise on the positioning of the peak center) to determine the center line of the brightness peak column as the diagonal line of the calendered rhombic node. Specifically, the calendered rhombic node is a rhombic texture unit formed by the calendering process (roller extrusion) of the non-woven fabric, which contains a central through-hole (pore) and surrounding fiber shrinkage ring (edge frame fiber beam), and its diagonal line direction is consistent with the calendering force transmission direction (45°, 135°). The directionality (45°, 135°) of the diagonal line of the calendered rhombic node is used to accurately extract the peak column through gray scale projection, solving the problem of difficult identification of the rhombic structure in the complex background, and the Gaussian fitting reduces the noise influence, so that the positioning error of the diagonal line center line is ≤0.5 pixels, providing an accurate reference for subsequent node segmentation; According to the designed spacing D of the calendering diamond nodes (i.e., the distance between the centers of adjacent nodes along the diagonal direction, determined by the production process parameters), each brightness peak column is divided into local windows of length 1.5D (ensuring that each window contains one complete node); within each local window, the local grayscale peak (candidate center point) and the grayscale valley (low grayscale area around the peak) in the ±0.3D range of its neighborhood are detected, and the peak-to-valley grayscale difference is calculated; when the peak-to-valley grayscale difference is ≥ the preset peak-to-valley difference threshold (the peak-to-valley difference threshold is calculated by taking the mean + 3 times the standard deviation of the peak-to-valley difference of 100 qualified nodes), the peak is determined to be the center point of the corresponding calendering diamond node, the coordinates of the center point in each local window in the spatial mapping coordinate system are recorded, and the grayscale mean G of the 3×3 neighborhood of the center point is calculated to obtain a center point set containing coordinates and grayscale means. Specifically, local window segmentation ensures that a single node is analyzed independently to avoid interference from adjacent nodes, and the peak-to-valley difference threshold effectively filters noise (such as false peaks caused by fiber lint), thereby improving the accuracy of center point detection; The center point is set as the center of the circle, and the circle is expanded outward in sequence according to a preset radial step size to form K sampling circles (K is the maximum number of expansions to ensure that the node edge is covered). The radial step size is the radius difference between adjacent sampling circles (which can be 1 pixel); the number of pixels N in each sampling circle is counted, and the grayscale value corresponding to each pixel is recorded (starting from a circle center, K circles are formed around this circle center, and the number of pixels N on the circle gradually increases). When the M-th sampling circle is traversed, the number of pixels on the circle with a grayscale value ≤ G (the grayscale mean of the 3×3 neighborhood of the center point) is marked as N1. When N1 ≥ N multiplied by a proportion threshold (the proportion threshold is obtained by collecting 100 sets of qualified non-woven fabric surface shrinkage ring grayscale samples, calculating the statistical relationship between the central neighborhood mean and the grayscale of the inner boundary of the ring, and fitting), the M-th sampling circle is determined as the inner boundary of the fiber shrinkage ring; The shrinkage ring width W of the calendered diamond node is obtained by calculating the product of the radial step length and M (W reflects the uniformity of the edge fiber distribution. If W is too large, it means that the edge fibers are loose, and if it is too small, it may be over-extruded). The node diagonal length L is obtained by calculating the Euclidean distance between the relative vertices on the diagonal. Specifically, along the diagonal line of the calendered diamond node (the center line of the brightness peak column), extending from the center point to both ends, the pixel position where the grayscale value first drops from ≥0.8G to ≤0.5G is detected as the relative vertex on the diagonal line. L includes two diagonals, L1 and L2, which are two diagonals along 45° and 135° in the calendered diamond node, respectively. In this example, the diagonal line L1 along 45° is selected as the node diagonal length L; Specifically, the step accurately positions the framework by using the directionality of the diamond nodes, objectively determines the characteristics by using the peak-valley difference, avoids the dependence on artificial experience, quantifies the parameters to provide a surface reference for joint analysis, and is a key input for subsequent defect determination. The step converts the microstructure characteristics of the calendered diamond nodes into calculable digital parameters, solving the problem of more qualitative description and less quantitative analysis in traditional detection.

[0021] Step four: Map the center point coordinates of the calendered diamond nodes in the non-woven fabric surface image to the non-woven fabric transmission image through the space mapping coordinate system to obtain a coordinate set of the center points on the non-woven fabric transmission image. Perform Delaunay triangulation on the coordinate set, and take the smallest rhombus quadrilateral formed by every four coordinates as a search window. Specifically, the search window is used to define a rhombus area of the detection range of the through-hole window, and the side length is determined by the design distance D of the adjacent calendered diamond nodes, such as 1.2D, to ensure covering the through-hole window and the surrounding frame. The through-hole window is a pore structure formed by the center of the calendered diamond node due to the extrusion of fibers to the edge in the calendering process, including the boundary of the pore itself (defined by the gray gradient change) and the surrounding fiber frame (associated with the number of beam stacks); Divide the search window into a center area occupying 40% of the search window area and a ring area (60% of the area can cover the ring-shaped sampling band) occupying the remaining 60% of the area. Calculate the gray mean value of the center area and the ring area, respectively. When the gray mean value of the center area minus the gray mean value of the ring area is greater than the preset gray difference threshold value (the gray difference threshold value is obtained by taking the mean value + 3 times the standard deviation of 100 qualified through-hole window samples), it is determined that the search window contains a through-hole window. The through-hole window is displayed as bright in the center area and dark in the surrounding area after transmission by the light source; Expand outward along the edge of the center area by one pixel at a time, and calculate the gray gradient of each adjacent pixel in the same direction. When the gray gradient of N2 consecutive pairs of adjacent pixels in the same direction is less than or equal to the preset gradient threshold value, stop the pixel expansion in that direction, and mark the coordinates of the first pixel in the N2 consecutive pairs of adjacent pixels as the boundary coordinates and the last coordinate as the frame coordinates. Specifically, each direction represents the direction of the expansion of all pixels on the edge of the center region, and due to the presence of defects, the final expanded contour can not be a contour equal to the shape of the center region. In this example, N2=5, and the value of N2 cannot be greater than the frame range of the open window (obtained from historical measurement data). Calculating the continuous gray gradient can ensure stable gradient change and avoid misjudgment caused by single-pixel noise. The multi-directional expansion can ensure the integrity of the boundary contour, and by marking the coordinates of the first pixel in the adjacent pixels as the boundary coordinates and the last coordinate as the frame coordinates, and then connecting all the boundary coordinates and all the frame coordinates, a ring-shaped sampling band is formed, which is the frame of the open window. When the pixel expansion in all directions stops, all the marked boundary coordinates are fitted as a closed contour, and the closed contour is formed as the open window boundary. The open window boundary is fitted by the minimum circumscribed diamond algorithm, the length of the short diagonal of the fitted minimum circumscribed diamond is calculated, and the aperture V of the open window is obtained. Specifically, the minimum circumscribed diamond algorithm is used to fit the open window boundary to obtain the minimum circumscribed diamond that can match the actual shape of the open window, and the aperture calculation error is ≤2%. The lengths of the two diagonals of the minimum circumscribed diamond are calculated, and in this example, the short diagonal is selected as the aperture index, which can intuitively reflect whether the aperture of the non-woven fabric is too large or too small. The region formed between the closed contour formed by all the frame coordinates and the open window boundary is constructed as a ring-shaped sampling band, and the ring-shaped sampling band is divided into four regions (a plane rectangular coordinate system is established with the center region of the search window as the origin, and the four regions are located in the four quadrants of the coordinate system). The fiber bundle texture period in different regions of the ring-shaped sampling band is extracted by a Gabor filter (with a wavelength of 8 pixels and a direction consistent with the frame direction of the region). The fiber bundle texture period is the periodic gray change of the fiber stack. The number of local maxima (each maximum corresponds to a layer of fiber bundle) of the Gabor filter response fiber bundle texture period in each region is counted, and the average of the number of local maxima in the four regions is taken as the number of frame bundle layers C. Specifically, this step uses a search window based on spatial mapping to solve the problem of open window positioning. By combining gray gradient and texture analysis, the size of the aperture and the fiber stacking density can be quantified at the same time. The internal structure of the non-woven fabric and the surface image form a complement, providing a synergistic basis for defect judgment. This step combines geometric constraints and image texture analysis to realize the automatic and high-precision extraction of the micro features of the open window.

[0022] Step five: calculate the ratio of the shrink ring width corresponding to each center point in the center point set on the non-woven fabric surface image to the diagonal length of the node to obtain K1. Specifically, K1 can reflect the proportion of the shrink ring width relative to the overall size of the node, and the physical meaning is the relative degree of surface fiber shrinkage. For example, if K1 is too large, it indicates that the shrink ring is too wide, and the edge fiber distribution is loose. Calculate the ratio of the aperture of the through window corresponding to each center point in the center point set on the non-woven fabric transmission image to the number of stacked layers of the frame beam to obtain K2. Specifically, K2 can reflect the balance between the size of the through window and the stacking density of the surrounding fibers. For example, if K2 is too large, it indicates that the aperture is too large and the fiber stacking is insufficient, which may pose a risk of fiber breakage. The center point coordinates corresponding to K1 and K2 and the values of K1 and K2 form a coupling vector F. Specifically, the coupling vector F obtained by combining K1 and K2 and the corresponding center point coordinates in the spatial mapping coordinate system in a fixed format contains the structural parameters of the non-woven fabric outside and inside at the same physical node, providing a complete basis for subsequent defect location and defect type determination through coordinate positioning.

[0023] Step six: the offset index is obtained by calculating the sum of the absolute value of the difference between K1 in the coupling vector and the reference value (the reference value is the designed ratio of the shrink ring width to the diagonal length of the node, which is directly determined by the calendering process design parameters) and the absolute value of the difference between K2 and the calibration constant K0 (based on the parameters C and V of one hundred qualified calendering rhombus nodes, the ratio of each qualified parameter C and V is calculated, and the arithmetic mean is obtained to get the calibration constant K0). Specifically, the reference value is directly obtained based on the process design parameters to ensure that the detection standard is consistent with the production target. K0 is obtained by statistical analysis of actual normal samples to automatically adapt to equipment characteristics and environmental differences, avoiding the deviation between theoretical values and actual production. The calculated offset index represents the comprehensive deviation of the surface features and internal features in the coupling vector relative to the reference value. The larger the value, the more serious the imbalance. A preset offset threshold is set (the offset threshold is the critical value for distinguishing between qualified and defective, which is determined by analyzing the distribution of the offset index of historical defect samples, for example, taking the upper limit value of the 95% confidence interval). If the offset index of the i-th coupling vector is greater than the offset threshold, the center point corresponding to the i-th coupling vector is marked as an imbalance defect (an imbalance defect is a node whose surface and internal features deviate from the threshold, corresponding to process problems such as uneven fiber distribution and calendering pressure fluctuation). Otherwise, it is marked as qualified. Step seven: count the number of center points marked as imbalance defects in the center point set of the non-woven fabric surface image, and divide the number of center points marked as imbalance defects by the total number of center points to obtain the defect proportion. A preset defect threshold is set, when the defect ratio is greater than the defect threshold, it is determined as a first-level risk, a control instruction is generated, and the coordinate set of the defect center point at the determination time is recorded; When the defect ratio is less than or equal to the defect threshold, it is determined as a qualified state.

[0024] Specifically, in the obtained non-woven fabric surface image, the number of defects on the non-woven fabric surface image is finally calculated through the above series of steps, for example, a non-woven fabric surface image contains 1000 center points (corresponding to 1000 calendered rhombus nodes), and the number of defects is 100, that is, there are 100 calendered rhombus nodes with imbalance defects in the non-woven fabric surface image, and the defect ratio is 10%. The defect threshold is the critical value for distinguishing between the qualified state and the first-level risk state, but the production requirements of non-woven fabrics of different grades need to be formulated with different defect thresholds, so the defect threshold can be determined by historical production data and quality standards. When the defect ratio is greater than the defect threshold, the non-woven fabric defect surface detection system generates a control instruction, such as reducing the production line speed, adjusting the calendering roller pressure, etc., and records the coordinates of these imbalance defects and the determination time, which facilitates manual positioning of the non-woven fabric with defects on the production line and analysis, and when the defect ratio is less than or equal to the defect threshold, the current production parameters are maintained.

[0025] Further, referring to Figure 2 The non-woven fabric surface defect detection system is used to implement the defect detection method of any one of the above, comprising: An image acquisition module, the image acquisition module comprises a first subunit, a second subunit and a matching light source, the first subunit is a first linear array camera arranged above the running direction of the non-woven fabric, the second subunit is a second linear array camera arranged below the running direction of the non-woven fabric, and the matching light source is a backlight source installed above the running direction of the non-woven fabric. The first subunit and the second subunit simultaneously collect the non-woven fabric surface image and the non-woven fabric transmission image through a double-camera synchronous triggering mechanism; A data extraction module, the data extraction module is used to construct a space mapping coordinate system for the collected non-woven fabric surface image and non-woven fabric transmission image, and extract the shrink ring width and node diagonal length of the calendered rhombus node according to the non-woven fabric surface image, and extract the through-window aperture and frame bundle stack number of the through-window according to the non-woven fabric transmission image; A defect determination module, the risk determination module is used to construct a coupling vector according to the extracted shrink ring width, node diagonal length, through-window aperture and frame bundle stack number, and calculate the offset index according to the coupling vector, and perform defect determination according to the offset index; A risk control module, the risk control module is used to count the results of defect determination to obtain the defect ratio, and determine the risk level of the non-woven fabric production line and generate a control instruction according to the defect ratio.

[0026] The above embodiments are used to illustrate the technical solutions of the present application but not to limit the present application. Although the present application is explained in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solutions of the present application can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present application.

Claims

1. A method for detecting surface defects of nonwoven fabrics, characterized in that: include: Acquire non-woven fabric surface images and non-woven fabric transmission images; The shrinkage loop width and diagonal length of the calendered diamond node are extracted based on the surface image of the non-woven fabric. The calendered diamond node is a diamond-shaped texture unit formed by the calendering process of the non-woven fabric, which includes a central air window and a surrounding fiber shrinkage loop. The diagonal direction of the calendered diamond node is consistent with the direction of calendering force transmission. The aperture of the hollow window and the number of frame bundle layers are extracted based on the transmission image of the non-woven fabric. The hollow window is a pore structure formed at the center of the calendering diamond node due to the fibers being squeezed to the edge during the calendering process. It includes the boundary of the pore itself and the surrounding fiber frame. The coupling vector is constructed based on the ratio of the shrinkage ring width to the node diagonal length, the aperture of the air-through window and the number of stacked layers of the frame bundle. The deviation index is obtained by calculating the sum of the difference values ​​based on the coupling vector, and the defect is judged by summing the deviation index according to the preset deviation threshold.

2. The method for detecting surface defects of nonwoven fabrics according to claim 1, wherein: Acquire non-woven fabric surface images and non-woven fabric transmission images, including: The surface image of the non-woven fabric is collected by a first line array camera above the running direction of the non-woven fabric, and the transmission image of the non-woven fabric is collected by a second line array camera below the running direction of the non-woven fabric; The spatial mapping relationship between the non-woven fabric surface image and the non-woven fabric transmission image is established through a preset checkerboard calibration plate, and a spatial mapping coordinate system is constructed based on the spatial mapping relationship.

3. The method for detecting surface defects of nonwoven fabrics according to claim 2, wherein: Construct a spatial mapping coordinate system based on the spatial mapping relationship, including: Synchronously acquiring a surface calibration image and a transmission calibration image of a checkerboard calibration plate based on the first line array camera and the second line array camera; The pixel coordinates of the checkerboard corner points in the surface calibration image and the transmission calibration image are extracted using the Harris corner detection algorithm to obtain the surface corner point set and the transmission corner point set respectively. Solve the homography matrix by the least square method, and construct the spatial mapping relationship between the transmission corner point set and the surface corner point set by multiplying the homography matrix with the transmission corner point set, and convert the spatial mapping relationship between the transmission corner point set and the surface corner point set into a spatial mapping table; The pixels of the non-woven fabric surface image are mapped to the checkerboard corner points of the surface calibration image, and the upper left corner pixel of the non-woven fabric surface image is set as the origin of the initial space mapping coordinate system, the running direction of the non-woven fabric is set as the X axis, and the width direction is set as the Y axis; The pixels of the non-woven fabric transmission image are mapped to the checkerboard corner points of the transmission calibration image and converted to the initial spatial mapping coordinate system through the spatial mapping table to obtain a spatial mapping coordinate system containing a one-to-one correspondence between the pixel coordinates of the non-woven fabric surface image and the pixel coordinates of the non-woven fabric transmission image.

4. The method for detecting surface defects of nonwoven fabrics according to claim 1, wherein: Extract the shrinkage loop width and node diagonal length of the calendered diamond node based on the non-woven fabric surface image, including: The nonwoven surface image was grayscale projected along the 45° and 135° diagonal directions of the nonwoven running direction and width direction, and the brightness peaks with a cross distribution of 45° and 135° were extracted. For each brightness peak, the center coordinate distribution of the peak was fitted by Gaussian fitting, and the center line of the brightness peak was determined as the diagonal line of the calendering diamond node. According to the designed spacing D of the rolled diamond nodes, each brightness peak column is divided into local windows of length 1.5D. Within each local window, the local grayscale peak and the grayscale valley within its ±0.3D range are detected, and the peak-to-valley grayscale difference is calculated. When the peak-to-valley grayscale difference is ≥ the preset peak-to-valley difference threshold, the peak is determined to be the center point of the corresponding rolled diamond node. The coordinates of the center point in each local window in the spatial mapping coordinate system are recorded, and the grayscale mean G of the 3×3 neighborhood of the center point is calculated to obtain a center point set containing coordinates and grayscale means. The center point is set as the center of the circle, and the sample is expanded outward in sequence according to the preset radial step size to form K sampling circles. The radial step size is the radius difference between adjacent sampling circles. The number of pixels N in each sampling circle is counted, and the grayscale value corresponding to each pixel is recorded. When traversing to the Mth sampling circle, the number of pixels on this circle with a grayscale value ≤ G is marked as N1. When N1 ≥ N multiplied by the proportion threshold, the Mth sampling circle is determined as the inner boundary of the fiber contraction ring. The shrinkage ring width W of the rolled diamond node is obtained by calculating the product of the radial step length and M, and the node diagonal length L is obtained by calculating the Euclidean distance between opposite vertices on the diagonal line.

5. The method for detecting surface defects of nonwoven fabrics according to claim 1, wherein: Extract the aperture of the transparent window and the number of frame bundle layers of the transparent window based on the non-woven fabric transmission image, including: The coordinates of the center points of the calendered diamond nodes in the non-woven fabric surface image are mapped to the non-woven fabric transmission image through the spatial mapping coordinate system to obtain the coordinate set of the center points on the non-woven fabric transmission image. The coordinate set is connected by Delaunay triangulation, and the minimum diamond quadrilateral formed by every four coordinates is taken as the search window; The search window is divided into a central area and an annular area, and the grayscale means of the central area and the annular area are calculated respectively. When the value obtained by subtracting the grayscale mean of the annular area from the grayscale mean of the central area is greater than a preset grayscale difference threshold, it is determined that the search window contains a transparent window; Expand outward pixel by pixel along the edge of the central area, calculate the grayscale gradient of adjacent pixels in each direction, and stop pixel expansion in that direction when the grayscale gradient of N2 consecutive pairs of adjacent pixels in the same direction is less than or equal to the preset gradient threshold. Mark the coordinates of the first pixel in the N2 consecutive pairs of adjacent pixels as the boundary coordinates, and mark the last coordinate as the border coordinates. When pixel expansion in all directions stops, the closed contour formed by all marked boundary coordinates is used as the transparent window boundary. The transparent window boundary is fitted using the minimum circumscribed diamond algorithm. The length of the short diagonal of the fitted minimum circumscribed diamond is calculated to obtain the transparent window aperture V. The area between the closed contour formed by all the frame coordinates and the boundary of the transparent window is constructed as a ring sampling band, which is then divided into four regions. The fiber bundle texture period in different regions of the ring sampling band is extracted using a Gabor filter. The number of local maxima of the Gabor filter response to the fiber bundle texture period in each region is counted, and the average of the number of local maxima in the four regions is taken as the frame bundle stacking number C.

6. The method for detecting surface defects of nonwoven fabrics according to claim 1, wherein: The coupling vector is constructed based on the ratio of the shrinkage ring width to the node diagonal length, the aperture of the air window, and the number of frame bundle stacks, including: Calculate the ratio of the shrinkage ring width corresponding to each center point in the center point set on the non-woven fabric surface image to the diagonal length of the node to obtain K1; Calculate the ratio of the aperture of the air-transmitting window to the number of stacked layers of the frame beam corresponding to each center point in the center point set on the non-woven transmission image to obtain K2; The center point coordinates corresponding to K1 and K2 and the values ​​of K1 and K2 constitute the coupling vector F.

7. The method for detecting surface defects of nonwoven fabrics according to claim 1, wherein: The offset index is obtained by calculating the sum of the difference values ​​based on the coupling vector, and the defect is determined based on the preset offset threshold and the offset threshold, including: The deviation index is obtained by calculating the sum of the absolute value of the difference between K1 and the reference value in the coupling vector and the absolute value of the difference between K2 and the calibration constant K0. The reference value is the designed ratio of the contraction ring width to the node diagonal length. A preset offset threshold is set. If the offset index of the i-th coupling vector is greater than the offset threshold, the center point corresponding to the i-th coupling vector is marked as an imbalance defect, otherwise it is marked as qualified. Count the number of imbalance defects and perform risk control based on the number of imbalance defects.

8. The method for detecting surface defects of nonwoven fabrics according to claim 7, wherein: Count the number of imbalance defects and perform risk control based on the number of imbalance defects, including: Counting the number of center points marked as imbalance defects in the center point set of the non-woven fabric surface image, dividing the number of center points marked as imbalance defects by the total number of center points to obtain the defect ratio; A defect threshold is preset. When the defect ratio is greater than the defect threshold, it is judged as a level one risk, a control instruction is generated, and the coordinate set of the judgment time and the defect center point is recorded.

9. A nonwoven fabric surface defect detection system, for implementing the defect detection method according to any one of claims 1 to 8, characterized in that: include: The image acquisition module includes a first subunit, a second subunit, and a supporting light source. The first subunit is a first linear array camera arranged above the running direction of the non-woven fabric, and the second subunit is a second linear array camera arranged below the running direction of the non-woven fabric. The supporting light source is a backlight installed above the running direction of the non-woven fabric. The first subunit and the second subunit use a dual-camera synchronous trigger mechanism to simultaneously capture the surface image and the transmission image of the non-woven fabric. A data extraction module is used to construct a spatial mapping coordinate system for the collected non-woven fabric surface image and non-woven fabric transmission image, and to extract the shrinkage ring width and node diagonal length of the calendered diamond node based on the non-woven fabric surface image, and to extract the aperture of the transparent window and the number of frame bundle stacking layers based on the non-woven fabric transmission image; The defect judgment module and the risk judgment module are used to construct a coupling vector based on the extracted shrinkage ring width, node diagonal length, aperture of the air window and number of frame bundle stacks, and calculate the offset index based on the coupling vector, and perform defect judgment based on the offset index; The risk control module is used to collect the results of defect determination, obtain the defect ratio, determine the risk level of the non-woven fabric production line based on the defect ratio, and generate control instructions.

Images