Intelligent sanitary towel production defect detection method based on multi-modal image analysis

Abstract

The invention belongs to the field of visual inspection, and provides a sanitary towel production defect intelligent detection method based on multi-modal image analysis, and the method comprises the following steps: laying a sanitary towel on a transparent plane, irradiating the sanitary towel with light, and obtaining a light-transmitting image of the irradiated sanitary towel; graying the light-transmitting image of the sanitary towel to obtain a grey-scale map of the sanitary towel; according to the sanitary napkin grey-scale map, obtaining a light-transmitting display grid, and obtaining a real light-transmitting area through the light-transmitting display grid; dividing the grey-scale map of the sanitary napkin into two parts to obtain a left grey-scale map and a right grey-scale map, and comparing real light-transmitting areas of the left grey-scale map and the right grey-scale map to obtain a side leakage area; and dividing the grey-scale map into two parts, and comparing the left grey-scale map with the right grey-scale map to obtain a thinness defect area. According to the intelligent detection method provided by the embodiment of the invention, whether the sanitary napkin structure has the side leakage defect or not can be detected through visual detection according to the light transmission information of the sanitary napkin.

Description

Technical Field

[0001] This invention belongs to the field of visual inspection, specifically relating to an intelligent detection method for sanitary napkin production defects based on multimodal image analysis. Background Technology

[0002] Currently, in the production and manufacturing process of sanitary napkins, in order to ensure product quality, the industry usually relies on image recognition technology, manual visual inspection, or traditional machine vision algorithms based on features such as color and shape contours for defect detection and quality control. These methods mainly focus on surface appearance defects of the product, such as obvious defects such as pattern printing offset, crease asymmetry, edge cutting error, and surface structure damage. Because these defects are highly visually discernible, they are easy to identify after being captured by a camera, making them suitable for online inspection systems on high-speed production lines.

[0003] However, sanitary napkins have a complex structure, typically consisting of multiple layers including a surface non-woven fabric, an absorbent layer, a diversion layer, and a bottom film. Their functionality primarily depends on the density of their internal structure and the evenness of their material distribution. When the absorbent layer has areas with sparse material, insufficient compression, or loose fiber arrangement, even if there is no obvious difference in appearance, these thin areas may still become the starting points for liquid to break through the absorption barrier and leak out through the edges during actual use, causing side leakage problems.

[0004] These latent defects, because they do not directly manifest as color abnormalities or changes in edge shape, often cannot be effectively detected by conventional image vision algorithms. Especially in high-speed production environments, thin, transparent products with intact appearances are easily misjudged as qualified products by traditional systems. Summary of the Invention

[0005] This invention aims to at least partially solve one of the technical problems in related technologies. Therefore, the objective of this invention is to propose an intelligent detection method for sanitary napkin manufacturing defects based on multimodal image analysis. This method can detect the presence of side leakage defects in the sanitary napkin structure through visual inspection using the light transmittance information of the sanitary napkin, especially those hidden problems that are difficult to detect through visual inspection.

[0006] To achieve the above objectives, this invention proposes an intelligent detection method for sanitary napkin production defects based on multimodal image analysis, the method comprising the following steps:

[0007] S100: Lay the sanitary napkin flat on a transparent surface and shine a light on the sanitary napkin to obtain a translucent image of the illuminated sanitary napkin;

[0008] S200: Convert the light-transmitted image of the sanitary napkin to grayscale to obtain a grayscale image of the sanitary napkin;

[0009] S300 obtains the light-transmitting display grid based on the grayscale image of the sanitary napkin, and obtains the actual light-transmitting area through the light-transmitting display grid;

[0010] S400, the grayscale of sanitary napkins Figure 1 The image is divided into two parts, a left grayscale image and a right grayscale image. The actual light-transmitting areas of the left grayscale image and the right grayscale image are compared to obtain the side leakage area.

[0011] According to the intelligent detection method of the present invention, the presence of side leakage defects in the structure of a sanitary napkin can be detected by visual inspection using the light transmittance information of the sanitary napkin.

[0012] Furthermore, in step S100, laying the sanitary napkin flat on a transparent surface and illuminating it with a light to obtain a translucent image of the illuminated sanitary napkin includes:

[0013] The sanitary napkin to be tested is laid flat on a transparent optical working platform. The platform is made of high-transmittance glass or acrylic sheet to ensure uniform light penetration from the bottom. During the laying process, the sanitary napkin is ensured to be wrinkle-free and non-overlapping, with its long axis parallel to the platform's coordinate axis for subsequent symmetrical image processing. Then, the top lighting and bottom transmission light source system is activated, using visible light LEDs with a wavelength range of 400–700 nm to excite the optical response of the sanitary napkin structure through vertical illumination. The bottom light source penetrates the sanitary napkin and forms a translucent image on the top imaging device, reflecting the differences in light density of the product's internal absorbing materials, fiber distribution, and compression structure. The top imaging device uses a high-resolution industrial camera or line scan camera to acquire complete translucent image data and save it as a raw image file for subsequent image processing.

[0014] Furthermore, in step S200, converting the light-transmitted image of the sanitary napkin to grayscale to obtain a grayscale image of the sanitary napkin includes:

[0015] The acquired translucent image is input into the image preprocessing software OpenCV for grayscale conversion. Image processing algorithms are then used to convert the original color image into a single-channel grayscale image, extracting the luminance component of each pixel as its grayscale value. The grayscale conversion is achieved using a weighted average method, a maximum value method, or the OpenCV standard grayscale transformation function, ensuring that the value of each pixel in the image is within the range of 0 to 255. A higher value indicates stronger light transmission, while a lower value indicates a denser structure or opacity. The resulting grayscale image of the sanitary napkin clearly reflects the structural translucency characteristics of the product under illumination, serving as the input basis for subsequent translucent grid analysis and defect identification.

[0016] Currently, in the manufacturing process of sanitary napkins, product quality control mainly relies on image recognition, manual inspection, or machine vision algorithms based on color and contour. However, these methods can only be used for obvious defects such as surface appearance, pattern integrity, or edge misalignment. They are difficult to effectively identify hidden defects caused by uneven distribution of absorbent material or insufficient structural compaction, especially local thin areas located at the edges or inside of the sanitary napkin. These areas are very likely to become the starting point for liquid leakage during product use, thus causing side leakage problems. In order to solve the above problems, this invention proposes step S300.

[0017] Furthermore, in step S300, obtaining the light-transmitting display grid based on the grayscale image of the sanitary napkin includes:

[0018] Divide the grayscale image into K grids, where K = 1800. Let sut(i) represent the grayscale value of the i-th grid in the grayscale image, where i takes the value [1, K], and K is the number of grids after dividing the grayscale image. Obtain the median ur of the grayscale values ​​of each grid in sut(i), obtain the average urm of the grayscale values ​​of each grid in sut(i), and obtain the maximum grayscale value of each grid in sut(i) and denote it as ura; where the grayscale value of a grid is the average of the grayscale values ​​of all pixels in the grid; calculate the equalized transmittance grayscale urp, where the calculation method for the equalized transmittance grayscale is as follows:

[0019]

[0020] Furthermore, grids with gray values ​​greater than urp are designated as light-transmitting grids, and grids with gray values ​​less than urp are designated as light-blocking grids.

[0021] Furthermore, in actual production, localized thin areas of the absorbent layer may not necessarily manifest as color or edge abnormalities. However, due to insufficient material density, light can penetrate more easily, resulting in higher grayscale values. By using URP (Ultraviolet Reduction Profile), grids exceeding URP in the grayscale image are identified as "transparent grids," indicating areas where the material may be sparse. Although these areas may not exhibit significant defects in appearance, they are the potential starting point for side leakage during sanitary napkin use.

[0022] The beneficial effects of this step are as follows: Traditional methods are usually based on fixed thresholds for image segmentation, which are easily affected by changes in ambient light and lighting angles and may fail. However, the urp defined in this invention is adaptively adjusted according to the actual distribution characteristics of each image, and can still maintain the consistency of discrimination when the overall brightness of the image changes. Furthermore, since this value integrates three statistical features, namely median, mean and extreme value, it not only ensures stable judgment of the overall image, but also guides the sensitive response to local high grayscale (thin) areas.

[0023] Furthermore, in step S300, obtaining the actual light-transmitting area through the light-transmitting display grid includes the following steps:

[0024] S301, let zj(j) represent the gray value of the j-th light-transmitting display grid, where the value of j is in the range of [1,H], and H is the number of light-transmitting display grids; zg(j) represent the gray value of the light-blocking grid that is closest to the center point of the j-th light-transmitting display grid; for each zg(j), obtain the median and the maximum value, and record the average of the median and the maximum value as zjxp;

[0025] S302, define an integer variable k, initialized to 1, and create two variables sutPTRa and sutPTRb, initialized to zero, for subsequent calculations and comparisons;

[0026] S303, create a blank sequence TYR1, denoted as the real light-transmitting grid sequence, and create a blank sequence TYR2, denoted as the blurred light-transmitting grid sequence;

[0027] S304, calculate the values ​​of sutPTRa and sutPTRb, where: sutPTRa is calculated as zg(k) - zj(k), and sutPTRb is calculated as [zg(k) + zjxp] / 2 - urp; compare the values ​​of sutPTRa and sutPTRb: if sutPTRa is greater than sutPTRb, add the grid number of zj(k) to the real light-transmitting grid sequence TYR1; if sutPTRa is less than or equal to sutPTRb, add the grid number of zg(k) to the blurred light-transmitting grid sequence TYR2;

[0028] S305, if the current variable k is less than H, then increment k by 1 and return to step S304; if the current variable k is not less than H, then jump to S306.

[0029] S306, obtain the median puy in the fuzzy light-transmitting grid sequence TYR2, add the gray values ​​greater than puy in the fuzzy light-transmitting grid sequence TYR2 to the real light-transmitting grid sequence TYR1; the area composed of the grids corresponding to the gray values ​​in the real light-transmitting grid sequence TYR1 is recorded as the real light-transmitting area.

[0030] The beneficial effects of this step are as follows: By analyzing grayscale differences, comparing neighboring blocking grids, and using a median compensation mechanism, regions with true physical thinness characteristics are accurately selected from the transparent grids. Specifically, this step establishes a local contrast relationship between transparent and blocking grids, avoiding the bias of simply using the average value of the entire image for judgment. The difference sutPTRa = zg(j) - zj(j) is defined as the degree of light transmission prominence of the transparent grid. Simultaneously, a reference value sutPTRb = ([zg(j) + zjxp] / 2 - urp) is calculated, combining the characteristic point value zjxp of the blocking grayscale and the balanced transparent grayscale urp to reflect the boundary stability of the grayscale distribution of the structure. When sutPTRa > sutPTRb, it indicates a significant difference in grayscale between the transparent grid and its neighboring blocking grids, and it can be determined as a true transparent grid. Its grid number is then added to the TYR1 sequence. Conversely, areas with poor light transmission are marked as fuzzy transparent grids and added to the TYR2 sequence. However, considering that some edge areas or fiber gradient areas may have gray-level perturbations, it is necessary to further filter out fuzzy grids with gray-level values ​​significantly higher than the median gray-level value puy in TYR2 and add them to TYR1 to reduce the risk of missed detection and enhance the model's discrimination tolerance. This real transparent area can be used to construct representative thin transparent candidate areas, providing accurate regional input for subsequent left-right image comparison and side leakage risk assessment. Compared with existing technologies, this method has the following significant advantages: by using a comparison model of "light transmission and adjacent blocking", it avoids excessive false detection and missed detection caused by simple threshold judgment; by introducing dynamic gray-level reference values ​​and compensation strategies, the recognition algorithm can adapt to different batches, lighting changes and material differences, and the identified areas reflect anomalies at the physical structure level, such as insufficient compaction of the absorption layer and thin material thickness, which are key sources of side leakage.

[0031] Because traditional technologies have not yet used light transmittance as a core criterion for judging the structural stability or defects of sanitary napkins, this physical characteristic lacks a modeling and quantification mechanism in existing quality inspection systems. Those skilled in the art mostly use grayscale and texture to identify visible surface defects, rather than establishing a standardized method for judging the structural integrity of products based on light transmittance distribution and symmetry. Therefore, existing testing systems suffer from low accuracy, vague judgment criteria, and difficulty in identifying asymmetric defects in complex, uneven structures. To address these issues, this invention proposes step S400.

[0032] Furthermore, in step S400, the grayscale of the sanitary napkin is... Figure 1 The process involves dividing the image into two parts, obtaining a left grayscale image and a right grayscale image. The actual light-transmitting areas of the left and right grayscale images are then compared to determine the side leakage area. This process includes the following steps:

[0033] S401, the acquired grayscale image of the sanitary napkin is vertically divided with the midpoint of the horizontal width as the axis of symmetry to obtain a left grayscale image and a right grayscale image, which respectively represent the left and right symmetrical areas of the product structure;

[0034] Based on the symmetry of sanitary napkins, the image of a sanitary napkin is vertically divided at the midpoint of its horizontal width to obtain the axis of symmetry. This axis of symmetry divides the grayscale image into a left grayscale image and a right grayscale image, which are used to compare the differences in the distribution of the real light-transmitting grids in their respective regions to identify potential structural asymmetry and side leakage risks. By comparing the real light-transmitting areas in the left and right grayscale images, the side leakage area is obtained: each grid in the left grayscale image is examined to determine whether it belongs to the real light-transmitting area.

[0035] S402, compare and match the actual light-transmitting area to obtain the side-transmitting area;

[0036] Specifically, for the actual light-transmitting area in the left grayscale image, check whether the corresponding position in the right grayscale image is also an actual light-transmitting area; if not, mark the left side of that area as a side-transmitting area.

[0037] Furthermore, for the actual light-transmitting area in the right grayscale image, check whether the corresponding position in the left grayscale image is also an actual light-transmitting area; if not, mark the right side of that area as a side-transmitting area.

[0038] S403, extract the regions in the left and right grayscale images that are in the same position and are both real light-transmitting grids, and record them as the same-transmitting regions. Extract the grayscale value sequences of these regions in the left and right grayscale images respectively to form the left same-transmitting sequence and the right same-transmitting sequence.

[0039] Specifically, obtain the grids in the left and right grayscale images that belong to the real light-transmitting area at the same position, and mark them as the same light-transmitting area. Record the grids of the same light-transmitting area in the left grayscale image as the left same light-transmitting grid, and the grids of the same light-transmitting area in the right grayscale image as the right same light-transmitting grid.

[0040] Let zyp(f) represent the gray value of the f-th left co-transparency grid in the left grayscale image; zhg(f) represents the gray value grid at the same position as the f-th left co-transparency grid in the right grayscale image, where f is the index and the value of f is [1, T], where T is the number of left co-transparency grids in the left grayscale image; create an empty sequence, import zyp(f) into the sequence in order and denote it as the left co-transparency sequence, create an empty sequence, import zhg(f) into the sequence in order and denote it as the right co-transparency sequence; obtain the average value of the left co-transparency sequence and denote it as zym.

[0041] S404, the light transmission oscillation range is obtained through the left same light transmission sequence, and the light transmission difference limit is calculated;

[0042] Specifically, obtain the upper quartile ZTJ and lower quartile ZTK of the left same-transmittance sequence; when an element is less than or equal to ZTK, it is recorded as a low-order transmittance oscillation range value; when an element is greater than ZTK and less than ZTJ, it is recorded as a mid-order transmittance oscillation range value; when an element is greater than or equal to ZTJ, it is recorded as a high-order transmittance oscillation range value; let the minimum value of the left same-transmittance sequence be SOU, the maximum value be MAX, and the average value be zym. Then, all low-order and mid-order transmittance oscillation ranges are obtained. The ranges of higher-order transmission oscillations are denoted as [SOU, ZTK], (ZTK, ZTJ), and [ZTJ, MAX], respectively; the difference between the average value zym of the left same transmission sequence and the minimum oscillation value SOU is denoted as the first oscillation gap fluctuation Fos1; the difference between the maximum oscillation point MAX and the average value zym of the left same transmission sequence is denoted as the second oscillation gap fluctuation Fos2; the difference between SOU and ZTK is denoted as the lower bound extension Fxer; the difference between MAX and ZTJ is denoted as the upper bound extension Rxer.

[0043] The smaller of Fos1 and Fos2 is used as the numerator, and the larger of Fos1 and Fos2 is used as the denominator. The ratio of the numerator to the denominator is used as the transmittance difference limit coefficient UYATR. When Fxer is less than or equal to Rxer, the transmittance difference limit UYATR is: Fxer × YTER. When Fxer is greater than Rxer, the transmittance difference limit UYATR is: Rxer × YTER. When the number of grids in the transmittance region is insufficient, the difference between the average transmittance grayscale urm and urp is used by default as the transmittance difference limit UYATR.

[0044] The principle of the transmittance difference limit (UYATR) is as follows: In this invention, the calculation principle of the transmittance difference limit (UYATR) is based on statistical modeling of the grayscale distribution of the real transmittance area in the left grayscale image. By constructing a grayscale sequence from the grayscale values ​​within the same transmittance area on the left, and calculating its minimum value SOU, maximum value MAX, average value zym, lower quartile ZTK, and upper quartile ZTJ, a three-segment distribution model of grayscale fluctuation is constructed: [SOU, ZTK] is defined as the low-order transmittance oscillation interval, (ZTK, ZTJ) as the middle-order interval, and [ZTJ, MAX] as the high-order interval. The oscillation difference between the average value and the boundary is calculated to obtain the first oscillation fluctuation Fos1 = zym - SOU and the second oscillation fluctuation Fos2 = MAX - zym. The distribution balance coefficient is constructed by the ratio of the smaller value to the larger value of the two. At the same time, the lower bound extension value Fxer = ZTK - SOU and the upper bound extension value Rxer = MAX - ZTJ are calculated to measure the low and high tolerance capabilities of the grayscale distribution. The final transmittance difference limit UYATR is determined according to the following rules: when Fxer is less than or equal to Rxer, UYATR = Fxer × YTER; otherwise, UYATR = Rxer × YTER. This limit value is used to determine whether the grayscale difference at the same position in the left and right grayscale images exceeds a reasonable structural deviation range. Once the grayscale difference value is greater than UYATR, it can be determined that there is a serious transmittance asymmetry in that area, which may constitute a structural thinning defect or a potential risk of side leakage. Through this calculation model, not only can different image grayscale distribution characteristics be dynamically adapted to improve the judgment accuracy, but the risk of misjudgment caused by fixed threshold settings is also avoided.

[0045] S405, obtaining the side leakage mesh through the limit of light transmittance difference;

[0046] Calculate the difference between zyp(f) and zhg(f) in sequence, and denote the absolute value of the difference between zyp(f) and zhg(f) as caz(f). Compare caz(f) with the transmittance difference limit UYATR. If caz(f) is greater than the transmittance difference limit UYATR, compare the size of zyp(f) and zhg(f). Mark the grid corresponding to the larger gray value between zyp(f) and zhg(f) as the side leakage grid.

[0047] S406, Obtain the side leakage area through the side leakage grid;

[0048] Obtain the number of side-leaking grids in the left and right grayscale images, and denote the number of side-leaking grids in the left grayscale image as TUEW and the number of side-leaking grids in the right grayscale image as YUEW. The area formed by the side-leaking grids in the left grayscale image is denoteed as the left leaking area, and the area formed by the side-leaking grids in the right grayscale image is denoteed as the right leaking area. Compare the number of side-leaking grids in the left and right grayscale images. If the number of side-leaking grids in the right grayscale image YUEW is greater than 2 × TUEW, it indicates that there is a defect on the right side of the sanitary napkin, and the right leaking area will cause side leakage during the use of the sanitary napkin. The right leaking area is denoteed as the side-leaking area. If TUEW is greater than 2 × YUEW, it indicates that there is a defect in the sanitary napkin and it will cause side leakage. If TUEW is greater than 2 × YUEW, it indicates that there is a defect in the left side of the sanitary napkin, and the left leaking area will cause side leakage during the use of the sanitary napkin. The left leaking area is denoteed as the side-leaking area.

[0049] The principle of this step is as follows: by calculating the difference between the gray values ​​of the real light-transmitting grid at the same position in the left grayscale image and the right grayscale image in turn, a difference sequence caz(f) is constructed, and then compared point by point with the light transmittance difference limit UYATR obtained above. When the grayscale difference caz(f) at a certain location is greater than UYATR, it indicates that there is a significant asymmetry in the light transmission performance of the left and right structures at that point, which is a potential thin area. At this time, it is further determined which grayscale value, zyp(f) or zhg(f), is larger, and the grid corresponding to the larger grayscale value is marked as the side leakage grid, because the structure of this side area is thinner, the light transmission is stronger, and it is more prone to liquid leakage. Then, the number of side leakage grids in the left and right grayscale images are counted separately and denoted as TUEW and YUEW. Based on the comparison results, the structural integrity is judged: if the number of side leakage grids on one side is more than twice that on the other side, it indicates that there is a significant structural imbalance on that side, which is prone to side leakage during use. The core basis of this judgment logic is the natural symmetry of the sanitary napkin structure design. Any large-scale asymmetric light transmission abnormality usually reflects functional defects such as uneven distribution of absorbent material, poor sealing, or insufficient compaction.

[0050] The beneficial effects of this step are as follows: by establishing a dynamic identification mechanism based on the linkage between grayscale difference and limit deviation, this step can achieve high sensitivity and high reliability in the location and judgment of side leakage defects, overcoming the technical bottlenecks of existing detection systems such as fuzzy identification of local structural latent defects and weak detection capability of asymmetric thin defects.

[0051] The reason this invention is not obvious is that it introduces "transmittance," a physical quantity originally used for evaluating the optical properties of materials, into the identification of structural defects in sanitary napkins. It constructs a complete algorithm system for quantifying transmittance grayscale features, dynamically judging grids, and comparing structural symmetry, achieving high-precision identification of functional latent defects that are difficult to identify using traditional visual systems. This method has not been revealed or inspired by existing technologies, specifically in the following aspects: Existing technologies mainly focus on superficial features such as color, edges, or texture for defect identification, while this invention, based on the response characteristics of the product's internal structure to transmitted light, establishes a direct correlation between grayscale values ​​and material density, proposing a method for identifying true transmittance areas. This method has a unique physical modeling foundation and represents a fundamental breakthrough in detection mechanisms. Traditional structural integrity judgments often rely on dimensional parameters such as dimensional deviations or edge misalignments. This invention, however, uses symmetrical grayscale difference modeling to dynamically generate the transmittance difference limit (UYATR), and then compares the grayscale distribution on both sides point by point, proposing an approach to measure internal structural consistency using grayscale symmetry, breaking through the limitation that what is invisible is unmeasurable in traditional detection methods. This invention employs a multi-level statistical fusion judgment to construct a multi-order oscillation judgment mechanism that differs from the simple threshold method. This mechanism considers local abnormal fluctuations and has an automatic compensation mechanism, significantly improving fault tolerance and detection robustness, something that conventional engineers would not readily conceive of. Furthermore, the industry generally believes that the light transmittance of sanitary napkins has no direct functional indicator significance and is not used as a core parameter for defect judgment. Therefore, this invention elevates it to an important dimension for identifying structural defects, breaking the limitation of traditional quality inspection that only judges defects based on what is visible. This represents a reverse inspiration and breakthrough from existing technological understanding.

[0052] The beneficial effects of this invention are as follows: By constructing a multimodal structural analysis model based on translucent grayscale images, it breaks through the traditional detection method that relies on surface image recognition. It can identify structural hidden defects caused by uneven compaction of the absorbent layer, insufficient material thickness, etc., and is especially suitable for locating potential side leakage sources such as the "thin zone" at the edge or inside of the sanitary napkin. It can stably identify abnormal areas under different lighting and image noise conditions. At the same time, the translucent difference limit UYATR model designed based on the symmetry analysis of the left and right grayscale images improves the sensitivity of the discrimination ability of structural asymmetric defects, and makes up for the problems of low recognition rate and fuzzy standards of existing detection systems for defects such as complex structural unevenness or compaction imbalance. Thus, it realizes a comprehensive upgrade of intelligent identification of sanitary napkin defects. Attached Figure Description

[0053] Figure 1 The diagram shows a flowchart of an intelligent detection method for sanitary napkin production defects based on multimodal image analysis. Detailed Implementation

[0054] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0055] Figure 1 The diagram shows a flowchart of an intelligent detection method for sanitary napkin production defects based on multimodal image analysis.

[0056] Reference Figure 1 This invention proposes an intelligent detection method for sanitary napkin production defects based on multimodal image analysis. The method includes the following steps:

[0057] S100: Lay the sanitary napkin flat on a transparent surface and shine a light on the sanitary napkin to obtain a translucent image of the illuminated sanitary napkin;

[0058] S200: Convert the light-transmitted image of the sanitary napkin to grayscale to obtain a grayscale image of the sanitary napkin;

[0059] S300 obtains the light-transmitting display grid based on the grayscale image of the sanitary napkin, and obtains the actual light-transmitting area through the light-transmitting display grid;

[0060] S400, the grayscale of sanitary napkins Figure 1 The image is divided into two parts, a left grayscale image and a right grayscale image. The actual light-transmitting areas of the left grayscale image and the right grayscale image are compared to obtain the side leakage area.

[0061] According to the intelligent detection method of the present invention, the presence of side leakage defects in the structure of a sanitary napkin can be detected by visual inspection using the light transmittance information of the sanitary napkin.

[0062] Furthermore, in step S100, laying the sanitary napkin flat on a transparent surface and illuminating it with a light to obtain a translucent image of the illuminated sanitary napkin includes:

[0063] The sanitary napkin to be tested is laid flat on a transparent optical working platform made of high-transmittance glass to ensure uniform light penetration from the bottom. During the laying process, the sanitary napkin is ensured to be wrinkle-free and non-overlapping, with its long axis parallel to the platform's coordinate axes for subsequent symmetrical image processing. Then, the top lighting and bottom transmission light source system are activated, using a visible light LED light source with a wavelength range of 600nm to excite the optical response of the sanitary napkin structure through vertical illumination. The bottom light source penetrates the sanitary napkin and forms a translucent image on the top imaging device, reflecting the differences in light density of the product's internal absorbing materials, fiber distribution, and compression structure. The top imaging device uses a high-resolution industrial camera, Vieworks VC-4MC-M180, to acquire complete translucent image data and save it as a raw image file for subsequent image processing.

[0064] Furthermore, in step S200, converting the light-transmitted image of the sanitary napkin to grayscale to obtain a grayscale image of the sanitary napkin includes:

[0065] The acquired translucent image is input into the image preprocessing software OpenCV for grayscale conversion. Image processing algorithms are then used to convert the original color image into a single-channel grayscale image, extracting the luminance component of each pixel as its grayscale value. The grayscale conversion is achieved using a weighted average method, a maximum value method, or the OpenCV standard grayscale transformation function, ensuring that the value of each pixel in the image is within the range of 0 to 255. A higher value indicates stronger light transmission, while a lower value indicates a denser structure or opacity. The resulting grayscale image of the sanitary napkin clearly reflects the structural translucency characteristics of the product under illumination, serving as the input basis for subsequent translucent grid analysis and defect identification.

[0066] Furthermore, in step S300, obtaining the light-transmitting display grid based on the grayscale image of the sanitary napkin includes:

[0067] Divide the grayscale image into K grids, where K = 1800. Let sut(i) represent the grayscale value of the i-th grid in the grayscale image, where i takes the value [1, K], and K is the number of grids after dividing the grayscale image. Obtain the median ur of the grayscale values ​​of each grid in sut(i), obtain the average urm of the grayscale values ​​of each grid in sut(i), and obtain the maximum grayscale value of each grid in sut(i) and denote it as ura; where the grayscale value of a grid is the average of the grayscale values ​​of all pixels in the grid; calculate the equalized transmittance grayscale urp, where the calculation method for the equalized transmittance grayscale is as follows:

[0068]

[0069] Furthermore, grids with gray values ​​greater than urp are designated as light-transmitting grids, and grids with gray values ​​less than urp are designated as light-blocking grids.

[0070] Furthermore, in step S300, obtaining the actual light-transmitting area through the light-transmitting display grid includes the following steps:

[0071] S301, let zj(j) represent the gray value of the j-th light-transmitting display grid, where the value of j is in the range of [1,H], and H is the number of light-transmitting display grids; zg(j) represent the gray value of the light-blocking grid that is closest to the center point of the j-th light-transmitting display grid; for each zg(j), obtain the median and the maximum value, and record the average of the median and the maximum value as zjxp;

[0072] S302, define an integer variable k, initialized to 1, and create two variables sutPTRa and sutPTRb, initialized to zero, for subsequent calculations and comparisons;

[0073] S303, create a blank sequence TYR1, denoted as the real light-transmitting grid sequence, and create a blank sequence TYR2, denoted as the blurred light-transmitting grid sequence;

[0074] S304, calculate the values ​​of sutPTRa and sutPTRb, where: sutPTRa is calculated as zg(k) - zj(k), and sutPTRb is calculated as [zg(k) + zjxp] / 2 - urp; compare the values ​​of sutPTRa and sutPTRb: if sutPTRa is greater than sutPTRb, add the grid number of zj(k) to the real light-transmitting grid sequence TYR1; if sutPTRa is less than or equal to sutPTRb, add the grid number of zg(k) to the blurred light-transmitting grid sequence TYR2;

[0075] S305, if the current variable k is less than H, then increment k by 1 and return to step S304; if the current variable k is not less than H, then jump to S306.

[0076] S306, obtain the median puy in the fuzzy light-transmitting grid sequence TYR2, add the gray values ​​greater than puy in the fuzzy light-transmitting grid sequence TYR2 to the real light-transmitting grid sequence TYR1; the area composed of the grids corresponding to the gray values ​​in the real light-transmitting grid sequence TYR1 is recorded as the real light-transmitting area.

[0077] Furthermore, in step S400, the grayscale of the sanitary napkin is... Figure 1 The process involves dividing the image into two parts, obtaining a left grayscale image and a right grayscale image. The actual light-transmitting areas of the left and right grayscale images are then compared to determine the side leakage area. This process includes the following steps:

[0078] S401, the acquired grayscale image of the sanitary napkin is vertically divided with the midpoint of the horizontal width as the axis of symmetry to obtain a left grayscale image and a right grayscale image, which respectively represent the left and right symmetrical areas of the product structure;

[0079] Based on the symmetry of sanitary napkins, the image of a sanitary napkin is vertically divided at the midpoint of its horizontal width to obtain the axis of symmetry. This axis of symmetry divides the grayscale image into a left grayscale image and a right grayscale image, which are used to compare the differences in the distribution of the real light-transmitting grids in their respective regions to identify potential structural asymmetry and side leakage risks. By comparing the real light-transmitting areas in the left and right grayscale images, the side leakage area is obtained: each grid in the left grayscale image is examined to determine whether it belongs to the real light-transmitting area.

[0080] S402, compare and match the actual light-transmitting area to obtain the side-transmitting area;

[0081] Specifically, for the actual light-transmitting area in the left grayscale image, check whether the corresponding position in the right grayscale image is also an actual light-transmitting area; if not, mark the left side of that area as a side-transmitting area.

[0082] Furthermore, for the actual light-transmitting area in the right grayscale image, check whether the corresponding position in the left grayscale image is also an actual light-transmitting area; if not, mark the right side of that area as a side-transmitting area.

[0083] S403, extract the regions in the left and right grayscale images that are in the same position and are both real light-transmitting grids, and record them as the same-transmitting regions. Extract the grayscale value sequences of these regions in the left and right grayscale images respectively to form the left same-transmitting sequence and the right same-transmitting sequence.

[0084] Specifically, obtain the grids in the left and right grayscale images that belong to the real light-transmitting area at the same position, and mark them as the same light-transmitting area. Record the grids of the same light-transmitting area in the left grayscale image as the left same light-transmitting grid, and the grids of the same light-transmitting area in the right grayscale image as the right same light-transmitting grid.

[0085] Let zyp(f) represent the gray value of the f-th left co-transparency grid in the left grayscale image; zhg(f) represents the gray value grid at the same position as the f-th left co-transparency grid in the right grayscale image, where f is the index and the value of f is [1, T], where T is the number of left co-transparency grids in the left grayscale image; create an empty sequence, import zyp(f) into the sequence in order and denote it as the left co-transparency sequence, create an empty sequence, import zhg(f) into the sequence in order and denote it as the right co-transparency sequence; obtain the average value of the left co-transparency sequence and denote it as zym.

[0086] S404, the light transmission oscillation range is obtained through the left same light transmission sequence, and the light transmission difference limit is calculated;

[0087] Specifically, obtain the upper quartile ZTJ and lower quartile ZTK of the left same-transmittance sequence; when an element is less than or equal to ZTK, it is recorded as a low-order transmittance oscillation range value; when an element is greater than ZTK and less than ZTJ, it is recorded as a mid-order transmittance oscillation range value; when an element is greater than or equal to ZTJ, it is recorded as a high-order transmittance oscillation range value; let the minimum value of the left same-transmittance sequence be SOU, the maximum value be MAX, and the average value be zym. Then, all low-order and mid-order transmittance oscillation ranges are obtained. The ranges of higher-order transmission oscillations are denoted as [SOU, ZTK], (ZTK, ZTJ), and [ZTJ, MAX], respectively; the difference between the average value zym of the left same transmission sequence and the minimum oscillation value SOU is denoted as the first oscillation gap fluctuation Fos1; the difference between the maximum oscillation point MAX and the average value zym of the left same transmission sequence is denoted as the second oscillation gap fluctuation Fos2; the difference between SOU and ZTK is denoted as the lower bound extension Fxer; the difference between MAX and ZTJ is denoted as the upper bound extension Rxer.

[0088] The smaller of Fos1 and Fos2 is used as the numerator, and the larger of Fos1 and Fos2 is used as the denominator. The ratio of the numerator to the denominator is used as the transmittance difference limit coefficient UYATR. When Fxer is less than or equal to Rxer, the transmittance difference limit UYATR is: Fxer × YTER. When Fxer is greater than Rxer, the transmittance difference limit UYATR is: Rxer × YTER.

[0089] S405, obtaining the side leakage mesh through the limit of light transmittance difference;

[0090] Calculate the difference between zyp(f) and zhg(f) in sequence, and denote the absolute value of the difference between zyp(f) and zhg(f) as caz(f). Compare caz(f) with the transmittance difference limit UYATR. If caz(f) is greater than the transmittance difference limit UYATR, compare the size of zyp(f) and zhg(f). Mark the grid corresponding to the larger gray value between zyp(f) and zhg(f) as the side leakage grid.

[0091] S406, Obtain the side leakage area through the side leakage grid;

[0092] Obtain the number of side-leaking grids in the left and right grayscale images, and denote the number of side-leaking grids in the left grayscale image as TUEW and the number of side-leaking grids in the right grayscale image as YUEW. The area formed by the side-leaking grids in the left grayscale image is denoteed as the left leaking area, and the area formed by the side-leaking grids in the right grayscale image is denoteed as the right leaking area. Compare the number of side-leaking grids in the left and right grayscale images. If the number of side-leaking grids in the right grayscale image YUEW is greater than 2 × TUEW, it indicates that there is a defect on the right side of the sanitary napkin, and the right leaking area will cause side leakage during the use of the sanitary napkin. The right leaking area is denoteed as the side-leaking area. If TUEW is greater than 2 × YUEW, it indicates that there is a defect in the sanitary napkin and it will cause side leakage. If TUEW is greater than 2 × YUEW, it indicates that there is a defect in the left side of the sanitary napkin, and the left leaking area will cause side leakage during the use of the sanitary napkin. The left leaking area is denoteed as the side-leaking area.

[0093] The implementation process is as follows:

[0094] Implementation Environment

[0095] Sample: 100 standard sanitary napkins (model A) from the market, including 20 samples with artificially created thinness defects, randomly placed; and 80 normal control samples.

[0096] Equipment: High-transmittance acrylic platform, adjustable visible light LED light source (600nm band), Vieworks VC-4MC-M180 industrial camera.

[0097] Software: OpenCV 4.5 + Python 3.8 to implement grayscale conversion, mesh analysis, symmetry analysis, difference judgment and other steps.

[0098] Steps for implementing human-caused defect samples:

[0099] S100, Transmitted Image Acquisition: Unified ambient light intensity 400 lux, image acquisition and saving.

[0100] S200, Grayscale: Converted to grayscale using OpenCV 4.5, pixel value range 0–255.

[0101] S300–S306 Extraction of True Transmitted Area: The grayscale image is divided into a grid of K=1800. Statistical values ​​are ur=125, urm=128, ura=200, and urp≈151 is calculated; approximately 400 grids are identified as translucent areas, and approximately 250 grids are further identified as true translucent areas.

[0102] S401–S406 Symmetry and Side Leakage Identification: After segmentation, the sequence length of the same permeability region is extracted, T≈180.

[0103] Statistical values: SOU = 110, ZTK = 130, ZTJ = 160, MAX = 210, zym = 145. Calculate Fos1 = 35, Fos2 = 65; YTER ≈ 0.538; Fxer = 20, Rxer = 50; UYATR ≈ 20 * 0.538 ≈ 10.8.

[0104] Calculate chaz(f), and the defective sample is identified as having approximately 30 leaky meshes, while the control sample has an average of 15.

[0105] Comparing the number of side leakage grids in the left and right grayscale images: TUEW = 13, YUEW = 29, YUEW > 2 × TUEW, it is determined that there is a side leakage defect on the right side.

[0106] Experimental Results and Comparisons:

[0107]

[0108] This embodiment demonstrates that by employing a series of algorithms based on light-transmitting grayscale grid analysis and symmetry difference discrimination, it is possible to identify latent structural defects and potential side leakage risks with higher accuracy than traditional vision algorithms, and significantly reduce false detection and false negative rates, thus possessing high industrial application value.

[0109] It should be noted that the logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0110] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0111] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0112] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," "counterclockwise," "axial," "radial," and "circumferential" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing this invention and simplifying the description, and are not intended to indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0113] Furthermore, the terms "first," "second," etc., used in the embodiments of this invention are for descriptive purposes only and should not be construed as indicating or implying relative importance, or implicitly specifying the number of technical features indicated in this embodiment. Therefore, features defined with terms such as "first" and "second" in the embodiments of this invention can explicitly or implicitly indicate that the embodiment includes at least one of those features. In the description of this invention, the word "multiple" means at least two or more, such as two, three, four, etc., unless otherwise explicitly specified in the embodiments.

[0114] In this invention, unless otherwise explicitly specified or limited in the embodiments, the terms "installation," "connection," "joining," and "fixing" appearing in the embodiments should be interpreted broadly. For example, a connection can be a fixed connection, a detachable connection, or an integral part; it can also be a mechanical connection, an electrical connection, etc. Of course, it can also be a direct connection, or an indirect connection through an intermediate medium, or it can be the internal communication of two components, or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific implementation.

[0115] In this invention, unless otherwise explicitly specified and limited, "above" or "below" the second feature can mean that the first feature is in direct contact with the second feature, or that the first feature is in indirect contact with the second feature through an intermediate medium. Furthermore, "above," "over," and "on top" of the second feature can mean that the first feature is directly above or diagonally above the second feature, or simply that the first feature is at a higher horizontal level than the second feature. "Below," "below," and "under" the second feature can mean that the first feature is directly below or diagonally below the second feature, or simply that the first feature is at a lower horizontal level than the second feature.

[0116] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for intelligent detection of defects in sanitary napkin production based on multimodal image analysis, characterized in that, The method includes the following steps: S100: Lay the sanitary napkin flat on a transparent surface and shine a light on the sanitary napkin to obtain a translucent image of the illuminated sanitary napkin; S200: Convert the light-transmitted image of the sanitary napkin to grayscale to obtain a grayscale image of the sanitary napkin; S300 obtains the light-transmitting display grid based on the grayscale image of the sanitary napkin, and obtains the actual light-transmitting area through the light-transmitting display grid; S400 divides the grayscale image of the sanitary napkin into two parts to obtain a left grayscale image and a right grayscale image. The actual light-transmitting areas of the left grayscale image and the right grayscale image are compared to obtain the side leakage area.

2. The intelligent detection method for sanitary napkin production defects based on multimodal image analysis according to claim 1, characterized in that, Step S100 includes: laying the sanitary napkin product to be tested flat on a transparent optical working platform, the platform being made of high-transmittance glass or acrylic sheet; then activating the top lighting and bottom transmission light source system, using a visible light LED light source with a wavelength range of 400-700nm, to excite the optical response of the sanitary napkin structure through vertical illumination; the bottom light source penetrates the sanitary napkin and forms a translucent image on the top imaging device.

3. The intelligent detection method for sanitary napkin production defects based on multimodal image analysis according to claim 1, characterized in that, Step S200 includes: inputting the translucent image into the image preprocessing software OpenCV and performing grayscale operation on it; using image processing algorithms to convert the original color image into a single-channel grayscale image and extracting the luminance component of each pixel as its grayscale value; wherein the grayscale processing is performed using the weighted average method, the maximum value method, or the OpenCV standard grayscale transformation function, so that the grayscale value of each pixel in the image is in the range of 0 to 255.

4. The intelligent detection method for sanitary napkin production defects based on multimodal image analysis according to claim 1, characterized in that, Step S300 includes: dividing the grayscale image into K grids, where K = 1800, and sut(i) represents the grayscale value of the i-th grid in the grayscale image, where i takes the value [1, K], and K is the number of grids after the grayscale image is divided; obtaining the median ur of the grayscale values ​​of each grid in sut(i), obtaining the average urm of the grayscale values ​​of each grid in sut(i), obtaining the maximum grayscale value of each grid in sut(i) and recording it as ura; wherein the grayscale value of the grid is the average of the grayscale values ​​of all pixels in the grid; calculating the equalized light transmission grayscale urp, wherein the calculation method for the equalized light transmission grayscale is: urp = ur × (ur + ura) / 2urm; and recording the grids with grayscale values ​​greater than urp as light-transmitting grids and the grids with grayscale values ​​less than urp as light-blocking grids.

5. The intelligent detection method for sanitary napkin production defects based on multimodal image analysis according to claim 4, characterized in that, Step S300 also includes: S301, let zj(j) represent the gray value of the j-th light-transmitting display grid, where the value of j is in the range of [1,H], and H is the number of light-transmitting display grids; zg(j) represent the gray value of the light-blocking grid that is closest to the center point of the j-th light-transmitting display grid; for each zg(j), obtain the median and the maximum value, and record the average of the median and the maximum value as zjxp; S302, define an integer variable k, initialized to 1, and create two variables sutPTRa and sutPTRb, initialized to zero, for subsequent calculations and comparisons; S303, create a blank sequence TYR1, denoted as the real light-transmitting grid sequence, and create a blank sequence TYR2, denoted as the blurred light-transmitting grid sequence; S304, calculate the values ​​of sutPTRa and sutPTRb, where: sutPTRa is calculated as zg(k) - zj(k), and sutPTRb is calculated as [zg(k) + zjxp] / 2 - urp; compare the values ​​of sutPTRa and sutPTRb: if sutPTRa is greater than sutPTRb, add the grid number of zj(k) to the real light-transmitting grid sequence TYR1; if sutPTRa is less than or equal to sutPTRb, add the grid number of zg(k) to the blurred light-transmitting grid sequence TYR2; S305, if the current variable k is less than H, then increment k by 1 and return to step S304; if the current variable k is not less than H, then jump to S306. S306, obtain the median puy in the fuzzy light-transmitting grid sequence TYR2, add the gray values ​​greater than puy in the fuzzy light-transmitting grid sequence TYR2 to the real light-transmitting grid sequence TYR1; the area composed of the grids corresponding to the gray values ​​in the real light-transmitting grid sequence TYR1 is recorded as the real light-transmitting area.

6. The intelligent detection method for sanitary napkin production defects based on multimodal image analysis according to claim 1, characterized in that, Step S400 includes: S401, the acquired grayscale image of the sanitary napkin is vertically divided with the midpoint of the horizontal width as the axis of symmetry to obtain a left grayscale image and a right grayscale image, which respectively represent the left and right symmetrical areas of the product structure; S402, compare and match the actual light-transmitting area to obtain the side-transmitting area; S403, extract the regions in the left and right grayscale images that are in the same position and are both real light-transmitting grids, and record them as the same-transmitting regions. Extract the grayscale value sequences of these regions in the left and right grayscale images respectively to form the left same-transmitting sequence and the right same-transmitting sequence. S404, the light transmission oscillation range is obtained through the left same light transmission sequence, and the light transmission difference limit is calculated; S405, obtaining the side leakage mesh through the limit of light transmittance difference; S406, obtain the side leakage area through the side leakage grid.

Images