Online monitoring method and system for embossing process of sanitary paper products based on machine vision

Abstract

The present invention belongs to the field of product monitoring technology and proposes a machine vision-based online monitoring method and system for the embossing process of sanitary paper products. Specifically, the method comprises: real-time acquisition of image data to obtain an embossed paper product image; preprocessing of the embossed paper product image to form a processed image; feature extraction of the processed image to obtain an embossed image feature vector; further dynamic evaluation of uniformity quality using the embossed image feature vector to form a real-time uniformity reference; and finally, determination of the embossing roller state based on the real-time uniformity reference. The present invention provides a machine vision-based online monitoring method and system for the embossing process of sanitary paper products. The method aims to introduce advanced image processing algorithms and alarm mechanisms to acquire embossed sanitary paper product images online at high frequency, effectively quantify the risk of embossing quality fluctuations, including uneven embossing effects caused by static electricity accumulation and microscopic damage, thereby setting an alarm mechanism to quickly respond to fluctuations in embossing quality.

Description

Technical Field

[0001] The present invention belongs to the technical field of product monitoring, and in particular relates to a method and system for online monitoring of an embossing process of sanitary paper products based on machine vision. Background Art

[0002] Embossing technology is widely used in toilet paper production, especially in the manufacture of high-end tissue paper, where embossing is an essential process. Embossing improves the comfort and absorbency of tissue paper. It also enhances the adhesion between multiple layers of tissue paper, preventing delamination. It also strengthens the overall strength of the tissue paper and reduces the risk of breakage during use. However, in actual production, the embossing process is susceptible to a series of negative cumulative effects, primarily due to static electricity generated by friction between the mold and the tissue paper. During the embossing operation, static electricity gradually accumulates due to repeated friction between the mold and the tissue paper, especially in extremely dry environments. The uneven charge distribution caused by this static effect not only affects the uniformity of contact between the tissue paper and the mold surface, but also causes the embossing process to deviate from the preset pressure and trajectory, negatively impacting the clarity and consistency of the pattern. Specifically, the risk of uneven contact between the toilet paper and the mold surface arises from the Coulomb force of attraction, which causes the toilet paper to adhere tightly to the mold surface in areas where the toilet paper and the mold have opposite charges. This charge attraction increases the contact pressure in localized embossed areas, resulting in a deeper embossed pattern in these areas. Conversely, in areas where the toilet paper and the mold have like charges, the Coulomb force of repulsion causes the toilet paper to float or suspend above the mold surface, making the embossed pattern appear faint or blurred in these areas. This uneven contact directly leads to significant quality differences between embossed areas, resulting in uneven embossing and, in turn, affecting the overall quality of the product. Severe static electricity accumulation can even trigger localized corona discharge, causing microscopic damage to the toilet paper surface. These damages gradually expand with subsequent use, making the toilet paper more susceptible to breakage, thereby affecting the overall strength and durability of the product. Traditional pressure sensing-based monitoring often fails to detect this static electricity accumulation. Therefore, in order to address the problems of uneven embossing effect and microscopic damage of sanitary paper products caused by static electricity accumulation, an online monitoring method for the embossing process of sanitary paper products based on machine vision is urgently needed to ensure the stability of the embossing process and the high quality of sanitary paper products. Summary of the Invention

[0003] The purpose of the present invention is to propose an online monitoring method and system for the embossing process of sanitary paper products based on machine vision to solve one or more technical problems existing in the prior art and at least provide a beneficial option or create conditions.

[0004] To achieve the above object, according to one aspect of the present invention, a method for online monitoring of the embossing process of sanitary paper products based on machine vision is provided, the method comprising the following steps:

[0005] S100, real-time acquisition of image data to obtain an embossed paper product image;

[0006] S200, pre-processing the embossed paper product image to form a processed image;

[0007] S300, performing feature extraction on the processed image to obtain an embossed image feature vector;

[0008] S400, dynamically evaluates the uniformity quality through the embossing image feature vector to form a real-time uniformity reference;

[0009] S500, judging the state of the embossing roller according to the real-time uniformity reference;

[0010] Furthermore, in step S100, the method for collecting image data in real time to obtain the embossed paper product image is: the camera for collecting image data is any one of a CCD camera and a CMOS camera; the camera is arranged above the paper outlet after the paper leaves the embossing roller, so that the optical axis of the camera is perpendicular to the paper surface; image collection is performed every 5-10 seconds, and the collected image is recorded as the embossed paper product image.

[0011] Furthermore, in step S200, the method for pre-processing the embossed paper product image to form a processed image is as follows: performing multi-stage filter processing on the embossed paper product image to achieve noise filtering, and performing grayscale processing, applying an adaptive histogram equalization algorithm to adjust the distribution of the grayscale value of the image to enhance the contrast of the overall image; and performing grayscale enhancement on the pixels of the pattern edge line in the image through an edge enhancement algorithm to form a processed image.

[0012] Furthermore, in step S300, feature extraction is performed on the processed image to obtain a feature vector of the embossing image. The method is as follows: a plurality of positions are selected on the embossing die cavity of the embossing roller as roller detection positions, and the distance between each roller detection position on the same busbar of the embossing roller is the same; each roller detection position is assigned a corresponding number, and for each roller detection position corresponding to any number: all embossed areas corresponding to the number in the processed image are identified using a deep learning algorithm, and the maximum value of the grayscale values ​​corresponding to each pixel in an embossing area is used as the embossing grayscale; the average value and the range value of the embossing grayscale corresponding to all embossing areas are recorded as the embossing uniformity and radial distance, respectively; and a binary group consisting of the embossing uniformity and radial distance is used as a positioning feature;

[0013] The sequence of positioning features of all roller detection positions under a processing graph is recorded as the embossing image feature vector.

[0014] Furthermore, in step S400, the method of dynamically evaluating the uniformity quality by using the embossing image feature vector to form a real-time uniformity reference is as follows: each time the embossing image feature vector is obtained is recorded as a monitoring point;

[0015] For any monitoring point, an empty sequence is constructed and recorded as the sub-average sequence Hru.Ls. If the embossing average of a positioning feature is greater than or equal to the median value of all embossing averages at that monitoring point, the positioning feature is added to the sequence Hru.Ls. The length of the sequence Hru.Ls is recorded as SuH. The positioning feature corresponding to the minimum amplitude distance in the sequence Hru.Ls is recorded as the instantaneous reference position.

[0016] The instantaneous reference positions within the monitoring point range constitute a set Tss, and the average two-tuple of all positioning features of Tss is calculated and recorded as the instantaneous level feature; the Euclidean distance between any positioning feature and the instantaneous level feature is calculated and recorded as the average amplitude threat degree Avrm;

[0017] The principle of the process of calculating and obtaining the instantaneous reference position is that the instantaneous reference position represents the roller detection position with the best embossing quality at a specific monitoring point; it is determined by comparing the embossing uniformity and radial distance of all roller detection positions at the same moment, and selecting the positioning feature with the smallest radial distance under the condition that the embossing uniformity is greater than or equal to the median value; this calculation method involves statistical models and real-time monitoring models, so that it can provide a real-time quality benchmark, monitor the embossing quality in real time, and quickly identify the roller detection position with the best embossing effect at a specific moment; it effectively handles the ambiguity and dynamics of the data in the positioning features, and provides an effective quality control means for online detection of sanitary paper products, thereby ensuring the accuracy and robustness of the embossing quality prediction of sanitary paper products.

[0018] Take any monitoring point as the current monitoring point: For any roller detection position, obtain the positioning feature with the maximum value of the average amplitude threat degree among all positioning features between the current monitoring point and the SuH×TD period in the reverse time direction, and record it as the reverse reference position; compare the positioning feature with the maximum value among the average amplitude threat degrees of the instantaneous reference position and the reverse reference position, and record the absolute value of the difference between the obtained positioning feature and the embossing uniformity of the roller detection position at the current monitoring point as the uniformity difference component Ovd;

[0019] The real-time homogeneity reference degree Rtue of any roller detection position is calculated based on the instantaneous reference position and the reverse reference position:

[0020]

[0021] Among them, i1 is the serial number of the monitoring point, Rtue i1 is the real-time homogeneity reference of the i1th monitoring point, Ovd i1 and Avrm i1is the average difference component and average amplitude threat degree of the i1th monitoring point, H.Avrm i1 and B.Avrm i1 The average amplitude threat degree of the inverse reference position and instantaneous reference position corresponding to the i1-th monitoring point respectively.

[0022] Since the real-time uniformity reference is calculated by processing the instantaneous reference position and the reverse reference position, it can effectively quantify the risk of significant quality differences between embossed areas directly caused by uneven contact conditions. However, due to excessive reliance on the instantaneous reference position at each monitoring point during the screening process of the reverse reference position, the sensitivity of the data analysis to the uniform amplitude threat is reduced, and the problem of insufficient accuracy of the real-time uniformity reference occurs, which easily leads to uneven embossing effects and affects the overall quality of the product. However, the existing technology cannot effectively compensate for this phenomenon of decreased sensitivity. In order to eliminate this influence, the present invention proposes a more preferred solution as follows.

[0023] Preferably, in step S400, the method for dynamically evaluating the uniformity quality by using the embossing image feature vector to form a real-time uniformity reference is as follows: set a time period as a monitoring period MOTM, set a total of m roller detection positions and n monitoring points in the current MOTM period, obtain an m×n-order two-dimensional positioning feature matrix Mri(i, j), record the embossing uniformity in the positioning feature input by the j-th monitoring point of the i-th roller detection position as Emave(i, j), abbreviated as E i,j , the matrix composed of embossing uniformity is recorded as ∑E i,j , the radial distance is Radis(i,j), abbreviated as R i,j , the matrix composed of radial distance is ∑R i,j , so the element corresponding to the two-dimensional positioning feature matrix Mri(i,j) is (E i,j ,R i,j );

[0024] In the two-dimensional position feature matrix Mri(i,j), the embossing increase of the jth monitoring point at the i-th roller detection position is recorded as EMcre(i,j)=(E i,j -E i,j-1 ) / TD, where j>1; for any roller detection position, draw a broken line graph of the embossing amplitude with respect to the monitoring point and record it as a time series graph, obtain the ratio of the number of monitoring points between any inflection point of the time series graph and the first inflection point in the reverse time direction to all monitoring points, and record the inverse of the obtained ratio as the steady-state parameter STpar corresponding to the monitoring point between the two inflection points. The steady-state parameter of any inflection point is 1;

[0025] The matrix ∑E i,jThe interquartile range of all elements in is the divergence parameter. For the i-th roller detection position, the proportion of all monitoring points whose radial distance is greater than the divergence parameter to the total number of monitoring points is the outlier rate; the m×1-order matrix composed of the outlier rates corresponding to m roller detection positions is the position outlier matrix Lomat. In the matrix ∑R i,j In the equation, if any of the directional distances is less than the divergence parameter, the monitoring point is defined as the overflow moment, and its number is recorded as n_k; the corresponding columns of each overflow moment are intercepted from Mri(i, j) to obtain an m×n_k order matrix, which is recorded as the overflow location feature matrix;

[0026] The columns of the two-tuples in the overflow positioning feature matrix are normalized: for the c-th monitoring point, the average values ​​of the embossing uniformity and radial distance are AV_Eme and AV_Eme respectively. c and AV_Rad c , and the standard deviations are SD_Eme c and SD_Rad c , the standardized embossing uniformity and widthwise distance at the cth monitoring point at the i-th roller detection position are Emave*(i,c)=(E i,c -AV_Eme c ) / SD_Emec,Radis*(i,c)=(R i,c -AV_Rad c ) / SD_Rad c , the new matrices composed of the standardized embossing uniformity and radial distance are ∑E* i,k and ∑R* i,k , ∑E* i,k and ∑R* i,k Subtract to get the deviation matrix ∑Ds i,k ,

[0027] Multiply the deviation matrix and its transposed matrix to obtain the eigenvector Chvor; calculate the real-time homogeneity reference degree Rthr of the m-th roller detection position at the current moment according to the overflow position characteristic matrix s :

[0028]

[0029] Where p is the cumulative variable, Derat s is the outlier rate of the s-th roller detection position, Chvor s is the element of the sth row of the eigenvector Chvor, n_k is the number of monitoring points in the overflow location feature matrix, EMcre s,p The embossing increase at the p-th monitoring point of the s-th roller detection position, ln() is a logarithmic function with the natural constant e as the base.

[0030] Furthermore, in step S500, the method for judging the state of the embossing roller based on the real-time uniformity reference degree is as follows: for any roller detection position, a sequence of real-time uniformity reference degrees obtained within one hour is recorded as a reference time series; a dynamic benchmark model of the embossing roller operating state within the reference time series is constructed: Bthr = MAthr + α·MSDthr; Bthr is the reference model value of any roller detection position, and MAthr is the 5-minute moving average of the roller detection position, where the 5-minute moving average of the roller detection position refers to the average value of each real-time uniformity reference degree within a 5-minute period from any moment to the reverse time direction;

[0031] MSDthr is the 5-minute moving standard deviation, which refers to the standard deviation of each real-time homogeneous reference within a 5-minute period from any moment to its reverse time direction;

[0032] α is the preset sensitivity adjustment factor, and its value range is [0.01, 2]. The difference between the real-time homogeneity reference degree and the reference model value at any roller detection position is recorded as the homogeneity deviation; the Z-score of each homogeneity deviation within 1 hour is calculated. If the Z-score is greater than 3, the corresponding homogeneous deviation is considered to be an abnormal homogeneous deviation; two integer-valued variables are preset, which are recorded as the deviation amount kd and the risk roller position threshold nov respectively; the Z-score is the value obtained by dividing the difference between a number and the mean by the standard deviation.

[0033] When the number of abnormal uniform deviations at any roller detection position is greater than the deviation amount, the roller detection position is recorded as a risk roller position; when the number of risk roller positions is greater than the risk roller position threshold, the embossing roller state is judged to be abnormal; otherwise, the embossing roller state is judged to be qualified.

[0034] Preferably, all undefined variables in the present invention, if not clearly defined, can be manually set thresholds.

[0035] The present invention also provides an online monitoring system for the embossing process of sanitary paper products based on machine vision. The online monitoring system for the embossing process of sanitary paper products based on machine vision comprises: a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the online monitoring method for the embossing process of sanitary paper products based on machine vision are implemented. The online monitoring system for the embossing process of sanitary paper products based on machine vision can be run on computing devices such as desktop computers, laptop computers, PDAs, and cloud data centers. The executable system may include, but is not limited to, a processor, a memory, and a server cluster. The processor executes the computer program and runs it in the following system units:

[0036] An image acquisition unit, used for acquiring image data in real time to obtain an embossed paper product image;

[0037] Image pre-processing unit, used to pre-process the embossed paper product image to form a processed image

[0038] An image feature extraction unit, configured to extract features from the processed image to obtain an embossed image feature vector;

[0039] A quality assessment unit, used to dynamically assess the uniformity quality through the embossing image feature vector to form a real-time uniformity reference;

[0040] The feedback judgment transmission unit is used to judge the state of the embossing roller based on the real-time uniformity reference degree.

[0041] The beneficial effects of the present invention are: the present invention provides an online monitoring method and system for the embossing process of sanitary paper products based on machine vision, which aims to introduce advanced image processing algorithms and alarm mechanisms to collect embossed sanitary paper product images online at high frequency, and effectively quantify the embossing balance performance, including the risk of fluctuation in embossing quality caused by uneven embossing effects and microscopic damage caused by static electricity accumulation, so as to set up an alarm mechanism to quickly respond to the fluctuation of embossing quality, thereby enhancing the accuracy of real-time monitoring of embossing quality and ensuring the overall product quality and performance level of sanitary paper products. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The above and other features of the present invention will become more apparent through a detailed description of the embodiments shown in conjunction with the accompanying drawings. The same reference numerals in the drawings of the present invention represent the same or similar elements. Obviously, the drawings described below are only some embodiments of the present invention. Those skilled in the art can derive other drawings based on these drawings without inventive work. In the drawings:

[0043] Figure 1 Shown is a flow chart of a method for online monitoring of the embossing process of sanitary paper products based on machine vision;

[0044] Figure 2 Shown is the structural diagram of the online monitoring system for the embossing process of sanitary paper products based on machine vision. DETAILED DESCRIPTION

[0045] The following will be combined with the embodiments and drawings to clearly and completely describe the concept, specific structure and technical effects of the present invention so as to fully understand the purpose, scheme and effect of the present invention. It should be noted that the embodiments and features in the embodiments of this application can be combined with each other unless there is a conflict.

[0046] like Figure 1 The figure shows the flow chart of the on-line monitoring method of the embossing process of sanitary paper products based on machine vision. Figure 1To illustrate the online monitoring method of the embossing process of sanitary paper products based on machine vision according to an embodiment of the present invention, the method includes the following steps:

[0047] S100, real-time acquisition of image data to obtain an embossed paper product image;

[0048] S200, pre-processing the embossed paper product image to form a processed image;

[0049] S300, performing feature extraction on the processed image to obtain an embossed image feature vector;

[0050] S400, dynamically evaluates the uniformity quality through the embossing image feature vector to form a real-time uniformity reference;

[0051] S500, judging the state of the embossing roller according to the real-time uniformity reference;

[0052] Furthermore, in step S100, the method for collecting image data in real time to obtain the embossed paper product image is: the camera for collecting image data is any one of a CCD camera and a CMOS camera; the camera is arranged above the paper outlet after the paper leaves the embossing roller, so that the optical axis of the camera is perpendicular to the paper surface; image collection is performed every 5 seconds, and the collected image is recorded as the embossed paper product image.

[0053] Furthermore, in step S200, the method for pre-processing the embossed paper product image to form a processed image is as follows: performing multi-stage filter processing on the embossed paper product image to achieve noise filtering, and performing grayscale processing, applying an adaptive histogram equalization algorithm to adjust the distribution of the grayscale value of the image to enhance the contrast of the overall image; and performing grayscale enhancement on the pixels of the pattern edge line in the image through an edge enhancement algorithm to form a processed image.

[0054] Furthermore, in step S300, feature extraction is performed on the processed image to obtain a feature vector of the embossing image. The method is as follows: a plurality of positions are selected on the embossing die cavity of the embossing roller as roller detection positions, and the distance between each roller detection position on the same busbar of the embossing roller is the same; each roller detection position is assigned a corresponding number, and for each roller detection position corresponding to any number: all embossed areas corresponding to the number in the processed image are identified using a deep learning algorithm, and the maximum value of the grayscale values ​​corresponding to each pixel in an embossing area is used as the embossing grayscale; the average value and the range value of the embossing grayscale corresponding to all embossing areas are recorded as the embossing uniformity and radial distance, respectively; and a binary group consisting of the embossing uniformity and radial distance is used as a positioning feature;

[0055] The sequence of positioning features of all roller detection positions under a processing graph is recorded as the embossing image feature vector.

[0056] The deep learning algorithm includes the YOLO target detection algorithm, the RPN target detection algorithm or the CNN convolutional neural network algorithm.

[0057] The embossed image feature vector can be obtained by performing feature extraction on the processed image using an image feature extraction model preset in the industrial imaging software Vision Pro; that is, the process of performing feature extraction on the processed image and obtaining the embossed image feature vector can be performed by the industrial imaging software Vision Pro.

[0058] Furthermore, in step S400, the embossing image feature vector is used to dynamically evaluate the uniformity quality, thereby forming a real-time uniformity reference. The method is as follows: each time the embossing image feature vector is obtained is recorded as a monitoring point; the range of the monitoring point is limited to within 30 minutes before the current time; and the embossing uniformity and radial distance corresponding to each lateral feature are obtained from the embossing image feature vector.

[0059] For any monitoring point, an empty sequence is constructed and recorded as the sub-average sequence Hru.Ls. If the embossing average of a positioning feature is greater than or equal to the median value of all embossing averages at that monitoring point, the positioning feature is added to the sequence Hru.Ls. The length of the sequence Hru.Ls is recorded as SuH. The positioning feature corresponding to the minimum amplitude distance in the sequence Hru.Ls is recorded as the instantaneous reference position.

[0060] The instantaneous reference positions within the monitoring point range constitute a set Tss, and the average two-tuple of all positioning features of Tss is calculated and recorded as the instantaneous level feature; the Euclidean distance between any positioning feature and the instantaneous level feature is calculated and recorded as the average amplitude threat degree Avrm;

[0061] Take any monitoring point as the current monitoring point: For any roller detection position, obtain the positioning feature with the maximum average amplitude threat degree among all positioning features between the current monitoring point and the SuH×TD period in the reverse time direction, and record it as the reverse reference position; where TD is the data collection interval; compare the positioning feature with the maximum average amplitude threat degree between the instantaneous reference position and the reverse reference position, and record the absolute value of the difference between the obtained positioning feature and the embossing uniformity of the roller detection position at the current monitoring point as the uniformity difference component Ovd;

[0062] The real-time homogeneity reference degree Rtue of any roller detection position is calculated based on the instantaneous reference position and the reverse reference position:

[0063]

[0064] Among them, i1 is the serial number of the monitoring point, Rtue i1 is the real-time homogeneity reference of the i1th monitoring point, Ovd i1 and Avrm i1is the average difference component and average amplitude threat degree of the i1th monitoring point, H.Avrm i1 and B.Avrm i1 The average threat level of the inverse reference position and the instantaneous reference position corresponding to the i1-th monitoring point, respectively. exp() represents the exponential function with the natural constant e as the base, lg() represents the logarithmic function with 10 as the base, and mean represents the symbol of the mean value function.

[0065] Preferably, in step S400, the method for dynamically evaluating the uniformity quality by using the embossing image feature vector to form a real-time uniformity reference is as follows: a time period is set as a monitoring period MOTM, MOTM∈[0.5,2] hours; monitoring is performed at equal intervals TD within the current MOTM period; TD is a measurement interval, i.e., a time interval for obtaining positioning features;

[0066] Assume that there are m roller detection positions and n monitoring points in the current MOTM period, and obtain an m×n-order two-dimensional positioning feature matrix Mri(i, j), where i and j are the serial numbers of the roller detection position and the monitoring point respectively. The embossing uniformity in the positioning feature input by the i-th roller detection position and the j-th monitoring point is recorded as Emave(i, j), which is abbreviated as E i,j , the matrix composed of embossing uniformity is recorded as ∑E i,j , the radial distance is Radis(i,j), abbreviated as R i,j , the matrix composed of radial distance is ∑R i,j , so the element corresponding to the two-dimensional positioning feature matrix Mri(i,j) is (E i,j ,R i,j ), the real-time homogeneous reference matrix Rthr(i,j) corresponds to Mri(i,j);

[0067] In the two-dimensional position feature matrix Mri(i,j), the embossing increase of the jth monitoring point at the i-th roller detection position is recorded as EMcre(i,j)=(E i,j -E i,j-1 ) / TD, where j>1; when j=1, that is, the embossing amplitude corresponding to all roller detection positions in the first column is 1, for any roller detection position, a broken line graph of the embossing amplitude with respect to the monitoring point is drawn and recorded as a time series graph, and the ratio of the number of monitoring points between any inflection point in the time series graph and the first inflection point in the reverse time direction to all monitoring points is obtained, and the inverse of the obtained ratio is recorded as the steady-state parameter STpar corresponding to the monitoring point between the two inflection points, and the steady-state parameter of any inflection point is 1; wherein the inflection points in the time series graph are the maximum and minimum values ​​of the embossing amplitude obtained when forming a time series variation sequence;

[0068] The matrix ∑E i,jThe interquartile range of all elements in is the divergence parameter. For the i-th roller detection position, the proportion of all monitoring points whose radial distance is greater than the divergence parameter to the total number of monitoring points is the outlier rate; the m×1-order matrix composed of the outlier rates corresponding to m roller detection positions is the position outlier matrix Lomat. In the matrix ∑R i,j In the above equation, if any radial distance is less than the divergence parameter, the monitoring point is defined as the overflow moment, and its number is recorded as n_k. This process is used to eliminate the monitoring point where the radial distance is located, that is, it is no longer used in subsequent calculations. The corresponding columns of each overflow moment are intercepted from Mri(i, j) to obtain an m×n_k-order matrix, which is recorded as the overflow location feature matrix.

[0069] The columns of the two-tuples in the overflow positioning feature matrix are normalized: for the c-th monitoring point, the average values ​​of the embossing uniformity and radial distance are AV_Eme and AV_Eme respectively. c and AV_Rad c , and the standard deviations are SD_Eme c and SD_Rad c , the standardized embossing uniformity and widthwise distance at the cth monitoring point at the i-th roller detection position are Emave*(i,c)=(E i,c -AV_Eme c ) / SD_Emec,Radis*(i,c)=(R i,c -AV_Rad c ) / SD_Rad c , the new matrices composed of the standardized embossing uniformity and radial distance are ∑E* i,k and ∑R* i,k , ∑E* i,k and ∑R* i,k Subtract to get the deviation matrix ∑Ds i,k , that is, the corresponding elements in the two matrices are subtracted to form a new matrix denoted as ∑Ds i,k ;

[0070] The eigenvector Chvor is obtained by multiplying the deviation matrix and its transposed matrix. Specifically, the m×m-order matrix obtained by multiplying the deviation matrix and its transposed matrix is ​​recorded as the characteristic square matrix. Since the characteristic square matrix is ​​a non-diagonal matrix, the m×1-order eigenvector Chvor is obtained;

[0071] The real-time homogeneity reference degree Rthr of the mth roller detection position at the current moment is calculated based on the overflow position feature matrix s :

[0072]

[0073] Where p is the cumulative variable, Derat sis the outlier rate of the s-th roller detection position, Chvor s is the element of the sth row of the eigenvector Chvor, n_k is the number of monitoring points in the overflow location feature matrix, EMcre s,p The embossing increase at the p-th monitoring point of the s-th roller detection position, ln() is a logarithmic function with the natural constant e as the base.

[0074] Furthermore, in step S500, the method for judging the state of the embossing roller based on the real-time uniformity reference degree is as follows: for any roller detection position, a sequence of real-time uniformity reference degrees obtained within 1 hour is recorded as a reference time series; a dynamic reference model of the operating state of the embossing roller within the reference time series is constructed: Bthr=MAthr+α·MSDthr; Bthr is the reference model value of any roller detection position, MAthr is the 5-minute moving average of the roller detection position, MSDthr is the 5-minute moving standard deviation, α is a sensitivity adjustment factor for controlling the sensitivity of the baseline, and its value is 1; the difference between the real-time uniformity reference degree and the reference model value at any roller detection position is recorded as a uniformity deviation; the Z-score of each uniformity deviation within 1 hour is calculated, and if the Z-score is greater than 3, the corresponding uniformity deviation is considered to be an abnormal uniformity deviation; two integer value variables are preset, which are recorded as the deviation amount kd and the risk roller position threshold nov; kd defaults to 3, and nov defaults to 1;

[0075] When the number of abnormal uniform deviations at any roller detection position is greater than the deviation amount, the roller detection position is recorded as a risk roller position; when the number of risk roller positions is greater than the risk roller position threshold, the embossing roller state is judged to be abnormal; otherwise, the embossing roller state is judged to be qualified.

[0076] When the embossing roller is in abnormal condition, an alternative embossing roller should be selected for replacement.

[0077] The embodiment of the present invention provides an online monitoring system for embossing process of sanitary paper products based on machine vision, such as Figure 2 Shown is a structural diagram of the online monitoring system for the embossing process of sanitary paper products based on machine vision of the present invention. The online monitoring system for the embossing process of sanitary paper products based on machine vision of this embodiment includes: a processor, a memory, and a computer program stored in the memory and capable of running on the processor. When the processor executes the computer program, the steps in the above-mentioned embodiment of the online monitoring method for the embossing process of sanitary paper products based on machine vision are implemented.

[0078] The system includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to run in the following units of the system:

[0079] An image acquisition unit, used for acquiring image data in real time to obtain an embossed paper product image;

[0080] An image preprocessing unit, configured to preprocess the embossed paper product image to form a processed image;

[0081] An image feature extraction unit, configured to extract features from the processed image to obtain an embossed image feature vector;

[0082] A quality assessment unit, used to dynamically assess the uniformity quality through the embossing image feature vector to form a real-time uniformity reference;

[0083] The feedback judgment transmission unit is used to judge the state of the embossing roller based on the real-time uniformity reference degree.

[0084] The machine vision-based online monitoring system for the embossing process of sanitary paper products can be run on computing devices such as desktop computers, laptops, PDAs, and cloud servers. The system can include, but is not limited to, a processor and memory. Those skilled in the art will appreciate that the examples described are merely illustrative of the machine vision-based online monitoring system for the embossing process of sanitary paper products and do not limit the entire system. The system can include more or fewer components than the example, or a combination of certain components, or different components. For example, the machine vision-based online monitoring system for the embossing process of sanitary paper products can also include input and output devices, network access devices, buses, and the like.

[0085] The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the operation system of the online monitoring system for the embossing process of sanitary paper products based on machine vision, and utilizes various interfaces and lines to connect various parts of the entire operation system of the online monitoring system for the embossing process of sanitary paper products based on machine vision.

[0086] The memory can be used to store the computer programs and / or modules. The processor implements the various functions of the machine vision-based sanitary paper embossing process online monitoring system by running or executing the computer programs and / or modules stored in the memory and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required for a function (such as a sound playback function, an image playback function, etc.); the data storage area can store data created based on the use of the mobile phone (such as audio data, a phone book, etc.). In addition, the memory can include a high-speed random access memory and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (SMC), a secure digital (SD) card, a flash card, at least one disk storage device, a flash memory device, or other volatile solid-state storage device.

[0087] Although the present invention has been described in considerable detail and with particularity with respect to several embodiments, it is not intended to limit the present invention to any of these details or embodiments or any particular embodiment, so as to effectively encompass the intended scope of the present invention. In addition, the present invention has been described above with respect to embodiments foreseen by the inventors for the purpose of providing a useful description, and those insubstantial modifications of the present invention that are not currently foreseen may still represent equivalent modifications of the present invention.

Claims

1. A machine vision-based online monitoring method for embossing process of sanitary paper products, characterized in that: The method comprises the following steps: S100, real-time acquisition of image data to obtain an embossed paper product image; S200, pre-processing the embossed paper product image to form a processed image; S300, performing feature extraction on the processed image to obtain an embossed image feature vector; S400, dynamically evaluates the uniformity quality through the embossing image feature vector to form a real-time uniformity reference; S500, judging the state of the embossing roller according to the real-time uniformity reference; In step S300, feature extraction is performed on the processed image to obtain a feature vector of the embossing image. The method is as follows: a plurality of positions are selected on the embossing die cavity of the embossing roller as roller detection positions, and the distances between the roller detection positions on the same busbar of the embossing roller are the same; each roller detection position is assigned a corresponding number; for each roller detection position corresponding to any number, a deep learning algorithm is used to identify all embossed areas corresponding to the number in the processed image; the maximum value of the grayscale values ​​corresponding to each pixel in an embossing area is used as the embossing grayscale; the average value and the range value of the embossing grayscale corresponding to all embossing areas are recorded as the embossing uniformity and radial distance, respectively; and a binary group consisting of the embossing uniformity and radial distance is used as a positioning feature. The sequence of positioning features of all roller detection positions under a processing graph is recorded as the embossing image feature vector; In step S400, the method for dynamically evaluating the uniformity quality by using the embossing image feature vector to form a real-time uniformity reference is as follows: each time the embossing image feature vector is obtained is recorded as a monitoring point; the range of the monitoring point is limited to the previous 30 minutes; For any monitoring point, an empty sequence is constructed and recorded as the sub-average sequence Hru.Ls. If the embossing average of a positioning feature is greater than or equal to the median value of all embossing averages at that monitoring point, the positioning feature is added to the sequence Hru.Ls. The length of the sequence Hru.Ls is recorded as SuH. The positioning feature corresponding to the minimum amplitude distance in the sequence Hru.Ls is recorded as the instantaneous reference position. The instantaneous reference positions within the monitoring point range constitute a set Tss, and the average two-tuple of all positioning features of Tss is calculated and recorded as the instantaneous horizontal feature; the Euclidean distance between any positioning feature and the instantaneous horizontal feature is calculated and recorded as the mean amplitude deviation Avrm; Take any monitoring point as the current monitoring point: For any roller detection position, obtain the positioning feature corresponding to the maximum average amplitude deviation among all positioning features between the current monitoring point and the SuH×TD period in the reverse time direction, and record it as the reverse reference position; where TD is the data collection interval; compare the positioning feature with the maximum average amplitude deviation between the instantaneous reference position and the reverse reference position, and record the absolute value of the difference between the obtained positioning feature and the embossing uniformity of the roller detection position at the current monitoring point as the uniformity difference component Ovd; The real-time homogeneity reference degree Rtue of any roller detection position is calculated based on the instantaneous reference position and the reverse reference position: ； Among them, i1 is the serial number of the monitoring point, Rtue i1 is the real-time homogeneity reference of the i1th monitoring point, Ovd i1 and Avrm i1 is the average difference and average amplitude deviation of the i1th monitoring point, H.Avrm i1 and B.Avrm i1 The average amplitude deviation of the inverse reference position and the instantaneous reference position corresponding to the i1th monitoring point respectively.

2. The online monitoring method of the embossing process of sanitary paper products based on machine vision is characterized in that: The method comprises the following steps: S100, real-time acquisition of image data to obtain an embossed paper product image; S200, pre-processing the embossed paper product image to form a processed image; S300, performing feature extraction on the processed image to obtain an embossed image feature vector; S400, dynamically evaluates the uniformity quality through the embossing image feature vector to form a real-time uniformity reference; S500, judging the state of the embossing roller according to the real-time uniformity reference; In step S300, feature extraction is performed on the processed image to obtain a feature vector of the embossing image. The method is as follows: a plurality of positions are selected on the embossing die cavity of the embossing roller as roller detection positions, and the distances between the roller detection positions on the same busbar of the embossing roller are the same; each roller detection position is assigned a corresponding number; for each roller detection position corresponding to any number, a deep learning algorithm is used to identify all embossed areas corresponding to the number in the processed image; the maximum value of the grayscale values ​​corresponding to each pixel in an embossing area is used as the embossing grayscale; the average value and the range value of the embossing grayscale corresponding to all embossing areas are recorded as the embossing uniformity and radial distance, respectively; and a binary group consisting of the embossing uniformity and radial distance is used as a positioning feature. The sequence of positioning features of all roller detection positions under a processing graph is recorded as the embossing image feature vector; In step S400, the uniformity quality is dynamically evaluated by the embossing image feature vector to form a real-time uniformity reference method as follows: set a time period as the monitoring period MOTM, set the current MOTM period to include m roller detection positions and n monitoring points, obtain an m×n order two-dimensional positioning feature matrix Mri(i, j), record the embossing uniformity in the positioning feature input by the jth monitoring point of the i-th roller detection position as E i,j , and its constituent matrix is ​​recorded as ∑E i,j , the radial distance is abbreviated as R i,j , and its constituent matrix is ​​recorded as ∑R i,j ; The embossing increase is calculated as EMcre(i,j)=(E i,j -E i,j-1 ) / TD, where j>1; for any roller detection position, draw a line graph of the embossing amplitude with respect to the monitoring point and record it as a time series graph, obtain the ratio of the number of monitoring points between any inflection point of the time series graph and the first inflection point in the reverse time direction to all monitoring points, and record the inverse of the obtained ratio as the steady-state parameter STpar corresponding to the monitoring point between the two inflection points; The matrix ∑R i,j The interquartile range of all elements in is the divergence parameter, and the ratio of all monitoring points whose radial distance to the roller detection position is greater than the divergence parameter to the total number of monitoring points is the outlier rate; the m×1-order matrix composed of the outlier rates corresponding to m roller detection positions is the position outlier matrix Lomat. In the matrix ∑R i,j In the equation, if a radial distance is less than the divergence parameter, the monitoring point is defined as an overflow moment, and its number is recorded as n_k; the corresponding columns of each overflow moment are intercepted from Mri(i, j) to obtain an m×n_k order matrix, which is recorded as the overflow location feature matrix; The columns of the two-tuple in the overflow positioning feature matrix are normalized: for the c-th monitoring point, the average values ​​of the embossing uniformity and radial distance are AV_Eme and AV_Eme respectively. c and AV_Rad c , and the standard deviations are SD_Eme c and SD_Rad c , the standardized embossing uniformity and widthwise distance at the cth monitoring point at the i-th roller detection position are Emave*(i,c)=(E i,c -AV_Eme c ) / SD_Emec,Radis*(i,c)=(R i,c -AV_Rad c ) / SD_Rad c , the new matrices composed of the standardized embossing uniformity and radial distance are ∑E* i,k and ∑R* i,k , ∑E* i,k and ∑R* i,k Subtract to get the deviation matrix ∑Ds i,k ; The characteristic vector Chvor is obtained by multiplying the deviation matrix and its transposed matrix. The real-time homogeneity reference degree is calculated based on the overflow position characteristic matrix and the outlier rate of the roller detection position.

3. The on-line monitoring method for embossing process of sanitary paper products based on machine vision according to claims 1 and 2, characterized in that: In step S100, the method for collecting image data in real time to obtain the embossed paper product image is: the camera for collecting image data is any one of a CCD camera and a CMOS camera; the camera is arranged above the paper outlet after the paper leaves the embossing roller, so that the optical axis of the camera is perpendicular to the paper surface; image collection is performed every 5-10 seconds, and the collected image is recorded as the embossed paper product image.

4. The on-line monitoring method for embossing process of sanitary paper products based on machine vision according to claim 1 and 2, characterized in that: In step S200, the method for preprocessing the embossed paper product image to form a processed image is as follows: performing multi-stage filter processing on the embossed paper product image to achieve noise filtering, and performing grayscale processing, applying an adaptive histogram equalization algorithm to adjust the distribution of the grayscale values ​​of the image to enhance the contrast of the overall image; and performing grayscale enhancement on the pixels of the pattern edge line in the image through an edge enhancement algorithm to form a processed image.

5. The on-line monitoring method for embossing process of sanitary paper products based on machine vision according to claim 1 and 2, characterized in that: In step S500, the method for judging the state of the embossing roller based on the real-time homogeneity reference degree is as follows: for any roller detection position, a sequence of real-time homogeneity reference degrees obtained within one hour is recorded as a reference time series; a dynamic benchmark model of the embossing roller operating state within the reference time series is constructed: Bthr = MAthr + α⋅MSDthr; Bthr is the reference model value of any roller detection position, MAthr is the 5-minute moving average of the roller detection position, MSDthr is the 5-minute moving standard deviation, α is the sensitivity adjustment factor, and the difference between the real-time homogeneity reference degree at any roller detection position and the reference model value is recorded as the homogeneity deviation; the Z-score of each homogeneity deviation within one hour is calculated; if the Z-score is greater than 3, the corresponding homogeneity deviation is considered to be an abnormal homogeneity deviation; two integer-valued variables are preset, which are recorded as the deviation amount kd and the risk roller position threshold nov. When the number of abnormal uniform deviations at any roller detection position is greater than the deviation amount, the roller detection position is recorded as a risk roller position; when the number of risk roller positions is greater than the risk roller position threshold, the embossing roller state is judged to be abnormal; otherwise, the embossing roller state is judged to be qualified.

6. The online monitoring system for embossing process of sanitary paper products based on machine vision is characterized by: The online monitoring system for the embossing process of sanitary paper products based on machine vision includes: a processor, a memory, and a computer program stored in the memory and runnable on the processor. When the processor executes the computer program, the steps in the online monitoring method for the embossing process of sanitary paper products based on machine vision according to any one of claims 1 to 5 are implemented. The online monitoring system for the embossing process of sanitary paper products based on machine vision runs on computing devices such as desktop computers, laptop computers, PDAs, and cloud data centers.

Images