A method for analyzing the pressure uniformity of an impression roller based on indentation image features

Abstract

The present application belongs to the technical field of image data processing, and particularly relates to a method for analyzing pressure uniformity of an impression roller based on indentation image features, which comprises the following steps: obtaining an indentation grayscale image; determining local texture complexity of a pixel point; determining neighborhood gradient direction consistency and real noise disturbance index of the pixel point; determining an adaptive gradient threshold value of the pixel point; obtaining a filtered indentation grayscale image, extracting indentation width features, and evaluating pressure uniformity of the impression roller. The present application analyzes the gray value features and gradient direction features of the pixel points in the neighborhood of the pixel point, extracts the indentation under the interference of wear texture, overcomes the difficulty of distinguishing the wear texture and the indentation by using a fixed gradient threshold value in the traditional anisotropic diffusion filtering algorithm, filters out the wear texture noise interference and extracts the indentation by using the adaptive gradient threshold value, and improves the accuracy of the evaluation of the pressure uniformity of the impression roller.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image data processing. More particularly, the present application relates to a method for analyzing the pressure uniformity of an impression roller based on indentation image features. BACKGROUND

[0002] In the intaglio printing industry, the impression roller, as a key component for transferring ink from the intaglio plate to the surface of the substrate, its pressure distribution uniformity directly determines the quality of the printed matter. Especially for the impression device using a flexible rubber roller, long-term operation will cause uneven wear on the surface of the rubber roller, leading to printing slip, inaccurate overprint or color difference problems. Traditional pressure detection usually relies on offline measurement of indentation width, which is inefficient and cannot reflect the production status in real time. In recent years, online detection technology based on machine vision has gradually emerged, which can infer the pressure state by processing image data. However, the surface of the impression roller is not ideal smooth, the texture of its own material and the splashed ink noise in the production environment will form high-frequency interference on the image. An existing algorithm for image denoising while preserving edge features is anisotropic diffusion filtering algorithm, which can remove noise in smooth areas while suppressing diffusion in edge areas to maintain structural information, and has good application prospects.

[0003] However, when the anisotropic diffusion filtering algorithm is used to smooth the image, its diffusion coefficient usually depends on a fixed gradient threshold value, which determines the algorithm's discrimination standard for edges and noise. In the analysis of the pressure uniformity of the impression roller, the wear degree of the surface of the impression roller is not uniform in the axial direction. Due to long-term uneven stress, some areas of the rubber roller are severely worn, and the surface texture is rough, while some areas are lightly worn, and the surface is relatively flat. If a fixed gradient threshold value is used, in the rough texture area, a low threshold value will cause the algorithm to misjudge it as an effective edge and stop diffusion, so that the texture noise remains and interferes with the subsequent quantification of the pressure distribution. In the flat surface area, a high threshold value may cause the weak edge to be excessively smoothed, losing the key details reflecting the pressure uniformity, affecting the accuracy of the evaluation results of the pressure uniformity of the impression roller. SUMMARY

[0004] To address the technical problem that traditional anisotropic diffusion filtering algorithms rely on fixed gradient thresholds, making it difficult to accurately extract texture noise and thus reducing the accuracy of pressure uniformity assessment for impression rollers, this invention provides a method for analyzing the pressure uniformity of impression rollers based on indentation image features. The method includes: acquiring an indentation image of the contact area between the impression roller and the substrate, and preprocessing the indentation image to obtain an indentation grayscale image; determining the local texture complexity of a pixel based on the grayscale entropy and standard deviation of grayscale values ​​within the pixel's neighborhood; determining the consistency of the gradient direction of the pixel's neighborhood gradient based on the gradient numerical components of the pixels in the horizontal and vertical directions, and determining the true noise perturbation index of the pixel based on the consistency of the gradient direction and the local texture complexity; determining an adaptive gradient threshold for the pixel based on the true noise perturbation index; performing anisotropic diffusion filtering on the indentation grayscale image using the adaptive gradient threshold to obtain a filtered indentation grayscale image; extracting indentation width features from the filtered indentation grayscale image; and evaluating the pressure uniformity of the impression roller based on the indentation width features.

[0005] This invention highlights the characteristics of wear texture by calculating local texture complexity and utilizing the grayscale information entropy and grayscale value standard deviation of pixels in the neighborhood of a pixel. By calculating the consistency of the gradient direction in the neighborhood, it analyzes the local directional disorder and irregular shape of the wear texture region, achieving accurate identification of texture noise. By combining local texture complexity and the consistency of the gradient direction in the neighborhood to calculate the true noise disturbance index and adaptively correcting the benchmark gradient threshold, it enhances the anti-interference ability of the anisotropic diffusion filtering algorithm in complex backgrounds. By generating a filtered indentation grayscale image based on the adaptive gradient threshold and extracting indentation width features and evaluating pressure uniformity, it achieves the evaluation of the pressure distribution of the embossing roller, accurately identifying pressure unevenness and improving the accuracy and robustness of the analysis.

[0006] Preferably, the acquisition of the indentation image of the contact area between the impression roller and the substrate includes: taking pictures of the strip-shaped indentation area formed by the contact between the impression roller and the substrate using an industrial line scan camera while the impression roller is stationary, thereby obtaining an indentation image.

[0007] Preferably, the method for obtaining the grayscale information entropy in the neighborhood of the pixel is as follows: selecting a region centered on the pixel... The neighborhood of a pixel is constructed, and the probability of occurrence of each gray level is counted within the neighborhood of the pixel. The expression is calculated using information entropy, and the result is used as the gray information entropy within the neighborhood of the pixel.

[0008] Preferably, the local texture complexity satisfies the expression:

[0009] In the formula, local texture complexity of the pixel point local texture complexity of the pixel point total number of pixel points in the neighborhood of the pixel point local texture complexity of the pixel point gray value of the i-th pixel point in the neighborhood of the pixel point average gray value of all pixel points in the neighborhood of the pixel point max-min normalization function.

[0010] The application realizes the evaluation of the local texture complexity by constructing a positive correlation function relationship between the product of the gray information entropy of the pixel points in the neighborhood of the pixel point and the gray value standard deviation term of the pixel points in the neighborhood of the pixel point, the gray information entropy normalization term reflects the degree of confusion of the pixel gray distribution in the neighborhood of the pixel point, and the gray value standard deviation term amplifies the gray fluctuation response of the wear texture, so that the wear texture region calculates a larger local texture complexity, thereby providing a reliable texture feature basis for the calculation of the real noise disturbance index.

[0011] Preferably, the neighborhood gradient direction consistency degree satisfies the expression:

[0012] ; in the expression, neighborhood gradient direction consistency degree of the pixel point total number of pixel points in the neighborhood of the pixel point and respectively, the gradient value component of the i-th pixel point in the neighborhood of the pixel point in the horizontal direction and the vertical direction to prevent the extreme value of the denominator being 0.

[0013] The application realizes the evaluation of the neighborhood gradient direction consistency degree based on the analysis of the gradient value components of the pixel points in the neighborhood of the pixel point in the horizontal direction and the vertical direction, so that the real indentation edge region calculates a larger neighborhood gradient direction consistency degree, and the wear texture noise region calculates a smaller value due to the disorder of the direction, thereby providing a reliable basis for distinguishing the indentation edge and the wear texture noise.

[0014] Preferably, the real noise disturbance index satisfies the expression:

[0015] ; in the expression, real noise disturbance index of the pixel point ​​​​​​​​For pixels The local texture complexity, For pixels The degree of consistency of the neighborhood gradient directions It is a natural exponential function.

[0016] This invention constructs a composite function that includes an inverse correlation term of neighborhood gradient direction consistency and a positive correlation term of local texture complexity, thereby achieving the evaluation of the true noise perturbation index. The inverse correlation term of neighborhood gradient direction consistency reduces the perturbation index of the true indentation edge, while the positive correlation term of local texture complexity enhances the perturbation index of wear texture interference. This results in a larger true noise perturbation index being calculated for the wear texture noise region, while the true indentation region maintains a lower perturbation index, thus effectively distinguishing between indentation edges and wear texture noise.

[0017] Preferably, the adaptive gradient threshold satisfies the expression:

[0018] In the formula, For pixels Adaptive gradient threshold As the global baseline gradient threshold, The sensitivity coefficient is adjusted based on the threshold. For pixels The true noise disturbance index.

[0019] This invention dynamically adjusts the global baseline gradient threshold by using the real noise perturbation index, thus achieving adaptive gradient thresholding. In areas with strong wear texture interference, the threshold is automatically increased to suppress false filtering, while in areas with obvious indentation features, a lower threshold is maintained to preserve the true response, ensuring the accuracy of the filtered indentation grayscale image.

[0020] Preferably, the method for obtaining the filtered indentation grayscale image is as follows: an anisotropic diffusion filtering algorithm is executed based on the gradient threshold after pixel adaptation to update the grayscale values ​​of the pixels in the indentation grayscale image, and the result is used as the filtered indentation grayscale image.

[0021] Preferably, the step of extracting the indentation width feature from the filtered indentation grayscale image includes: performing binarization processing on the filtered indentation grayscale image to obtain a binarized indentation grayscale image, extracting the upper and lower boundary coordinates of the indentation strips in the binarized indentation grayscale image, and calculating the indentation width of each column in the indentation image based on the upper and lower boundary coordinates.

[0022] Preferably, the evaluation of the pressure uniformity of the impression roller includes: calculating the standard deviation of the indentation width of all columns in the indentation image, and in response to the standard deviation being greater than a preset uniformity threshold, the pressure distribution of the impression roller is not uniform.

[0023] The beneficial effects of the present application are that:

[0024] The present application solves the technical problem that the traditional anisotropic diffusion filtering algorithm is easy to misjudge the surface wear texture noise as the indentation in the environment of the indentation image of the impression roller, and leads to filtering distortion, by introducing an adaptive gradient threshold correction mechanism based on multi-feature fusion.

[0025] The present application establishes the internal mapping relationship between the local texture complexity and the wear texture region by analyzing the neighborhood gray information entropy and gray value standard deviation of the pixel points, and on this basis, further fuses the consistency degree of the neighborhood gradient direction, tightly couples the index reflecting the gray distribution and the index reflecting the gradient direction, and realizes the accurate identification of the wear texture region.

[0026] The present application can calculate the real noise disturbance index matching the local gray feature of each pixel point, and adaptively correct the reference gradient threshold, realize the accurate matching of the threshold and the local gray feature of the pixel point, automatically increase the threshold to filter out the noise for the wear texture interference region, and keep the threshold to completely retain for the real indentation region. Finally, the present application improves the accuracy and anti-interference ability of the indentation image processing of the impression roller, enhances the robustness of the pressure uniformity analysis, can timely and accurately identify the uneven pressure defects, and provides a solid technical support for the maintenance of the impression roller and the printing quality control. BRIEF DESCRIPTION OF DRAWINGS

[0027] Figure 1 The flowchart of the pressure uniformity analysis method of the impression roller based on the indentation image features in the present application is schematically shown;

[0028] Figure 2 The comparison chart of the measurement results of the indentation after the indentation gray image is processed by the fixed gradient threshold of the prior art and the adaptive gradient threshold of the present application. DETAILED DESCRIPTION

[0029] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0030] The specific embodiments of the present application will be described in detail below with reference to the drawings.

[0031] The present application discloses a pressure uniformity analysis method of an impression roller based on indentation image features, referring to Figure 1 , comprising steps S001 to S005, specifically:

[0032] S001: Collect the indentation image of the contact area between the impression roller and the substrate, and preprocess the indentation image to obtain the grayscale image of the indentation.

[0033] Specifically, with the impression roller stationary, an industrial line scan camera is used to photograph the strip-shaped indentation area formed by the contact between the impression roller and the substrate. The collected indentation image is then converted into a single-channel grayscale image to obtain the indentation grayscale image.

[0034] S002: Determine the local texture complexity of a pixel based on the grayscale information entropy and grayscale value standard deviation within the pixel's neighborhood.

[0035] It should be noted that, since traditional anisotropic diffusion filtering algorithms rely on fixed gradient thresholds, it is difficult to take into account the differentiated smoothing requirements of different wear areas on the surface of the impression roller. This leads to problems such as incomplete removal of texture noise or over-smoothing of effective indentation edges when processing areas with different wear textures, thereby reducing the accuracy of indentation width feature extraction. Based on the fact that severely worn areas usually exhibit highly disordered grayscale distribution and violent grayscale value oscillations in the local grayscale image of the indentation, this invention combines the grayscale information entropy and grayscale standard deviation in the neighborhood of the pixel to determine the local texture complexity of the pixel, which is used to comprehensively characterize the degree of grayscale distribution disorder and grayscale oscillation amplitude in the region where the pixel is located.

[0036] Specifically, selecting from the pixel as the center. The neighborhood of a pixel is constructed, with a total of 25 pixels in the neighborhood. The occurrence probability of each gray level is counted in the neighborhood of the pixel, and the gray information entropy in the neighborhood of the pixel is calculated using the information entropy calculation expression. In other embodiments, the implementer can set the size of the neighborhood window according to the actual implementation situation.

[0037] Specifically, the local texture complexity satisfies the expression:

[0038] ;

[0039] In the formula, For pixels The local texture complexity, For pixels Gray-scale information entropy within the neighborhood, The total number of pixels in the neighborhood of a given pixel. For pixels Within the neighborhood The grayscale value of each pixel For pixels The average grayscale value of all pixels in the neighborhood. It is the maximum-minimum normalization function.

[0040] In the formula, The larger the value, the more pixels there are. The more chaotic and disordered the grayscale value distribution of pixels within a neighborhood, the more likely the pixels are to be affected. The greater the likelihood that the neighboring region belongs to a worn texture area, the more likely the pixel is to be affected. The greater the local texture complexity, the better. The larger the value, the more pixels there are. The more drastic the fluctuation in the grayscale value of a pixel within its neighborhood, the more significant the pixel's performance. The greater the likelihood that a neighboring region contains high-contrast noise or worn textures, the more likely the pixel is to be affected. The greater the local texture complexity, the better.

[0041] S003: Based on the gradient numerical components of pixels in the horizontal and vertical directions within the neighborhood of a pixel, determine the degree of consistency of the gradient direction in the neighborhood of the pixel. Combining the degree of consistency of the gradient direction in the neighborhood of the pixel with the local texture complexity, determine the true noise perturbation index of the pixel.

[0042] It should be noted that because the effective edges in an indentation grayscale image—that is, the boundary between the indented and non-indented areas—have a high grayscale change rate and information entropy, the pixels in the effective edge regions of the indentation grayscale image also possess significant local texture complexity. This makes it difficult to distinguish between the effective edges and wear textures in the indentation grayscale image, which may lead to the erroneous treatment of key indentation edges as noise during subsequent anisotropic diffusion, resulting in distorted indentation width measurements. Based on image morphological characteristics, wear texture regions exhibit random spatial distribution, with gradient directions in their neighborhoods displaying chaotic isotropic characteristics. In contrast, effective edges in an indentation grayscale image have clear structural orientation, with gradient directions in their neighborhoods exhibiting high consistency. Therefore, this invention combines the numerical components of the gradient vector in the horizontal and vertical directions within the pixel's neighborhood to determine the consistency of the pixel's neighborhood gradient directions and the true noise perturbation index, thereby characterizing the probability that a pixel belongs to a true wear texture region.

[0043] Specifically, the consistency of gradient directions in the neighborhood satisfies the expression:

[0044] ;

[0045] In the formula, For pixels The degree of consistency of the neighborhood gradient directions The total number of pixels in the neighborhood of a given pixel. and Each pixel Within the neighborhood Gradient value components of a pixel point in horizontal and vertical directions, To prevent the minimum value of the denominator being 0, the minimum value of the denominator being 0 is set to 0.001 in this embodiment, and in other embodiments, the implementer can set it according to the actual implementation.

[0046] In the formula, the greater the value of the gradient direction consistency of the neighborhood of the pixel point is, the greater the gradient direction consistency of the pixel point is. The greater the value of the gradient direction consistency of the neighborhood of the pixel point is, the greater the gradient direction consistency of the pixel point is. The stronger the consistency of the gradient component direction of the pixel points in the neighborhood is, the greater the gradient direction consistency of the pixel point is. The greater the credibility of the gradient direction of all the pixel points in the neighborhood tending to be consistent is, the greater the gradient direction consistency of the pixel point is. The greater the gradient direction consistency of the neighborhood of the pixel point is.

[0047] Further, the real noise disturbance index satisfies the expression:

[0048] ;

[0049] In the formula, the greater the value of the real noise disturbance index of the pixel point is, the greater the real noise disturbance index of the pixel point is. The real noise disturbance index of the pixel point, The local texture complexity of the pixel point, The gradient direction consistency of the neighborhood of the pixel point, The natural exponential function. The greater the value of the real noise disturbance index of the pixel point is, the greater the real noise disturbance index of the pixel point is. The greater the credibility of the gradient direction of all the pixel points in the neighborhood tending to be consistent is, the greater the real noise disturbance index of the pixel point is. The greater the gradient direction consistency of the neighborhood of the pixel point is, the greater the real noise disturbance index of the pixel point is. The greater the value of the real noise disturbance index of the pixel point is, the greater the real noise disturbance index of the pixel point is.

[0050] The greater the credibility of the gradient direction of all the pixel points in the neighborhood tending to be consistent is, the greater the real noise disturbance index of the pixel point is. The greater the gradient direction consistency of the neighborhood of the pixel point is, the greater the real noise disturbance index of the pixel point is. The greater the credibility of the gradient direction of all the pixel points in the neighborhood tending to be consistent is, the greater the real noise disturbance index of the pixel point is. The greater the gradient direction consistency of the neighborhood of the pixel point is, the greater the real noise disturbance index of the pixel point is. The greater the credibility of the gradient direction of all the pixel points in the neighborhood tending to be consistent is, the greater the real noise disturbance index of the pixel point is. The greater the gradient direction consistency of the neighborhood of the pixel point is, the greater the real noise disturbance index of the pixel point is. The greater the credibility of the gradient direction of all the pixel points in the neighborhood tending to be consistent is, the greater the real noise disturbance index of the pixel point is. The greater the gradient direction consistency of the neighborhood of the pixel point is, the greater the real noise disturbance index of the pixel point is.

[0051] S004: According to the real noise disturbance index of the pixel point, determine the gradient threshold of the pixel point after self-adaption.

[0052] It should be noted that after obtaining the real noise disturbance index of the pixel point, the pixel point global reference gradient threshold is corrected according to the real noise disturbance index of the pixel point. The traditional anisotropic diffusion filtering algorithm usually presets a globally fixed gradient threshold for all pixel points. The real noise disturbance index is used to correct the global reference gradient threshold of the pixel point, so that the anisotropic diffusion filtering algorithm can more accurately filter out noise.

[0053] Specifically, the adaptive gradient threshold satisfies the expression:

[0054] ;

[0055] In the formula, is the real noise disturbance index of the pixel point , the adaptive gradient threshold, is a global reference gradient threshold, which is set to 30 in this embodiment, and in other embodiments, the implementer can set it according to the actual implementation, for example, when the surface of the impression roller is seriously worn and the image background noise is large, the global reference gradient threshold can be appropriately increased to enhance the noise removal ability of the anisotropic diffusion filtering, when the positioning accuracy of the indentation edge is required to be high and the weak pressure change characteristics need to be preserved, the global reference gradient threshold can be appropriately reduced to prevent the effective edge from being excessively smoothed, is a threshold adjustment sensitivity coefficient, which is set to 1.5 in this embodiment, and in other embodiments, the implementer can set it according to the actual implementation, for example, when the surface of the impression roller is complex and it is difficult to distinguish the texture noise from the edge, the threshold adjustment sensitivity coefficient can be appropriately increased to strengthen the suppression effect of the texture noise, when the surface of the impression roller is relatively flat or the stability of the algorithm is required to be high, the threshold adjustment sensitivity coefficient can be appropriately reduced to avoid the loss of edge information caused by excessive adjustment, is the real noise disturbance index of the pixel point .

[0056] In the formula, The larger the real noise disturbance index of the pixel point , the more likely it is that the region where the pixel point and its neighborhood are located belongs to the wear texture region, and the larger the adaptive gradient threshold of the pixel point , which ensures that the corresponding texture information of the pixel point can be completely filtered out.

[0057] S005: Anisotropic diffusion filtering is performed on the indentation grayscale image using the adaptive gradient threshold of the pixel point, to obtain a filtered indentation grayscale image, and the indentation width feature is extracted according to the filtered indentation grayscale image, and the pressure uniformity of the impression roller is evaluated based on the indentation width feature.

[0058] Specifically, the pressure uniformity of the impression roller is evaluated, comprising:

[0059] The indentation grayscale image is filtered by using an anisotropic diffusion filtering algorithm with preset time step, iteration number and conduction function. In each iteration filtering process, the gradient threshold of the pixel point after self-adaption is substituted into the pre-selected conduction function to calculate the conduction coefficient, and the conduction coefficient is used to update the grayscale value of the pixel point of the indentation grayscale image until the preset iteration number is reached, and the obtained result is taken as the filtered indentation grayscale image.

[0060] The filtered indentation grayscale image is binarized to obtain a binarized indentation grayscale image, the upper and lower boundary coordinates of the indentation strip in the binarized indentation grayscale image are extracted, and the indentation width of each column in the indentation image is calculated according to the upper and lower boundary coordinates.

[0061] It should be noted that the anisotropic diffusion filtering algorithm will traverse each column of pixel points in the binarized indentation grayscale image, and lock the uppermost white pixel point in each column of pixel points as the upper boundary and the lowermost white pixel point as the lower boundary. Therefore, for any column of pixel points, the upper and lower boundary coordinates exist in pairs, and if there is a break in the middle of the column of pixel points, i.e. no indentation, the indentation width of the column of pixel points is 0. This ensures that the number of upper and lower boundary coordinates of all columns of pixel points is always consistent and cannot be matched.

[0062] Based on the indentation width of all columns in the indentation image, the standard deviation of the indentation width of all columns in the indentation image is calculated, and in response to the standard deviation being greater than a preset uniformity threshold, the pressure distribution of the impression roller is uneven.

[0063] As Figure 2As shown, the figure intuitively shows the comparison of the measurement results of the indentation after the indentation grayscale image is processed by the prior art using a fixed gradient threshold and the present application using an adaptive gradient threshold. The horizontal axis of the figure is the axial position of the impression roller, and the vertical axis is the indentation width value. The horizontal dotted line running through the horizontal axis represents the actual value of the indentation width. The curve in the figure shows a significant high-frequency and violent oscillation and is greatly deviated from the horizontal dotted line, which corresponds to the measurement result after the prior art uses a fixed gradient threshold. Since the prior art cannot dynamically adjust the filtering strength according to the local texture features, the wear texture is misjudged as the edge or the real edge is blurred, so that the calculated width value shows great discreteness. On the contrary, the other curve in the figure closely fits the horizontal dotted line, has a smooth trend and a very small fluctuation amplitude, which corresponds to the measurement result after the present application uses an adaptive gradient threshold. This shows that the present application can filter out background wear interference and accurately lock the indentation boundary by modifying the fixed gradient threshold with the real noise disturbance index, so that the measurement result maintains high accuracy and stability in the entire axial range.

Claims

1. A method of impression roller pressure uniformity analysis based on indentation image features, characterized by, The method comprises the following steps: acquiring an indentation image of a contact area between the impression roller and the printing substrate, and preprocessing the indentation image to obtain an indentation grayscale image; determining a local texture complexity of a pixel point according to a grayscale information entropy and a grayscale value standard deviation in a neighborhood of the pixel point; determining a neighborhood gradient direction consistency of the pixel point according to gradient numerical components of the pixel points in the horizontal and vertical directions in the neighborhood of the pixel point, and determining a real noise disturbance index of the pixel point in combination with the neighborhood gradient direction consistency and the local texture complexity of the pixel point; determining an adaptive gradient threshold of the pixel point according to the real noise disturbance index of the pixel point; performing an anisotropic diffusion filtering on the indentation grayscale image by using the adaptive gradient threshold of the pixel point to obtain a filtered indentation grayscale image, extracting an indentation width feature according to the filtered indentation grayscale image, and evaluating the pressure uniformity of the impression roller based on the indentation width feature; The true noise disturbance index satisfies the following expression: In the formula, For pixels The true noise disturbance index. For pixels The local texture complexity, For pixels The degree of consistency of the neighborhood gradient directions It is a natural exponential function; The adaptive gradient threshold satisfies the expression: ; in the expression, is a pixel point the adaptive gradient threshold, is a global reference gradient threshold, is a threshold adjustment sensitivity coefficient; the method for obtaining the filtered indentation grayscale image comprises: performing an anisotropic diffusion filtering algorithm based on the adaptive gradient threshold of the pixel point to realize the updating of the grayscale value of the pixel point of the indentation grayscale image, and taking the obtained result as the filtered indentation grayscale image; the method for extracting the indentation width feature according to the filtered indentation grayscale image comprises: performing a binarization processing on the filtered indentation grayscale image to obtain a binarized indentation grayscale image, extracting upper and lower boundary coordinates of an indentation strip in the binarized indentation grayscale image, and calculating the indentation width of each column in the indentation image according to the upper and lower boundary coordinates; the method for evaluating the pressure uniformity of the impression roller comprises: calculating a standard deviation of the indentation widths of all columns in the indentation image, and determining that the pressure distribution of the impression roller is uneven in response to the standard deviation being greater than a preset uniformity threshold.

2. The impression roll pressure uniformity analysis method based on indentation image features according to claim 1, characterized in that, The method for acquiring the indentation image of the contact area between the impression roller and the printing substrate comprises: capturing a strip-shaped indentation area formed by the contact between the impression roller and the printing substrate by using an industrial line array camera in a stationary state of the impression roller to obtain the indentation image.

3. The method of embossing roll pressure uniformity analysis based on indentation image features according to claim 1, characterized in that, The method for obtaining the gray information entropy in the neighborhood of the pixel point is: taking the pixel point as the center, selecting The neighborhood of the pixel point is constructed, the occurrence probability of each gray level in the neighborhood of the pixel point is counted, and the calculated result is taken as the gray information entropy in the neighborhood of the pixel point by using an information entropy calculation expression.

4. The method of embossing roll pressure uniformity analysis based on indentation image features according to claim 1 or 3, characterized in that, The local texture complexity satisfies the expression: ； wherein, is the local texture complexity of the pixel point is the gray information entropy in the neighborhood of the pixel point is the total number of pixel points in the neighborhood of the pixel point is the gray value of the pixel point is the mean value of the gray values of all pixel points in the neighborhood of the pixel point is the max-min normalization function.​​​​​ 5. The impression roll pressure uniformity analysis method based on indentation image features according to claim 1, characterized in that, The neighborhood gradient direction consistency satisfies the expression: ； wherein, is the degree of consistency of the gradient direction of the neighborhood of the pixel point , is the total number of pixel points in the neighborhood of the pixel point , are the gradient value components in the horizontal direction and in the vertical direction, respectively, of the th pixel point in the neighborhood of the pixel point , is a minimum value to prevent the denominator from being zero.

Images