Package printing quality real-time monitoring method

Abstract

The invention discloses a real-time monitoring method for package printing quality, and belongs to the technical field of image processing, and the method comprises the following steps: S1, collecting a package image when printing is completed, carrying out the data cleaning of the package image, and generating a package detection image; s2, inputting the package detection image into an image characterization processor to obtain an edge characterization coefficient and a local characterization coefficient, and generating a characterization group of the package detection image; and S3, completing comparison between the package detection image and the sample image of the package printing quality real-time monitoring method by using the representation group, and determining a printing quality monitoring result of the package detection image. By monitoring the image parameters in the printing process in real time, the printing quality problem can be found and solved in time, and the percent of pass and the production efficiency of printed matter are improved.

Description

Technical Field

[0001] The invention belongs to the technical field of image processing, and in particular relates to a real-time monitoring method for packaging printing quality. Background Art

[0002] In the packaging printing industry, printing quality directly affects the aesthetics, durability and market acceptance of products. Traditional printing quality control methods usually rely on manual sampling, which is not only inefficient but also difficult to fully cover all printed products and is prone to missing quality problems. Therefore, it is particularly important to develop a method that can monitor printing quality in real time. Summary of the invention

[0003] In order to solve the above problems, the present invention proposes a real-time monitoring method for packaging printing quality.

[0004] The technical solution of the present invention is: a method for real-time monitoring of packaging printing quality comprises the following steps: S1. Collect the packaging image after printing, clean the packaging image data, and generate the packaging inspection image; S2, inputting the packaging inspection image into an image representation processor, obtaining edge representation coefficients and local representation coefficients, and generating a representation group of the packaging inspection image; S3. Use the representation group to compare the packaging inspection image with the packaging example image, and determine the printing quality monitoring result of the packaging inspection image.

[0005] Furthermore, S2 includes the following sub-steps: S21, inputting the edge of the packaging inspection image into the first representation unit of the image representation processor to obtain an edge representation coefficient; S22, inputting the remaining pixel points of the packaging inspection image except the edge into the second representation unit, and obtaining the local representation coefficient by using the edge representation coefficient; S23. Obtain a representation group of the packaging inspection image according to the edge representation coefficient and the local representation coefficient.

[0006] The beneficial effect of the above further scheme is: in the present invention, the fast nearest neighbor search algorithm matcher (FLANNMatcher) is a tool for feature matching in OpenCV, which can be used to calculate the similarity between images. When using FLANNMatcher to calculate image similarity, it is usually necessary to first use a feature extraction algorithm to extract the feature points of the image. Because the packaging detection image is collected in real time, there may still be noise after data cleaning, and the packaging example image is generally more standard. Therefore, the present invention constructs an image characterization processor to complete the extraction of feature points (i.e., characterization groups). The image characterization processor designed by the present invention contains two pixel processing models, a first characterization unit and a second characterization unit. The edge is processed by the first characterization unit with an activation function to obtain an edge characterization coefficient; the edge characterization coefficient is brought into the calculation of the local characterization coefficient, which can better integrate the pixel data of all pixel points of the packaging detection image to obtain a more representative characterization group.

[0007] Further, in S21, the expression of the first characterization unit B1 is: ; In the formula, q1 represents the edge strength of the first pixel in the edge of the packaging detection image, Indicates the edge of the package detection image The edge strength of each pixel, Indicates the edge of the package detection image The edge strength of each pixel, q H represents the edge strength of the Hth pixel in the edge of the package detection image, H represents the total number of pixels on the edge of the package detection image, σ(·) represents the activation function, max(·) represents the maximum value function, and α represents the learning rate of the first representation unit.

[0008] The beneficial effect of the above further solution is that: in the present invention, the edge strength is essentially the amplitude of the edge point gradient, which reflects the prominence of the image edge. The learning rate determines the step size or speed of the model weight update.

[0009] Further, in S22, the expression of the second characterization unit B2 is: ; In the formula, s k represents the maximum singular value of the Hessian matrix corresponding to the remaining k-th pixel, K represents the total number of remaining pixels in the package detection image except the edge, H represents the total number of pixels at the edge of the package detection image, and u k Represents the pixel value of the remaining k-th pixel, represents the pixel value of the hth pixel in the edge of the package detection image, cos(·) represents the Euclidean distance function between two pixel values, and e represents the exponent.

[0010] The beneficial effect of the above further scheme is: in the present invention, the Hessian matrix is ​​an important tool for describing the local gradient change of pixel points. In image processing, the Hessian matrix can be obtained by performing Gaussian smoothing on the image and then calculating its second-order partial derivative. The eigenvalues ​​of the Hessian matrix reflect the gradient change of the local area of ​​the pixel point and can be used to analyze the pixel data except the edge.

[0011] Furthermore, in S22, the calculation formula of the local characterization coefficient β is: ; Wherein, b represents the edge representation coefficient, B2 represents the second representation unit, and max(·) represents the maximum value function.

[0012] Further, S23 includes the following sub-steps: S231, calculating the overall control valve according to the edge characterization coefficient and the local characterization coefficient; S232. Pixel points that satisfy the pixel group characterization conditions constructed by the overall control valve are taken as the characterization group.

[0013] Furthermore, in S231, the calculation formula of the overall control valve φ is: ; In the formula, β represents the local representation coefficient, b represents the edge representation coefficient, and rand(0,1) represents the generation of a random number between 0 and 1.

[0014] Furthermore, in S232, the expression of the pixel group characterization condition is: ; In the formula, u j represents the pixel value of the jth pixel in the package detection image, It represents the mean pixel value of all pixels in the D field of the j-th pixel in the package inspection image, and U represents the mean pixel value of all pixels in the package inspection image.

[0015] Furthermore, in S3, a fast nearest neighbor search algorithm matcher is used to calculate the matching similarity between the representation group of the packaging inspection image and the packaging example image. If the matching similarity is greater than or equal to a threshold, the printing quality monitoring result of the packaging inspection image is qualified, otherwise it is unqualified.

[0016] The beneficial effect of the above further scheme is: in the present invention, the fast nearest neighbor search algorithm matcher is an algorithm based on the approximate nearest neighbor search, which can significantly improve the matching speed while maintaining a high matching accuracy, and is particularly suitable for large-scale data sets and real-time applications. Therefore, the present invention adopts this matcher in consideration of the large number of pixels in the image.

[0017] The beneficial effects of the present invention are as follows: the present invention takes into account the situation that there are many image pixels, adopts a fast nearest neighbor search algorithm matcher as an image matching tool, and also takes into account that the fast nearest neighbor search algorithm matcher has requirements for input pixel points, so the real-time collected packaging inspection image is subjected to characterization group extraction to ensure that more matching-valued pixel points are obtained, thereby reducing interference in real-time collection; the present invention can timely discover and solve printing quality problems by real-time monitoring of image parameters in the printing process, thereby improving the qualified rate and production efficiency of printed products. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 The figure is a flow chart of the real-time monitoring method for packaging printing quality. DETAILED DESCRIPTION

[0019] The embodiments of the present invention will be further described below in conjunction with the accompanying drawings.

[0020] like Figure 1 As shown, the present invention provides a method for real-time monitoring of packaging printing quality, comprising the following steps: S1. Collect the packaging image after printing, clean the packaging image data, and generate the packaging inspection image; S2, inputting the packaging inspection image into an image representation processor, obtaining edge representation coefficients and local representation coefficients, and generating a representation group of the packaging inspection image; S3. Use the representation group to compare the packaging inspection image with the packaging example image, and determine the printing quality monitoring result of the packaging inspection image.

[0021] In this embodiment of the present invention, S2 includes the following sub-steps: S21, inputting the edge of the packaging inspection image into the first representation unit of the image representation processor to obtain an edge representation coefficient; S22, inputting the remaining pixel points of the packaging inspection image except the edge into the second representation unit, and obtaining the local representation coefficient by using the edge representation coefficient; S23. Obtain a representation group of the packaging inspection image according to the edge representation coefficient and the local representation coefficient.

[0022] In the present invention, the fast nearest neighbor search algorithm matcher (FLANNMatcher) is a tool for feature matching in OpenCV, which can be used to calculate the similarity between images. When using FLANNMatcher to calculate image similarity, it is usually necessary to first use a feature extraction algorithm to extract the feature points of the image. Because the packaging detection image is collected in real time, there may still be noise after data cleaning, and the packaging example image is generally more standard. Therefore, the present invention constructs an image characterization processor to complete the extraction of feature points (i.e., characterization groups). The image characterization processor designed by the present invention contains two pixel processing models, a first characterization unit and a second characterization unit. The edge is processed by the first characterization unit with an activation function to obtain an edge characterization coefficient; the edge characterization coefficient is brought into the calculation of the local characterization coefficient, which can better integrate the pixel data of all pixel points of the packaging detection image to obtain a more representative characterization group.

[0023] In the embodiment of the present invention, in S21, the expression of the first characterization unit B1 is: ; In the formula, q1 represents the edge strength of the first pixel in the edge of the package detection image, Indicates the edge of the package detection image The edge strength of each pixel, Indicates the edge of the package detection image The edge strength of each pixel, q H represents the edge strength of the Hth pixel in the edge of the package detection image, H represents the total number of pixels on the edge of the package detection image, σ(·) represents the activation function, max(·) represents the maximum value function, and α represents the learning rate of the first representation unit.

[0024] In the present invention, edge strength is essentially the magnitude of the gradient of the edge point, which reflects the prominence of the edge of the image. The learning rate determines the step size or speed of the model weight update.

[0025] In the embodiment of the present invention, in S22, the expression of the second characterization unit B2 is: ; In the formula, s k represents the maximum singular value of the Hessian matrix corresponding to the remaining k-th pixel, K represents the total number of remaining pixels in the package detection image except the edge, H represents the total number of pixels at the edge of the package detection image, and u k Represents the pixel value of the remaining k-th pixel, represents the pixel value of the hth pixel in the edge of the package detection image, cos(·) represents the Euclidean distance function between two pixel values, and e represents the exponent.

[0026] In the present invention, the Hessian matrix is ​​an important tool for describing the local gradient changes of pixels. In image processing, the Hessian matrix can be obtained by performing Gaussian smoothing on the image and then calculating its second-order partial derivative. The eigenvalues ​​of the Hessian matrix reflect the gradient changes of the local area of ​​the pixel and can be used to analyze the pixel data except the edge.

[0027] In the embodiment of the present invention, in S22, the calculation formula of the local characterization coefficient β is: ; Wherein, b represents the edge representation coefficient, B2 represents the second representation unit, and max(·) represents the maximum value function.

[0028] In this embodiment of the present invention, S23 includes the following sub-steps: S231, calculating the overall control valve according to the edge characterization coefficient and the local characterization coefficient; S232. Pixel points that satisfy the pixel group characterization conditions constructed by the overall control valve are taken as the characterization group.

[0029] In the embodiment of the present invention, in S231, the calculation formula of the overall control valve φ is: ; In the formula, β represents the local representation coefficient, b represents the edge representation coefficient, and rand(0,1) represents the generation of a random number between 0 and 1.

[0030] In the embodiment of the present invention, in S232, the expression of the pixel group characterization condition is: ; In the formula, u j represents the pixel value of the jth pixel in the package detection image, It represents the mean pixel value of all pixels in the D field of the j-th pixel in the package inspection image, and U represents the mean pixel value of all pixels in the package inspection image.

[0031] In an embodiment of the present invention, in S3, a fast nearest neighbor search algorithm matcher is used to calculate the matching similarity between the representation group of the packaging inspection image and the packaging example image. If the matching similarity is greater than or equal to a threshold, the printing quality monitoring result of the packaging inspection image is qualified, otherwise it is unqualified.

[0032] In the present invention, the fast nearest neighbor search algorithm matcher is an algorithm based on the approximate nearest neighbor search, which can significantly improve the matching speed while maintaining a high matching accuracy. It is particularly suitable for large-scale data sets and real-time applications. Therefore, the present invention adopts this matcher in consideration of the large number of pixels in the image.

[0033] Those skilled in the art will appreciate that the embodiments described herein are intended to help readers understand the principles of the present invention, and should be understood that the protection scope of the present invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific variations and combinations that do not deviate from the essence of the present invention based on the technical revelations disclosed by the present invention, and these variations and combinations are still within the protection scope of the present invention.

Claims

1. A real-time monitoring method for packaging printing quality, It is characterized in that The following steps are involved: S1. Collect the packaging image after printing, clean the packaging image data, and generate the packaging inspection image; S2, inputting the packaging inspection image into an image representation processor, obtaining edge representation coefficients and local representation coefficients, and generating a representation group of the packaging inspection image; S3. Use the representation group to compare the packaging inspection image with the packaging example image, and determine the printing quality monitoring result of the packaging inspection image.

2. The method for real-time monitoring of packaging printing quality according to claim 1, It is characterized in that The S2 comprises the following sub-steps: S21, inputting the edge of the packaging inspection image into the first representation unit of the image representation processor to obtain an edge representation coefficient; S22, inputting the remaining pixel points of the packaging inspection image except the edge into the second representation unit, and obtaining the local representation coefficient by using the edge representation coefficient; S23. Obtain a representation group of the packaging inspection image according to the edge representation coefficient and the local representation coefficient.

3. The method for real-time monitoring of packaging printing quality according to claim 2, It is characterized in that In S21, the first characterization unit B 1 The expression is: ; In the formula, q 1 Indicates the edge strength of the first pixel in the edge of the package detection image. Indicates the edge of the package detection image The edge strength of each pixel, Indicates the edge of the package detection image The edge strength of each pixel, q H represents the edge strength of the Hth pixel in the edge of the package detection image, H represents the total number of pixels on the edge of the package detection image, σ(·) represents the activation function, max(·) represents the maximum value function, and α represents the learning rate of the first representation unit.

4. The method for real-time monitoring of packaging printing quality according to claim 2, It is characterized in that In S22, the second characterization unit B 2 The expression is: ; In the formula, s k represents the maximum singular value of the Hessian matrix corresponding to the remaining k-th pixel, K represents the total number of remaining pixels in the package detection image except the edge, H represents the total number of pixels at the edge of the package detection image, and u k Represents the pixel value of the remaining k-th pixel, represents the pixel value of the hth pixel in the edge of the package detection image, cos(·) represents the Euclidean distance function between two pixel values, and e represents the exponent.

5. The method for real-time monitoring of packaging printing quality according to claim 2, It is characterized in that In S22, the calculation formula of the local characterization coefficient β is: ; In the formula, b represents the edge representation coefficient, B 2 represents the second representation unit, and max(·) represents the maximum value function.

6. The method for real-time monitoring of packaging printing quality according to claim 2, It is characterized in that The S23 comprises the following sub-steps: S231, calculating the overall control valve according to the edge characterization coefficient and the local characterization coefficient; S232. Pixel points that satisfy the pixel group characterization conditions constructed by the overall control valve are taken as the characterization group.

7. The method for real-time monitoring of packaging printing quality according to claim 6, It is characterized in that In S231, the calculation formula of the overall control valve φ is: ; In the formula, β represents the local representation coefficient, b represents the edge representation coefficient, and rand(0,1) represents the generation of a random number between 0 and 1.

8. The method for real-time monitoring of packaging printing quality according to claim 6, It is characterized in that In S232, the expression of the pixel group characterization condition is: ; In the formula, u j represents the pixel value of the jth pixel in the package detection image, It represents the mean pixel value of all pixels in the D field of the j-th pixel in the package inspection image, and U represents the mean pixel value of all pixels in the package inspection image.

9. The method for real-time monitoring of packaging printing quality according to claim 1, It is characterized in that In S3, a fast nearest neighbor search algorithm matcher is used to calculate the matching similarity between the representation group of the packaging inspection image and the packaging example image. If the matching similarity is greater than or equal to a threshold, the printing quality monitoring result of the packaging inspection image is qualified, otherwise it is unqualified.

Images