Machine vision method and system for adaptive ink supply optimization in gravure printing of decorative paper

Abstract

The invention relates to the field of machine vision, and discloses a machine vision method and system for self-adaptive ink supply optimization in gravure printing of decorative paper. The method first collects an image sequence of the anilox surface morphology of an anilox roller for gravure printing of decorative paper to be optimized, obtains initial three-dimensional point cloud data, and obtains a three-dimensional model of smooth dots; then extracts dot morphological parameters, constructs an ink supply optimization model to predict the initial ink supply; finally, starts gravure printing and collects visual images of the anilox roller surface in real time, calculates the overall ink uniformity and defect rate, and corrects the initial ink supply accordingly; the invention can dynamically adjust the ink supply according to the actual printing state of the dots, thereby effectively improving the printing quality and efficiency of gravure printing of decorative paper.

Description

Technical Field

[0001] The present invention relates to the field of machine vision, and more particularly to a machine vision method and system for optimizing the adaptive ink supply amount in gravure printing of decorative paper. Background Art

[0002] With the continuous development of decorative paper gravure printing technology, precise control of ink supply has become a key factor in ensuring printing quality and efficiency. Traditional ink supply control mainly relies on manual experience, which is difficult to adapt to changes in different printing needs, and has low adjustment accuracy and efficiency. To solve this problem, some researchers have begun to use machine vision and machine learning technology to optimize the control process of ink supply.

[0003] The Chinese patent with publication number CN117240977A discloses a method and system for controlling the ink supply of a printing press, which uses image processing and machine learning technology to extract and analyze multi-scale semantic features of the image to be printed, and predicts the ink gap width through a decoder, thereby realizing automatic adjustment of the ink supply. This method can reduce manual operations and improve adjustment accuracy and efficiency. However, this method mainly analyzes the semantic features of the printed image and fails to consider the influence of the surface morphology of the anilox roller on the ink transfer and inking quality. Its applicability in decorative paper gravure printing needs further verification.

[0004] Most existing technologies focus on the analysis of printed images, while ignoring the intrinsic relationship between the microscopic morphology of the anilox roller surface and the ink supply and inking quality in gravure printing of decorative paper. The geometric parameters of the anilox roller surface, such as the opening area, depth, and volume, directly determine the ink storage and transfer capacity, but how to quantitatively characterize these parameters and build an ink supply optimization model based on them is still a problem to be solved. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art, the present invention provides a machine vision method and system for adaptively optimizing ink supply in gravure printing of decorative paper, which can accurately reconstruct the three-dimensional morphology of the textured surface, quantitatively extract dot morphology parameters, construct an ink supply optimization model, and realize dynamic visual monitoring and adaptive correction of ink supply during the printing process, thereby significantly improving the printing quality and efficiency of gravure printing of decorative paper.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] Machine vision approach for adaptive ink delivery optimization in gravure printing of decorative paper, including:

[0008] Collecting a sequence of anilox surface morphology images of an anilox roller for gravure printing of decorative paper to be optimized, and obtaining initial three-dimensional point cloud data of the anilox surface; obtaining a smooth dot three-dimensional model based on the initial three-dimensional point cloud data; extracting dot morphology parameters based on the smooth dot three-dimensional model, and obtaining first dot morphology parameters;

[0009] A set of anilox surface morphology image sequences of anilox rollers for gravure printing of different decorative papers are collected, and a set of dot morphology parameters is obtained according to the set of anilox surface morphology image sequences; an ink supply optimization model is constructed according to the dot morphology parameter set; and an initial ink supply is predicted according to the first dot morphology parameter and the ink supply optimization model;

[0010] With an initial ink supply, gravure printing is started, the surface visual image of the anilox roller is collected in real time, the local area image blocks of each dot on the surface of the anilox roller are extracted, and the inking uniformity and defect rate of each dot are calculated; according to the inking uniformity of each dot, an overall inking uniformity index U is generated; according to the defect rate of each dot, an overall defect rate index R is generated; if the overall inking uniformity index U is less than the preset uniformity threshold θ1, or the overall defect rate index R is greater than the preset defect rate threshold θ2, the correction of the initial ink supply is triggered, and the corrected ink supply value is obtained.

[0011] Furthermore, the obtaining of initial three-dimensional point cloud data of the textured surface includes: splicing and fusing the textured surface topography image sequence, and calculating the spatial coordinates of each pixel point based on the spliced ​​and fused textured surface topography image sequence to obtain the initial three-dimensional point cloud data of the textured surface.

[0012] Furthermore, obtaining a smooth dot three-dimensional model according to the initial three-dimensional point cloud data includes:

[0013] Correcting the initial three-dimensional point cloud data to the z-axis direction of the mesh normal to obtain corrected initial three-dimensional point cloud data; performing smoothing filtering on the corrected initial three-dimensional point cloud data to obtain smoothed initial three-dimensional point cloud data;

[0014] The smoothed initial 3D point cloud data is reconstructed, and the dot surface is fitted using the Poisson reconstruction algorithm to obtain a smooth dot 3D model composed of continuous triangular facets.

[0015] Furthermore, obtaining a smooth dot three-dimensional model composed of continuous triangular facets includes:

[0016] The smoothed initial three-dimensional point cloud data is used as the input of the Poisson reconstruction algorithm to establish an octree spatial index structure; the octree spatial index structure includes m2 leaf nodes;

[0017] For each leaf node O of the octree spatial index structure (j), calculate the leaf node O (j) The normal vectors of all the initial 3D point cloud data in the leaf node O are obtained. (j) The mean normal vector N (j) ; Among them, 1≤j≤m2;

[0018] The normal vector mean N based on the octree leaf node (j) , construct Poisson's equation;

[0019] Solve Poisson's equation to obtain the scalar field function;

[0020] Triangulate the scalar field function, extract the isosurface, and generate a three-dimensional mesh model composed of a large number of triangular facets;

[0021] The three-dimensional mesh model is processed to obtain the final smooth dot three-dimensional model.

[0022] Furthermore, the first dot morphology parameters include dot opening area S, dot depth D and dot volume V;

[0023] The extracting of dot morphological parameters based on the smooth dot three-dimensional model includes:

[0024] Calculate the maximum curvature value k1 and the minimum curvature value k2 of each point on the surface of the smooth dot 3D model;

[0025] The k1 and k2 of each point on the surface of the smooth dot 3D model are combined into a cluster feature vector v=[k1,k2], and each point on the surface of the smooth dot 3D model is classified according to the cluster feature vector v to extract the dot area and the non-dot area;

[0026] For the extracted dot area, fit the dot contour curve, calculate the area enclosed by the dot contour curve, and obtain the dot opening area S;

[0027] Extract the Z coordinate value set of all points from the dot area ,from Search for the minimum value point Z of the Z coordinate value min ; Extract the Z coordinate value set of all points from the non-dot area ,calculate The mean of all Z coordinate values ​​in , the minimum point Z min With the mean The difference is taken as the dot depth D;

[0028] All triangular facets in the dot area are extracted, and the spatial coordinates of the triangular facets are obtained according to the spatial coordinates of each pixel point in the initial three-dimensional point cloud data; based on the spatial coordinates of the triangular facets, the dot volume V is calculated.

[0029] Furthermore, the calculating of the dot volume V includes:

[0030] Extract all triangles in the dot area and mark them as ; m1 is the total number of triangles in the dot area;

[0031] For Triangles , extract the spatial coordinates of its three vertices , ;in, ;

[0032] Calculate the triangle patch based on the spatial coordinates of the three vertices Area ;

[0033] Space coordinates and area based on three vertices , calculate the triangle patch Volume contribution ;

[0034] The volume contribution of all triangles is summed up to get the mesh volume V.

[0035] Furthermore, the extracting of the local area image blocks of each dot on the surface of the anilox roller comprises:

[0036] The surface visual image of the anilox roller is segmented into dot areas and non-dot areas by using an image segmentation algorithm based on region growth.

[0037] The minimum bounding rectangle of each dot area is extracted, and the image area within the minimum bounding rectangle is used as the local area image block of the dot.

[0038] Furthermore, the calculation of the ink uniformity and defect rate of each dot includes:

[0039] Binarization is performed on the local area image block of each dot to obtain a dot binary image;

[0040] Calculate the standard deviation of the grayscale value of the foreground pixel in the dot binary image to obtain the ink uniformity of the dot; the ink uniformity is inversely proportional to the standard deviation of the grayscale value of the foreground pixel;

[0041] The pixels with holes or missing in the foreground pixels are marked as defective dot pixels, the number of defective dot pixels is counted, and the ratio of defective dot pixels to the total number of foreground pixels is calculated to obtain the defective dot rate.

[0042] Furthermore, the correction of the initial ink supply amount includes:

[0043] According to the overall ink uniformity index and the overall defect rate index, the ink supply correction value is calculated; according to the ink supply correction value, the initial ink supply is corrected to obtain the corrected ink supply value.

[0044] A machine vision system for optimizing the amount of ink supplied in gravure printing of decorative paper, which is used to implement the above-mentioned machine vision method for optimizing the amount of ink supplied in gravure printing of decorative paper, the system comprising:

[0045] 3D model building module: used to collect the image sequence of the anilox surface of the gravure printing anilox roller of the decorative paper to be optimized, and obtain the initial 3D point cloud data of the anilox surface; and obtain the smooth dot 3D model based on the initial 3D point cloud data;

[0046] Dot parameter extraction module: based on the smooth dot three-dimensional model, extract the dot morphological parameters to obtain the first dot morphological parameters;

[0047] Ink supply prediction module: used to collect a set of anilox surface morphology image sequences of anilox rollers for gravure printing of different decorative papers, obtain a set of dot morphology parameters according to the set of anilox surface morphology image sequences; construct an ink supply optimization model according to the dot morphology parameter set; predict the initial ink supply according to the first dot morphology parameter and the ink supply optimization model;

[0048] Ink supply correction module: used to adopt the initial ink supply, start gravure printing, collect the surface visual image of the anilox roller in real time, extract the local area image block of each dot on the surface of the anilox roller, and calculate the inking uniformity and defect rate of each dot; generate the overall inking uniformity index U according to the inking uniformity of each dot; generate the overall defect rate index R according to the defect rate of each dot; if the overall inking uniformity index U is less than the preset uniformity threshold θ1, or the overall defect rate index R is greater than the preset defect rate threshold θ2, the correction of the initial ink supply is triggered to obtain the corrected ink supply value.

[0049] An electronic device comprises a memory, a central processing unit and a computer program stored in the memory and executable on the central processing unit, wherein the central processing unit implements the above-mentioned machine vision method for optimizing the adaptive ink supply amount in gravure printing of decorative paper when executing the computer program.

[0050] A computer-readable storage medium stores a computer program, and when the computer program is executed, the machine vision method for optimizing the adaptive ink supply amount in gravure printing of decorative paper is implemented.

[0051] Compared with the prior art, the present invention has the following beneficial effects:

[0052] The present invention can obtain high-precision three-dimensional morphological information of the anilox surface, accurately characterize the geometric characteristics of the dots by constructing a smooth three-dimensional model of dots and extracting dot morphological parameters, and provide a reliable data basis for ink supply optimization. The ink supply optimization model is constructed using historical printing data, and the optimal initial ink supply is predicted in combination with the current dot morphological parameters, thereby avoiding the blindness of the traditional empirical setting method and improving the scientificity and accuracy of ink usage prediction. In the gravure printing process, by real-time acquisition of the visual image of the anilox roller surface and analysis of the dot inking uniformity and defect rate, printing defects are discovered in time, triggering the online correction of the ink supply, and realizing the dynamic closed-loop optimization control of the ink supply. The ink supply is adaptively adjusted, while ensuring the printing quality, ink waste is reduced, and printing costs are reduced. At the same time, the need for manual adjustment of ink and rework correction is reduced, and the automation and production efficiency of decorative paper gravure printing are improved. The introduction of machine vision methods makes the optimization of ink supply more intelligent, with higher precision and faster speed than the traditional manual adjustment method, which improves the printing quality of decorative paper as a whole and enhances the visual beauty and added value of the product. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0054] Figure 1 The principle flow chart of the machine vision method for optimizing the adaptive ink supply amount in gravure printing of decorative paper in the present invention;

[0055] Figure 2 A flow chart of a method for obtaining initial three-dimensional point cloud data of a textured surface in a machine vision method for optimizing adaptive ink supply in gravure printing of decorative paper according to the present invention;

[0056] Figure 3 A flow chart of a method for obtaining a smooth dot three-dimensional model according to initial three-dimensional point cloud data in a machine vision method for optimizing the adaptive ink supply amount in gravure printing of decorative paper according to the present invention;

[0057] Figure 4 A flow chart of a method for fitting a dot surface by a Poisson reconstruction algorithm in a machine vision method for optimizing the adaptive ink supply amount in gravure printing of decorative paper according to the present invention to obtain a smooth dot three-dimensional model composed of continuous triangular facets;

[0058] Figure 5A flow chart of a method for extracting local area image blocks of each dot in a machine vision method for optimizing the adaptive ink supply amount in gravure printing of decorative paper according to the present invention;

[0059] Figure 6 A flow chart of a method for calculating the inking uniformity and defect rate of each dot in a machine vision method for optimizing the adaptive ink supply amount in gravure printing of decorative paper according to the present invention;

[0060] Figure 7 This is a functional module diagram of a machine vision system for adaptively optimizing ink supply in gravure printing of decorative paper in the present invention. DETAILED DESCRIPTION

[0061] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0062] Example 1

[0063] See also Figure 1 As shown, this embodiment provides a machine vision method for optimizing the adaptive ink supply amount in gravure printing of decorative paper, including:

[0064] Step S1000, collecting a sequence of anilox surface topography images of an anilox roller for gravure printing of decorative paper to be optimized, and obtaining initial three-dimensional point cloud data of the anilox surface; obtaining a smooth dot three-dimensional model based on the initial three-dimensional point cloud data; extracting dot morphology parameters based on the smooth dot three-dimensional model, and obtaining first dot morphology parameters;

[0065] Furthermore, step S1000 includes:

[0066] Step S1100, collecting a sequence of anilox surface topography images of an anilox roller for gravure printing of decorative paper to be optimized, and obtaining initial three-dimensional point cloud data of the anilox surface;

[0067] Furthermore, if Figure 2 As shown, step S1100 includes:

[0068] Step S1110, scanning and imaging the surface of the gravure printing anilox roller of the decorative paper to be optimized, and acquiring a sequence of multiple frames of anilox surface topography images;

[0069] Step S1120, splicing and fusing the textured surface topography image sequence;

[0070] Step S1130, based on the spliced ​​and fused textured surface topography image sequence, calculate the spatial coordinates of each pixel point to obtain initial three-dimensional point cloud data of the textured surface.

[0071] Specifically, the purpose of collecting the surface morphology of the decorative paper gravure printing anilox roller in step S1100 is to obtain its three-dimensional geometric information, so as to provide the original data basis for the subsequent reconstruction of the three-dimensional model of the dots and the extraction of morphological parameters. The surface morphology of the anilox roller usually has a fine and complex three-dimensional structure, and the traditional two-dimensional imaging method is difficult to accurately characterize its spatial geometric characteristics. Therefore, it is necessary to use a special three-dimensional imaging technology to collect the spatial morphology information of the anilox roller surface.

[0072] In step S1110, the surface of the anilox roller is scanned and imaged to obtain a continuous multi-frame image sequence. Scanning imaging usually adopts non-contact three-dimensional imaging methods such as structured light and laser, by projecting a specific pattern of structured light or laser beam onto the surface of the object, and calculating the three-dimensional coordinates of the surface according to the change of reflected light. Through continuous scanning, a series of dense surface morphology images can be obtained, and each frame of the image reflects the three-dimensional shape of a local area on the surface of the object. During the scanning process, attention should be paid to keeping the relative position of the imaging device and the surface of the anilox roller stable, and there should be a certain degree of overlap between adjacent frames in the image sequence to facilitate subsequent image stitching.

[0073] Step S1120 performs stitching and fusion processing on the acquired image sequence. Since a single-frame image can only cover a limited area on the surface of the anilox roller, multiple frames of images need to be stitched together to obtain a complete surface morphology. The image stitching process needs to find the correspondence between adjacent image frames, which is usually achieved through methods such as feature point matching and external parameter estimation. The stitched image may have obvious seams and discontinuities, and the local areas need to be smoothly transitioned and fused to make the overall morphology after stitching more natural and continuous. Image fusion can be based on methods such as weighted averaging and multi-resolution decomposition to improve the smoothness and consistency of the overall morphology while maintaining detailed features.

[0074] Step S1130 calculates the three-dimensional point cloud coordinates from the stitched and fused image sequence. According to the imaging principle, each pixel corresponds to a three-dimensional coordinate point on the surface, and its spatial position can be calculated through information such as pixel grayscale, camera parameters, and optical path. Point cloud data is a set of discrete three-dimensional coordinate points that intuitively represents the spatial shape of the surface of an object. By calculating pixel by pixel, the three-dimensional point cloud data of the surface morphology of the anilox roller is finally obtained. Although the point cloud data has certain noise and missing, it basically reflects the macroscopic geometric shape of the surface and provides important raw data for subsequent model reconstruction. The greater the point cloud density, the richer the surface detail information contained, but it also means an increase in the amount of calculation, and a balance needs to be struck between accuracy and efficiency.

[0075] In summary, step S1100 obtains an image sequence of the surface morphology of the anilox roller through three-dimensional scanning imaging technology, and finally obtains three-dimensional point cloud data reflecting the surface geometric features through image stitching, fusion and three-dimensional coordinate calculation. This step obtains the original data of the three-dimensional shape of the anilox roller, which intuitively reflects the fine and complex dot structure in gravure printing, and is an important data basis for subsequent dot three-dimensional modeling, parameter extraction and even ink transfer analysis. This three-dimensional acquisition method overcomes the limitations of traditional two-dimensional imaging and provides a new way for accurate analysis and intelligent optimization of gravure printing quality.

[0076] Step S1200, obtaining a smooth dot 3D model based on the initial 3D point cloud data;

[0077] Furthermore, if Figure 3 As shown, step S1200 includes:

[0078] Step S1210, correcting the initial three-dimensional point cloud data to the z-axis direction of the mesh normal to obtain corrected initial three-dimensional point cloud data;

[0079] Step S1220, performing smoothing filtering on the corrected initial three-dimensional point cloud data to obtain smoothed initial three-dimensional point cloud data;

[0080] Step S1230, performing surface reconstruction on the smoothed initial three-dimensional point cloud data, fitting the dot surface by using a Poisson reconstruction algorithm, and obtaining a smooth dot three-dimensional model composed of continuous triangular facets.

[0081] Furthermore, if Figure 4 As shown, step S1230 includes:

[0082] Step S1231, using the smoothed initial three-dimensional point cloud data as the input of the Poisson reconstruction algorithm to establish an octree spatial index structure; the octree spatial index structure includes m2 leaf nodes;

[0083] Step S1232, for each leaf node O of the octree spatial index structure (j) , calculate the leaf node O (j) The normal vectors of all the initial 3D point cloud data in the leaf node O are obtained. (j) The mean normal vector N (j) ; Among them, 1≤j≤m2;

[0084] Step S1233: based on the coordinate information of the initial three-dimensional point cloud data and the normal vector mean N of the octree leaf node (j) , construct Poisson's equation;

[0085] Step S1234, solving Poisson's equation to obtain a scalar field function;

[0086] Step S1235, triangulating the scalar field function, extracting isosurfaces, and generating a three-dimensional mesh model composed of a large number of triangular facets;

[0087] Step S1236, mesh simplification, smoothing and detail processing are performed on the three-dimensional mesh model to obtain a final smooth mesh point three-dimensional model.

[0088] Specifically, the Poisson reconstruction algorithm is an implicit surface reconstruction method. Its core idea is to convert the three-dimensional point cloud data into the boundary conditions of the Poisson equation, obtain the spatial scalar field by solving the Poisson equation, and then extract the isosurface of the scalar field to finally obtain the reconstructed three-dimensional model surface. In order to accelerate the solution of the Poisson equation, it is necessary to discretize the three-dimensional space and establish an octree spatial index structure. Each node of the octree represents a cubic space area, and the entire three-dimensional space is recursively divided into eight sub-cubes to form a hierarchical tree structure. The root node of the octree corresponds to the largest cube containing all the three-dimensional point cloud data, and the leaf node corresponds to the smallest scale cube unit. Each leaf node contains all the three-dimensional point cloud data in the cube unit. Through the octree structure, the local space area where any three-dimensional point is located can be quickly located, thereby accelerating the solution process of the Poisson equation.

[0089] For each leaf node of the octree, the normal vectors of all three-dimensional points in the node are calculated, and these normal vectors are averaged to obtain the average normal vector of the leaf node. The normal vector represents the orientation of the surface at a certain point and is an important information to characterize the geometric features of the local surface. By averaging the normal vectors of the three-dimensional points in the leaf node, the overall orientation trend of the surface in the local area can be obtained. The Poisson equation is constructed using the coordinate information of the three-dimensional points and the average normal vector of the leaf node. The Poisson equation is a common partial differential equation and is widely used in physics, engineering and other fields. The three-dimensional point cloud data is regarded as the boundary condition of the Poisson equation. By solving the Poisson equation, a spatial scalar field function can be obtained, which corresponds to the reconstructed surface in three-dimensional space. The numerical optimization algorithm is used to solve the Poisson equation to obtain the spatial scalar field function. Commonly used numerical optimization algorithms include Gaussian elimination method, conjugate gradient method, etc. The value of the scalar field function in three-dimensional space is closely related to the shape of the reconstructed surface.

[0090] The obtained scalar field function is triangulated, specific isosurfaces are extracted, and a three-dimensional mesh model composed of a large number of triangular facets is generated. Triangulation is the process of discretizing the scalar field isosurfaces into triangular facets. The generated three-dimensional mesh model can better approximate the continuous surface. The generated three-dimensional mesh model is further post-processed, including mesh simplification, smoothing and detail processing. Mesh simplification is to reduce the number of triangular facets while maintaining the surface shape to improve the subsequent calculation efficiency. Mesh smoothing is to make the mesh surface smoother and smoother by adjusting the position of the triangular facet vertices, reducing noise and jaggedness. Detail processing is to restore or enhance the high-frequency details of the model surface to make the reconstructed three-dimensional model more realistic.

[0091] The Poisson reconstruction algorithm can effectively transform scattered 3D point cloud data into a continuous and smooth 3D model surface. The algorithm makes full use of the position and normal vector information of the 3D points, obtains the spatial scalar field function by solving the partial differential equation, and then extracts the isosurface to obtain the reconstructed 3D model. The Poisson reconstruction algorithm has strong robustness and tolerance to noise, and can process irregular and incomplete point cloud data to generate a 3D model with correct topology and smooth surface.

[0092] For example, taking a gravure printing dot as an example, the three-dimensional point cloud data of its surface is obtained by laser scanning. The point cloud data is irregularly distributed, the local density is uneven, and there is a certain amount of noise. It is difficult to obtain satisfactory results by directly reconstructing the surface of these original point cloud data. However, through the Poisson reconstruction algorithm, the point cloud data is first preprocessed, and the subsequent calculation is accelerated through the octree spatial index. Then the average normal vector of the point cloud is calculated in each local area to construct the Poisson equation. The spatial scalar field function is obtained by solving the Poisson equation, and then the isosurface is extracted to generate a three-dimensional mesh model. Finally, after mesh simplification, smoothing and detail processing, a three-dimensional model of dots with smooth surface and realistic geometric shape is obtained. This model not only maintains the overall shape characteristics of the original point cloud data, but also reasonably completes and smoothes the surface details, providing a high-quality three-dimensional model foundation for the subsequent extraction of dot morphological parameters.

[0093] In summary, the introduction of Poisson reconstruction algorithm into 3D point cloud processing in screen printing can effectively overcome the defects of original data and generate a smooth, complete and accurate 3D model of dots, laying a solid model foundation for subsequent morphological parameter extraction and ink supply optimization. Poisson reconstruction algorithm is a robust, efficient and intelligent 3D reconstruction method with broad application prospects in reverse engineering, computer vision, medical imaging and other fields.

[0094] In step S1210, the initial three-dimensional point cloud data is corrected and transformed to a point where the normal direction of the textured surface is the positive direction of the Z axis. The textured surface normal here refers to the average normal vector direction of the textured surface, which characterizes the spatial orientation of the textured surface as a whole. By aligning the coordinate system of the point cloud data with the textured surface normal, the influence of the tilt of the textured surface can be eliminated, which facilitates subsequent processing. The corrected point cloud data is more consistent with the textured surface in terms of spatial position and orientation.

[0095] Step S1220 performs smoothing filtering on the corrected point cloud data. Three-dimensional point cloud data usually contains certain noise and outliers. These noise points may come from the errors of the imaging device itself, or defects such as contamination and damage on the textured surface. The purpose of smoothing filtering is to remove or reduce these noise points to make the point cloud data smoother. Common point cloud smoothing filtering methods include moving least squares, bilateral filtering, Gaussian filtering, etc. Through smoothing filtering, the quality of point cloud data can be effectively improved, providing more reliable input for subsequent surface reconstruction.

[0096] In summary, step S1200 converts the initial three-dimensional point cloud data into a high-quality smooth dot three-dimensional model through point cloud correction, smoothing and surface reconstruction. This step fully considers the defects and limitations of the original point cloud data, and adopts a robust and intelligent processing method. While ensuring the overall shape characteristics of the dots, the local details are reasonably repaired and smoothed. The dot three-dimensional model obtained after processing in step S1200 has accurate geometric shape and smooth and complete surface, which is an important model basis for subsequent dot morphology parameter extraction and ink supply optimization. This step method is reliable and the mathematical principle is rigorous, which is of great significance to improving the quality of decorative paper gravure printing.

[0097] Step S1300 , extracting dot morphological parameters based on the smoothed dot three-dimensional model to obtain first dot morphological parameters, wherein the first dot morphological parameters include dot opening area S, dot depth D and dot volume V.

[0098] Further, step S1300 includes:

[0099] Step S1310, calculating the maximum curvature value k1 and the minimum curvature value k2 of each point on the surface of the smoothed dot 3D model;

[0100] Step S1320, k1 and k2 of each point on the surface of the smooth dot 3D model are combined into a cluster feature vector v=[k1, k2], and each point on the surface of the smooth dot 3D model is classified according to the cluster feature vector v to extract the dot area and the non-dot area;

[0101] Step S1330, fitting a dot contour curve to the extracted dot area, calculating the area enclosed by the dot contour curve, and obtaining a dot opening area S;

[0102] Step S1340: extract the Z coordinate value set of all points from the dot area ,from Search for the minimum value point Z of the Z coordinate value min ; Extract the Z coordinate value set of all points from the non-dot area ,calculate The mean of all Z coordinate values ​​in , the minimum point Z min With the mean The difference is taken as the dot depth D;

[0103] In a three-dimensional model, Z represents the height (or depth) coordinate of a point. In the dot area and non-dot area, all points have their own Z coordinate value.

[0104] For example, assume the model is as follows:

[0105] Network area Coordinate value collection: ;

[0106] Non-network area Coordinate value collection: ;

[0107] Calculate the mean of non-dot areas :

[0108] ;

[0109] Find the outlet area Minimum value of coordinates :

[0110] ;

[0111] Calculate dot depth :

[0112] .

[0113] Step S1350, extract all triangular facets in the dot area, obtain the spatial coordinates of the triangular facets according to the spatial coordinates of each pixel point in the initial three-dimensional point cloud data; calculate the dot volume V based on the spatial coordinates of the triangular facets.

[0114] Specifically, step S1300 extracts the morphological parameters of the dot area, including the opening area, depth and volume, etc., based on the reconstructed smooth dot 3D model. These parameters characterize the geometric morphological characteristics of the dots and are key factors affecting ink transfer and printing quality. Through quantitative parameter description, the corresponding relationship between the dot structure and the printing quality can be established, providing an important basis for the subsequent optimization of ink supply.

[0115] In steps S1310 and S1320, the dot area is extracted by using the curvature features of different areas of the dot surface. Curvature is an important geometric quantity that characterizes the local curvature of the surface. By calculating the principal curvature value of the three-dimensional point cloud, the shape characteristics of each point on the surface can be characterized. Specifically, for each surface point, its principal curvature in two orthogonal directions is calculated, corresponding to the maximum and minimum curvature of the surface, respectively, denoted as k1 and k2. The dot area usually corresponds to the concave part of the surface, and the curvature is large in all directions; while the non-dot area is relatively flat and has a small curvature value. Therefore, the two-dimensional feature vector composed of k1 and k2 can be used as the classification basis of the points, and a clustering algorithm (such as k-means) is used to divide all points into two categories: dot area and non-dot area. This method cleverly utilizes the second-order differential geometric feature of curvature to adaptively extract irregularly distributed and differently shaped dot surfaces. Compared with the traditional grayscale threshold segmentation method, this method can handle more complex and variable dot structures and has better robustness.

[0116] In step S1330, a contour curve is fitted in the extracted dot area, and the area enclosed by the curve is calculated as the opening area of ​​the dot. The size of the dot opening directly affects the amount of ink transferred and is a key geometric parameter that determines the printing concentration. Traditionally, the opening area is estimated by threshold segmentation of the two-dimensional dot image, but in the three-dimensional model, the spatial contour curve of the dot opening can be fitted more accurately. Specifically, the least squares curve fitting can be applied to the boundary points of the dot area to obtain a smooth and closed spatial curve, and then the area of ​​the area enclosed by the curve is calculated by curve integration. Compared with the area estimation of the two-dimensional contour, this method takes into account the actual shape of the dot opening in three-dimensional space, improves the accuracy of area calculation, and has obvious advantages, especially for dots with inclined and irregular openings.

[0117] In steps S1340 and S1350, the depth and volume parameters of the dots are calculated respectively. The dot depth characterizes the depth of the grooves on the gravure surface and affects the ink volume of the dots; the dot volume directly reflects the total amount of ink that can be accommodated in the dot grooves. The two together determine the transfer volume of the ink, which in turn affects the concentration and layer thickness of the printing. In the three-dimensional model, the height coordinates of each point in the dot area and the non-dot area can be easily obtained, and the minimum height of the dot area and the average height of the non-dot area can be obtained respectively. The difference between the two is the depth parameter of the dot. This method avoids the trouble of manually calibrating the reference surface and truly reflects the actual depth of the dots. For the calculation of the dot volume, the volume contribution of each triangular facet in the dot area is counted one by one, and they are added up to obtain the total dot volume. The discrete representation of triangular facets can handle dot grooves of arbitrary shapes without being restricted by their internal structure, and the calculation is simple and universal.

[0118] In summary, step S1300 extracts morphological parameters such as the opening area, depth and volume of the dot area based on the reconstructed three-dimensional model of the dots. The main curvature feature is used to realize the adaptive extraction of the dot area, the opening area is calculated by spatial curve fitting and integration, the dot depth is obtained by using height information, and the dot volume is estimated by discretizing the triangular patch. This step makes full use of the three-dimensional geometric information, improves the accuracy and robustness of the dot parameter extraction, and provides a reliable morphological basis for the subsequent ink transfer analysis. By quantitatively characterizing the dot structural characteristics, a mapping relationship is established between the printing quality problem and the three-dimensional morphology of the dots, which is of great significance for understanding the ink transfer mechanism, optimizing the gravure printing process, and improving the quality of printed products. It also expands the application of computer vision in image printing quality analysis.

[0119] Further, step S1350 includes:

[0120] Step S1351, extract all triangles in the dot area and mark them as ; m1 is the total number of triangles in the dot area;

[0121] Step S1352, for Triangles , extract the spatial coordinates of its three vertices , ;

[0122] Step S1353, calculate the triangle patch based on the spatial coordinates of the three vertices Area ;

[0123] Step S1354, based on the spatial coordinates and areas of the three vertices , calculate the triangle patch Volume contribution ;

[0124] The calculation of triangular patches Volume contribution include:

[0125] Set triangle patch The unit normal vector is ,but It is expressed as:

[0126]

[0127] in:

[0128] Represents a vertex To origin The directed distance in the normal direction projection on;

[0129] Represents a vertex To origin The directed distance in the normal direction projection on;

[0130] Represents a vertex To origin The directed distance in the normal direction projection on;

[0131] Unit Normal Vector It can be calculated by the cross product of two edge vectors of the triangle:

[0132]

[0133] Step S1355: sum up the volume contributions of all triangular facets to obtain the dot volume V.

[0134] Specifically, the surface of the smooth dot 3D model is composed of a large number of triangular patches, and the geometric information (vertex coordinates, normal vector) of each triangular patch is known. The total volume of the dot area can be obtained by adding up the volume contribution of each triangular patch. The area of ​​the triangular patch is obtained by Heron's formula, which only depends on the length of the three sides of the triangle. The length of the three sides can be calculated by the distance formula between two points. Therefore, the area of ​​the triangular patch can be easily calculated with only the coordinates of the three vertices of the triangle.

[0135] The volume contribution of a triangular facet depends on the area, vertex coordinates and unit normal vector. Its physical meaning is that the projection of the directed distance from the triangle vertex to the origin in the normal direction multiplied by the area, and then divided by 3, is equivalent to 1 / 3 of the volume of a triangular prism. Considering each triangular facet as a triangular prism, it is natural to accumulate the volumes of these triangular prisms to obtain a discrete representation of the total volume of the dots. Step S1350 makes full use of the known information of the triangular mesh on the surface of the three-dimensional model, and obtains the volume parameters of the dot area by accumulating the volume contribution of each triangular facet. This method has clear ideas, simple and efficient calculations, and the transformation from discrete models to continuous parameters is natural and reasonable, laying the foundation for the subsequent optimization of ink supply.

[0136] Step S2000, collecting a set of anilox surface topography image sequences of anilox rollers for gravure printing of different decorative papers, obtaining a set of dot morphology parameters according to the set of anilox surface topography image sequences; constructing an ink supply optimization model according to the set of dot morphology parameters; and predicting an initial ink supply according to the first dot morphology parameter and the ink supply optimization model;

[0137] Furthermore, step S2000 includes:

[0138] Step S2100, obtaining a dot morphology parameter set according to the textured surface topography image sequence set;

[0139] Step S2200, constructing an ink supply optimization model according to the dot morphology parameter set;

[0140] Further, step S2200 includes:

[0141] Step S2110, taking the dot morphology parameter set as input and the ink supply amount as output, to construct a training sample set of the ink supply amount optimization model;

[0142] Step S2120: Based on the training sample set, a Gaussian radial basis kernel function is used as the kernel function of the support vector machine to construct an ink supply optimization model.

[0143] Specifically, when constructing an ink supply optimization model, it is necessary to first prepare a training sample set. The training sample consists of two parts: the input is a dot morphology parameter set, including dot opening area S, dot depth D, and dot volume V; the output is the corresponding ink supply Q. These training samples can be obtained by extracting dot morphology parameters and measuring ink supply for different anilox rollers.

[0144] Support Vector Machine (SVM) is a commonly used machine learning algorithm, especially suitable for small sample, nonlinear classification and regression problems. The core idea of ​​SVM is to construct the optimal classification hyperplane or regression hyperplane in high-dimensional space to maximize the interval between samples or predicted values ​​of different categories and the true value. In the ink supply optimization model, the Gaussian radial basis kernel function is used to map the dot morphology parameters to the high-dimensional feature space, and then the optimal regression hyperplane is constructed in the space to achieve the nonlinear mapping from the dot morphology parameters to the ink supply. The Gaussian radial basis kernel function can map the original feature space to the infinite-dimensional feature space, thereby enhancing the nonlinear expression ability of SVM.

[0145] When training the SVM model, two important parameters need to be optimized: the kernel function scale parameter σ and the penalty coefficient C. σ controls the nonlinear mapping ability of the kernel function. The larger σ is, the wider the Gaussian function is, and the smoother the feature space after mapping is. On the contrary, the smaller σ is, the sharper the feature space after mapping is. The penalty coefficient C controls the complexity of the SVM model and the tolerance for misclassified samples. The larger C is, the more complex the SVM model is, and the lower the tolerance for misclassified samples is. On the contrary, the smaller C is, the lower the complexity of the SVM model is, and the higher the tolerance for misclassified samples is.

[0146] In order to obtain the optimal σ and C, the cross-validation method is used to optimize the parameters. Common cross-validation methods include K-fold cross-validation and leave-one-out cross-validation. Taking K-fold cross-validation as an example, the training sample set is randomly divided into K subsets of similar size. One of the subsets is selected as the validation set each time, and the remaining K-1 subsets are used as training sets. The SVM model is trained and the model performance is evaluated on the validation set. Repeat K times so that each subset has the opportunity to be used as a validation set. Finally, the average of the K validation results is taken as the generalization performance indicator of the model. Through methods such as grid search or random search, different combinations of σ and C are traversed, and the parameter combination with the best generalization performance is selected to obtain the final ink supply optimization model.

[0147] The construction of the ink supply optimization model integrates the techniques of dot morphology parameter extraction, support vector machine regression and cross-validation, fully explores the intrinsic relationship between the three-dimensional morphology of dots and ink transfer, and provides a strong model support for the precise control of ink supply. The model has the advantages of strong nonlinear expression ability, good generalization performance, and applicability to small sample data, laying a solid foundation for the realization of intelligent and refined control of gravure printing.

[0148] Step S2300: predicting the initial ink supply amount according to the first dot morphology parameter and the ink supply amount optimization model.

[0149] Specifically, step S2000 collects and analyzes a large amount of surface morphology data of the decorative paper gravure printing anilox roller, builds an ink supply optimization model, and predicts the initial ink supply based on the dot morphology parameters of the anilox roller to be optimized. This step aims to use historical data to mine the intrinsic relationship between dot structure and ink transfer, establish a quantitative ink supply optimization strategy, and provide an initial value reference for subsequent dynamic adjustment.

[0150] In step S2100, the surface morphology image sequences of different anilox rollers are processed to obtain a comprehensive and rich set of dot morphology parameters. Specifically, the sub-steps of step S1000 are sequentially executed for the image sequence of each anilox roller to extract key parameters such as dot opening area, depth, and volume, and the parameters of each anilox roller are summarized into a large parameter set. This parameter set covers a variety of screen types, dot layout methods, and plate-making processes, and is highly representative and diverse. Through the accumulation of massive morphological parameter data, the distribution characteristics of the dot structure in the gravure printing of decorative paper can be fully characterized, laying a data foundation for the construction of subsequent optimization models.

[0151] Step S2200 uses the dot morphology parameter set to construct an ink supply optimization model through data-driven methods such as machine learning. The model aims to reveal the quantitative relationship between the dot geometry and the ink transfer effect, and provide a theoretical basis and calculation method for setting the ink supply. The key to building an optimization model is feature selection and algorithm design. First, from the massive dot morphology parameters, key parameters closely related to ink transfer are extracted as model inputs, such as the statistical characteristics of dot opening area, depth, and volume distribution. Then, a suitable machine learning algorithm, such as support vector regression and neural network, is used to train the ink supply prediction model. The output of the model is a quantitative value of the ink supply, which reflects the ideal state of ink transfer under a specific dot structure. Through model training, a nonlinear mapping relationship between dot morphology parameters and ink supply is established, providing a powerful tool for subsequent parameter optimization.

[0152] In step S2300, the constructed ink supply optimization model is used in combination with the dot morphology parameters of the anilox roller to be optimized obtained in step S1000 to predict the initial ink supply under the anilox roller. Specifically, the parameters such as the dot opening area, depth, and volume of the anilox roller to be optimized are input into the optimization model, and the theoretical optimal value of the ink supply is obtained through forward calculation, which is used as the initial reference value for subsequent dynamic adjustment. This prediction process can be regarded as an example application of the optimization model, which concretizes the abstract mathematical relationship into the actual production scenario. Since the optimization model integrates the empirical knowledge of a large amount of historical data, it can give a reasonable reference value of the ink supply under the new anilox roller parameters, reducing the cost of manual trial and error and improving production efficiency. At the same time, the initial prediction value is not fixed, and it will be dynamically adjusted according to feedback in the subsequent printing process to adapt to the fluctuation of the actual ink transfer state.

[0153] A simple example of predicting the initial ink supply is as follows: Assume that after the anilox roller to be optimized is processed in step S1000, the dot opening area S=1000μm 2， Dot depth D = 30 μm, dot volume V = 2 × 10 4 μm 3 These parameters are input into the trained ink supply optimization model, and the initial ink supply Q0 = 1.5 ml / m is obtained through forward calculation of the model. 2 This quantitative ink supply prediction value provides an optimized initial state for the subsequent printing production of the anilox roller, reduces the time cost of blind debugging, and improves production efficiency and print quality.

[0154] In summary, step S2000 constructs a quantitative optimization model between dot morphology parameters and ink supply through in-depth mining of historical data and machine learning modeling, and predicts the initial ink supply based on the measured parameters of the anilox roller to be optimized. This data-driven optimization method makes full use of the empirical knowledge contained in massive production data, and applies it to new production instances through technical means such as machine learning. It has strong intelligence and practicality. The initial optimization of ink supply is conducive to shortening process debugging time, reducing trial and error costs, and improving the production efficiency and print quality of decorative paper gravure printing. At the same time, this method provides reliable initial values ​​for subsequent dynamic feedback control, and is an important link in realizing intelligent and precise printing production.

[0155] Step S3000, using the initial ink supply, starting gravure printing, collecting the surface visual image of the anilox roller in real time, extracting the local area image blocks of each dot on the surface of the anilox roller, and calculating the inking uniformity and defect rate of each dot; generating an overall inking uniformity index U according to the inking uniformity of each dot; generating an overall defect rate index R according to the defect rate of each dot; if the overall inking uniformity index U is less than the preset uniformity threshold θ1, or the overall defect rate index R is greater than the preset defect rate threshold θ2, then triggering the correction of the initial ink supply, and obtaining the corrected ink supply value.

[0156] Furthermore, step S3000 includes:

[0157] Step S3100, performing dot area segmentation on the surface visual image of the anilox roller to extract local area image blocks of each dot;

[0158] Furthermore, if Figure 5 As shown, step S3100 includes:

[0159] Step S3110, using a region growing-based image segmentation algorithm to segment the surface visual image of the anilox roller into halftone dot areas and non-halftone dot areas;

[0160] Step S3120 , extracting the minimum bounding rectangle of each halftone dot area, and using the image area within the minimum bounding rectangle as the local area image block of the halftone dot.

[0161] Specifically, after starting the gravure printing equipment, the initial ink supply predicted by the ink supply optimization model is set to the ink supply system of the gravure printing equipment. As the printing process begins, the anilox roller starts to rotate and continuously obtains ink from the ink tank and transfers the ink to the surface of the substrate. In order to monitor the printing quality in real time, a high-speed camera is used to image the surface of the anilox roller. The imaging frequency of the high-speed camera must match the rotation speed of the anilox roller to ensure that the dynamic changes of the anilox surface can be fully recorded.

[0162] The acquired visual image may have some problems such as noise, distortion and uneven lighting, which require image preprocessing. When preprocessing the visual image of the anilox roller surface, median filtering and other methods are first used to remove noise interference in the image and improve image quality. Then, histogram equalization, Laplace sharpening and other operations are used to enhance image contrast and edge details. Finally, geometric correction technology is used to eliminate perspective distortion and lens distortion of the image, so that the image can restore the geometric shape of the real scene. The preprocessed image provides a good input for subsequent dot area segmentation.

[0163] The region growing method is a region-based image segmentation algorithm. Its basic idea is to start from a set of seed pixels and gradually merge neighboring pixels with similar properties to the seed pixels into the region where the seed pixels are located according to the pixel similarity criterion until they can no longer be expanded. In the dot area segmentation, the center position of the dot or the edge position of the dot can be selected as the initial seed pixel, and then the region growth is performed based on the similarity measurement criteria of grayscale, texture and other features, and finally a complete dot area is obtained. The region growing method can effectively overcome the interference caused by image noise and uneven lighting, and obtain a closed and semantically complete dot area.

[0164] In order to further improve the accuracy of dot area segmentation, it is necessary to perform morphological post-processing on the segmentation results. The opening operation (corrosion followed by expansion) can eliminate the noise area with a smaller area, making the dot area cleaner and more complete. The closing operation (expansion followed by corrosion) can smooth the dot boundary, fill the small holes in the dot area, and obtain a more regular dot area. Morphological processing enhances the robustness and adaptability of dot area segmentation.

[0165] After obtaining the segmented dot area, it is necessary to extract the local area image block of each dot for subsequent feature calculation and analysis. A commonly used method is to calculate the minimum bounding rectangle of the dot area, that is, the rectangle with the smallest area that can completely surround the dot area. The minimum bounding rectangle can not only determine the position and size of the dot, but also exclude the background area around the dot to obtain a local image block containing only the dot itself. When extracting the minimum bounding rectangle, the rotating caliper algorithm, rectangle measurement and other methods can be used to quickly and efficiently calculate the optimal bounding rectangle.

[0166] The extraction of the local area image block of the dot lays the foundation for the subsequent printing quality analysis. By focusing on the local image of each dot, the interference between the dots can be eliminated, and the ink uniformity and incompleteness of the dots can be evaluated more accurately. At the same time, the size of the local image block of the dot can be flexibly adjusted according to the analysis requirements to adapt to quality inspections of different scales and precisions. The local area image block of the dot is the key to achieving a refined evaluation of the printing quality on the surface of the anilox roller.

[0167] Take a typical visual image of the surface of anilox roller as an example. The image contains a large number of dot areas and non-dot areas. The shapes and sizes of dots are different and unevenly distributed. First, the image is subjected to median filtering denoising and histogram equalization enhancement to make the contrast between the dot area and the background area more obvious. Then, the center position of the dot is used as the seed point, and the region growing method is used for image segmentation. The dot area is expanded based on the grayscale similarity criterion to obtain the preliminary dot segmentation result. Then, the segmentation result is subjected to morphological processing by opening and closing operations to eliminate small noise areas and smooth the dot boundaries. Finally, the minimum circumscribed rectangle of each dot area is extracted to obtain a local image block containing the complete dot shape and texture. Through the above steps, the original visual image of the surface of anilox roller is converted into a series of dot local area image blocks. Each image block focuses on a single dot, which provides accurate and reliable input for the subsequent calculation of ink uniformity and defect rate. The analysis method based on local area image blocks not only improves the accuracy of printing quality evaluation, but also greatly reduces the computational complexity, and has good practical value.

[0168] Through dot area segmentation and local image block extraction, printing quality analysis moves from overall to local, from qualitative to quantitative, providing new ideas and methods for the refined quality control of the anilox roller surface.

[0169] Step S3200, calculating the ink uniformity and defect rate of each halftone dot according to the local area image block of each halftone dot;

[0170] Furthermore, if Figure 6 As shown, step S3200 includes:

[0171] Step S3210, binarizing the local area image block of each halftone dot to obtain a halftone dot binarized image;

[0172] Step S3220, calculating the standard deviation of the grayscale values ​​of the foreground pixels in the dot binary image to obtain the ink uniformity of the dots; the ink uniformity is inversely proportional to the standard deviation of the grayscale values ​​of the foreground pixels;

[0173] Step S3230, marking pixels with holes or missing in the foreground pixels as defective dot pixels, counting the number of defective dot pixels, calculating the ratio of defective dot pixels to the total number of foreground pixels, and obtaining the defective dot rate.

[0174] Specifically, binarization is the process of converting a grayscale image into an image containing only black and white pixel values, with the goal of separating the target area of ​​interest from the background area. In the binarization of the image block in the local area of ​​the halftone dots, the global threshold method or the local adaptive threshold method is usually used. The global threshold method selects the threshold between the two peaks as the segmentation threshold for binarization based on the bimodal characteristics of the image grayscale histogram. The local adaptive threshold method takes into account the differences in grayscale distribution in different areas of the image, and calculates a local threshold for each pixel based on the grayscale statistical characteristics within the pixel neighborhood, thereby obtaining a more accurate binarization result. The binarization process simplifies the image block in the local area of ​​the halftone dots into black and white, highlighting the printing area of ​​the halftone dots.

[0175] Ink uniformity is used to measure the consistency of ink distribution inside the dot. Ideally, the ink inside the dot should be evenly distributed, and the grayscale value should change smoothly. In order to quantify the ink uniformity, the standard deviation of the grayscale value of the foreground pixel in the dot binary image can be calculated. The standard deviation reflects the degree of dispersion of the pixel grayscale value around the average value. The larger the standard deviation, the more uneven the ink distribution inside the dot; conversely, the smaller the standard deviation, the more uniform the ink distribution inside the dot. By setting the threshold of ink uniformity, defective dots with uneven ink can be quickly screened out.

[0176] Dot defect refers to defects such as holes and missing pixels that appear in the dot printing area, which will lead to a decrease in the quality of the printed image. In order to detect and quantify dot defects, holes or missing pixels can be searched in the foreground pixels of the dot binary image. Hole pixels refer to background pixels (black) that appear inside the foreground pixels, while missing pixels refer to foreground pixels that are missing at the edge of the dot. The dot defect rate is obtained by counting the number of defective pixels and calculating their proportion to the total number of foreground pixels. The higher the defect rate, the worse the dot printing quality is, and it is necessary to adjust the printing parameters or replace the anilox roller in time.

[0177] The calculation of the ink uniformity and defect rate based on the image block of the local area of ​​the dots can not only accurately detect printing defects, but also greatly improve the analysis efficiency. By focusing on the local characteristics of each dot, mutual interference between dots is avoided, and a more accurate and reliable evaluation result is obtained. At the same time, this method is easy to automate and engineer. By setting parameters such as the ink uniformity threshold and the defect rate threshold, online monitoring and real-time early warning of the printing quality of the anilox roller surface can be achieved, so as to timely discover and solve printing problems and improve production efficiency and product quality. In addition, this method also provides data support for the design optimization of the anilox roller. By analyzing the distribution characteristics of the ink uniformity and defect rate on the surface of different anilox rollers, the parameters such as the shape, size, and arrangement of the dots can be optimized to improve the printing suitability and service life of the anilox roller.

[0178] Ink uniformity and defect rate are two key indicators that characterize the quality of dot printing. The calculation method is simple and efficient, with a high degree of automation. They have broad application prospects in gravure printing quality control and are of great significance for improving the quality and competitiveness of printed products such as decorative paper and aluminum foil paper.

[0179] Step S3300, generates an overall ink uniformity index U according to the ink uniformity of each dot; generates an overall defect rate index R according to the defect rate of each dot; if the overall ink uniformity index U is less than the preset uniformity threshold θ1, or the overall defect rate index R is greater than the preset defect rate threshold θ2, then the correction of the initial ink supply amount is triggered to obtain the corrected ink supply amount value.

[0180] Further, step S3300 includes:

[0181] Step S3310, combining the ink uniformity of the local area image block of each dot to generate an overall ink uniformity index U;

[0182] Step S3320, comprehensively analyzing the defect rates of the local area image blocks of each grid point to generate an overall defect rate index R;

[0183] Step S3330, preset a uniformity threshold θ1 and a defect rate threshold θ2, if U<θ1 or R>θ2, then trigger the correction of the initial ink supply amount, otherwise, no correction is triggered.

[0184] The correction of the initial ink supply amount includes:

[0185] According to the overall ink uniformity index and the overall defect rate index, the ink supply correction value is calculated; according to the ink supply correction value, the initial ink supply is corrected to obtain the corrected ink supply value.

[0186] The calculating of the ink supply correction value according to the overall ink uniformity index and the overall defect rate index comprises:

[0187]

[0188] in:

[0189] Indicates the ink supply correction value, which is the dependent variable of the formula and represents the adjustment amount to the initial ink supply. When it is positive, increase the ink supply; when When it is negative, reduce the ink supply. The value reflects the magnitude of the adjustment.

[0190] It represents the overall ink uniformity index, which is obtained by comprehensively analyzing the ink uniformity of the local area image block of each dot. The value range of is [0,1]. The closer to 1, the more uniform the overall ink application. It is the target value for ink uniformity.

[0191] It represents the overall defect rate index, which is calculated by comprehensively analyzing the defect rates of the local area image blocks of each dot. The value range of is [0,1]. The closer it is to 0, the lower the overall defect rate. is the target value of defect rate.

[0192] represents the dot opening area, which can be calculated in step S1330. and They are the minimum and maximum values ​​of the dot opening area respectively.

[0193] This is the initial ink supply amount.

[0194] is the weight of the overall inking uniformity index.

[0195] is the weight of the overall defect rate indicator.

[0196] It is a scale factor that controls the overall correction amplitude. The larger α is, the larger the correction amplitude is; and vice versa.

[0197] It is a regulatory factor that adjusts the combined effect of the two indicators, ink uniformity and defect rate. The larger β is, the stronger the combined effect of the two indicators is, and vice versa.

[0198] is the amplitude of the sine function. The sine term is introduced to consider the periodic effect of the dot opening area on the ink supply. When the area is medium, the sine term is 0 and no additional adjustment is made. When the area is too large or too small, the sine term provides positive or negative corrections respectively.

[0199] It is the sensitivity coefficient of ink uniformity, which controls the sensitivity of ink uniformity deviating from the target value.

[0200] It is the sensitivity coefficient of the defect rate, which controls the sensitivity of the defect rate to the target value. The larger the value, the more sensitive it is to the deviation from the target value, and the correction amplitude increases sharply with the increase of the deviation.

[0201] , , , , , and is a hyperparameter in the formula, which can be set empirically or optimized based on training data. The formula comprehensively considers the influence of factors such as overall ink uniformity, overall defect rate, and dot opening area on the ink supply, and constructs a nonlinear correction model through the combination of square root form and sine function. When the ink unevenness increases or the defect rate increases, the correction value increases accordingly, indicating that the ink supply needs to be adjusted to improve the printing quality. At the same time, the dot opening area is a periodic factor, and sinusoidal modulation is applied to the correction value, which reflects the periodic influence of area change on ink transfer.

[0202] In general, the formula can quantitatively guide the dynamic optimization and adjustment of ink supply according to printing quality evaluation indicators and key parameters, which helps to improve the quality stability and consistency of decorative paper gravure printing. The multiple hyperparameters introduced in the formula provide flexible adjustment freedom, which can be adapted and optimized according to actual process requirements and empirical data, and has strong adaptability and robustness.

[0203] Specifically, step S3300 dynamically adjusts the ink supply based on the obtained initial ink supply for printing by analyzing the actual printing effect, thereby achieving fine optimization control of printing quality. This step focuses on two quality evaluation indicators: ink uniformity and dot defect rate, quantitatively evaluating the printing quality from the two perspectives of ink distribution uniformity and dot structure integrity, and accordingly correcting the ink supply to continuously improve the printing effect.

[0204] In step S3310, the ink uniformity of the local area of ​​each dot is comprehensively analyzed to generate an overall ink uniformity index. Specifically, for each dot, an image block within its local neighborhood is extracted, and the grayscale mean and standard deviation within the image block are calculated. The grayscale mean reflects the overall ink concentration level of the local area, and the standard deviation measures the discrete degree of ink distribution. The combination of the two can more comprehensively characterize the uniformity of local ink application. The local uniformity index of each dot is weighted averaged to obtain the overall ink uniformity of the entire printing area. The setting of the weight takes into account factors such as the area and position of the dot, and highlights the influence of key areas. The overall uniformity index quantifies the overall ink quality of the actual printing effect, providing a basis for the subsequent adjustment of the ink supply.

[0205] Step S3320 quantitatively evaluates the integrity of the dot structure and generates an overall defect rate index. Dot defect is caused by insufficient ink supply, wear of the anilox roller, and other reasons, such as the lack of dot outline and breakage, and is an important factor affecting the quality of printed products. The lower the defect rate, the more complete the dot structure. The defect rate of each dot is weighted averaged to obtain the overall defect rate index of the entire printing area. Similar to the uniformity index, the weight of the defect rate index also takes into account factors such as the geometric characteristics and spatial distribution of the dots. The overall defect rate quantitatively reflects the overall integrity of the dot structure and is another key indicator for evaluating printing quality.

[0206] Based on the overall ink uniformity index and the overall defect rate index, the ink supply correction value is calculated. This correction value reflects the deviation between the actual printing quality and the ideal target, and is used to dynamically adjust the initial supply so that the printing state continuously approaches the optimal state. Specifically, the target thresholds of the uniformity index and the defect rate index are set respectively. When the actual index value deviates from the target threshold, the correction of the ink supply is triggered. For example, when the uniformity index is lower than the uniformity threshold, it means that the ink is uneven and the ink supply needs to be appropriately increased; and when the defect rate index is higher than the defect rate threshold, it means that the dots are seriously defective and the ink supply needs to be appropriately reduced.

[0207] Based on the ink supply correction value, the initial ink supply is corrected to obtain the corrected value. The corrected ink supply value is sent to the ink supply system of the gravure printing equipment, and the real-time dynamic control of the ink supply is achieved by adjusting the pump speed, valve opening, etc. The ink supply system automatically adjusts the rate and volume of ink delivered to the surface of the anilox roller according to the feedback correction value, thereby achieving the purpose of optimizing the printing quality. This is a closed-loop feedback control process. The quality indicators obtained through visual image analysis guide the optimization of ink supply parameters, forming a mutual feedback adjustment mechanism between gravure printing process parameters and printing quality.

[0208] It is worth mentioning that the dynamic feedback optimization strategy adopted in step S3300 closely combines the printing quality evaluation with the ink supply parameter adjustment to form a closed-loop control. This dynamic adjustment based on real-time feedback can continuously monitor and improve the printing quality, effectively deal with various interferences and fluctuations in the production process, and keep the printing in the best state. In addition, the definitions of the uniformity index and the defect rate index are scientific, the calculation is simple, and it is easy to implement in engineering. The combination of the two can comprehensively judge the printing quality and provide a reliable basis for the optimization of ink supply.

[0209] In general, step S3300 establishes a dynamic optimization method for ink supply for gravure printing of decorative paper. By obtaining real-time feedback on the quality of printed products and dynamically adjusting the ink supply parameters, continuous improvement and enhancement of printing quality and efficiency can be achieved on the basis of initial optimization. This method makes full use of measurable indicators such as uniformity and defect rate, converts qualitative quality judgment into quantitative parameter correction, and improves the accuracy and reliability of ink supply optimization. At the same time, dynamic adjustment based on feedback gives printing control adaptability, enabling it to respond to real-time changes in production status. Closed-loop optimization between printing quality and ink supply is a key measure to achieve intelligent and precise printing control, and is of great significance for improving the quality of decorative paper printing products, improving production efficiency and material utilization.

[0210] Example 2

[0211] This embodiment provides a machine vision system for optimizing the adaptive ink supply amount in gravure printing of decorative paper based on the embodiment 1, such as Figure 7 As shown, including:

[0212] 3D model building module: used to collect the image sequence of the anilox surface of the gravure printing anilox roller of the decorative paper to be optimized, and obtain the initial 3D point cloud data of the anilox surface; and obtain the smooth dot 3D model based on the initial 3D point cloud data;

[0213] Dot parameter extraction module: based on the smooth dot three-dimensional model, extract the dot morphological parameters to obtain the first dot morphological parameters;

[0214] Ink supply prediction module: used to collect a set of anilox surface morphology image sequences of anilox rollers for gravure printing of different decorative papers, obtain a set of dot morphology parameters according to the set of anilox surface morphology image sequences; construct an ink supply optimization model according to the dot morphology parameter set; predict the initial ink supply according to the first dot morphology parameter and the ink supply optimization model;

[0215] Ink supply correction module: used to adopt the initial ink supply, start gravure printing, collect the surface visual image of the anilox roller in real time, extract the local area image block of each dot on the surface of the anilox roller, and calculate the inking uniformity and defect rate of each dot; generate the overall inking uniformity index U according to the inking uniformity of each dot; generate the overall defect rate index R according to the defect rate of each dot; if the overall inking uniformity index U is less than the preset uniformity threshold θ1, or the overall defect rate index R is greater than the preset defect rate threshold θ2, the correction of the initial ink supply is triggered to obtain the corrected ink supply value.

[0216] In the three-dimensional model building module, the initial three-dimensional point cloud data of the textured surface is obtained, including:

[0217] Step S1110, scanning and imaging the surface of the gravure printing anilox roller of the decorative paper to be optimized, and acquiring a sequence of multiple frames of anilox surface topography images;

[0218] Step S1120, splicing and fusing the textured surface topography image sequence;

[0219] Step S1130, based on the spliced ​​and fused textured surface topography image sequence, calculate the spatial coordinates of each pixel point to obtain initial three-dimensional point cloud data of the textured surface.

[0220] In the three-dimensional model construction module, obtaining a smooth dot three-dimensional model according to the initial three-dimensional point cloud data includes:

[0221] Step S1210, correcting the initial three-dimensional point cloud data to the z-axis direction of the mesh normal to obtain corrected initial three-dimensional point cloud data;

[0222] Step S1220, performing smoothing filtering on the corrected initial three-dimensional point cloud data to obtain smoothed initial three-dimensional point cloud data;

[0223] Step S1230, performing surface reconstruction on the smoothed initial three-dimensional point cloud data, fitting the dot surface by using a Poisson reconstruction algorithm, and obtaining a smooth dot three-dimensional model composed of continuous triangular facets.

[0224] The step S1230 includes:

[0225] Step S1231, using the smoothed initial three-dimensional point cloud data as the input of the Poisson reconstruction algorithm to establish an octree spatial index structure; the octree spatial index structure includes m2 leaf nodes;

[0226] Step S1232, for each leaf node O of the octree spatial index structure (j) , calculate the leaf node O (j) The normal vectors of all the initial 3D point cloud data in the leaf node O are obtained. (j) The mean normal vector N (j) ; Among them, 1≤j≤m2;

[0227] Step S1233: based on the coordinate information of the initial three-dimensional point cloud data and the normal vector mean N of the octree leaf node (j) , construct Poisson's equation;

[0228] Step S1234, solving Poisson's equation to obtain a scalar field function;

[0229] Step S1235, triangulating the scalar field function, extracting isosurfaces, and generating a three-dimensional mesh model composed of a large number of triangular facets;

[0230] Step S1236, mesh simplification, smoothing and detail processing are performed on the three-dimensional mesh model to obtain a final smooth mesh point three-dimensional model.

[0231] In the dot parameter extraction module, the first dot morphological parameters include dot opening area S, dot depth D and dot volume V; the dot morphological parameter extraction to obtain the first dot morphological parameters includes:

[0232] Step S1310, calculating the maximum curvature value k1 and the minimum curvature value k2 of each point on the surface of the smoothed dot 3D model;

[0233] Step S1320, k1 and k2 of each point on the surface of the smooth dot 3D model are combined into a cluster feature vector v=[k1, k2], and each point on the surface of the smooth dot 3D model is classified according to the cluster feature vector v to extract the dot area and the non-dot area;

[0234] Step S1330, fitting a dot contour curve to the extracted dot area, calculating the area enclosed by the dot contour curve, and obtaining a dot opening area S;

[0235] Step S1340: extract the Z coordinate value set of all points from the dot area ,from Search for the minimum value point Z of the Z coordinate value min ; Extract the Z coordinate value set of all points from the non-dot area ,calculate The mean of all Z coordinate values ​​in , the minimum point Z min With the mean The difference is taken as the dot depth D;

[0236] Step S1350, extract all triangular facets in the dot area, obtain the spatial coordinates of the triangular facets according to the spatial coordinates of each pixel point in the initial three-dimensional point cloud data; calculate the dot volume V based on the spatial coordinates of the triangular facets.

[0237] The step S1350 includes:

[0238] Step S1351, extract all triangles in the dot area and mark them as ; m1 is the total number of triangles in the dot area;

[0239] Step S1352, for Triangles , extract the spatial coordinates of its three vertices , ;

[0240] Step S1353, calculate the triangle patch based on the spatial coordinates of the three vertices Area ;

[0241] Step S1354, based on the spatial coordinates and areas of the three vertices , calculate the triangle patch Volume contribution ;

[0242] Step S1355: sum up the volume contributions of all triangular facets to obtain the dot volume V.

[0243] In the ink supply prediction module, the ink supply optimization model is constructed according to the dot morphology parameter set, including:

[0244] Step S2110, taking the dot morphology parameter set as input and the ink supply amount as output, to construct a training sample set of the ink supply amount optimization model;

[0245] Step S2120: Based on the training sample set, a Gaussian radial basis kernel function is used as the kernel function of the support vector machine to construct an ink supply optimization model.

[0246] In the ink supply correction module, the local area image block of each dot on the surface of the anilox roller is extracted, including:

[0247] Step S3110, using a region growing-based image segmentation algorithm to segment the surface visual image of the anilox roller into halftone dot areas and non-halftone dot areas;

[0248] Step S3120 , extracting the minimum bounding rectangle of each halftone dot area, and using the image area within the minimum bounding rectangle as the local area image block of the halftone dot.

[0249] In the ink supply correction module, the calculation of the ink uniformity and defect rate of each dot according to the local area image block of each dot includes:

[0250] Step S3210, binarizing the local area image block of each halftone dot to obtain a halftone dot binarized image;

[0251] Step S3220, calculating the standard deviation of the grayscale values ​​of the foreground pixels in the dot binary image to obtain the ink uniformity of the dots; the ink uniformity is inversely proportional to the standard deviation of the grayscale values ​​of the foreground pixels;

[0252] Step S3230, marking pixels with holes or missing in the foreground pixels as defective dot pixels, counting the number of defective dot pixels, calculating the ratio of defective dot pixels to the total number of foreground pixels, and obtaining the defective dot rate.

[0253] The correction of the initial ink supply amount includes:

[0254] According to the overall ink uniformity index and the overall defect rate index, the ink supply correction value is calculated; according to the ink supply correction value, the initial ink supply is corrected to obtain the corrected ink supply value.

[0255] Example 3

[0256] This embodiment discloses an electronic device, which may include one or more processors and one or more memories, wherein the memories store computer readable codes, and when the computer readable codes are executed by the one or more processors, the machine vision method for optimizing the adaptive ink supply amount in gravure printing of decorative paper as described above may be executed.

[0257] The method or system according to the embodiment of the present application can also be implemented with the help of the architecture of an electronic device. The electronic device may include a bus, one or more CPUs, a read-only memory (ROM), a random access memory (RAM), a communication port connected to a network, an input / output component, a hard disk, etc. The storage device in the electronic device, such as a ROM or a hard disk, can store the machine vision method for adaptively optimizing the ink supply amount in the gravure printing of decorative paper provided in the present application. The machine vision method for adaptively optimizing the ink supply amount in the gravure printing of decorative paper may, for example, include: collecting a sequence of anilox surface morphology images of the gravure printing anilox roller of the decorative paper to be optimized, and obtaining the initial three-dimensional point cloud data of the anilox surface; obtaining a smooth three-dimensional model of dots based on the initial three-dimensional point cloud data; extracting dot morphology parameters based on the smooth three-dimensional model of dots to obtain the first dot morphology parameters; collecting a set of anilox surface morphology image sequences of different decorative paper gravure printing anilox rollers, and obtaining a dot morphology parameter set based on the anilox surface morphology image sequence set; constructing an ink supply optimization model based on the dot morphology parameter set; and extracting the dot morphology parameters based on the first dot morphology. The parameter and ink supply optimization model are used to predict the initial ink supply; the initial ink supply is used to start gravure printing, the surface visual image of the anilox roller is collected in real time, the local area image blocks of each dot on the surface of the anilox roller are extracted, and the inking uniformity and defect rate of each dot are calculated; according to the inking uniformity of each dot, the overall inking uniformity index U is generated; according to the defect rate of each dot, the overall defect rate index R is generated; if the overall inking uniformity index U is less than the preset uniformity threshold θ1, or the overall defect rate index R is greater than the preset defect rate threshold θ2, the correction of the initial ink supply is triggered to obtain the corrected ink supply value.

[0258] Furthermore, the electronic device may also include a user interface. Of course, the architecture disclosed in the present invention is only exemplary, and when implementing different devices, one or more components in the electronic device disclosed in the present invention may be omitted according to actual needs.

[0259] Example 4

[0260] This embodiment discloses a computer-readable storage medium, on which computer-readable instructions are stored. When the computer-readable instructions are executed by a processor, the machine vision method for optimizing the adaptive ink supply amount in the gravure printing of decorative paper according to the embodiment of the present application can be executed. The storage medium includes, but is not limited to, for example, volatile memory and / or non-volatile memory. For example, the volatile memory may include a random access memory (RAM) and a cache memory (cache). For example, the non-volatile memory may include a read-only memory (ROM), a hard disk, a flash memory, etc.

[0261] In addition, according to the implementation manner of the present application, the process described in the above reference flowchart can be implemented as a computer software program. For example, the present application provides a non-transitory machine-readable storage medium, the non-transitory machine-readable storage medium stores machine-readable instructions, the machine-readable instructions can be executed by a processor to execute instructions corresponding to the method steps provided in the present application, for example: collecting a sequence of anilox surface morphology images of the gravure printing anilox roller of the decorative paper to be optimized, and obtaining initial three-dimensional point cloud data of the anilox surface; obtaining a smooth three-dimensional model of dots based on the initial three-dimensional point cloud data; extracting dot morphology parameters based on the smooth three-dimensional model of dots to obtain first dot morphology parameters; collecting a set of anilox surface morphology image sequences of gravure printing anilox rollers of different decorative papers, and obtaining a set of dot morphology parameters based on the set of anilox surface morphology image sequences; The method comprises the following steps: a first ink supply optimization model is constructed based on the first dot morphology parameter set; an initial ink supply is predicted based on the first dot morphology parameter and the ink supply optimization model; an initial ink supply is started based on the initial ink supply, and the surface visual image of the anilox roller is collected in real time, and the local area image blocks of each dot on the surface of the anilox roller are extracted to calculate the inking uniformity and defect rate of each dot; an overall inking uniformity index U is generated based on the inking uniformity of each dot; an overall defect rate index R is generated based on the defect rate of each dot; if the overall inking uniformity index U is less than the preset uniformity threshold value θ1, or the overall defect rate index R is greater than the preset defect rate threshold value θ2, the correction of the initial ink supply is triggered to obtain the corrected ink supply value. When the computer program is executed by a central processing unit (CPU), the above functions defined in the method of the present application are executed.

[0262] The methods, systems, and devices of the present application may be implemented in many ways. For example, the methods, systems, and devices of the present application may be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above order of steps for the method is for illustration only, and the steps of the method of the present application are not limited to the order specifically described above unless otherwise specified. In addition, in some embodiments, the present application may also be implemented as programs recorded in a recording medium, which include machine-readable instructions for implementing the method according to the present application. Therefore, the present application also covers recording media storing programs for executing the method according to the present application.

[0263] In addition, the parts of the above-mentioned technical solutions provided in the embodiments of the present application that are consistent with the implementation principles of the corresponding technical solutions in the prior art are not described in detail to avoid excessive redundancy.

[0264] The specific implementation modes as described above further describe the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation mode of the present invention and is not intended to limit the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A machine vision method for adaptive ink supply optimization in gravure printing of decorative paper, characterized in that: The method comprises: Collecting a sequence of anilox surface morphology images of an anilox roller for gravure printing of decorative paper to be optimized, and obtaining initial three-dimensional point cloud data of the anilox surface; obtaining a smooth dot three-dimensional model based on the initial three-dimensional point cloud data; extracting dot morphology parameters based on the smooth dot three-dimensional model, and obtaining first dot morphology parameters; A set of anilox surface morphology image sequences of anilox rollers for gravure printing of different decorative papers are collected, and a set of dot morphology parameters is obtained according to the set of anilox surface morphology image sequences; an ink supply optimization model is constructed according to the dot morphology parameter set; and an initial ink supply is predicted according to the first dot morphology parameter and the ink supply optimization model; The initial ink supply is adopted, the gravure printing is started, the surface visual image of the anilox roller is collected in real time, the local area image block of each dot on the surface of the anilox roller is extracted, and the inking uniformity and defect rate of each dot are calculated; the overall inking uniformity index U is generated according to the inking uniformity of each dot; the overall defect rate index R is generated according to the defect rate of each dot; if the overall inking uniformity index U is less than the preset uniformity threshold value θ1, or the overall defect rate index R is greater than the preset defect rate threshold value θ2, the correction of the initial ink supply is triggered, and the corrected ink supply value is obtained; The calculation of the ink uniformity and defect rate of each dot includes: Binarization is performed on the local area image block of each dot to obtain a dot binary image; Calculate the standard deviation of the grayscale value of the foreground pixel in the dot binary image to obtain the ink uniformity of the dot; the ink uniformity is inversely proportional to the standard deviation of the grayscale value of the foreground pixel; The pixels with holes or missing in the foreground pixels are marked as defective dot pixels, the number of defective dot pixels is counted, and the ratio of defective dot pixels to the total number of foreground pixels is calculated to obtain the defective dot rate.

2. The machine vision method for adaptively optimizing ink supply in gravure printing of decorative paper according to claim 1, characterized in that: The obtaining of the initial three-dimensional point cloud data of the textured surface comprises: splicing and fusing the textured surface topography image sequence, and calculating the spatial coordinates of each pixel point based on the spliced ​​and fused textured surface topography image sequence to obtain the initial three-dimensional point cloud data of the textured surface.

3. The machine vision method for adaptively optimizing ink supply in gravure printing of decorative paper according to claim 1, characterized in that: The step of obtaining a smooth dot 3D model according to the initial 3D point cloud data comprises: Correcting the initial three-dimensional point cloud data to the z-axis direction of the mesh normal to obtain corrected initial three-dimensional point cloud data; performing smoothing filtering on the corrected initial three-dimensional point cloud data to obtain smoothed initial three-dimensional point cloud data; The smoothed initial 3D point cloud data is reconstructed, and the dot surface is fitted using the Poisson reconstruction algorithm to obtain a smooth dot 3D model composed of continuous triangular facets.

4. The machine vision method for adaptively optimizing ink supply in gravure printing of decorative paper according to claim 3, characterized in that: The obtaining of a smooth dot three-dimensional model composed of continuous triangular facets comprises: The smoothed initial three-dimensional point cloud data is used as the input of the Poisson reconstruction algorithm to establish an octree spatial index structure; the octree spatial index structure includes m2 leaf nodes; For each leaf node O of the octree spatial index structure (j) , calculate the leaf node O (j) The normal vectors of all the initial 3D point cloud data in the leaf node O are obtained. (j) The mean normal vector N (j) ; Among them, 1≤j≤m2; The normal vector mean N based on the octree leaf node (j) , construct Poisson's equation; Solve Poisson's equation to obtain the scalar field function; Triangulate the scalar field function, extract the isosurface, and generate a three-dimensional mesh model composed of a large number of triangular facets; The three-dimensional mesh model is processed to obtain the final smooth dot three-dimensional model.

5. The machine vision method for adaptively optimizing ink supply in gravure printing of decorative paper according to claim 2, characterized in that: The first dot morphology parameters include dot opening area S, dot depth D and dot volume V; The extracting of dot morphological parameters based on the smooth dot three-dimensional model includes: Calculate the maximum curvature value k1 and the minimum curvature value k2 of each point on the surface of the smooth dot 3D model; The k1 and k2 of each point on the surface of the smooth dot 3D model are combined into a cluster feature vector v=[k1,k2], and each point on the surface of the smooth dot 3D model is classified according to the cluster feature vector v to extract the dot area and the non-dot area; For the extracted dot area, fit the dot contour curve, calculate the area enclosed by the dot contour curve, and obtain the dot opening area S; Extract the Z coordinate value set of all points from the dot area ,from Search for the minimum value point Z of the Z coordinate value min ; Extract the Z coordinate value set of all points from the non-dot area ,calculate The mean of all Z coordinate values ​​in , the minimum point Z min With the mean The difference is taken as the dot depth D; All triangular facets in the dot area are extracted, and the spatial coordinates of the triangular facets are obtained according to the spatial coordinates of each pixel point in the initial three-dimensional point cloud data; based on the spatial coordinates of the triangular facets, the dot volume V is calculated.

6. The machine vision method for adaptively optimizing ink supply in gravure printing of decorative paper according to claim 5, characterized in that: The calculation of the dot volume V comprises: Extract all triangles in the dot area and mark them as ; m1 is the total number of triangles in the dot area; For Triangles , extract the spatial coordinates of its three vertices , ;in, ; Calculate the triangle patch based on the spatial coordinates of the three vertices Area ; Space coordinates and area based on three vertices , calculate the triangle patch Volume contribution ; The volume contribution of all triangles is summed up to get the mesh volume V.

7. The machine vision method for adaptively optimizing ink supply in gravure printing of decorative paper according to claim 1, characterized in that: The method of extracting the local area image blocks of each dot on the surface of the anilox roller comprises: The surface visual image of the anilox roller is segmented into dot areas and non-dot areas by using an image segmentation algorithm based on region growth. The minimum bounding rectangle of each dot area is extracted, and the image area within the minimum bounding rectangle is used as the local area image block of the dot.

8. The machine vision method for adaptively optimizing ink supply in gravure printing of decorative paper according to claim 1, characterized in that: The correction of the initial ink supply amount includes: According to the overall ink uniformity index and the overall defect rate index, the ink supply correction value is calculated; according to the ink supply correction value, the initial ink supply is corrected to obtain the corrected ink supply value.

9. A machine vision system for optimizing the amount of ink supply in gravure printing of decorative paper, which is used to implement the machine vision method for optimizing the amount of ink supply in gravure printing of decorative paper according to any one of claims 1 to 8, characterized in that: The system comprises: 3D model building module: used to collect the image sequence of the anilox surface of the gravure printing anilox roller of the decorative paper to be optimized, and obtain the initial 3D point cloud data of the anilox surface; and obtain the smooth dot 3D model based on the initial 3D point cloud data; Dot parameter extraction module: based on the smooth dot three-dimensional model, extract the dot morphological parameters to obtain the first dot morphological parameters; Ink supply prediction module: used to collect a set of anilox surface morphology image sequences of anilox rollers for gravure printing of different decorative papers, obtain a set of dot morphology parameters according to the set of anilox surface morphology image sequences; construct an ink supply optimization model according to the dot morphology parameter set; predict the initial ink supply according to the first dot morphology parameter and the ink supply optimization model; Ink supply correction module: used to adopt the initial ink supply, start gravure printing, collect the surface visual image of the anilox roller in real time, extract the local area image block of each dot on the surface of the anilox roller, and calculate the inking uniformity and defect rate of each dot; generate the overall inking uniformity index U according to the inking uniformity of each dot; generate the overall defect rate index R according to the defect rate of each dot; if the overall inking uniformity index U is less than the preset uniformity threshold θ1, or the overall defect rate index R is greater than the preset defect rate threshold θ2, the correction of the initial ink supply is triggered to obtain the corrected ink supply value.

Images