Color management method and system for digital printing of PVC floor film

Abstract

The present application provides a color management method and system for digital printing of PVC floor films, belonging to the field of intelligent printing color management technology. The method includes: obtaining a historical printing data set, including RGB pigment color, printing environment, multi-scale material texture encoding, and corresponding CMYK color data; training an adaptive color convolutional neural network model based on this. When applied, real-time printing RGB pigment color, printing environment, and multi-scale material texture encoding data are obtained, and the initial CMYK color data is predicted by the trained model, and then the simulated annealing algorithm is used to optimize the solution to obtain the target CMYK color data. The present application comprehensively utilizes color and texture information, and improves the color reproduction accuracy and efficiency of digital printing of PVC floor films through adaptive model training, real-time prediction, and optimization solution.

Description

Technical Field

[0001] The present application relates to the technical field of intelligent printing color management, and in particular to a method and system for digital printing color management of PVC floor films. Background Art

[0002] With the continuous development of digital printing technology, the printing of PVC floor film has gradually shifted from traditional gravure printing to digital inkjet printing. Digital inkjet printing has the advantages of personalized customization, small batch production, and rapid response to market demand, and has received widespread attention in the industry. However, due to the particularity of PVC floor film materials and the complexity of digital printing processes, how to achieve high-precision and high-quality color reproduction on PVC floor film has always been a technical problem that needs to be solved urgently.

[0003] In the prior art, digital printing of PVC floor films usually adopts the traditional IC color management method, which realizes color conversion and restoration by constructing the mapping relationship between RGB and CMYK color spaces. However, the traditional IC color management method is mainly aimed at ordinary printing materials such as paper, and it is difficult to fully consider the material properties and texture characteristics of PVC floor films, resulting in low color restoration accuracy and large color difference, which is difficult to meet the requirements of high-quality printing.

[0004] In addition, most existing color management methods rely on manual experience and repeated test printing, which requires a lot of time and material costs. During the printing process, due to the influence of factors such as the printing environment and material characteristics, color consistency is difficult to ensure, resulting in unstable printing quality. At the same time, existing methods are difficult to achieve real-time dynamic color management and optimization, and cannot meet the needs of fast and flexible production.

[0005] Therefore, how to develop an intelligent and adaptive digital printing color management method and system based on the material characteristics and texture features of PVC floor film to achieve high-precision and high-quality color reproduction and improve printing efficiency and stability is a technical problem that needs to be urgently solved in this field. Summary of the invention

[0006] The present application aims to solve at least one of the technical problems in the related art to a certain extent. To this end, one purpose of the present application is to propose a color management method and method for digital printing of PVC floor film, which improves the accuracy of printed colors and the authenticity of texture details.

[0007] One aspect of the present application provides a method for color management of digital printing of PVC floor film.

[0008] The method includes: obtaining historical printing pigment RGB pigment color data, historical printing environment data, historical multi-scale material texture encoding data for characterizing material texture, and corresponding historical CMYK color data; taking the historical printing pigment RGB pigment color data, historical printing environment data and historical multi-scale material texture encoding data as input, and taking the corresponding historical CMYK color data as output, training a first adaptive color convolutional neural network model; obtaining real-time printing pigment RGB pigment color data, real-time printing environment data and real-time multi-scale material texture encoding data, and predicting first CMYK color data based on the first adaptive color convolutional neural network model; initializing a simulated annealing algorithm, and performing a simulated annealing algorithm on the first CMYK color data to obtain second CMYK color data.

[0009] One aspect of the present application provides a PVC floor film digital printing color management system, including: a data acquisition module, which acquires historical printing pigment RGB pigment color data, historical printing environment data, historical multi-scale material texture encoding data for characterizing material texture, and corresponding historical CMYK color data; a model training module, which takes the historical printing pigment RGB pigment color data, historical printing environment data and historical multi-scale material texture encoding data as input, and takes the corresponding historical CMYK color data as output, to train a first adaptive color convolutional neural network model; a real-time prediction module, which acquires real-time printing pigment RGB pigment color data, real-time printing environment data and real-time multi-scale material texture encoding data, and predicts first CMYK color data based on the first adaptive color convolutional neural network model; an annealing solution module, which initializes a simulated annealing algorithm, and performs a simulated annealing algorithm on the first CMYK color data to obtain second CMYK color data.

[0010] One aspect of the present application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in the color management method for digital printing of PVC floor films are implemented.

[0011] One aspect of the present application provides a readable storage medium, characterized in that the readable storage medium stores a computer program, and the computer program is suitable for loading by a processor to execute the steps in the PVC floor film digital printing color management method.

[0012] According to the exemplary embodiment of the present application, by constructing a PVC floor film digital printing color management method and system, there are at least the following beneficial effects:

[0013] By obtaining historical printing pigment RGB pigment color data, historical printing environment data, historical multi-scale material texture encoding data used to characterize material texture, and corresponding historical CMYK color data, and using the comprehensive color texture discreteness formula to screen historical data, we fully considered the impact of color attributes and texture complexity on sample quality, and adaptively obtained high-quality, comprehensive training data sets, laying a good data foundation for subsequent model training.

[0014] An adaptive color convolutional neural network model is used for color mapping. An adaptive convolution kernel is introduced in the convolution layer, which can dynamically adjust the weight of the convolution kernel according to the characteristics of the material texture, effectively improving the accuracy of color mapping. At the same time, the loss function considers both the color restoration loss and the material texture feature loss, ensuring the accuracy of color restoration and the fidelity of texture details.

[0015] Based on the trained adaptive color convolutional neural network model, real-time printing pigment RGB pigment color data, real-time printing environment data and real-time multi-scale material texture encoding data are obtained, and the corresponding CMYK color data can be quickly predicted, realizing real-time color management and improving the efficiency of digital printing.

[0016] The predicted CMYK color data is further optimized using the simulated annealing algorithm. The color difference between the predicted color and the target color is used as the energy function. The optimal CMYK color combination is found through simulated annealing, which further improves the accuracy and stability of color reproduction.

[0017] The color management method and system for digital printing of PVC floor films provided by the present invention fully utilize color and texture information through adaptive model training, real-time prediction and optimization solution, and significantly improve the color reproduction accuracy and efficiency of digital printing of PVC floor films. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a schematic diagram of the process of the color management method of digital printing of PVC floor film according to an exemplary embodiment of the present application;

[0019] Figure 2 It is a schematic diagram of a process of selecting historical printing pigment RGB pigment color data according to an exemplary embodiment of the present application;

[0020] Figure 3 It is a schematic diagram of an optimal CMYK color combination output process according to an exemplary embodiment of the present application;

[0021] Figure 4 A schematic diagram of a PVC floor film digital printing color management system of the present invention;

[0022] Figure 5It is a schematic diagram of the structure of an electronic device provided by an embodiment of the present application;

[0023] Figure 6 It is a schematic diagram of the structure of a computer-readable storage medium provided by an embodiment of the present application. DETAILED DESCRIPTION

[0024] In order to better understand the present application, a more detailed description will be made of various aspects of the present application with reference to the accompanying drawings. It should be understood that these detailed descriptions are only descriptions of exemplary embodiments of the present application, and are not intended to limit the scope of the present application in any way. Throughout the specification, the same reference numerals refer to the same elements. The expression "and / or" includes any and all combinations of one or more of the associated listed items.

[0025] In the accompanying drawings, the size, dimensions and shapes of the elements have been slightly adjusted for ease of illustration. The drawings are for illustration only and are not strictly drawn to scale. As used herein, the terms "substantially", "approximately" and similar terms are used as terms of approximation, not as terms of degree, and are intended to illustrate the inherent deviations in measurements or calculations that will be recognized by those of ordinary skill in the art. In addition, in this application, the order in which the steps are described does not necessarily represent the order in which these processes occur in actual operation, unless otherwise specified or can be derived from the context.

[0026] It should also be understood that expressions such as "include", "including", "have", "contain" and / or "comprising" are open rather than closed expressions in this specification, which indicate the presence of the stated features, elements and / or components, but do not exclude the presence of one or more other features, elements, components and / or combinations thereof. In addition, when expressions such as "at least one of..." appear after a list of listed features, they modify the entire list of features rather than just the individual elements in the list. In addition, when describing embodiments of the present application, "may" is used to mean "one or more embodiments of the present application". And, the term "exemplary" is intended to refer to an example or illustration.

[0027] Unless otherwise specified, all words (including engineering terms and scientific and technological terms) used in this article have the same meaning as those commonly understood by ordinary technicians in the field to which this application belongs. It should also be understood that, unless clearly stated in this application, words defined in common dictionaries should be interpreted as having the same meaning as their meaning in the context of the relevant technology, and should not be interpreted in an idealized or overly formal sense.

[0028] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0029] Example 1

[0030] like Figure 1 As shown, a color management method for digital printing of PVC floor film, the method is applied to a digital printing system and may include:

[0031] Step S100: Acquire historical printing pigment RGB pigment color data, historical printing environment data, historical multi-scale material texture encoding data for characterizing material texture, and corresponding historical CMYK color data;

[0032] Step S200: using historical printing pigment RGB pigment color data, historical printing environment data and historical multi-scale material texture encoding data as input and corresponding historical CMYK color data as output to train a first adaptive color convolutional neural network model;

[0033] Step S300: acquiring real-time printing pigment RGB pigment color data, real-time printing environment data and real-time multi-scale material texture encoding data, and predicting first CMYK color data based on a first adaptive color convolutional neural network model;

[0034] Step S400: Initialize the simulated annealing algorithm, and perform the simulated annealing algorithm on the first CMYK color data to obtain second CMYK color data.

[0035] Step S100: Acquire historical printing pigment RGB pigment color data, historical printing environment data, historical multi-scale material texture encoding data for characterizing material texture, and corresponding historical CMYK color data;

[0036] like Figure 2 As shown, a large number of historical RGB pigment color data of different batches of printing pigments and corresponding historical CMYK color data and corresponding environmental data, as well as historical multi-scale material texture encoding data for characterizing material textures are collected in advance.

[0037] Collects historical RGB pigment color data of a large number of different batches of printing pigments and corresponding historical CMYK color data and corresponding environmental data.

[0038] In particular, the historical printing pigment RGB pigment color data is converted into HSI data, and combined with the multi-scale material texture encoding data, the comprehensive color texture discreteness of each sample is calculated; specifically, the historical printing pigment RGB color data is converted into HSI color space using the following formula:

[0039]

[0040] Calculate the weighted Euclidean distance between each historical printing pigment sample and the overall mean in the HSI space, the formula is as follows: in, Respectively represent the mean of all samples in the HSI space, w H ,w S ,w I is the weight coefficient of the corresponding attribute, which is used to adjust the contribution of different attributes to color differences.

[0041] It should be noted that w H ,w S ,w I The weight coefficient of each attribute is preliminarily set based on the experience and knowledge of domain experts. For example, the importance of hue, saturation, and brightness to color differences can be judged based on the characteristics of the material texture, and a larger or smaller weight can be assigned accordingly. The weight coefficient can also be used as a hyperparameter to search for the optimal value within a certain range using an optimization algorithm (such as grid search, random search, Bayesian optimization, etc.). By evaluating the data screening effect under different weight combinations, a set of weight coefficients with the best performance is selected.

[0042] Multi-scale Gabor filtering: The material texture image corresponding to each historical printing pigment sample is filtered using Gabor filters of different scales (σ), directions (θ) and wavelengths (λ). By adjusting these parameters, the characteristics of the material texture can be extracted at different scales and directions.

[0043] Feature encoding: For each scale and direction of the Gabor filter response, extract its amplitude and phase information and encode it. Common encoding methods include binary encoding, grayscale encoding, etc. The encoded feature can be represented as a high-dimensional vector.

[0044] Multi-scale feature concatenation: Concatenate the coded features at different scales and directions into a long vector, that is, obtain a multi-scale material texture coding vector. The dimension m of the vector is equal to the sum of the feature dimensions at all scales and directions.

[0045] Texture coding entropy calculation: For each sample's multi-scale material texture coding vector, calculate the normalized frequency distribution of its elements. Then, the texture coding entropy is calculated using the Shannon entropy formula: Among them, t ijRepresents the normalized frequency of the jth coding feature of the i-th sample. The larger the entropy value, the more uniform the texture coding distribution of the sample and the higher the texture complexity. Through the above steps, we can obtain the multi-scale material texture coding vector and texture coding entropy corresponding to each historical printing pigment sample. This information can be used together with color attributes to calculate the comprehensive color texture discreteness and perform data screening.

[0046] In practical applications, the parameter range and combination of Gabor filters can be flexibly selected according to the specific material texture characteristics and printing requirements. In addition, other texture feature extraction and encoding methods, such as wavelet transform and Fourier transform, can also be explored to further optimize the effect of multi-scale material texture encoding.

[0047] Define the comprehensive color texture discreteness, the formula is as follows: This formula multiplies the weighted color distance by the texture coding entropy, exponentially enhancing the effect of texture complexity on discreteness.

[0048] Calculate the mean of the comprehensive color texture dispersion and standard deviation σ C According to the 3σ principle, filter the historical printing pigment RGB pigment color data and set the upper threshold and lower threshold When C i >θ u or C i <θ l When , the comprehensive discreteness of the i-th sample is judged to be abnormal and is eliminated; otherwise, the sample data quality is considered to be good and is included in the training set.

[0049] Through the above steps, this embodiment uses a comprehensive color texture discreteness formula, fully considers the influence of color attributes and texture complexity on the discreteness of samples, and adaptively screens out high-quality and comprehensive historical printing pigment data, laying a good data foundation for subsequent adaptive color convolutional neural network model training, and effectively improving the color restoration performance and printing effect.

[0050] The process of acquiring historical multi-scale material texture encoding data used to characterize material texture is as follows:

[0051] S101. Use Gabor filter to encode material texture. Gabor filter is a linear filter used for texture analysis. Its mathematical expression is:

[0052] The Gabor filter can be expressed as: G(x, y, θ, λ, ψ, σ,

[0053] γ)=exp(((x')^2+γ^2(y')^2) / 2σ^2)*cos(2π(x' / λ)+ψ);

[0054] Wherein, (x', y') is the coordinate in the rotation coordinate system, which is defined as: (x', y') = (xcosθ + ysinθ, xsinθ + ycosθ);

[0055] x and y represent the spatial coordinates of the image, θ represents the direction angle of the filter, represents the direction of the filter, λ represents the wavelength of the filter, represents the periodicity, ψ represents the phase shift of the filter, σ represents the standard deviation, represents the scale of the filter, and γ represents the aspect ratio, represents the aspect ratio of the Gaussian envelope.

[0056] S102 can encode material textures at different scales and directions by adjusting the parameters σ, θ, and λ of the Gabor filter;

[0057] Scale (σ): By changing the value of σ, the scale of the filter can be controlled to capture texture features at different scales. Larger σ values ​​are used to capture macroscopic features, while smaller σ values ​​are used to capture microscopic features.

[0058] Direction (θ): By adjusting the θ value, you can capture texture features in different directions. Wavelength (λ): By adjusting the λ value, you can capture texture features at different frequencies.

[0059] The Gabor feature maps of different scales and directions are combined to obtain a multi-scale coding representation of material texture.

[0060] Simply put, when observing a material image, you can find that the material presents different characteristics at different scales. For example, at a larger scale, you can see the overall direction and distribution of the material; at a smaller scale, you can see the details of the material, such as the thickness, straightness, etc. of the grain. In order to fully capture the characteristics of the material texture, it is necessary to analyze at different scales. This is like using magnifying glasses of different magnifications to observe the material: Low-power magnifying glass (large scale): observe the overall distribution and direction of the material, and capture the macroscopic characteristics of the material. Medium-power magnifying glass (medium scale): observe the regional characteristics of the material, and capture the changes in the material in different regions. High-power magnifying glass (small scale): observe the details of the material and capture the fine characteristics of the material grain.

[0061] By extracting the texture features of materials at different scales, a comprehensive description of the texture of the material can be obtained. These features may include: the direction and orientation of the material, the thickness and density of the texture of the material, the degree of curvature of the texture of the material, and the distribution and changes of the material in different regions. The process of extracting these features can be achieved through various image processing and analysis techniques, such as Fourier transform, wavelet transform, gray level co-occurrence matrix, etc. These technologies can convert the texture information in the image into digital features, which is convenient for subsequent analysis and processing.

[0062] Exemplarily, the parameters of the Gabor filter are selected: scale parameter σ, direction parameter θ and wavelength parameter λ. Among them, σ controls the spatial scale of the Gaussian kernel function, θ represents the direction of the filter, and λ represents the wavelength of the sine wave. For each σ, multiple combinations of θ and λ are selected. For example, σ takes 2, 4, 8, 16, θ takes 0°, 45°, 90°, 135°, and λ takes different sine wave wavelengths. A series of Gabor filters are generated using parameter combinations, and convolution operations are performed on the material texture image to obtain filter response images at multiple scales and directions. The mean and standard deviation of each response image are calculated to generate a multi-scale texture coding feature vector, and the feature vectors are combined into a multi-scale texture coding matrix as the input of the first adaptive color convolutional neural network model.

[0063] Through multi-scale texture feature extraction, a comprehensive and detailed material texture description can be obtained, providing important reference information for subsequent adaptive color mapping.

[0064] Step S200: Using historical printing pigment RGB pigment color data, historical printing environment data and historical multi-scale material texture encoding data as input and corresponding historical CMYK color data as output, a first adaptive color convolutional neural network model is trained.

[0065] In the exemplary implementation of the present application, RGB pigment color data represents the color data of the image to be printed, and CMYK color data represents the color data to be printed by the device. Traditional color conversion methods usually use a fixed mapping relationship to map RGB colors to CMYK colors. However, this fixed mapping relationship cannot adapt well to the texture characteristics of different materials, resulting in unsatisfactory printing effects. The purpose of introducing adaptive color mapping CNN in this embodiment is to dynamically adjust the mapping relationship from RGB to CMYK according to the texture characteristics of the material to obtain a more accurate and realistic printing effect.

[0066] Taking historical printing pigment RGB pigment color data, historical printing environment data and historical multi-scale material texture encoding data as input and corresponding historical CMYK color data as output, a first adaptive color convolutional neural network model is trained;

[0067] Among them, an adaptive convolution kernel is introduced into the convolution layer of the first adaptive color convolutional neural network model, and the weight of the convolution kernel is dynamically adjusted according to the characteristics of the material texture.

[0068] The adaptive convolution kernel can be expressed as: W_a(i, j) = W(i, j)*(1+α*M(i, j) / T)

[0069] Among them, W_a represents the adaptive convolution kernel, W represents the original convolution kernel, T represents the material texture feature, α represents the adaptive coefficient, and the adaptive coefficient controls the degree of adaptation of the convolution kernel weight to the material texture feature. The larger the α, the greater the weight adjustment. You can choose a suitable α value according to the experimental results, for example, α = 0.1, and T represents the temperature coefficient.

[0070] i and j represent the row index and column index in the convolution kernel respectively.

[0071] Through adaptive convolution kernels, CNN can adaptively adjust the color mapping relationship according to the characteristics of material texture and improve the accuracy of color restoration.

[0072] At the same time, in order to further improve the stability and robustness of the adaptive convolution kernel, the material texture feature M(i, j) can be normalized:

[0073] W_a(i,j)=W(i,j)*(1+α*(M(i,j)M_mean) / (M_std*β))

[0074] Among them, M_mean represents the mean of the material texture feature, M_std represents the standard deviation of the material texture feature, and β represents the temperature coefficient, which is used to control the smoothness of the adaptive adjustment. The larger the β, the smoother the adjustment. You can choose a suitable β value according to the experimental results, for example, β = 1, and the meanings of the other symbols are the same as before. During the training process, α and β can be set as learnable parameters to let the neural network optimize itself.

[0075] By normalizing the material texture features, they can be mapped to a smaller range, reducing their impact on the adaptive convolution kernel and improving the stability and generalization ability of the network. The first adaptive color convolutional neural network model training uses a multi-task loss function, taking into account both color restoration loss and material texture feature loss.

[0076] The color restoration loss uses L1 loss or L2 loss to measure the difference between the predicted CMYK color and the true value.

[0077] The material texture feature loss uses feature reconstruction loss to measure the difference between the material texture features extracted by CNN and Gabor coding.

[0078] The total loss function can be expressed as: L_total=λ_1*L_color+λ_2*L_texture*T.

[0079] Among them, L_color represents the color restoration loss, L_texture represents the material texture feature loss, and λ_1 and λ_2 are balance coefficients. CNN is trained using RGBCMYK data with multi-scale material texture encoding, and the network parameters are optimized through the back propagation algorithm.

[0080] During the training process, the weights of the adaptive convolution kernels are dynamically adjusted so that CNN can adaptively adjust the color mapping relationship according to the characteristics of the material texture.

[0081] Among them, the color restoration loss L_color can use the mean square error (MSE) or the mean absolute error (MAE) to measure the difference between the predicted CMYK value and the true value, and the texture feature reconstruction loss L_texture can be measured by the MSE or MAE between the predicted material texture coding features and the real coding features extracted by the Gabor filter.

[0082] By optimizing the multi-task loss function, CNN can simultaneously consider the accuracy of color restoration and material texture features.

[0083] A convolutional neural network (CNN) model is used for color conversion prediction, with the input being the RGB color data in the design file and the output being the first CMYK color data.

[0084] S300: Acquire real-time printing pigment RGB pigment color data, real-time printing environment data and real-time multi-scale material texture encoding data, and predict first CMYK color data based on a first adaptive color convolutional neural network model.

[0085] like Figure 3 As shown, Figure 3 A flow chart of the optimal CMYK color combination output of this embodiment.

[0086] Step S400 initializes the simulated annealing algorithm and processes the first CMYK color data using the simulated annealing algorithm to obtain second CMYK color data. Specifically, the following steps are included:

[0087] S401 initializes simulated annealing algorithm parameters:

[0088] Set the key parameters of the simulated annealing algorithm. First, the initial annealing temperature is T0, which determines the probability of the simulated annealing algorithm accepting inferior solutions in the early stage. Set T0 = 2000, which is high enough to allow the algorithm to have a large exploration space in the early stage.

[0089] Set the cooling rate α, which controls the speed of temperature drop. The smaller the α value, the faster the temperature drops and the faster the algorithm converges, but it may also miss the global optimal solution. Take a more balanced value α = 0.95.

[0090] Set the minimum annealing temperature Tmin. When the temperature drops to Tmin, the annealing algorithm is considered to have converged and the iteration can be stopped. Here, Tmin=0.01 is set, which is low enough to allow the algorithm to stop when it is close to the optimal solution.

[0091] Finally, there is the upper limit of the number of consecutive iterations without improvement, maxStab le. If no new solution is accepted for maxStab le consecutive iterations, the algorithm is considered to have converged. Here, maxStab le = 200, which is large enough to prevent the algorithm from stopping prematurely.

[0092] S402 uses the first adaptive color convolutional neural network model to predict the first CMYK color data as the initial solution S0 of the simulated annealing algorithm.

[0093] Use the trained first adaptive color convolutional neural network model to predict the target RGB color to be printed and get an initial CMYK color combination. This combination may not be perfect, but it is a good starting point and is closer to the optimal solution than the randomly selected initial value. This initial CMYK combination is recorded as S0, which will serve as the starting point of the simulated annealing algorithm.

[0094] S403 defines the energy function E(S) as the CI EDE2000 color difference between the predicted CMYK color and the target RGB color:

[0095] The simulated annealing algorithm requires an energy function to evaluate the quality of the current solution. In the color optimization problem of the present invention, the goal is to find a CMYK color combination that is as close as possible to the target RGB color after conversion to RGB color.

[0096] Therefore, the present invention defines the energy function E(S) as the color difference between the RGB color corresponding to the current solution S and the target RGB color. The smaller the color difference, the closer the current solution is to the optimal solution.

[0097] Specifically, the present invention first converts the CMYK value in the current solution S into an RGB value, recorded as RGB_predicted. Then, the CIEDE2000 color difference formula is used to calculate the color difference ΔE between RGB_predicted and the target RGB value. CIEDE2000 is an advanced color difference calculation method that takes into account the human eye's perception characteristics of color differences. Finally, ΔE is directly used as the value of the energy function, that is, E(S) = ΔE.

[0098] Through the energy function defined in this way, the simulated annealing algorithm can effectively evaluate the quality of each candidate solution and continuously converge to a better solution. In each iteration, the algorithm will accept the solution with a lower energy function value, that is, the CMYK combination with a smaller color difference, thereby gradually approaching the optimal printing color scheme.

[0099] This energy function design based on CI EDE2000 color difference fully considers the color perception characteristics of the human eye and can accurately evaluate the closeness between the printed color and the target color. At the same time, it also abstracts the essence of the color optimization problem, allowing the simulated annealing algorithm to efficiently search for the optimal solution in the solution space, avoiding a lot of unnecessary calculations.

[0100] It should be noted that the CIEDE2000 color difference formula is a color difference calculation method recommended by the International Commission on Illumination (CIE) in 2001, which aims to improve previous color difference formulas such as CIE76 and CIE94. It is developed based on a large amount of visual experimental data and color science theory, and has been applied in many fields. It is an existing technology, and the details will not be introduced in detail here.

[0101] S404 sets initial parameters:

[0102] Before starting the iteration, set the initial parameters: current annealing temperature T: assign the initial annealing temperature T0 to T, that is, T=T0, current solution S: assign the initial solution S0 predicted by the CNN model to S, that is, S=S0, number of iterations iter, and the initial value is 0.

[0103] The number of consecutive no-improvement stab le is used to record how many consecutive iterations have not accepted a new solution. The initial value is 0.

[0104] S405 starts the iteration, and a while loop is used to control the iteration. The loop condition is that the current annealing temperature T must be greater than the minimum annealing temperature Tmin, and the number of consecutive no-improvement stab le must be less than the upper limit maxStab le. This means that the iteration will continue until the annealing temperature drops to the minimum, or no new solution is accepted for a long time, and the optimal CMYK color combination output is obtained. The specific steps are as follows:

[0105] S4051 Perturbation Current Solution

[0106] In order to explore the solution space, a new candidate solution S' is generated based on the current solution S.

[0107] The specific steps are as follows:

[0108] First, generate a random integer δ in the range [-10, 10].

[0109] Then, for each component Si (i=1, 2, 3, 4, corresponding to CMYK respectively) of the current solution S, δ is added to Si to obtain a new component value S'i.

[0110] Finally, make sure that each S'i is in the range [0, 255], since CMYK values ​​range from 0 to 255. If S'i is outside this range, clamp it to the boundary of the range.

[0111] Through such perturbations, a new candidate solution S' is obtained.

[0112] S4052 calculates the energy E(S') of the new solution S':

[0113] Evaluate the quality of the new solution S', that is, calculate its energy function E(S').

[0114] First, convert the CMYK values ​​in the new solution S' into RGB values, recorded as RGB_predicted'. Then, use the CIEDE2000 color difference formula to calculate the color difference ΔE' between RGB_predicted' and the target RGB value. Finally, ΔE' is used as the energy value of the new solution, that is, E(S') = ΔE'.

[0115] S4053 calculates the energy difference.

[0116] The difference between the energy E(S') of the new solution and the energy E(S) of the current solution is calculated and recorded as ΔE, that is, ΔE = E(S') - E(S).

[0117] According to the value of ΔE, the quality change of the new solution relative to the current solution is determined:

[0118] If ΔE<0, it means that the new solution has lower energy and better quality, and the new solution should be accepted.

[0119] If ΔE ≥ 0, it means that the new solution has higher energy and worse quality, and the new solution should be accepted with a certain probability, which depends on the size of ΔE and the current annealing temperature T.

[0120] S4054 determines whether to accept the new solution S':

[0121] According to the value of ΔE, determine whether to accept the new solution S' as the current solution:

[0122] If ΔE<0, directly accept the new solution and update the current solution S to S'.

[0123] If ΔE≥0, the new solution is accepted with a probability P. The calculation formula of probability P is P=exp(-ΔE / T), where exp is a natural exponential function. This formula indicates that the larger ΔE is, the smaller P is; the larger T is, the larger P is. In other words, when the new solution is much worse than the current solution, the present invention rarely accepts it; but when the temperature is very high, the present invention has a greater probability of accepting a worse solution, which helps the algorithm to jump out of the local optimum.

[0124] In order to decide whether to accept the new solution based on the probability P, a random number r is generated in the range [0, 1]. If the random number r < probability P, the new solution is accepted and the current solution S is updated to S'; otherwise, the new solution is rejected and the current solution S is kept unchanged.

[0125] S4055 Update continuous no improvement times stabl:

[0126] If the new solution is accepted, it means that a better solution has been found, and the number of consecutive times of no improvement is reset to 0.

[0127] If the new solution is rejected, it means that the new solution has not been accepted for the consecutive number of no-improvement stab le+1 times. At this time, the value of the consecutive number of no-improvement stab le is increased by 1.

[0128] The value of the number of consecutive no-improvement stab le reflects the current convergence of the annealing algorithm. If the number of consecutive no-improvement stab le continues to increase, it means that the annealing algorithm may have fallen into a local optimum and needs to jump out of the current area to explore a new solution space.

[0129] S4056 updates the annealing temperature T and the number of iterations iter:

[0130] Annealing temperature T: Update T in an exponential cooling manner, that is, T = α*T, where α is the cooling rate. This means that the temperature will drop to α times the original value each time.

[0131] Iteration number iter: Add 1 to the value of iter to indicate that one iteration has been completed.

[0132] The decrease of annealing temperature T controls the convergence speed of annealing algorithm. When T is very high, the algorithm tends to freely explore the solution space; when annealing temperature T is very low, the algorithm tends to search for the optimal solution locally. The value of iteration number iter records the total number of iterations of the algorithm, which reflects the computational cost of the algorithm.

[0133] When the while loop ends, the final solution S of the simulated annealing algorithm is obtained. This S is the optimal CMYK color combination found, that is, the second CMYK color data, which is used as the output result of the algorithm.

[0134] This optimal combination S has been repeatedly explored and optimized by the algorithm. The RGB color it corresponds to is very close to the target RGB color and can well restore the desired printing color.

[0135] Example 2

[0136] This embodiment is based on the embodiment 1, and the same parts as the embodiment 1 are not described in detail here. Figure 4 Shown

[0137] A PVC floor film digital printing color management system includes: a data acquisition module 510, which acquires historical printing pigment RGB pigment color data, historical printing environment data, historical multi-scale material texture encoding data for characterizing material texture, and corresponding historical CMYK color data; a model training module 520, which takes the historical printing pigment RGB pigment color data, historical printing environment data and historical multi-scale material texture encoding data as input and the corresponding historical CMYK color data as output to train a first adaptive color convolutional neural network model; a real-time prediction module 530, which acquires real-time printing pigment RGB pigment color data, real-time printing environment data and real-time multi-scale material texture encoding data, and predicts first CMYK color data based on the first adaptive color convolutional neural network model; an annealing solution module 540, which initializes a simulated annealing algorithm, and performs a simulated annealing algorithm on the first CMYK color data to obtain second CMYK color data.

[0138] Example 3

[0139] Figure 5 1 is a schematic diagram of the structure of an electronic device provided by an embodiment of the present application. Figure 5 As shown, according to another aspect of the present application, an electronic device 600 is also provided. The electronic device 600 may include one or more processors and one or more memories. The memories store computer readable codes, and when the computer readable codes are executed by one or more processors, the color management method for digital printing of PVC floor film as described above can be executed.

[0140] The method or system according to the embodiment of the present application can also be used by Figure 5 The electronic device architecture shown in FIG. Figure 5As shown, the electronic device 600 may include a bus 601, one or more CPUs 602, a read-only memory (ROM) 603, a random access memory (RAM) 604, a communication port 605 connected to a network, an input / output component 606, a hard disk 607, etc. The storage device in the electronic device 600, such as the ROM 603 or the hard disk 607, may store the PVC floor film digital printing color management method provided in the present application. The color management method for digital printing of PVC floor film may, for example, include: the method is applied to a digital printing system, characterized in that it includes: obtaining historical printing pigment RGB pigment color data, historical printing environment data, historical multi-scale material texture encoding data for characterizing material texture, and corresponding historical CMYK color data; using historical printing pigment RGB pigment color data, historical printing environment data and historical multi-scale material texture encoding data as input, and using corresponding historical CMYK color data as output, to train a first adaptive color convolutional neural network model; obtaining real-time printing pigment RGB pigment color data, real-time printing environment data and real-time multi-scale material texture encoding data, and predicting first CMYK color data based on the first adaptive color convolutional neural network model; initializing a simulated annealing algorithm, and performing a simulated annealing algorithm on the first CMYK color data to obtain second CMYK color data. Furthermore, the electronic device 600 may also include a user interface 608. Of course, Figure 5 The architecture shown is only exemplary and can be omitted according to actual needs when implementing different devices. Figure 5 One or more components of an electronic device are shown.

[0141] Example 4

[0142] Figure 6 Schematic diagram of a computer-readable storage medium structure provided by an embodiment of the present application. Figure 6 As shown, a computer-readable storage medium 700 according to an embodiment of the present application is shown. Computer-readable instructions are stored on the computer-readable storage medium 700. When the computer-readable instructions are executed by the processor, the PVC floor film digital printing color management method according to the embodiment of the present application described with reference to the above figures can be executed. The storage medium 700 includes, but is not limited to, for example, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory (cache). Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0143] In addition, according to the implementation mode of the present application, the process described with reference to the flowchart above can be implemented as a computer software program. For example, the present application provides a non-temporary machine-readable storage medium, the non-temporary 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 by the present application, for example: the method is applied to a digital printing system, characterized in that it includes: obtaining historical printing pigment RGB pigment color data, historical printing environment data, historical multi-scale material texture encoding data for characterizing material texture, and corresponding historical CMYK color data; taking the historical printing pigment RGB pigment color data, historical printing environment data and historical multi-scale material texture encoding data as input, and taking the corresponding historical CMYK color data as output, training a first adaptive color convolutional neural network model; obtaining real-time printing pigment RGB pigment color data, real-time printing environment data and real-time multi-scale material texture encoding data, and predicting first CMYK color data based on the first adaptive color convolutional neural network model; initializing a simulated annealing algorithm, and performing a simulated annealing algorithm on the first CMYK color data to obtain second CMYK color data. When the computer program is executed by a central processing unit (CPU), the above functions defined in the method of the present application are performed.

[0144] The methods, apparatuses, and devices of the present application may be implemented in many ways. For example, the methods, apparatuses, 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.

[0145] 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.

[0146] 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 color management method for digital printing of PVC floor film, the method is applied to a digital printing system, characterized in that: include: Obtain historical printing pigment RGB pigment color data, historical printing environment data, historical multi-scale material texture encoding data for characterizing material texture, and corresponding historical CMYK color data; Taking historical printing pigment RGB pigment color data, historical printing environment data and historical multi-scale material texture encoding data as input and corresponding historical CMYK color data as output, a first adaptive color convolutional neural network model is trained; In the convolution layer of the first adaptive color convolutional neural network model, an adaptive convolution kernel is introduced, and the weight of the convolution kernel is dynamically adjusted according to the characteristics of the material texture. The adaptive convolution kernel determination process is as follows: W_a(x,y)=W(x,y)*(1+α*(M(x,y)M_mean) / (M_std*β)) Where M_mean represents the mean of the material texture feature, M_std represents the standard deviation of the material texture feature, W_a(x, y) is the adaptive convolution kernel of the xth row and yth column, W(x, y) is the adaptive convolution kernel of the original xth row and yth column, α is the adaptive coefficient, M(x, y) is the material texture feature value at the xth row and yth column, and β is the temperature coefficient, which is used to control the smoothness of the adaptive adjustment; Acquire real-time printing pigment RGB pigment color data, real-time printing environment data and real-time multi-scale material texture encoding data, and predict first CMYK color data based on a first adaptive color convolutional neural network model; Initializing a simulated annealing algorithm, and performing a simulated annealing algorithm on the first CMYK color data to obtain second CMYK color data, The step of obtaining the historical printing pigment RGB pigment color data also includes: Convert historical printing pigment RGB pigment color data into HSI data; Calculate the weighted Euclidean distance between each historical print pigment sample in the HSI data and the overall mean ; For historical printing pigment RGB pigment color data, extract the corresponding multi-scale material texture encoding vector; Calculate the texture coding entropy for each sample ; Defining the comprehensive color texture dispersion , Calculate the mean of the comprehensive color texture dispersion and standard deviation ,according to Principle, filter historical printing pigment RGB pigment color data.

2. The method for digital printing color management of PVC floor film according to claim 1, characterized in that: The historical multi-scale material texture encoding data used to characterize material texture is obtained in the following process: Use Gabor filter to encode material texture; by adjusting the parameters σ, θ and λ of the Gabor filter, σ is the standard deviation, indicating the scale of the filter, θ is the direction angle of the filter, indicating the direction of the filter, λ is the wavelength of the filter, indicating periodicity, the material texture can be encoded at different scales and directions.

3. The method for digital printing color management of PVC floor film according to claim 2, characterized in that: The first adaptive color convolutional neural network model training adopts a multi-task loss function, which considers color restoration loss and material texture feature loss at the same time. The material texture feature loss adopts feature reconstruction loss to measure the difference between the material texture features extracted by CNN and Gabor coding. The total loss function is: L_total=λ_1*L_color+λ_2*L_texture*T; Among them, L_color represents the color restoration loss, L_texture represents the material texture feature loss, and λ_1 and λ_2 are balance coefficients.

4. The method for digital printing color management of PVC floor film according to claim 1, characterized in that: The initializing the simulated annealing algorithm and performing the simulated annealing algorithm processing on the first CMYK color data to obtain the second CMYK color data includes: Initialize simulated annealing algorithm parameters; The first CMYK color data predicted by the first adaptive color convolutional neural network model is used as the initial solution S0 of the simulated annealing algorithm; The energy function E(S) is defined as the CIEDE2000 color difference between the predicted CMYK color and the target RGB color; Set initial parameters; A while loop is used to control iteration. The loop condition is that the current annealing temperature T1 is greater than the minimum annealing temperature Tmin, and the number of consecutive no improvements stable is less than the upper limit maxStable.

5. A PVC floor film digital printing color management system, characterized in that: include: A data acquisition module, which acquires historical printing pigment RGB pigment color data, historical printing environment data, historical multi-scale material texture encoding data for characterizing material texture, and corresponding historical CMYK color data; A model training module, which takes historical printing pigment RGB pigment color data, historical printing environment data and historical multi-scale material texture encoding data as input and takes corresponding historical CMYK color data as output to train a first adaptive color convolutional neural network model; A real-time prediction module is used to obtain real-time printing pigment RGB pigment color data, real-time printing environment data and real-time multi-scale material texture encoding data, and predict the first CMYK color data based on the first adaptive color convolutional neural network model; In the convolution layer of the first adaptive color convolutional neural network model, an adaptive convolution kernel is introduced, and the weight of the convolution kernel is dynamically adjusted according to the characteristics of the material texture. The adaptive convolution kernel determination process is as follows: W_a(i,j)=W(i,j)*(1+α*(M(i,j)M_mean) / (M_std*β)) Where M_mean represents the mean of the material texture feature, M_std represents the standard deviation of the material texture feature, W_a(x, y) represents the adaptive convolution kernel of the xth row and yth column, W(x, y) represents the adaptive convolution kernel of the original xth row and yth column, α is the adaptive coefficient, M(x, y) represents the material texture feature value at the xth row and yth column, and β is the temperature coefficient, which is used to control the smoothness of the adaptive adjustment; The annealing solution module initializes the simulated annealing algorithm and processes the first CMYK color data with the simulated annealing algorithm to obtain the second CMYK color data; before obtaining the historical printing pigment RGB pigment color data, the method also includes: Convert historical printing pigment RGB pigment color data into HSI data; Calculate the weighted Euclidean distance between each historical print pigment sample in the HSI data and the overall mean ; For historical printing pigment RGB pigment color data, extract the corresponding multi-scale material texture encoding vector; Calculate the texture coding entropy for each sample ; Defining the comprehensive color texture dispersion , Calculate the mean of the comprehensive color texture dispersion and standard deviation ,according to Principle, filter historical printing pigment RGB pigment color data.

6. An electronic device, characterized in that: The invention comprises a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps in the color management method for digital printing of PVC floor film as claimed in any one of claims 1 to 4 are implemented.

7. A readable storage medium, characterized in that: The readable storage medium stores a computer program, and the computer program is suitable for being loaded by a processor to execute the steps in the color management method for digital printing of PVC floor films according to any one of claims 1 to 4.

Images