Method and system for self-adaptive regulation and control of printed ink color based on machine vision

Abstract

The application discloses a printed matter ink color self-adaptive regulation method and system based on machine vision, relates to the technical field of printed matter ink color regulation, and comprises the following steps: acquiring to-be-printed image data and corresponding post-process process parameters; collecting process image data and constructing a post-process color deviation mapping relationship; reversely mapping expected product color values into pre-compensation target color values of a printing stage; obtaining ink amount pre-bias data; generating ink color regulation instructions and performing dynamic printing by referring to the ink color regulation instructions; and collecting and updating process image data in real time and correcting the ink amount pre-bias data. The application solves the technical problems that the existing printing regulation scheme is difficult to predict color deviation caused by post-processes, lacks an online closed-loop correction mechanism, and it is difficult to control product ink color difference and color consistency is poor, realizes printing color pre-compensation and online closed-loop correction, and improves the technical effects of product ink consistency and color control precision after post-processes.

Description

Technical Field

[0001] This invention relates to the field of ink color control technology for printed materials, and specifically to a method and system for adaptive ink color control of printed materials based on machine vision. Background Technology

[0002] Existing ink color control methods mostly only manage the color of the printing press output plate, and cannot predict color shifts caused by subsequent processing steps such as lamination and drying, lacking effective pre-compensation methods. At the same time, traditional control schemes mostly adopt offline sampling inspection mode, lacking continuous online data feedback and closed-loop optimization mechanisms, making it difficult to dynamically correct color drift during the production process. This easily leads to problems such as large ink color deviations in batch finished products and poor color consistency among products in the same batch, making it difficult to meet the production control requirements of high-precision printing products.

[0003] Existing printing control solutions struggle to predict color shifts caused by subsequent processes and lack online closed-loop correction mechanisms, resulting in significant technical challenges in controlling ink color differences and achieving poor color consistency in finished products. Summary of the Invention

[0004] This application provides a machine vision-based method and system for adaptive control of ink color in printed materials, which addresses the technical problems of existing printing control schemes, such as difficulty in predicting color shifts caused by subsequent processes, lack of online closed-loop correction mechanisms, difficulty in controlling ink color difference in finished products, and poor color consistency.

[0005] In view of the above problems, this application provides a machine vision-based method and system for adaptive control of ink color in printed materials.

[0006] The first aspect of this application provides a machine vision-based method for adaptive control of ink color in printed materials, the method comprising:

[0007] The process involves: acquiring the image data to be printed and the corresponding process parameters for subsequent processes; collecting process image data of the sample printed material before and after the execution of the subsequent processes using a visual sampling module, and constructing a subsequent color deviation mapping relationship based on the process image data; mapping the expected finished color value of the image data to be printed to a pre-compensation target color value for the printing stage according to the subsequent color deviation mapping relationship; comparing the pre-compensation target color value with the standard printing color value to obtain ink volume pre-bias data; generating ink color control instructions based on the ink volume pre-bias data, and performing dynamic printing execution with reference to the ink color control instructions; and during the dynamic printing execution process, collecting updated process image data under the current process in real time, and correcting the ink volume pre-bias data based on the updated process image data.

[0008] A second aspect of this application provides a machine vision-based adaptive ink color control system for printed materials, the system comprising:

[0009] The system includes the following modules: a process parameter acquisition module for acquiring image data to be printed and corresponding process parameters for subsequent processes; a mapping relationship construction module for acquiring process image data of sample printed materials before and after the execution of the subsequent processes through a visual sampling module, and constructing a subsequent color deviation mapping relationship based on the process image data; a pre-compensation target color value acquisition module for inversely mapping the expected finished color value of the image data to be printed to the pre-compensation target color value for the printing stage according to the subsequent color deviation mapping relationship; an ink volume pre-bias data acquisition module for comparing the pre-compensation target color value with the standard printing color value to obtain ink volume pre-bias data; a dynamic printing execution module for generating ink color control instructions based on the ink volume pre-bias data, and performing dynamic printing execution with reference to the ink color control instructions; and an ink volume pre-bias data correction module for acquiring updated process image data under the current process in real time during dynamic printing execution, and correcting the ink volume pre-bias data based on the updated process image data.

[0010] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0011] The process involves acquiring the image data to be printed and the corresponding process parameters for subsequent processes; collecting process image data of sample printed materials before and after the execution of the subsequent processes, and constructing a color deviation mapping relationship for subsequent processes based on the process image data; inversely mapping the expected finished color value of the image data to be printed to a pre-compensation target color value for the printing stage; comparing the pre-compensation target color value with the standard printing color value to obtain ink volume pre-bias data; generating ink color control instructions and performing dynamic printing execution with reference to the ink color control instructions; and collecting updated process image data under the current process in real time and correcting the ink volume pre-bias data based on the updated process image data. This achieves the technical effect of realizing pre-compensation and online closed-loop correction of printing color, improving the consistency of ink color and color control accuracy of the finished product after subsequent processes. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 A schematic flowchart of a machine vision-based adaptive ink color control method for printed materials provided in an embodiment of this application;

[0014] Figure 2 A schematic diagram of the structure of a machine vision-based adaptive ink color control system for printed materials provided in an embodiment of this application.

[0015] Explanation of reference numerals in the attached figures: Process parameter acquisition module 10, mapping relationship construction module 20, pre-compensation target color value acquisition module 30, ink volume pre-bias data acquisition module 40, dynamic printing execution module 50, ink volume pre-bias data correction module 60. Detailed Implementation

[0016] This application provides a machine vision-based method and system for adaptive control of ink color in printed materials, which addresses the technical problems of existing printing control schemes, such as difficulty in predicting color shifts caused by subsequent processes, lack of online closed-loop correction mechanisms, difficulty in controlling ink color difference in finished products, and poor color consistency.

[0017] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0018] Example 1, as Figure 1 As shown, this application provides a machine vision-based adaptive ink color control method for printed materials, the method comprising:

[0019] Step S100: Obtain the image data to be printed and the corresponding process parameters for subsequent processes.

[0020] Specifically, the process involves acquiring the image data to be printed and the corresponding downstream process parameters, reading the image output from the printing design end and performing layered analysis to distinguish brand logos, text, and pattern areas within the image, extracting image information such as color gamut and standard color values ​​for each area, and simultaneously collecting parameters such as the type of downstream processes (lamination, hot stamping, etc.), processing temperature, processing pressure, and lamination material thickness corresponding to the order. It also collects substrate attribute data such as the material and ink absorption limit of the current batch of substrates. The image data and the matching downstream process parameters are integrated into an input dataset for subsequent construction of downstream color deviation mapping relationships.

[0021] Step S200: Using a visual sampling module, process image data of the sample printed material before and after the subsequent process is performed, and construct a subsequent color deviation mapping relationship based on the process image data.

[0022] Specifically, the image acquisition and mapping relationship is constructed by using a visual sampling module. First, standard color card images are printed on the same substrate. The visual sampling module is used to collect the first process image data before the subsequent processing and the second process image data after the processing. The color values ​​of corresponding color blocks in the two sets of images are compared one by one to obtain the color value offset data of each color block under the action of subsequent processing. Then, the standard color card color value, substrate attributes, and subsequent process parameters are used as inputs, and the color value offset data is used as outputs to build a subsequent color deviation mapping relationship that can characterize the color change law brought about by subsequent processing.

[0023] Step S300: Based on the subsequent color deviation mapping relationship, the expected finished color value of the image data to be printed is reverse mapped to the pre-compensated target color value for the printing stage.

[0024] Specifically, relying on the established back-end color deviation mapping relationship, the pre-compensation target color value for the printing stage is solved in reverse. First, the acquired image data to be printed is analyzed in layers to identify the brand logo, text, and pattern areas in the image. The visual saliency index of each area is calculated. Areas with indices higher than the preset threshold are defined as key color areas, and the remaining areas are defined as non-key color areas. The color values ​​of the key color areas that meet the expected finished product color value limit and the average color value of the non-key color areas are extracted as the overall expected finished product color value. The back-end color deviation mapping relationship is called to query the expected color value changes of each key color area after the corresponding back-end process. The expected color value changes are reversed and the key color areas are given a higher weight coefficient in the calculation process. Finally, the pre-compensation target color value corresponding to the printing stage is obtained.

[0025] Step S400: Compare the pre-compensated target color value with the standard printing color value to obtain ink volume pre-bias data.

[0026] Specifically, the pre-compensated target color value obtained by inverse mapping is compared with the preset standard printing color value item by item, and the color value difference between each color channel is calculated. Based on the color value difference, the target ink layer thickness data required for the corresponding area is obtained by inverse conversion. It is determined whether the target ink layer thickness exceeds the ink absorption limit value corresponding to the current substrate. If it exceeds the limit, the color value deviation corresponding to the non-critical color area is compressed according to the preset ratio while keeping the desired finished color value unchanged, and the corresponding ink layer thickness value is corrected. Finally, the ink volume pre-offset data adapted to the full-width is converted and output.

[0027] Step S500: Generate ink color control instructions based on the ink volume pre-bias data, and perform dynamic printing in accordance with the ink color control instructions.

[0028] Specifically, the target ink layer thickness data corresponding to each area of ​​the printing area is obtained by inversion calculation based on the ink volume pre-bias data. First, it is determined whether the target ink layer thickness exceeds the ink absorption limit of the current substrate. If it exceeds the limit, the ink volume pre-bias data corresponding to the non-critical color area is compressed according to a preset ratio while maintaining the desired finished color value. Then, the ink color control command for zoned ink supply is generated based on the corrected ink volume pre-bias data. The control command for distinguishing the ink supply of critical color areas and non-critical color areas is sent to the printing equipment. The printing equipment adjusts the ink fountain output according to the ink color control command to complete the dynamic printing operation of the whole plate.

[0029] Step S600: During the dynamic printing process, the updated process image data of the current process is collected in real time, and the ink volume pre-bias data is corrected according to the updated process image data.

[0030] Specifically, throughout the entire process of dynamic printing by the printing equipment according to the ink color control instructions, the visual sampling module continuously collects the update process image data of the printed semi-finished product in real time, converts the image into the corresponding current measured color value, calculates the real-time deviation between the current measured color value and the pre-compensated target color value, and identifies the distribution position of the deviation on the printing surface if the real-time deviation exceeds the allowable error range. When the deviation is concentrated in the critical color area, a larger first gain coefficient is used to amplify the deviation amplitude, and the ink volume compensation increment is calculated to adjust the ink volume pre-bias data. If the deviation only exists in the non-critical color area and does not exceed the preset tolerance range, the existing ink volume pre-bias data is not changed. Finally, the ink color control instructions are updated according to the adjusted ink volume pre-bias data to realize the dynamic closed-loop correction of ink volume during the printing process.

[0031] In one possible implementation, step S200 further includes:

[0032] Step S210: Print standard color card images on the same substrate, and collect the first process image data of the standard color card images before the subsequent process is executed and the second process image data after the execution is executed through the visual sampling module.

[0033] Step S220: Compare the corresponding color blocks of the first process image data and the second process image data block by block to obtain the color value offset data of each color block under the action of the subsequent process.

[0034] Step S230: Using the color value data of the standard color card image, the substrate attribute data, and the process parameters of the subsequent process as input, and the color value offset data as output, establish the subsequent color offset mapping relationship.

[0035] Specifically, a substrate consistent with the production order is selected, and a standard color card image containing multiple sets of gradient standard color blocks is printed on its surface. After printing, the standard color card is photographed completely using a visual sampling module to collect the first process image data before subsequent processing and store the original color value information of each color block. Then, the substrate carrying the standard color card is sent to the corresponding subsequent processing such as lamination and hot stamping to complete the processing. After processing, the visual sampling module is used again to collect the image of the entire standard color card with the same shooting angle and light source parameters to obtain the second process image data after the subsequent processing is completed. The shooting environment parameters are consistent when collecting the two sets of images to eliminate the acquisition error caused by external light and shooting distance, ensuring that the color blocks of the images before and after can be accurately matched one by one.

[0036] The first and second process image data are pixel aligned and registered. The color block regions in the two sets of images are matched one by one according to the color card arrangement order. The Lab or CMYK standard color value of each matched color block is extracted. The difference of each color channel value before and after processing of the same color block is calculated item by item. The color channel differences of each set are integrated and summarized to obtain the complete color value offset data of each color block affected by subsequent processes such as lamination and hot stamping, and stored separately.

[0037] The original Lab color values ​​of standard color cards, substrate material and ink absorption limit, lamination / hot stamping temperature, rolling pressure, and film thickness are extracted as multi-dimensional input features. The obtained color value offsets of each color card are used as the model output labels. First, Min-Max normalization is performed on all input features to eliminate dimensional interference. Gradient Boosting Regression Tree (GBDT) is selected as the basic algorithm to construct a mapping model. The normalized multi-dimensional features are used as independent variables and the color value offset data is used as the dependent variable to divide the training set and validation set for iterative training. By adjusting hyperparameters such as learning rate and tree depth, the loss error between predicted and measured offsets is continuously reduced. After the model loss converges and reaches the target, the model weights are saved to construct a one-to-one mapping relationship between input parameters and subsequent color offset. In actual calculation, the color values ​​of the area to be printed, the parameters of the current batch of substrate, and the subsequent process parameters of the order are input. The trained GBDT model can directly output accurate expected color value offsets, providing a quantitative calculation basis for reverse solving of printing pre-compensation target color values.

[0038] In one possible implementation, step S300 further includes:

[0039] Step S310: Extract the key color regions from the image data to be printed.

[0040] Step S320: Call the subsequent color offset mapping relationship to query the expected color value change of the key color area after processing by the subsequent process parameters.

[0041] Step S330: Based on the expected color value change, perform reverse offset calculation on the expected finished color value of the key color area to obtain the pre-compensated target color value.

[0042] Specifically, the image data to be printed is first split into layers, dividing it into three independent regions: brand logo, text, and pattern. For each segmented region, three basic features are extracted: saturation, region pixel ratio, and edge contrast. After assigning fixed weights to each feature, the weighted sum is used to obtain the visual saliency value of the region. The visual saliency value is compared with a system preset threshold. Regions with saliency values ​​higher than the threshold are classified as key color regions, and those lower than the threshold are classified as non-key color regions. After the division, all pixels in the key color regions are traversed, and color values ​​that fall within the expected finished product color value range are selected and extracted separately. At the same time, the arithmetic mean of all pixel color values ​​in the non-key color regions is calculated as the unified color value for this type of region, thus completing the extraction of key color regions and corresponding color value data.

[0043] The extracted key color regions' original expected finished color values, current batch substrate attribute parameters, and all subsequent process parameters such as lamination and hot stamping are integrated into the model input vector. After input feature normalization, the vector is imported into the trained GBDT post-processing color shift mapping model. The model performs forward inference based on the weight parameters converged during internal training and outputs the offset of each color value in the Lab channel under the input conditions. This offset is the expected color value change that the key color regions will produce after processing by the corresponding subsequent processes. The complete output of the independent color shift results for each key color region is used for subsequent reverse compensation calculations.

[0044] First, the color calculation adopts the Lab color space. Let the expected finished color value of the key color area be (L0, a0, b0). The expected color value offset obtained by the mapping model is (ΔL, Δa, Δb). The offset represents the change of the finished color value after subsequent processing compared with the printed color value. The reverse cancellation operation logic is adopted: the pre-compensation target color value in the printing stage needs to offset the color deviation generated later. Therefore, subtraction operation is performed on the three channels L, a, and b respectively. The pre-compensation target color values ​​are L1=L0−ΔL, a1=a0−Δa, and b1=b0−Δb. At the same time, according to the claim limitation rules, the key color area is assigned a higher weight coefficient than the non-key color area. The complete and unattenuated offset offset value is directly used in the calculation without compressing the compensation range. Finally, after the calculation of each channel is completed, the pre-compensation target color value corresponding to the key color area is obtained. This value is used to compare with the standard printed color value to obtain the ink volume pre-bias data.

[0045] In one possible implementation, step S310 further includes:

[0046] Step S311: Perform layered analysis on the image data to be printed to identify the brand logo area, text area and pattern area contained therein.

[0047] Step S312: Calculate the visual saliency index for different regions, mark the regions with visual saliency indices higher than the preset threshold as key color regions, and mark the remaining regions as non-key color regions.

[0048] Step S313: For the key color areas, extract color value data that conforms to the expected finished product color value range; for the non-key color areas, extract the average color value as the expected finished product color value.

[0049] Specifically, the image data to be printed is loaded and subjected to layered analysis. If it is a vector image, the built-in layer information of the file is directly read to complete the initial layering. If it is a bitmap image, the virtual layer is decomposed through pixel clustering. Then, Canny edge detection is used to extract the closed contour of the image. The SIFT feature matching algorithm is used to compare with the pre-built brand logo template library. Regions with feature similarity exceeding a preset threshold are identified as brand logo regions. The MSER maximum stable extremum region algorithm is used to extract the connected components of the characters. The text regions are obtained by combining the character aspect ratio and gray-level gradient features. Finally, the gray-level co-occurrence matrix (GLCM) is used to extract texture features from the remaining image regions to divide them into decorative pattern regions. At the same time, the boundary mask, pixel coordinates and layer number of each type of region are recorded to complete the image region classification and recognition, providing data support for subsequent partitioning and visual saliency calculation.

[0050] For the identified brand logo area, text area, and pattern area, three types of features are extracted: average saturation, pixel occupancy ratio, and edge gradient contrast. Min-Max normalization is applied to these three features to eliminate dimensional differences. A weighted summation model is used to calculate the visual saliency index: Visual saliency S = w1×S1 + w2×S2 + w3×S3, where S1 is the normalized saturation, S2 is the normalized area occupancy ratio, and S3 is the normalized edge contrast, with w1, w2, and w3 being pre-defined weighting coefficients. The calculated visual saliency index for each area is compared with a preset threshold. Areas with index values ​​higher than the threshold are selected and marked as key color areas, while areas with visual saliency indices less than or equal to the threshold are marked as non-key color areas. The area classification results and corresponding coordinate information are saved simultaneously, providing a basis for color pre-compensation calculations for each area.

[0051] The process iterates through the Lab color space values ​​of all pixels within the critical color region. An interval filtering method is used to determine whether each pixel's color value falls within the preset upper and lower limits of the desired final color value. All pixel colors within the interval are retained to form a set of valid color values ​​for the critical color region. For non-critical color regions, the Lab channel color values ​​of all pixels within the region are extracted. The arithmetic mean of the L, a, and b channel pixel colors is calculated separately. The three average values ​​are combined to obtain the average color value of the region, which is then used as the desired final color value for the non-critical color region. Finally, the valid color values ​​after filtering for the critical color region and the average desired color value for the non-critical color region are stored separately for subsequent reverse pre-compensation calculations.

[0052] In one possible implementation, step S310 further includes:

[0053] When calculating the pre-compensated target color value, the weight coefficient of the key color region is given higher than that of the non-key color region.

[0054] Specifically, when calculating the pre-compensated target color value, differentiated weighting coefficients are configured, and weights are assigned to key color areas. Non-critical color areas are assigned weights ,satisfy The calculation is performed using the Lab color space, and the formula is as follows: In the formula The desired color value of the finished product for the region. For the expected color value offset brought about by the subsequent process, w selects the corresponding area weight; the key color area uses a high weight to fully perform the color offset reverse cancellation without compressing the compensation range, ensuring the color reproduction accuracy of core areas such as brand and text, while the non-key color area adopts a lower weight to appropriately reduce the compensation intensity and weaken the compensation input of secondary pattern areas, and solves the pre-compensation target color value corresponding to the two types of areas respectively.

[0055] In one possible implementation, step S500 further includes:

[0056] Step S510: Obtain the target ink layer thickness data by inverting the ink volume pre-bias data.

[0057] Step S520: Determine whether the target ink layer thickness data exceeds the current ink absorption limit of the substrate.

[0058] Step S530: If the target ink layer thickness data exceeds the upper limit of ink absorption of the current substrate, then, while keeping the desired finished color value unchanged, the ink volume pre-bias data corresponding to the non-critical color area is compressed according to a preset ratio, and the ink color control instruction is generated based on the compressed ink volume pre-bias data.

[0059] Specifically, multiple sets of gradient ink prototyping calibration experiments were first conducted. Samples with multiple gradient ink layer thicknesses were set on a substrate consistent with production. CMYK ink color values ​​corresponding to different ink layer thicknesses were collected. After collection, ink layer thickness was used as the independent variable and Lab measured color value as the dependent variable. A multivariate polynomial fitting algorithm was used to construct a calibration mapping model between ink layer thickness and color response. After the model was constructed, the pre-biased data of ink volume in each zone was substituted into the model and combined with the ink absorption parameters of the substrate to perform inverse numerical solution. The initial target ink layer thickness of each zone of the printing area was obtained by inversion. The solution results were then compared with the preset maximum and minimum allowable ink layer thickness thresholds of the substrate. Values ​​exceeding the range were subjected to amplitude limiting processing. Finally, the verified target ink layer thickness data of each zone was output, providing a quantitative basis for generating ink color control instructions.

[0060] The system retrieves the pre-calibrated upper limit of ink absorption and the minimum effective ink layer thickness threshold of the current batch of substrates. It then iterates through the target ink layer thickness data corresponding to each partition of the printing area and performs numerical verification on a partition-by-partition basis using an interval judgment algorithm. The target ink layer thickness of each partition is compared with the upper and lower limit thresholds. When the target ink layer thickness of a partition is greater than the upper limit of ink absorption of the substrate, the target ink layer thickness of that partition is forcibly corrected to the upper limit of ink absorption to prevent printing defects such as ink accumulation and bleed-through. When the target ink layer thickness of a partition is lower than the minimum effective ink layer thickness threshold, it is uniformly corrected to the minimum effective ink layer thickness. The target ink layer thickness within the threshold range remains unchanged. The system outputs the target ink layer thickness dataset after boundary constraint correction, which is used for subsequent conversion to generate ink color control instructions.

[0061] Based on the threshold determination results, if the ink layer thickness data of any zone exceeds the ink absorption limit specified by the current substrate, in order to strictly ensure the constant expected color value of the final product and not damage the color compensation accuracy of key color areas, a zone-specific ink volume compression algorithm is activated: the ink volume pre-bias data of the key color area remains unchanged, and only the ink volume pre-bias data of the non-key color area is subjected to proportional compression correction. The compression calculation formula is constructed by using the system's preset ink volume compression ratio coefficient k (0 < k < 1). The corrected ink volume pre-bias value = original non-key area ink volume pre-bias value × k, and is weakened proportionally. The ink compensation range in non-critical areas reduces the overall ink layer thickness load, causing the inverted target ink layer thickness to fall back to within the upper limit of ink absorption of the substrate. After completing the ink volume data compression and correction, the constant ink volume pre-bias data of critical color areas is integrated with the compressed ink volume pre-bias data of non-critical color areas. Through a preset ink volume-equipment parameter mapping algorithm, it is uniformly converted into standard control instructions for dot opening and ink supply speed that can be recognized by the printing equipment. Finally, complete zone ink color control instructions are generated, which not only avoids printing defects such as ink accumulation and poor drying, but also maximizes the overall color consistency of the finished product.

[0062] In one possible implementation, step S600 further includes:

[0063] Step S610: Collect the update process image data in real time through the visual sampling module, and convert the update process image data into the current measured color value.

[0064] Step S620: Calculate the real-time deviation between the current measured color value and the pre-compensated target color value.

[0065] Step S630: When the real-time deviation is outside the allowable error range, the ink volume pre-bias data is incrementally adjusted according to the real-time deviation, and the ink color control command is updated according to the incrementally adjusted ink volume pre-bias data.

[0066] Specifically, during the online production stage where printing ink color control commands are implemented and the ink layer on the printing plate is dynamically updated, the system relies on the high-definition visual sampling module on the equipment to capture images of the printing plate of the substrate in real time at a preset acquisition frame rate, acquiring image data of the update process. Before acquisition, the system automatically completes the light source brightness calibration, dynamic white balance correction, and lens geometric distortion correction of the visual module to eliminate hardware acquisition errors and ensure the accuracy of image color reproduction and image integrity. Subsequently, the original RGB values ​​of the pixels in the effective printing area of ​​the image are extracted, and the RGB values ​​acquired by the equipment are first converted to the CIE XYZ intermediate color space through a calibration matrix. The conversion formula is: In the formula: M is the 3×3 color conversion matrix obtained by the visual module through color chart calibration in advance; R, G, and B are the normalized original color values ​​of the three channels of the pixel obtained by visual sampling; X, Y, and Z represent the tristimulus values ​​in the CIE XYZ color space after conversion; and then the Lab color value is calculated from XYZ based on the CIE standard conversion formula: Piecewise functions: Where: L is the lightness channel color value, a is the red-green color value, and b is the yellow-blue color value; , , The standard white point is represented by the tristimulus reference value, using the D50 standard illuminator parameters; t is the ratio of each channel's tristimulus value to the white point reference value; the system combines a preset area mask to accurately filter noise, background, and invalid pixels, and calculates the arithmetic mean of the L, a, and b channel values ​​of all valid pixels in both key and non-key color areas, accurately outputting the current measured color value for each partition, forming a standardized measured color dataset, providing accurate raw data support for subsequent online color difference calculation and closed-loop iterative correction.

[0067] Using the Lab color space as a unified calculation benchmark, the current measured color values ​​of each partition are read. and pre-compensated target color value First, calculate the channel differences for the three channels: brightness, red-green hue, and yellow-blue hue. , , Then, the comprehensive color difference value is calculated using the CIEDE2000 color difference formula to obtain the real-time color deviation between the measured color value and the pre-compensated target color value. The independent deviation components of each monochrome channel are retained simultaneously, and the deviation results corresponding to key color areas and non-key color areas are extracted. The channel deviation data and comprehensive color difference data are output in their entirety to determine whether it is necessary to start the closed-loop adjustment of ink volume parameters.

[0068] The calculated overall color difference is compared with the single-channel deviation component and the preset color difference allowable threshold. If the real-time deviation exceeds the allowable error range, a proportional incremental adjustment algorithm is used to correct the parameters. The incremental correction amount is constructed based on the color difference value of each channel. Differentiated adjustment coefficients are configured for key color areas and non-key color areas. A larger adjustment coefficient is used to reduce the color deviation in key color areas, while a relatively gentle adjustment coefficient is used in non-key color areas. The original ink volume pre-bias data is superimposed and calculated according to the correction increment to complete the incremental adjustment. The target ink layer thickness corresponding to the adjusted ink volume pre-bias data is re-verified to meet the upper limit constraint of ink absorption of the substrate. After the verification is passed, a new ink color control command is recalculated based on the updated ink volume pre-bias data and sent to the printing execution mechanism to realize the online closed-loop correction of printing color.

[0069] In one possible implementation, step S630 further includes:

[0070] Step S631: Identify the distribution position of the real-time deviation within the printing area. When the real-time deviation is concentrated in the key color area, the deviation amplitude is amplified using the first gain coefficient, the ink volume compensation increment is calculated, and the ink volume pre-bias data is updated based on the ink volume compensation increment.

[0071] Step S632: When the real-time deviation only appears in the non-critical color area and does not exceed the preset tolerance range, maintain the current ink volume pre-bias data unchanged.

[0072] Specifically, the system iterates through the real-time deviation data corresponding to each zone of the printed area, identifies the coordinates of deviations exceeding the allowable error range, and determines the type of area to which the deviation belongs. If the identified out-of-tolerance real-time deviations are concentrated in key color areas, the first gain coefficient is activated. Furthermore, the first gain coefficient is greater than the gain coefficient used in non-critical areas, according to the calculation formula. Solve for the ink volume compensation increment, where Represents the ink volume compensation increment, where E is the real-time comprehensive color difference value for the partition; the calculated ink volume compensation increment is then added to the original ink volume pre-bias data for the current partition to satisfy... ,in To adjust the pre-offset data for ink volume, This is the incrementally updated ink volume pre-bias data; after the update is completed, the corresponding target ink layer thickness is simultaneously checked to see if it exceeds the upper limit constraint of ink absorption of the substrate. After the check is compliant, the updated ink volume pre-bias data is output to generate new ink color control instructions, which prioritizes accelerating the color difference convergence speed of key color areas and ensures the color accuracy of core areas.

[0073] The system identifies the location of real-time deviations exceeding the ideal standard within the printing area, determines that all deviation points are distributed in non-critical color regions, and then compares the overall color difference of each point with the preset tolerance threshold for non-critical color regions one by one. If all real-time deviations do not exceed the tolerance range, the color difference of the secondary pattern area is determined to be within the acceptable range. There is no need to start the ink volume adjustment process; the system directly maintains the current ink volume pre-bias data of all zones unchanged and continues to use the existing ink color control instructions to control the ink supply of the printing equipment. This reduces unnecessary frequent parameter adjustments, avoids printing condition fluctuations caused by small color difference fluctuations, and ensures stable operation of the printing production process.

[0074] In one possible implementation, step S600 further includes:

[0075] During the continuous production of the same order, the final product image data of the completed subsequent processes are collected intermittently.

[0076] The final product image data is compared with the expected product color value to obtain the actual color deviation data.

[0077] The difference between the actual color deviation data and the color value offset data is calculated to obtain the mapping residual data, and the parameters of the subsequent color deviation mapping relationship are fine-tuned using the mapping residual data.

[0078] Specifically, in the batch production stage of continuously producing the same printing order, the system triggers the sampling process according to a pre-set sampling time interval or a fixed product quantity interval. It uses a visual sampling module to acquire images of the finished products after drying, lamination and other downstream processes. Before the acquisition operation is executed, the system performs light source correction, white balance and distortion correction of the visual module to eliminate acquisition errors. During the acquisition process, it locks the effective printing area of ​​the finished product, removes invalid edge areas and areas covered by auxiliary materials, and continuously acquires multiple sets of finished product image data. The acquired finished product images are converted into the actual measured color values ​​of the finished product in the Lab color space to track the color drift phenomenon caused by downstream processes and provide a long-term finished product color benchmark for subsequent iteration optimization of ink volume pre-bias parameters.

[0079] The acquired final product image data is masked and segmented to separate key color regions from non-key color regions. The measured Lab color values ​​of the finished product in each region are calculated using a standard conversion algorithm from RGB to CIE XYZ and then to Lab. The expected color values ​​of the finished product corresponding to each region are retrieved. The CIEDE2000 color difference calculation formula is used to perform comparison calculations for each region, calculating the individual deviation components and the comprehensive color difference values ​​of the lightness channel, red-green channel, and yellow-blue channel. The channel deviation and comprehensive color difference information of each region are integrated to construct the actual color deviation dataset corresponding to each region. This quantifies the color shift caused by subsequent processes and provides data support for continuous optimization of ink volume pre-compensation parameters.

[0080] In the unified Lab color space, the actual color shift components of the finished product in each zone ( , , ) and the color value offset component of the subsequent process predicted by the system ( , , Precise matching and difference calculation are performed channel-by-channel and partition-by-partition, using the residual calculation formula. , , The residual components of brightness, red-green, and yellow-blue channels for each zone are obtained, where: , , The actual color deviation of the brightness, red-green hue, and yellow-blue hue obtained from the actual measured values ​​of the finished products in each zone; , , Predict color cast based on the brightness, red-green hue, and yellow-blue hue output by the subsequent color cast mapping model; , , Corresponding sequentially to the brightness channel residuals, red-green channel residuals, and yellow-blue channel residuals, all partition residuals are integrated to construct a mapped residual dataset; a mean squared error loss function is then constructed based on the residuals. In the formula: The overall loss function value represents the overall prediction error of the model; N represents the total number of valid printed partitions participating in the calculation; i is the partition number. , , The three-channel residuals correspond to the i-th partition are used; gradient descent optimization algorithm is employed for parameter fine-tuning, and the loss function is solved relative to all updatable parameters within the mapping model. gradient ,in For the loss function with respect to the parameters The gradient represents the optimal direction for parameter adjustment; a preset learning rate is introduced. Control the step size of each correction and execute the parameter update formula. In the formula: The original model parameters before fine-tuning; These are the model parameters after iterative updates; The learning rate is used to limit the magnitude of a single parameter adjustment to avoid oscillation and divergence. The parameters are slightly corrected according to the above rules to continuously reduce the error between the model's predicted color deviation and the actual color deviation of the finished product. The color deviation mapping relationship of the subsequent process is dynamically optimized to achieve online closed-loop self-optimization of the printing color pre-compensation model.

[0081] A comparative experiment was conducted in a batch color box offset printing production scenario. Printed semi-finished products underwent hot air drying and BOPP lamination to obtain the final product. Production trials were simultaneously conducted using both the traditional ink color control scheme and the machine vision-based adaptive ink color control method of this invention. Color data for key graphic areas and non-key background areas were collected at two key points: immediately after printing and after lamination. The CIEDE2000 comprehensive color difference was calculated. The experimental results are shown in Table 1.

[0082] Table 1. Comparison of color ΔE (CIEDE2000) at different stages under different control schemes

[0083] Key Image and Text Area Just off the printing press 2.92 1.03 ≤1.5 Key Image and Text Area Coated finished products 4.18 1.26 ≤1.5 Non-critical background area Just off the printing press 2.47 1.21 ≤2.0 Non-critical background area Coated finished products 3.54 1.52 ≤2.0

[0084] It can be seen that traditional solutions can only control the color in the printing off-press stage. After the subsequent lamination process, the color difference increases significantly, and the key areas exceed the allowable color difference threshold. This invention achieves pre-compensation by constructing a color deviation mapping relationship in the subsequent process, and is combined with online closed-loop correction in production. The color difference is always maintained within the standard threshold range throughout the entire process from the printing semi-finished product to the final product. It can effectively offset the systematic color deviation caused by drying and lamination, and suppress color drift in the continuous production process. This verifies the control advantages of this method in high-precision printing mass production scenarios.

[0085] Example 2, based on the same inventive concept as the machine vision-based adaptive ink color control method for printed materials in the foregoing examples, such as... Figure 2 As shown, this application provides a machine vision-based adaptive ink color control system for printed materials. The system and method embodiments in this application are based on the same inventive concept. The system includes:

[0086] The process parameter acquisition module 10 is used to acquire the image data to be printed and the corresponding process parameters of the subsequent process.

[0087] The mapping relationship construction module 20 is used to collect process image data of the sample printed material before and after the execution of the subsequent process through the visual sampling module, and to construct the subsequent color deviation mapping relationship based on the process image data.

[0088] The pre-compensation target color value acquisition module 30 is used to reverse map the expected finished color value of the image data to be printed to the pre-compensation target color value in the printing stage according to the subsequent color deviation mapping relationship.

[0089] The ink volume pre-bias data acquisition module 40 is used to compare the pre-compensated target color value with the standard printing color value to obtain ink volume pre-bias data.

[0090] The dynamic printing execution module 50 is used to generate ink color control instructions based on the ink volume pre-bias data, and to perform dynamic printing execution with reference to the ink color control instructions.

[0091] The ink volume pre-bias data correction module 60 is used to collect updated process image data in real time during the dynamic printing process and correct the ink volume pre-bias data based on the updated process image data.

[0092] Furthermore, the system is also used to implement the following functions:

[0093] A standard color card image is printed on the same substrate. A visual sampling module collects the first process image data of the standard color card image before the subsequent process and the second process image data after the process. The corresponding color blocks of the first process image data and the second process image data are compared block by block to obtain the color value offset data of each color block under the action of the subsequent process. The color value data of the standard color card image, the attribute data of the substrate, and the process parameters of the subsequent process are used as inputs, and the color value offset data is used as outputs to establish the subsequent color deviation mapping relationship.

[0094] Furthermore, the system is also used to implement the following functions:

[0095] Extract the key color regions from the image data to be printed; call the subsequent color offset mapping relationship to query the expected color value change of the key color regions after processing by the subsequent process parameters; perform reverse offset calculation on the expected finished color value of the key color regions according to the expected color value change to obtain the pre-compensated target color value.

[0096] Furthermore, the system is also used to implement the following functions:

[0097] The image data to be printed is analyzed in layers to identify the brand logo area, text area and pattern area contained therein; the visual salience index of different areas is calculated respectively, and the areas with visual salience index higher than the preset threshold are marked as key color areas, and the remaining areas are marked as non-key color areas; for the key color areas, color value data that meets the expected finished product color value limit range is extracted, and for the non-key color areas, the average color value is extracted as the expected finished product color value.

[0098] Furthermore, the system is also used to implement the following functions:

[0099] When calculating the pre-compensated target color value, the weight coefficient of the key color region is given higher than that of the non-key color region.

[0100] Furthermore, the system is also used to implement the following functions:

[0101] The target ink layer thickness data is obtained by inverting the ink volume pre-bias data; it is determined whether the target ink layer thickness data exceeds the ink absorption limit of the current substrate; if the target ink layer thickness data exceeds the ink absorption limit of the current substrate, then, while keeping the desired finished color value unchanged, the ink volume pre-bias data corresponding to the non-critical color area is compressed according to a preset ratio, and the ink color control instruction is generated based on the compressed ink volume pre-bias data.

[0102] Furthermore, the system is also used to implement the following functions:

[0103] The update process image data is collected in real time by a visual sampling module and converted into the current measured color value; the real-time deviation between the current measured color value and the pre-compensated target color value is calculated; when the real-time deviation is outside the allowable error range, the ink volume pre-bias data is incrementally adjusted according to the real-time deviation, and the ink color control command is updated according to the incrementally adjusted ink volume pre-bias data.

[0104] Furthermore, the system is also used to implement the following functions:

[0105] The distribution location of the real-time deviation within the printing area is identified. When the real-time deviation is concentrated in the critical color area, the deviation amplitude is amplified using a first gain coefficient to calculate the ink volume compensation increment, and the ink volume pre-bias data is updated based on the ink volume compensation increment. When the real-time deviation only appears in the non-critical color area and does not exceed the preset tolerance range, the current ink volume pre-bias data is maintained unchanged.

[0106] Furthermore, the system is also used to implement the following functions:

[0107] During the continuous production of the same order, the final finished product image data of the completed subsequent process is collected intermittently; the final finished product image data is compared with the expected finished product color value to obtain the actual color deviation data; the difference between the actual color deviation data and the color value deviation data is calculated to obtain the mapping residual data, and the parameters of the subsequent color deviation mapping relationship are fine-tuned using the mapping residual data.

[0108] It should be noted that the order of the embodiments described above is for descriptive purposes only and does not represent the superiority or inferiority of the embodiments. Specific embodiments of this specification have been described above. Furthermore, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0109] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0110] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application intends to include such modifications and variations.

Claims

1. A machine vision-based adaptive ink color control method for printed materials, characterized in that, The method includes: Obtain the image data to be printed and the corresponding process parameters for subsequent processes; The visual sampling module collects process image data of the sample printed material before and after the subsequent process, and constructs a subsequent color deviation mapping relationship based on the process image data. Based on the back-end color offset mapping relationship, the expected finished color value of the image data to be printed is inversely mapped to the pre-compensated target color value in the printing stage; By comparing the pre-compensated target color value with the standard printing color value, ink volume pre-bias data is obtained; Ink color control instructions are generated based on the ink volume pre-bias data, and dynamic printing is performed in accordance with the ink color control instructions; During the dynamic printing process, the updated process image data of the current process is collected in real time, and the ink volume pre-bias data is corrected based on the updated process image data.

2. The machine vision-based adaptive ink color control method for printed materials as described in claim 1, characterized in that, Constructing a post-processing color shift mapping relationship based on the process image data includes: Standard color card images are printed on the same substrate. The visual sampling module collects the first process image data of the standard color card images before the subsequent process is executed, and the second process image data after the execution is executed. By comparing the corresponding color blocks of the first process image data and the second process image data block by block, the color value offset data of each color block under the action of the subsequent process is obtained. Using the color value data of the standard color card image, the substrate attribute data, and the process parameters of the subsequent process as input, and the color value offset data as output, a subsequent color offset mapping relationship is established.

3. The machine vision-based adaptive ink color control method for printed materials as described in claim 1, characterized in that, Based on the aforementioned post-processing color offset mapping relationship, the desired finished color value of the image data to be printed is inversely mapped to the pre-compensated target color value for the printing stage, including: Extract the key color regions from the image data to be printed; Call the subsequent color offset mapping relationship to query the expected color value change of the key color area after processing by the subsequent process parameters; Based on the expected color value change, the desired finished color value of the key color area is reversed and offset to obtain the pre-compensated target color value.

4. The machine vision-based adaptive ink color control method for printed materials as described in claim 3, characterized in that, Extracting key color regions from the image data to be printed includes: The image data to be printed is analyzed in layers to identify the brand logo area, text area and pattern area contained therein; Calculate the visual saliency index for different regions, mark regions with visual saliency indices higher than a preset threshold as key color regions, and mark the remaining regions as non-key color regions; For the key color areas, extract color value data that conforms to the expected finished product color value range; for the non-key color areas, extract the average color value as the expected finished product color value.

5. The machine vision-based adaptive ink color control method for printed materials as described in claim 3, characterized in that, When calculating the pre-compensated target color value, the weight coefficient of the key color region is given higher than that of the non-key color region.

6. The machine vision-based adaptive ink color control method for printed materials as described in claim 4, characterized in that, Based on the ink volume pre-bias data, an ink color control command is generated, including: The target ink layer thickness data is obtained by inversion based on the ink volume pre-bias data; Determine whether the target ink layer thickness data exceeds the current ink absorption limit of the substrate; If the target ink layer thickness data exceeds the ink absorption limit of the current substrate, then while keeping the desired finished color value unchanged, the ink volume pre-bias data corresponding to the non-critical color area is compressed according to a preset ratio, and the ink color control instruction is generated based on the compressed ink volume pre-bias data.

7. The machine vision-based adaptive ink color control method for printed materials as described in claim 4, characterized in that, During dynamic printing execution, real-time acquisition of updated process image data for the current process is performed, and the ink volume pre-bias data is corrected based on the updated process image data, including: The update process image data is collected in real time by a visual sampling module and converted into the current measured color value. Calculate the real-time deviation between the current measured color value and the pre-compensated target color value; When the real-time deviation is outside the allowable error range, the ink volume pre-bias data is incrementally adjusted based on the real-time deviation, and the ink color control command is updated based on the incrementally adjusted ink volume pre-bias data.

8. The machine vision-based adaptive ink color control method for printed materials as described in claim 7, characterized in that, Incremental adjustments are made to the ink volume pre-bias data based on the real-time deviation, including: The distribution position of the real-time deviation within the printing area is identified. When the real-time deviation is concentrated in the key color area, the deviation amplitude is amplified by the first gain coefficient, the ink volume compensation increment is calculated, and the ink volume pre-bias data is updated based on the ink volume compensation increment. When the real-time deviation only occurs in the non-critical color area and does not exceed the preset tolerance range, the current ink volume pre-bias data remains unchanged.

9. The machine vision-based adaptive ink color control method for printed materials as described in claim 2, characterized in that, The method further includes: During the continuous production of the same order, the final finished product image data after the subsequent process has been completed is collected intermittently; The final product image data is compared with the expected product color value to obtain the actual color deviation data; The difference between the actual color deviation data and the color value offset data is calculated to obtain the mapping residual data, and the parameters of the subsequent color deviation mapping relationship are fine-tuned using the mapping residual data.

10. A machine vision-based adaptive ink color control system for printed materials, characterized in that, The system is used to implement the machine vision-based adaptive ink color control method for printed materials according to any one of claims 1-9, the system comprising: The process parameter acquisition module is used to acquire the image data to be printed and the corresponding process parameters for subsequent processes; The mapping relationship construction module is used to collect process image data of the sample printed material before and after the execution of the subsequent process through the visual sampling module, and to construct the subsequent color deviation mapping relationship based on the process image data; The pre-compensation target color value acquisition module is used to reverse map the expected finished color value of the image data to be printed into the pre-compensation target color value in the printing stage according to the subsequent color deviation mapping relationship. The ink volume pre-bias data acquisition module is used to compare the pre-compensated target color value with the standard printing color value to obtain ink volume pre-bias data; The dynamic printing execution module is used to generate ink color control instructions based on the ink volume pre-bias data, and to perform dynamic printing execution with reference to the ink color control instructions. The ink volume pre-bias data correction module is used to collect updated process image data in real time during the dynamic printing process and correct the ink volume pre-bias data based on the updated process image data.

Images