A high speed household towel printing control method and system
By using deep learning technology to accurately locate and correct lighting deviations, the problem of inaccurate color detection and control caused by uneven lighting in high-speed household tissue printing has been solved, thereby improving the uniformity of printed colors and the pass rate.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- KINGDOM AUTO CONTROL TECH LTD CHANGSHA
- Filing Date
- 2025-12-30
- Publication Date
- 2026-06-02
Smart Images
Figure CN121469146B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of paper towel printing technology, and in particular to a high-speed household paper towel printing control method and system. Background Technology
[0002] In the tissue printing industry, color control in high-speed household tissue printing directly affects the uniformity and pass rate of printed colors. With increasing production line speeds, real-time feedback and dynamic adjustment of printed colors have become crucial for ensuring product quality. As a high-frequency consumer good, household printed tissues face ever-increasing demands from consumers for consistent and stable printed colors. However, in high-speed production scenarios, color deviations are instantaneous and cumulative, easily leading to large-scale production of substandard products.
[0003] In the current scenario, uneven lighting distribution in the workshop leads to inaccurate color deviation detection and control. Specifically, during high-speed production, the printing images captured by the camera are affected by uneven lighting intensity distribution, resulting in local color distortion. This causes the detection data such as the RGB values and color gamut parameters of the printed color to deviate from the actual characteristics, which in turn leads to the linkage control module outputting incorrect inkjet parameter adjustment commands, ultimately causing inconsistent printing colors in batches of household paper towels.
[0004] In existing technologies, solutions to this pain point mainly include traditional machine vision color detection methods and inkjet control schemes with fixed parameters. However, these methods have drawbacks: First, traditional machine vision color detection relies on manually set color feature thresholds, which cannot adapt to the dynamic fluctuations in image color caused by changes in lighting, making it difficult to accurately extract the true color features of the printed product. Second, inkjet control schemes with fixed parameters do not have an adaptive adjustment mechanism designed for environmental interference factors such as uneven lighting, and the detection accuracy cannot meet the actual application requirements of color uniformity in household tissue printing under high-speed production.
[0005] Deep learning technology has core advantages such as autonomous learning of complex features, adaptive adaptation to dynamic scenes, and accurate modeling of high-dimensional data. It has shown significant application potential in similar technical scenarios such as color detection and dynamic control under environmental interference in industrial production. However, existing technologies have not deeply integrated it with the detection and control deviations caused by uneven lighting in the color control of high-speed household paper towel printing. Specifically, no suitable deep learning solution has been designed for the characteristic that the optical characteristics of printed colors of household paper towels are easily affected by lighting, which makes it impossible to accurately solve the problem of inaccurate color detection and control caused by uneven lighting.
[0006] Therefore, there is an urgent need for a high-speed, precise color control method and system for household tissue printing based on deep learning, to reduce the interference of uneven lighting on color detection results, improve the accuracy of color deviation recognition and the effect of inkjet parameter adjustment, and meet the requirements of color uniformity and pass rate in household tissue printing. Summary of the Invention
[0007] In view of this, the present invention aims to provide a high-speed household tissue printing control method and system to solve the problem that the traditional method has poor feedback adjustment effect of printing inkjet parameters on household tissues under light interference scenarios.
[0008] A method for controlling printing on high-speed household tissues includes:
[0009] A1: Collect images of household printed tissues, standard color templates for household tissue printing, and production line speed signals, and perform preprocessing to obtain preprocessed printed tissue images, preprocessed tissue printing template images, and preprocessed production line speed data.
[0010] A2: Based on the preprocessed printed tissue paper image and the preprocessed tissue paper printing template image, calculate the tissue paper natural color area mask, the basic illumination color deviation of the natural color area, the multi-scale color features of the printed image, and the multi-scale color features of the template image.
[0011] A3: Based on the mask of the tissue's natural color area, the basic lighting color deviation of the natural color area, the multi-scale color features of the printed image, and the multi-scale color features of the template image, the lighting compensation weight is calculated to obtain the corrected color features.
[0012] A4: Based on the corrected color characteristics and the mask of the tissue's natural color area, complete the printing color restoration and natural color area calibration to obtain a printing color image that eliminates light interference;
[0013] A5: Extract fusion features based on the printed color image after eliminating light interference, the pre-processed tissue printing template image, and the pre-processed production line running speed time series data;
[0014] A6: Based on the fusion characteristics, calculate the initial inkjet parameter adjustment values, and then complete the process constraint regularization to obtain the inkjet parameter adjustment values;
[0015] A7: Adjust the inkjet parameters to control the printhead to perform inkjet operation.
[0016] Furthermore, step A1 also includes:
[0017] A11: Images of household printed tissues are acquired using a high-speed industrial camera. The data type is two-dimensional pixel array data, which includes the grayscale values of the RGB three-color channels of the printed area and the acquisition timestamp. The images of the household printed tissues are then subjected to Gaussian filtering for noise reduction and linear interpolation to unify the resolution, resulting in a pre-processed printed tissue image.
[0018] A12: Collect the standard color template for household paper towel printing. The data type is three-dimensional color matrix data, which includes the RGB standard values of each standard color block and CIE Lab color gamut coordinates. The standard color template for household paper towel printing is subjected to linear normalization and three times standard difference constant removal to obtain the cleaned paper towel printing template image.
[0019] A13: The production line running speed signal is acquired by an incremental photoelectric encoder. The data type is a time-series electrical signal, which includes instantaneous speed value, sampling time interval, and equipment shaft code value. The production line running speed signal is subjected to mean filtering and linear fitting to fill in missing values, so as to obtain the pre-processed production line running speed time-series data.
[0020] A14: Based on the pre-processed printed tissue paper image, the cleaned tissue paper printing template data, and the pre-processed production line running speed time series data, the timestamp matching method is used to align the multi-source data to obtain the pre-processed printed tissue paper image, the pre-processed tissue paper printing template image, and the pre-processed production line running speed data.
[0021] Furthermore, step A2 also includes:
[0022] Based on the preprocessed printed tissue paper image and the preprocessed tissue paper printing template image, the tissue paper natural color region mask is obtained through semantic segmentation and illumination mask filtering function, and then the basic illumination color deviation of the natural color region is obtained through natural color region color statistics calculation.
[0023] Based on the tissue paper's natural color region mask, invalid regions are masked in the preprocessed printed tissue paper image and the preprocessed tissue paper printing template image. Then, a multi-scale feature extraction network is used to extract features from the two masked images to obtain multi-scale color features of the printed image and multi-scale color features of the template image.
[0024] It should be further explained that in the high-speed printing scenario of household paper towels, the natural color of the paper towels is relatively light and is easily affected by changes in ambient light. At the same time, high-speed production requires image processing to be both accurate and efficient. Light interference can cause the color comparison between the printed paper towel image and the template image to deviate, directly affecting the accuracy of subsequent color feature extraction, and thus leading to inaccurate inkjet parameter adjustment.
[0025] In step A2 of this invention, the first step is to calculate a mask for the natural color region of the tissue paper based on the preprocessed printed tissue paper image and the preprocessed tissue paper printing template image. This is achieved by combining a semantic segmentation convolutional neural network with a light mask filtering function. The semantic segmentation convolutional neural network is responsible for initially locating the natural color region of the tissue paper, while the light mask filtering function optimizes the location result for changes in lighting conditions. This solves the problem of inaccurate region location caused by lighting interference, which cannot be addressed by a single segmentation method. This dual filtering ensures accurate separation of the natural color region and the printed region, laying the foundation for subsequent elimination of interference from the natural color region and focusing on the color analysis of the printed area. Next, the actual average color value within the masked region is calculated based on the tissue paper's natural color region mask and compared with the factory-calibrated baseline value for the natural color of the tissue paper to obtain the basic lighting color deviation of the natural color region. This quantifies the impact of lighting on the natural color of the tissue paper, as the lighting deviation of the natural color is directly superimposed on the printed color. Only by quantifying this deviation can the underlying lighting color deviation be effectively addressed. Only by addressing these issues can we effectively eliminate lighting interference in subsequent feature extraction. Finally, multi-scale convolution is used to process the printed tissue image and the template image with the natural color area shielded, respectively. Batch normalization and pre-encoding functions are combined to process the basic lighting color deviation of the original image and the natural color area. The two results are then fused to obtain multi-scale color features of the printed image and the template image. Multi-scale convolution is used to capture the color features of printed patterns of different sizes to ensure the comprehensiveness of feature extraction. Batch normalization is used to stabilize the feature distribution and avoid data fluctuations affecting feature quality. The pre-encoding function incorporates the basic lighting color deviation of the natural color area into the feature processing process to compensate for lighting interference. The cooperation of each step ensures that the extracted color features can accurately reflect the color information of the print itself and have strong resistance to lighting interference, effectively solving the feature distortion problem caused by lighting interference in high-speed printing scenarios.
[0026] Existing technologies for extracting color features from printed tissue paper typically perform global color feature extraction on the entire printed tissue paper image and template image without specifically separating and processing the natural color region. They often use only a single semantic segmentation method to locate the natural color region without incorporating illumination filtering optimization. Furthermore, they largely ignore the impact of illumination deviations in the natural color region on the printed features, employing only simple global illumination compensation methods, which cannot accurately match the illumination interference characteristics of the tissue paper printing scene. Compared to existing technologies, the advantages of this invention are: First, through the synergistic effect of semantic segmentation convolutional neural networks and illumination mask filtering functions, accurate localization of the natural color region of the tissue paper is achieved, solving the problem of significant illumination interference in region localization in existing technologies and providing a reliable foundation for subsequent accurate processing. Second, the basic illumination color deviation of the natural color region is specifically quantified and integrated into the feature processing of the pre-encoding function, achieving targeted illumination interference compensation rather than the global compensation of existing technologies, significantly improving the anti-interference ability and accuracy of color features. Simultaneously, the extracted multi-scale color features of the printed image and template image can directly provide high-quality data support for subsequent color deviation calculations, further improving the reliability of the entire inkjet parameter adjustment scheme.
[0027] Furthermore, step A3 also includes:
[0028] Based on the multi-scale color features of the printed image and the template image, the color feature deviation between the printed image and the template image is calculated. Then, combined with the basic illumination color deviation of the original color region, the illumination compensation weight is calculated through a deep learning network, and the illumination color deviation feature and the actual printed color difference feature are calculated in sequence. Based on the multi-scale color features of the printed image, the illumination color deviation feature, and the actual printed color difference feature, the feature is calibrated through a feature alignment mechanism to obtain the corrected color features.
[0029] It should be further explained that in the high-speed printing scenario of household paper towels, the lighting environment is prone to fluctuation and high-speed production can easily cause slight shifts in the printing position. The original color features will be superimposed with lighting interference deviations and feature misalignment deviations. If they are directly used for subsequent inkjet parameter adjustments, it will lead to distortion of the parameter adjustment basis, and ultimately cause quality problems such as printing color deviation and pattern misalignment.
[0030] This invention extracts corrected color features: First, color feature deviations are calculated based on multi-scale color features of the printed image and the template image, providing basic deviation data for subsequent correction and clarifying the core differences between the original features. Next, combining the basic illumination color deviation of the original color region with the aforementioned color feature deviations, illumination compensation weights are calculated using global average pooling, multilayer perceptron, and attention mechanisms. These weights are then used to separate illumination color deviation features from the actual printed color difference features. This step, through precise adaptation of the illumination compensation weights, specifically filters out the interference of illumination fluctuations on color features, preventing illumination interference from being misjudged as color deviations in the printed material itself, ensuring the true color difference. The actual color difference features of the printed image can accurately reflect the actual color problem of the printed image. Finally, after removing the illumination color deviation features from the multi-scale color features of the printed image, it is fused with the multi-scale color features of the template image processed by the feature alignment function and the actual color difference features of the printed image to obtain the corrected color features. The feature alignment function can correct the feature misalignment problem caused by the slight shift in the printing position during high-speed production through the synergistic effect of cosine similarity and softmax function, so that the fused features are highly matched in spatial dimension. In the whole process, the corrected color features not only eliminate the influence of illumination interference, but also solve the feature misalignment problem, while completely preserving the color details and difference information of the printed image itself.
[0031] Existing technologies for color feature correction mostly employ a single illumination compensation method, filtering illumination interference solely through global parameter adjustments. This fails to consider the specificity of illumination deviations in the natural color areas of tissue paper printing scenarios, nor does it specifically address feature misalignment issues caused by high-speed production. While some technologies consider both illumination and misalignment correction, their step-by-step, independent processing leads to residual interference or fusion distortion in the corrected features. Compared to existing technologies, the corrected color features offer three advantages: First, by combining the illumination compensation weight with the basic illumination color deviation of the natural color area, targeted elimination of illumination interference is achieved, resulting in stronger anti-interference capabilities for the corrected features, surpassing the global compensation effect of existing technologies. Second, the integration of a feature alignment function specifically addresses feature misalignment issues in high-speed production, ensuring spatial matching of the corrected features and preventing distortion in subsequent deviation calculations due to feature misalignment. Third, the corrected color features directly reflect the true color state of the print, providing crucial assurance for the accuracy of subsequent inkjet parameter adjustments and significantly reducing the print defect rate caused by feature distortion.
[0032] Furthermore, step A4 also includes:
[0033] First, the corrected color features are input into a 1×1 convolutional layer. The feature dimensions of the corrected color features are adjusted by the 1×1 convolutional layer. Then, the Sigmoid function is input to obtain the color adjustment coefficients.
[0034] Subsequently, the color adjustment coefficients are multiplied pixel by pixel with the multi-scale color features of the printed image to obtain a preliminary corrected image;
[0035] Finally, a pixel-level replacement operation is used to perform a full pixel traversal of the preliminary correction image based on the tissue paper color area mask. Specifically, for areas in the preliminary correction image where the tissue paper color area mask is marked as 0, their color in the preliminary correction image is retained; for areas in the preliminary correction image where the tissue paper color area mask is marked as 1, their pixel values are replaced with the factory-calibrated tissue paper color reference value, resulting in a printed color image that eliminates light interference.
[0036] Furthermore, step A5 also includes:
[0037] A51: Based on the printed color image after eliminating light interference, the pre-processed tissue paper printing template image, and the pre-processed production line running speed time series data, extract the spatial features of color deviation and the time series features of production speed.
[0038] A52: Based on the spatial characteristics of color deviation and the temporal characteristics of production speed, the spatiotemporal heterogeneous features are adaptively fused to obtain the fused features.
[0039] It should be further explained that the production line operating speed fluctuates dynamically, and speed changes directly affect the inkjet rhythm of the printhead and the accuracy of the printing position; the spatial characteristics of color deviation and the temporal characteristics of production speed are heterogeneous data, and their feature distribution and dimensional attributes differ significantly. Direct fusion can easily lead to imbalance or distortion of feature information. It is necessary to ensure that the fused features can synchronously reflect the spatial color state and temporal speed changes, so as to provide a comprehensive basis for subsequent parameter adjustments.
[0040] Firstly, in step A51, for the printed color image after eliminating illumination interference and the pre-processed tissue paper printing template image, this invention uses a combination of convolutional layers with a kernel size of 3×3 and convolutional layers with a kernel size of 1×1. First, 1×1 convolution is used to achieve feature dimensionality reduction and channel fusion, and then 3×3 convolution is used to extract fine basic features of color deviation. At the same time, a spatial attention function is introduced to enhance the features after convolution. The spatial attention function focuses on the significant areas of printed color deviation and suppresses irrelevant interference information such as tissue paper texture. The resulting spatial features of color deviation can accurately reflect color differences in spatial dimensions, solving the problem that a single convolutional layer cannot distinguish between effective deviations and invalid interference. For the pre-processed production line operating speed time series data, a bidirectional gated loop unit is used to extract the production speed time series features, capturing the historical trend of speed changes and predicting subsequent changes. Then, time normalization is combined to eliminate the dimensional differences and fluctuation noise of the speed data, ensuring that the extracted production speed time series features can stably represent the speed change pattern in the time series dimension.
[0041] Subsequently, in step A52, speed gating weights are calculated based on the temporal features of production speed using a multilayer perceptron and a sigmoid function. This allows for dynamic adaptation to speed fluctuations—when speed changes drastically, the proportion of fusion of the temporal features of production speed is automatically increased; when the speed is stable, the focus is on preserving the details of the color deviation spatial features. Then, the adaptive fusion of the color deviation spatial features and the temporal features of production speed is achieved through Hadamard product and element-wise addition operations to obtain the fused features.
[0042] Throughout the entire chain, step A51 provides high-quality heterogeneous basic features for fusion, while step A52 solves the problem of imbalance in heterogeneous feature fusion through a dynamic gating mechanism. Together, they ensure that the fused features can not only fully preserve the spatial details of color deviation, but also respond to the temporal changes in production speed, providing both spatial color basis and temporal speed basis for subsequent inkjet parameter adjustments.
[0043] Existing technologies for feature extraction and fusion in this scenario often employ simple splicing or fixed-weight fusion methods, failing to consider the heterogeneity of spatial features of color deviation and temporal features of production speed. They cannot dynamically adjust fusion weights based on speed fluctuations, resulting in fused features that cannot synchronously adapt to spatial and temporal changes. In contrast, this invention innovatively designs speed-gated weights to achieve adaptive fusion of heterogeneous features, breaking the limitations of fixed weights or simple splicing in existing technologies. This allows fused features to dynamically adapt to speed fluctuations in high-speed production while maintaining the integrity of spatial color and temporal speed information. The resulting fused features can directly provide comprehensive and accurate core data for subsequent inkjet parameter adjustments, significantly improving the reliability and adaptability of the entire parameter adjustment scheme.
[0044] Furthermore, step A6 also includes:
[0045] Based on the fusion features, they are enhanced using a feature residual enhancement mechanism to obtain the fusion features with residual enhancement. Based on the fusion features with residual enhancement, the initial inkjet parameter adjustment value is calculated. Combined with the printing process constraint parameters, the initial inkjet parameter adjustment value is subjected to process constraint regularization. After speed adaptation calibration, the inkjet parameter adjustment value is obtained.
[0046] It is worth noting that in the high-speed printing scenario for household tissues, there is a dual interference from uneven lighting and high-speed linkage. Uneven lighting intensity distribution can cause local color distortion in the printed images captured by the camera, causing color-related detection data to deviate from the actual characteristics. At the same time, high-speed production requires parameter adjustments to be real-time and dynamically adaptable. If the parameter adjustment command is incorrect due to feature distortion, it will directly lead to inconsistent colors in batches of household tissue printing. The core difficulty lies in the fact that traditional methods often calculate parameters based only on the original color features, without considering the interference of lighting distortion on the features, resulting in distorted parameter adjustment basis. Moreover, the parameter calculation does not take into account process constraints and dynamic speed changes, which can easily lead to problems such as parameters being out of range or poor adaptability, ultimately causing batch quality defects.
[0047] To address the aforementioned issues, residual enhancement is first applied to the fused features. Residual connections preserve the core information of the original fused features, while multilayer perceptrons, layer normalization, and attention mechanisms further enhance the feature representation capabilities, particularly focusing on key color deviation areas affected by lighting. Feature enhancement filters out interference from lighting distortion, ensuring subsequent parameter calculations are based on true and accurate color features. Next, initial inkjet parameter adjustment values are calculated based on the enhanced fused features. Global average pooling compresses the feature dimension to improve computational efficiency, and a fully connected layer maps features to parameters. Speed-gated weights are introduced to dynamically respond to changes in production line speed, resolving parameter adaptability issues caused by speed fluctuations in high-speed scenarios. Subsequently, through… Process constraint regularization constrains the initial parameters, combining them with process parameters such as the upper limit of ink volume in the tissue printing area and the printhead movement step size. Simultaneously, the constraint weights are further optimized using the current production line speed to prevent parameters from exceeding physical limits. Finally, speed adaptation combines the constrained parameters with the production line speed time-series data. Time normalization ensures the stability of speed information, and basic calibration is performed using the tissue printing baseline inkjet parameters. Ultimately, inkjet parameter adjustment values adapted for high-speed production are obtained. This approach solves feature errors caused by illumination distortion, ensures the practical feasibility of parameters through process constraints, and achieves dynamic real-time adjustment through speed adaptation. The synergy of these three aspects ensures accurate and reliable inkjet parameter adjustment values, avoiding batch printing color inconsistencies caused by parameter errors.
[0048] Existing technologies for adjusting inkjet parameters in this scenario fail to consider dynamic speed changes, employing fixed parameters or simple linear adjustments. This fails to adapt to speed fluctuations in high-speed production, easily leading to batch color inconsistencies. In contrast, this invention combines residual enhancement with an attention mechanism to enhance feature extraction's resistance to lighting interference, ensuring parameter calculations are based on true color features and significantly reducing parameter error rates caused by feature distortion. Secondly, it innovatively introduces a process constraint regularization mechanism, integrating physical constraints such as ink volume limits and printhead step size into parameter calculations. Simultaneously, it dynamically adjusts constraint weights based on the current speed, ensuring parameters meet both theoretical requirements and practical feasibility, avoiding printing quality issues caused by improper parameters. Thirdly, through speed adaptation and benchmark parameter calibration, it achieves dynamic real-time parameter adjustment, adapting to the demands of high-speed production and further improving the accuracy and stability of inkjet parameter adjustment. Ultimately, it achieves precise inkjet parameter adjustment based on real-time visual feedback, resolving batch printing color inconsistencies caused by uneven lighting in high-speed production.
[0049] Furthermore, step A7 also includes:
[0050] A71: The inkjet parameter adjustment value is converted into an inkjet parameter control signal adapted to the mechanical actuator through a general digital-to-analog converter;
[0051] A72: The inkjet parameter control signal is received through the printhead drive unit, and the opening and closing degree of the solenoid valve of each inkjet channel is adjusted in real time by the constant current drive circuit to control the ink output. At the same time, the ink output is realized in real time through the ink output detection element.
[0052] A73: The printhead displacement adjustment unit adjusts the lateral movement step size and longitudinal height of the printhead according to the inkjet parameter control signal, using a stepper motor drive combined with pulse width modulation technology. At the same time, the displacement positioning is calibrated by a grating ruler to ensure that the relative position of the printhead and the tissue printing area matches the inkjet parameter adjustment value.
[0053] This invention also discloses a high-speed household tissue paper printing control system, comprising:
[0054] Image acquisition module: Acquires images of household printed tissues, standard color templates for household tissue printing, and production line speed signals, and performs preprocessing to obtain preprocessed printed tissue images, preprocessed tissue printing template images, and preprocessed production line speed data;
[0055] Image feature extraction module: Based on the preprocessed printed tissue image and the preprocessed tissue printing template image, calculate the tissue's natural color area mask, the basic illumination color deviation of the natural color area, the multi-scale color features of the printed image, and the multi-scale color features of the template image;
[0056] Illumination feature compensation module: Based on the mask of the natural color area of the tissue paper, the basic illumination color deviation of the natural color area, the multi-scale color features of the printed image, and the multi-scale color features of the template image, the illumination compensation weight is calculated to obtain the corrected color features;
[0057] Light interference elimination module: Based on the corrected color features and the mask of the original color area of the tissue paper, the module completes the restoration of printed color and calibration of the original color area to obtain a printed color image that eliminates light interference.
[0058] Feature fusion module: Extracts fusion features based on the printed color image after eliminating light interference, the pre-processed tissue printing template image, and the pre-processed production line running speed time series data;
[0059] Inkjet parameter adjustment module: Based on the fusion characteristics, calculate the initial inkjet parameter adjustment value, and then complete the process constraint regularization to obtain the inkjet parameter adjustment value;
[0060] Inkjet control module: Controls the printhead to perform inkjet operation based on the inkjet parameter adjustment values.
[0061] Compared with the prior art, the beneficial effects of the present invention are as follows:
[0062] (1) This invention solves the problem of poor real-time feedback control of printing color caused by uneven lighting in high-speed household tissue printing. Its core idea is to start from data acquisition and preprocessing, through color feature correction, light interference elimination, and multi-dimensional feature fusion, and finally achieve precise inkjet parameter adjustment and printhead control. Compared with the traditional detection and fixed parameter control scheme that relies on fixed thresholds, this invention can adaptively cope with workshop lighting fluctuations, avoid detection deviation caused by image color distortion, and restore the true color characteristics of the printing. By dynamically adapting to changes in production line speed and process constraints, the output inkjet parameters are more in line with actual production needs, effectively eliminating the problems of inconsistent batch colors and inconsistent colors with template colors, realizing high-quality feedback control in the tissue printing production process, greatly improving the uniformity of printing color and product qualification rate, and providing reliable support for quality assurance of household tissue printing.
[0063] (2) In response to the problem that the natural color of household paper towels is easily affected by light during high-speed printing, resulting in color comparison deviation, inaccurate feature extraction, and inaccurate inkjet parameters, this invention first combines semantic segmentation and light filtering methods to accurately locate and optimize the natural color area of the paper towel, separating the natural color from the printed area; then calculates the deviation between the actual color of the natural color and the reference value to quantify the influence of light; finally, multi-scale technology is used to extract the printed color features, combined with batch normalization and pre-coding to compensate for light deviation and stabilize features; in terms of technical effect, dual positioning avoids interference from the natural color, deviation quantification makes light compensation more accurate, and subsequent processing makes the extracted features more resistant to interference and reflects the true color of the printed color, completely solving the problem of feature distortion caused by light, providing a reliable basis for subsequent inkjet parameter adjustment, and greatly improving the color consistency and product qualification rate of high-speed printing.
[0064] (3) In the case of light environment fluctuations and slight offset of printing position caused by high-speed production in the high-speed printing of household paper towels, the original color features are superimposed with light interference deviation and feature misalignment deviation. Directly using inkjet parameter adjustment is easy to cause distortion, resulting in quality problems such as color deviation and pattern misalignment. This invention calculates the multi-scale color feature deviation between the printing and the template image, and calculates the light compensation weight by combining the light deviation of the original color area to separate the light and the true color difference features. The feature alignment function is used to correct the feature misalignment, effectively eliminates light interference, corrects feature misalignment, and makes the color features accurately reflect the true state of printing, improves the accuracy of inkjet parameter adjustment, and reduces the defect rate.
[0065] (4) To address the issue of printing effects caused by dynamic fluctuations in production line speed, and the problem that color deviation spatial features and speed temporal features are prone to imbalance or distortion when directly fused as heterogeneous data, this invention uses a combination of 3×3 and 1×1 convolutional layers to process color images and enhance spatial attention to obtain features that reflect spatial color differences. Speed temporal features are extracted through a bidirectional gated loop unit and noise is eliminated by time normalization. Then, a multilayer perceptron and Sigmoid are used to calculate dynamic gate weights. Combined with Hadamard product and element-wise addition, heterogeneous features are adaptively fused, irrelevant interference is eliminated, speed patterns are stably represented, and speed fluctuations are dynamically adapted. This allows the fused features to fully retain spatial details and respond to temporal changes, providing a comprehensive basis for inkjet parameter adjustment in terms of spatial color and temporal speed. This breaks through the limitations of existing technologies and improves the reliability and adaptability of parameter adjustment schemes.
[0066] (5) In the inkjet parameter control stage, in response to the dual interference of uneven illumination and high-speed linkage in high-speed printing of household paper towels, this invention first performs residual enhancement processing on the fusion features, and combines multilayer perceptron, layer normalization and attention mechanism to focus on the key color deviation areas affected by illumination, and filter the interference to ensure the authenticity and accuracy of the features; then, based on the enhanced features, the initial inkjet parameter adjustment value is calculated, and speed gating weight is introduced to dynamically respond to speed changes. At the same time, process constraint regularization is combined with physical constraints such as ink volume limit, printhead step size and speed optimization constraint weight; finally, the parameters are dynamically adjusted in real time through speed adaptation and benchmark parameter calibration, reducing the parameter error rate caused by feature distortion, ensuring that the parameters meet the theoretical requirements and actual executability, adapting to the high-speed production linkage requirements, improving the accuracy and stability of inkjet parameter adjustment, and finally solving the problem of inconsistent batch printing colors caused by uneven illumination in high-speed production. Attached Figure Description
[0067] Figure 1 A flowchart illustrating a high-speed household tissue printing control method provided by the present invention;
[0068] Figure 2 This is an example drawing of a printed paper towel provided by the present invention;
[0069] Figure 3 This invention provides a mask for the natural color area of the tissue paper. Detailed Implementation
[0070] The present invention will be further described below with reference to the accompanying drawings, but this is not intended to limit the present invention in any way. Any modifications or substitutions made based on the teachings of the present invention shall fall within the protection scope of the present invention.
[0071] Example 1: A method for controlling high-speed printing on household tissues, such as... Figure 1 As shown, it includes the following steps:
[0072] A1: Acquire images of household printed tissues, standard color templates for household tissue printing, and production line speed signals, and perform preprocessing to obtain preprocessed printed tissue images, preprocessed tissue printing template images, and preprocessed production line speed data, including:
[0073] A11: Images of household printed tissues are acquired using a high-speed industrial camera. The data type is two-dimensional pixel array data, which includes the grayscale values of the RGB three-color channels of the printed area and the acquisition timestamp. The images of the household printed tissues are then subjected to Gaussian filtering for noise reduction and linear interpolation to unify the resolution, resulting in a pre-processed printed tissue image.
[0074] A12: Collect the standard color template for household paper towel printing. The data type is three-dimensional color matrix data, which includes the RGB standard values of each standard color block and CIE Lab color gamut coordinates. The standard color template for household paper towel printing is subjected to linear normalization and three times standard difference constant removal to obtain the cleaned paper towel printing template image.
[0075] A13: The production line running speed signal is acquired by an incremental photoelectric encoder. The data type is a time-series electrical signal, which includes instantaneous speed value, sampling time interval, and equipment shaft code value. The production line running speed signal is subjected to mean filtering and linear fitting to fill in missing values, so as to obtain the pre-processed production line running speed time-series data.
[0076] A14: Based on the pre-processed printed tissue paper image, the cleaned tissue paper printing template data, and the pre-processed production line running speed time series data, the timestamp matching method is used to align the multi-source data to obtain the pre-processed printed tissue paper image, the pre-processed tissue paper printing template image, and the pre-processed production line running speed data.
[0077] A2: Based on the preprocessed printed tissue image and the preprocessed tissue printing template image, calculate the tissue's natural color region mask, the basic illumination color deviation of the natural color region, the multi-scale color features of the printed image, and the multi-scale color features of the template image. The calculation method is as follows:
[0078] ;
[0079] ;
[0080] ;
[0081] ;
[0082] ;
[0083] in, Mask the area of the tissue paper in its natural color, such as Figure 3 As shown, For semantic segmentation convolutional neural networks, The image of the pre-processed printed tissue paper, such as Figure 2 As shown, For Hadama accumulation, This is a function for filtering illumination masks. This represents the actual average color value of the tissue paper within the masked area. This represents the total number of pixels within the mask area. These are the pixels with x-coordinate and y-coordinate in the preprocessed printed tissue image. The pixel with x-coordinate and y-coordinate. This represents the base lighting color deviation in the original color area. The standard value for the natural color of the paper towels as specified by the manufacturer. For multi-scale color features of printed images, For multi-scale convolution, To sum element by element, For pre-encoding function, For batch normalization, For the multi-scale color features of the template image, This is the pre-processed tissue paper printing template image.
[0084] The illumination mask filtering function is calculated as follows:
[0085]
[0086] in, For the Sigmoid function, It is a one-dimensional convolutional layer. This is the grayscale image corresponding to the preprocessed printed tissue image. The grayscale threshold for the natural color of the tissue paper. It is an exponential function. Calculate the Euclidean distance. The RGB base value for the natural color of the tissue paper;
[0087] The calculation method of the pre-encoding function is as follows:
[0088] ;
[0089] in, This is the input to the pre-encoding function. For GeLU functions, It is a multilayer perceptron.
[0090] A3: Based on the mask of the tissue's natural color area, the basic lighting color deviation of the natural color area, the multi-scale color features of the printed image, and the multi-scale color features of the template image, the lighting compensation weight is calculated to obtain the corrected color features. The calculation method is as follows:
[0091] ;
[0092] ;
[0093] ;
[0094] ;
[0095] ;
[0096] in, The color feature deviation between the printed image and the template image, For illumination compensation weights, For global average pooling, For the calculation of attention mechanism, This refers to the characteristics of color deviation under illumination. This reflects the true color difference characteristics of printed materials. The corrected color characteristics, The feature alignment function is calculated as follows:
[0097] ;
[0098] in, For the Softmax function, This is for calculating cosine similarity.
[0099] A4: Based on the corrected color characteristics and the tissue paper's natural color area mask, complete the printing color restoration and natural color area calibration to obtain a printing color image that eliminates lighting interference, including:
[0100] First, the corrected color features are input into a 1×1 convolutional layer. The feature dimensions of the corrected color features are adjusted by the 1×1 convolutional layer. Then, the Sigmoid function is input to obtain the color adjustment coefficients.
[0101] Subsequently, the color adjustment coefficients are multiplied pixel by pixel with the multi-scale color features of the printed image to obtain a preliminary corrected image;
[0102] Finally, a pixel-level replacement operation is used to perform a full pixel traversal of the preliminary correction image based on the tissue paper color area mask. Specifically, for areas in the preliminary correction image where the tissue paper color area mask is marked as 0, their color in the preliminary correction image is retained; for areas in the preliminary correction image where the tissue paper color area mask is marked as 1, their pixel values are replaced with the factory-calibrated tissue paper color reference value, resulting in a printed color image that eliminates light interference.
[0103] A5: Based on the printed color image after eliminating light interference, the pre-processed tissue paper printing template image, and the pre-processed production line running speed time series data, extract fusion features, including:
[0104] A51: Based on the printed color image after eliminating light interference, the pre-processed tissue printing template image, and the pre-processed production line operating speed time-series data, extract the spatial features of color deviation and the time-series features of production speed. The calculation method is as follows:
[0105] ;
[0106] ;
[0107] in, For color deviation spatial features, This is a convolutional layer with a kernel size of 3×3. This is a convolutional layer with a kernel size of 1×1. To eliminate light interference in the printing color image, Let be the spatial attention function. For production speed timing characteristics, It is a bidirectional gated loop unit. This is the preprocessed production line operating speed time series data. For time normalization operation;
[0108] A52: Based on the spatial characteristics of color deviation and the temporal characteristics of production speed, adaptive fusion of spatiotemporal heterogeneous features is performed to obtain fused features. The calculation method is as follows:
[0109] ;
[0110] ;
[0111] in, For speed gating weights, This is a feature of fusion.
[0112] Specifically, for high-speed printing scenarios of household paper towels with high speed fluctuations and complex textures, this invention also provides an enhanced heterogeneous feature extraction calculation method to replace the calculation method of production speed temporal features in step A51, specifically including:
[0113] ;
[0114] in, It is a bidirectional gated recurrent unit with self-attention mechanism. This is a gated loop unit.
[0115] A6: Based on the fusion characteristics, calculate the initial inkjet parameter adjustment values, then perform process constraint regularization to obtain the inkjet parameter adjustment values. The calculation method is as follows:
[0116] ;
[0117] ;
[0118] ;
[0119] ;
[0120] in, The fusion features are those resulting from residual enhancement. For layer normalization, These are the initial inkjet parameter adjustment values. It is a fully connected layer. These are the initial inkjet parameter adjustment values after constraints. These are the initial inkjet parameter adjustment values. This is the upper limit of ink volume in the printed area of the tissue paper. The maximum absolute value of the initial inkjet parameter adjustment value. The nozzle movement step size, To round to the nearest integer, For splicing operations, This refers to the current production line operating speed in the preprocessed production line operating speed time series data. These are the inkjet parameter adjustment values. These are the baseline inkjet parameters for paper towel printing.
[0121] In this embodiment, the parameters of the neural network modules involved in step A6, such as layer normalization, multilayer perceptron, attention mechanism, and fully connected layer, are set as follows to adapt to the parameter calculation requirements of the high-speed printing scenario for household paper towels:
[0122] Layer normalization module: The input dimension is consistent with the feature dimension of the fused features, set to 128 dimensions;
[0123] Attention mechanism module: A self-attention structure is adopted, with 8 attention heads and 16 feature dimensions for keys, queries, and values. Dropout is used during attention calculation with a dropout probability of 0.1 to suppress overfitting. Finally, attention-enhanced features with the same dimensions as the input are obtained by concatenating the outputs of each attention head and mapping them through a 1×1 convolution.
[0124] The multilayer perceptron module in residual augmentation has an input dimension of 128 and adopts a structure of two fully connected layers plus an activation function. The first fully connected layer has an output dimension of 256 and uses GeLU as the activation function. The second fully connected layer has an output dimension of 128 and no output activation function. Both fully connected layers use L2 regularization with a weight decay coefficient of 1e-4.
[0125] The multilayer perceptron module in process constraint regularization has a 34-dimensional input dimension and adopts a two-layer fully connected structure with an activation function. The first layer has a 64-dimensional output dimension and the second layer has a 1-dimensional output dimension. The activation function is GeLU, and L2 regularization of 1e-4 is also set.
[0126] Fully connected layer module: The input dimension is 128-dimensional, and the output dimension is set to 4-dimensional (corresponding to the 4 dimensions of inkjet parameter adjustment: main ink volume, auxiliary ink volume, horizontal step size correction, and vertical height correction); it adopts a two-layer fully connected layer structure, with the middle hidden layer having a dimension of 64-dimensional and the activation function being GeLU, and the output layer having no activation function.
[0127] A7: Control the printhead to perform inkjet operation based on the inkjet parameter adjustment values, including:
[0128] A71: The inkjet parameter adjustment value is converted into an inkjet parameter control signal adapted to the mechanical actuator through a general digital-to-analog converter;
[0129] A72: The inkjet parameter control signal is received through the printhead drive unit, and the opening and closing degree of the solenoid valve of each inkjet channel is adjusted in real time by the constant current drive circuit to control the ink output. At the same time, the ink output is realized in real time through the ink output detection element.
[0130] A73: The printhead displacement adjustment unit adjusts the lateral movement step size and longitudinal height of the printhead according to the inkjet parameter control signal, using a stepper motor drive combined with pulse width modulation technology. At the same time, the displacement positioning is calibrated by a grating ruler to ensure that the relative position of the printhead and the tissue printing area matches the inkjet parameter adjustment value.
[0131] The specific training methods for the neural network modules involved in this invention (including semantic segmentation convolutional neural networks, multi-scale convolutional networks, bidirectional gated recurrent units, multilayer perceptrons, and attention mechanism-related modules, etc.) are as follows:
[0132] The training data consists of preprocessed printed tissue images (A1 step), tissue printing template images, production line speed time-series data, and corresponding annotation information. These are divided into training, validation, and test sets according to a preset ratio for model training and performance verification. A combined loss strategy is used: cross-entropy loss is used for the natural color region segmentation task, and mean squared error loss is used for regression tasks involving color feature bias and illumination compensation weights. The total loss function is obtained through weighted summation to simultaneously optimize the output accuracy of each module. The Adam optimizer is selected, with an initial learning rate set to 1e-4. A cosine annealing strategy is used to dynamically adjust the learning rate, while a weight decay mechanism is introduced to suppress overfitting. The training process employs a mini-batch stochastic gradient descent strategy, inputting batch data into each neural network module for forward propagation. After obtaining the prediction results, the total loss value is calculated, and the network parameters of each module are updated using a backpropagation algorithm. An early stopping mechanism is introduced during training, using the validation set loss as a monitoring indicator. Training is terminated when the validation set loss does not significantly decrease after a preset number of consecutive rounds to avoid model overfitting.
[0133] Example 2: The present invention also discloses a high-speed household tissue paper printing control system, comprising:
[0134] Image acquisition module: Acquires images of household printed tissues, standard color templates for household tissue printing, and production line speed signals, and performs preprocessing to obtain preprocessed printed tissue images, preprocessed tissue printing template images, and preprocessed production line speed data;
[0135] Image feature extraction module: Based on the preprocessed printed tissue image and the preprocessed tissue printing template image, calculate the tissue's natural color area mask, the basic illumination color deviation of the natural color area, the multi-scale color features of the printed image, and the multi-scale color features of the template image;
[0136] Illumination feature compensation module: Based on the mask of the natural color area of the tissue paper, the basic illumination color deviation of the natural color area, the multi-scale color features of the printed image, and the multi-scale color features of the template image, the illumination compensation weight is calculated to obtain the corrected color features;
[0137] Light interference elimination module: Based on the corrected color features and the mask of the original color area of the tissue paper, the module completes the restoration of printed color and calibration of the original color area to obtain a printed color image that eliminates light interference.
[0138] Feature fusion module: Extracts fusion features based on the printed color image after eliminating light interference, the pre-processed tissue printing template image, and the pre-processed production line running speed time series data;
[0139] Inkjet parameter adjustment module: Based on the fusion characteristics, calculate the initial inkjet parameter adjustment value, and then complete the process constraint regularization to obtain the inkjet parameter adjustment value;
[0140] Inkjet control module: Controls the printhead to perform inkjet operation based on the inkjet parameter adjustment values.
[0141] It should be noted that the sequence numbers of the above embodiments of the present invention are merely for descriptive purposes and do not represent the superiority or inferiority of the embodiments. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, apparatus, article, or method. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.
[0142] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.
[0143] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.
Claims
1. A method for controlling high-speed printing on household tissues, characterized in that, Includes the following steps: A1: Collect images of household printed tissues, standard color templates for household tissue printing, and production line speed signals, and perform preprocessing to obtain preprocessed printed tissue images, preprocessed tissue printing template images, and preprocessed production line speed data. A2: Based on the preprocessed printed tissue paper image and the preprocessed tissue paper printing template image, calculate the tissue paper natural color area mask, the basic illumination color deviation of the natural color area, the multi-scale color features of the printed image, and the multi-scale color features of the template image. Step A2 includes: Based on the preprocessed printed tissue paper image and the preprocessed tissue paper printing template image, the tissue paper natural color region mask is obtained through semantic segmentation and illumination mask filtering function, and then the basic illumination color deviation of the natural color region is obtained through natural color region color statistics calculation. Based on the tissue paper's natural color region mask, invalid regions are masked in the preprocessed printed tissue paper image and the preprocessed tissue paper printing template image. Then, feature extraction is performed through a multi-scale feature extraction network to obtain multi-scale color features of the printed image and multi-scale color features of the template image. A3: Based on the mask of the tissue's natural color area, the basic lighting color deviation of the natural color area, the multi-scale color features of the printed image, and the multi-scale color features of the template image, the lighting compensation weight is calculated to obtain the corrected color features. Step A3 includes: Based on the multi-scale color features of the printed image and the template image, the color feature deviation between the printed image and the template image is calculated. Then, combined with the basic illumination color deviation of the original color region, the illumination compensation weight is calculated through a deep learning network, and the illumination color deviation feature and the actual printed color difference feature are calculated in sequence. Based on the multi-scale color features of the printed image, the illumination color deviation feature, and the actual printed color difference feature, the feature alignment mechanism is used to perform feature calibration to obtain the corrected color features. A4: Based on the corrected color characteristics and the mask of the tissue's natural color area, complete the printing color restoration and natural color area calibration to obtain a printing color image that eliminates light interference; A5: Extract fusion features based on the printed color image after eliminating light interference, the pre-processed tissue printing template image, and the pre-processed production line running speed time series data; A6: Based on the fusion characteristics, calculate the initial inkjet parameter adjustment values, and then complete the process constraint regularization to obtain the inkjet parameter adjustment values; A7: Adjust the inkjet parameters to control the printhead to perform inkjet operation.
2. The high-speed household tissue paper printing control method according to claim 1, characterized in that, Step A1 includes: A11: Images of household printed tissues are acquired using a high-speed industrial camera. The data type is two-dimensional pixel array data, which includes the grayscale values of the RGB three-color channels of the printed area and the acquisition timestamp. The images of the household printed tissues are then subjected to Gaussian filtering for noise reduction and linear interpolation to unify the resolution, resulting in a pre-processed printed tissue image. A12: Collect the standard color template for household paper towel printing. The data type is three-dimensional color matrix data, which includes the RGB standard values of each standard color block and CIE Lab color gamut coordinates. The standard color template for household paper towel printing is subjected to linear normalization and three times standard difference constant removal to obtain the cleaned paper towel printing template image. A13: The production line running speed signal is acquired by an incremental photoelectric encoder. The data type is a time-series electrical signal, which includes instantaneous speed value, sampling time interval, and equipment shaft code value. The production line running speed signal is subjected to mean filtering and linear fitting to fill in missing values, so as to obtain the pre-processed production line running speed time-series data. A14: Based on the pre-processed printed tissue paper image, the cleaned tissue paper printing template data, and the pre-processed production line running speed time series data, the timestamp matching method is used to align the multi-source data to obtain the pre-processed printed tissue paper image, the pre-processed tissue paper printing template image, and the pre-processed production line running speed data.
3. The high-speed household tissue paper printing control method according to claim 1, characterized in that, The A4 step includes: First, the corrected color features are input into a 1×1 convolutional layer. The feature dimensions of the corrected color features are adjusted by the 1×1 convolutional layer. Then, the Sigmoid function is input to obtain the color adjustment coefficients. Subsequently, the color adjustment coefficients are multiplied pixel by pixel with the multi-scale color features of the printed image to obtain a preliminary corrected image; Finally, a pixel-level replacement operation is used to perform a full pixel traversal of the preliminary correction image based on the tissue paper color area mask. Specifically, for areas in the preliminary correction image where the tissue paper color area mask is marked as 0, their color in the preliminary correction image is retained; for areas in the preliminary correction image where the tissue paper color area mask is marked as 1, their pixel values are replaced with the factory-calibrated tissue paper color reference value, resulting in a printed color image that eliminates light interference.
4. The high-speed household tissue paper printing control method according to claim 3, characterized in that, Step A5 includes: A51: Based on the printed color image after eliminating light interference, the pre-processed tissue paper printing template image, and the pre-processed production line running speed time series data, extract the spatial features of color deviation and the time series features of production speed. A52: Based on the spatial characteristics of color deviation and the temporal characteristics of production speed, the spatiotemporal heterogeneous features are adaptively fused to obtain the fused features.
5. The high-speed household tissue paper printing control method according to claim 4, characterized in that, Step A6 includes: Based on the fusion features, they are enhanced using a feature residual enhancement mechanism to obtain the fusion features with residual enhancement. Based on the fusion features with residual enhancement, the initial inkjet parameter adjustment value is calculated. Combined with the printing process constraint parameters, the initial inkjet parameter adjustment value is subjected to process constraint regularization. After speed adaptation calibration, the inkjet parameter adjustment value is obtained.
6. The high-speed household tissue paper printing control method according to claim 5, characterized in that, Step A7 includes: A71: The inkjet parameter adjustment value is converted into an inkjet parameter control signal adapted to the mechanical actuator through a general digital-to-analog converter; A72: The inkjet parameter control signal is received through the printhead drive unit, and the opening and closing degree of the solenoid valve of each inkjet channel is adjusted in real time by the constant current drive circuit to control the ink output. At the same time, the ink output is realized in real time through the ink output detection element. A73: The printhead displacement adjustment unit adjusts the lateral movement step size and longitudinal height of the printhead according to the inkjet parameter control signal, using a stepper motor drive combined with pulse width modulation technology. At the same time, the displacement positioning calibration is performed by a grating ruler to ensure that the relative position of the printhead and the tissue printing area matches the inkjet parameter adjustment value.
7. A high-speed household tissue paper printing control system, characterized in that, include: Image acquisition module: Acquires images of household printed tissues, standard color templates for household tissue printing, and production line speed signals, and performs preprocessing to obtain preprocessed printed tissue images, preprocessed tissue printing template images, and preprocessed production line speed data; Image feature extraction module: Based on the preprocessed printed tissue image and the preprocessed tissue printing template image, calculate the tissue's natural color area mask, the basic illumination color deviation of the natural color area, the multi-scale color features of the printed image, and the multi-scale color features of the template image; Illumination feature compensation module: Based on the mask of the natural color area of the tissue paper, the basic illumination color deviation of the natural color area, the multi-scale color features of the printed image, and the multi-scale color features of the template image, the illumination compensation weight is calculated to obtain the corrected color features; Light interference elimination module: Based on the corrected color features and the mask of the original color area of the tissue paper, the module completes the restoration of printed color and calibration of the original color area to obtain a printed color image that eliminates light interference. Feature fusion module: Extracts fusion features based on the printed color image after eliminating light interference, the pre-processed tissue printing template image, and the pre-processed production line running speed time series data; Inkjet parameter adjustment module: Based on the fusion characteristics, calculate the initial inkjet parameter adjustment value, and then complete the process constraint regularization to obtain the inkjet parameter adjustment value; Inkjet control module: controls the printhead to perform inkjet operation according to the inkjet parameter adjustment value; so as to realize the high-speed household tissue printing control method as described in any one of claims 1-6.