Digital printing equipment color dynamic calibration system based on multi-modal data fusion

Abstract

The invention relates to the field of digital printing equipment, and discloses a multi-modal data fusion digital printing equipment color dynamic calibration system which comprises a data acquisition and preprocessing module, a tensor modeling and optimization module, a causal reasoning module, a feedback control module, a cross-equipment mapping module and a self-diagnosis module. The method comprises the following steps that multi-modal data of equipment and environment are collected in real time, intelligent analysis is carried out, and control parameters are dynamically adjusted, so that the problem that a traditional color management method lacks real-time feedback and dynamic adaptive capacity is solved, and the stability and accuracy of the color calibration process are ensured. According to the method, color, image, environment and equipment operation data are collected, a multi-modal tensor is constructed, sparse feature compression is carried out, color deviation reasoning and adjustment quantity generation are realized in combination with causal structure learning, and a self-diagnosis mechanism and closed-loop control are assisted, so that real-time monitoring of the equipment operation state and dynamic adjustment of the color state are realized.

Description

Technical Field

[0001] The present invention relates to the field of digital printing equipment, and in particular to a multimodal data fusion digital printing equipment color dynamic calibration system. Background Art

[0002] Traditional color management solutions for digital printing equipment rely on statically configured color calibration models and pre-established strategies for adjusting device operating parameters. During the continuous operation of actual printing jobs, the device's color output often experiences dynamic shifts due to factors such as changing environmental conditions and fluctuations in the device's own status. Because traditional systems lack the ability to perceive and model the environment and device status in real time during operation, they struggle to adjust control parameters in a timely manner, leading to problems such as poor color stability and low output consistency.

[0003] Some existing solutions attempt to introduce feedback adjustment mechanisms or color calibration methods based on image monitoring, but the following technical limitations are common: on the one hand, the types of data collected by the system are limited and cannot fully reflect the state changes of the printing equipment during operation, especially the lack of the ability to integrate multimodal data; on the other hand, there is a lack of a modeling mechanism based on causal relationships, and the system often relies on empirical rules or static models for adjustment, making it difficult to accurately identify and dynamically reason about the causes of color deviations.

[0004] Furthermore, existing methods generally fail to consider unified, structured modeling of environmental parameters, equipment operating parameters, image data, and color data. This, to a certain extent, limits the model's expressiveness and generalization capabilities under complex working conditions. Furthermore, traditional calibration systems often rely on manual intervention to perform color correction, lacking automation and failing to meet the color stability and process robustness requirements of current high-frequency, high-precision printing operations.

[0005] Therefore, the present invention proposes a digital printing equipment color dynamic calibration system based on multimodal data fusion to address the deficiencies of the prior art. Summary of the Invention

[0006] In response to the shortcomings of the existing technology, the present invention provides a digital printing equipment color dynamic calibration system with multimodal data fusion, which solves the problem of lack of real-time feedback and dynamic adaptability in traditional color management methods.

[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions: a digital printing equipment color dynamic calibration system with multimodal data fusion, the system comprising the following modules:

[0008] The data acquisition and preprocessing module is used to collect multimodal raw data from digital printing equipment to perform denoising, missing value filling, numerical normalization and time alignment operations to generate multimodal preprocessed data;

[0009] A tensor modeling and optimization module is used to construct the multimodal preprocessed data into a tensor expression of uniform dimension, and perform tensor feature optimization based on the importance of each modal data to obtain a sparse feature structure;

[0010] A causal reasoning module constructs a causal graph model based on the causal relationship between the variables contained in the sparse feature structure, identifies the key factors affecting the color deviation, and outputs the reasoning results of the deviation cause;

[0011] A feedback control module combines the inference result with the current color state and the preset color target to generate corresponding equipment adjustment parameters for dynamically adjusting the operating state of the digital printing equipment;

[0012] A cross-device mapping module is used to establish a modal structure mapping relationship between different types of digital printing devices and migrate the tensor expression structure and causal model on the existing device to the target device, so that the feedback control module maintains the continuity of the calibration function in heterogeneous devices;

[0013] The self-diagnosis module is used to collect various operating status data and environmental variables during operation, generate equipment health assessment results based on multi-dimensional indicators, and trigger the feedback control module to perform parameter correction and control strategy update when an abnormal equipment operating status is detected.

[0014] Preferably, the data acquisition and preprocessing module includes the following steps:

[0015] Acquiring multimodal original data from a digital printing device, wherein the multimodal original data includes color data, image data, environmental parameter data, and device operation parameter data;

[0016] Performing a time alignment operation on the multimodal raw data based on a unified time reference benchmark;

[0017] performing noise processing on the time-aligned data, wherein the noise processing includes a filtering operation based on a local statistical method;

[0018] In the case of missing data, the missing values ​​are filled using a numerical completion method based on interpolation functions;

[0019] The processed multimodal data is normalized according to a unified numerical range, and the normalization is a linear normalization mapping.

[0020] Preferably, the tensor modeling and optimization module includes the following steps:

[0021] Receive pre-processed multimodal data and construct a tensor representation structure of unified dimensions based on the structural characteristics of each modal data in time, space and semantic dimensions;

[0022] After the tensor is constructed, the influence of each mode in color deviation modeling is analyzed, and the corresponding importance weight is assigned to each mode accordingly;

[0023] The tensor structure is subjected to feature compression processing according to the weights of each modality to extract a sparse feature structure with strong representation ability and low data redundancy.

[0024] Preferably, the importance weight is determined based on an attention allocation mechanism, which dynamically adjusts the weight of each modality according to the response strength and sensitivity of different modalities in target prediction.

[0025] Preferably, the causal reasoning module comprises the following steps:

[0026] It receives modal variables in the sparse feature structure as inference input, learns causal structures based on the joint statistical properties of the variables, and constructs a causal graph model containing multiple nodes and directed connections.

[0027] In the causal graph, each node represents a modal variable, and each edge represents a causal dependency relationship between two variables. The causal graph structure is determined by a data-driven structure search algorithm.

[0028] After the causal graph is constructed, the target color deviation result is inferred and estimated in combination with the observed variables, and the corresponding color deviation cause information is output through the conditional probability propagation mechanism.

[0029] Preferably, the feedback control module includes the following steps:

[0030] Receive the color deviation inference result and the current color state provided by the causal reasoning module, and compare them with the preset target color state;

[0031] Calculating an adjustment amount to perform color correction based on a difference between a current color state and a target color state;

[0032] Inputting the adjustment amount into the adjustment coefficient matrix for weighted adjustment to generate device control parameters;

[0033] The generated device control parameters are used to dynamically adjust the operating status of the digital printing device. The adjustments include inkjet control, ink flow rate, voltage signal and temperature parameters, which are used to adjust the output color of the device to a preset target value.

[0034] Preferably, the cross-device mapping module includes the following steps:

[0035] Receive the tensor representation structure and causal model from the source device and extract the modal structure information related to the target device's operating environment;

[0036] Construct a mapping function for inter-device conversion, adjusting the tensor structure constructed in the source device in terms of dimension and semantics to adapt to the modal characteristics of the target device;

[0037] The modeling parameters and control logic in the source device are converted into a structural representation suitable for the target device according to the mapping function, so as to be used continuously after the model migration.

[0038] Preferably, the cross-device mapping module further comprises the following steps:

[0039] After completing the model structure mapping, the migrated model is adaptively adjusted using a small amount of sample data collected from the target device;

[0040] The adaptive adjustment dynamically corrects control parameters based on an error feedback mechanism, so that the migrated model can meet the color calibration accuracy requirements on the target device and remain consistent with the operating data of the target device.

[0041] Preferably, the self-diagnosis module comprises the following steps:

[0042] During the operation of digital printing equipment, multiple operating status data and environmental parameters including temperature, humidity, equipment load and ink viscosity are collected in real time;

[0043] Based on the preset health assessment function, the collected data is calculated in multiple dimensions to generate the current health assessment results of the device;

[0044] The health assessment result is compared with the set threshold. If the assessment result is lower than the threshold, it is determined that the device has a potential abnormal risk, and the feedback control module is automatically triggered to dynamically correct the current device operating parameters and control strategy to ensure that the color calibration process is continuous and stable.

[0045] The present invention also provides a method for dynamic color calibration of a digital printing device using multimodal data fusion, the method comprising the following steps:

[0046] S1. Acquire multimodal original data of a digital printing device, wherein the multimodal original data includes color data, image data, environmental parameter data, and device operation parameter data;

[0047] S2. performing time alignment, noise removal, missing complementation, and normalization processing on the multimodal raw data to generate preprocessed data with a unified structure;

[0048] S3. Constructing a multimodal tensor structure based on the preprocessed data, and performing sparse feature compression after setting weights for each mode;

[0049] S4. Causal structure learning is performed based on sparse feature structure, a causal graph model is constructed, and color deviation is inferred and estimated in combination with observed variables;

[0050] S5. Compare the inference result with the target color state to generate the adjustment amount required for color correction;

[0051] S6. Adjusting the equipment operating parameters according to the adjustment amount, dynamically correcting the control logic of the digital printing equipment for closed-loop control of the color state.

[0052] The present invention provides a digital printing equipment color dynamic calibration system with multimodal data fusion.

[0053] It has the following beneficial effects:

[0054] 1. By constructing a multimodal tensor structure and performing sparse feature compression, this solution achieves efficient fusion of multimodal data within a unified structure, ensuring consistent representation of all types of data within the model and improving the accuracy of subsequent causal modeling and reasoning. This approach avoids modeling biases caused by dimensional mismatches or semantic inconsistencies between modal data, enhancing the system's ability to analyze color state changes.

[0055] 2. By introducing a causal structure learning method, this paper not only captures the correlation between variables but also establishes a causal graph model that reflects the control mechanism, making the reasoning of color deviation more interpretable and targeted. The construction process of this causal model integrates the distribution of observed variables and conditional independence analysis, effectively supporting the subsequent generation of adjustment variables and the updating of control strategies, adapting to the needs of dynamic color regulation under the influence of multiple factors.

[0056] 3. This invention utilizes a cross-device mapping module design to support the migration and adaptation of tensor structures and causal models across different devices, significantly enhancing the system's versatility across heterogeneous printing devices. By adjusting the dimensionality and semantics of the mapping function, the continuity and consistency of the color calibration model is maintained across devices with significantly different hardware configurations, thereby reducing the resource consumption of duplicate modeling.

[0057] 4. This invention uses a self-diagnostic module to monitor the device's operating status and environmental parameters in real time, quantifying the device's health based on a health assessment function. When an anomaly is detected, the feedback control module automatically implements corrective measures. This mechanism ensures that the device maintains a stable color calibration process even under non-ideal operating conditions, effectively reducing the risk of delays in fault intervention.

[0058] 5. By constructing a closed-loop control structure, the system automatically adjusts control parameters based on inferred color deviations, achieving dynamic correction and continuous calibration of color conditions. This solution offers strong responsiveness during continuous printing, making it suitable for handling color output fluctuations under complex and variable operating conditions, thereby enhancing the overall stability and reliability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 This is a system architecture diagram of the present invention;

[0060] Figure 2 Flow chart of the method of the present invention. DETAILED DESCRIPTION

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

[0062] See also Figure 1 The embodiment of the present invention provides a digital printing equipment color dynamic calibration system based on multimodal data fusion, the system comprising the following modules:

[0063] The data acquisition and preprocessing module is used to collect multimodal raw data from digital printing equipment to perform denoising, missing value filling, numerical normalization and time alignment operations to generate multimodal preprocessed data;

[0064] In this embodiment, during system operation, the data acquisition and preprocessing module acquires multimodal raw data from the digital printing device in real time. This raw data includes, but is not limited to, color data output by the device, image data generated during device operation, environmental parameters related to the device's operating environment, such as temperature and humidity, and device operating parameters such as inkjet flow rate, voltage signal, and workload. These data modalities have significant source variations, non-uniform scales, and inconsistent sampling frequencies, necessitating standardization through preprocessing.

[0065] After data acquisition is complete, the module first performs time alignment on the multimodal data. This time alignment uses a unified reference timestamp to synchronize data from different modalities, addressing issues related to data sampling delays and frequency deviations. Preferably, an interpolation strategy or dynamic time warping method is employed to preserve the original data features while ensuring alignment accuracy. After alignment, all modal data is reorganized into a data structure based on a unified timeline to support subsequent tensor modeling.

[0066] After completing time alignment, the system further processes noise in the data. Noise processing methods include filtering operations based on local statistical characteristics, such as median filtering, weighted moving average, or wavelet-based denoising. Specific filtering strategies are dynamically selected based on modal characteristics. For example, a small window median filter is preferred for color data to remove outliers, while an exponential moving average can be used for environmental parameter data to smooth out changing trends.

[0067] In the case of missing values ​​in the data set, the system uses interpolation functions to fill in the missing values. Preferably, the interpolation methods include but are not limited to linear interpolation, spline interpolation, and neighborhood weighted interpolation. In modal data with missing points in time, the interpolation function uses the time neighborhood as a benchmark, combines existing observations to generate estimated values, and maintains the continuity of data changes. For cases where some modalities are severely missing, a regression model can be constructed in combination with other modal variables to assist in predicting missing values, thereby improving the accuracy of the completion.

[0068] After noise processing and missing value filling are completed, all modal data will be normalized. Normalization uses a linear normalization mapping strategy to map the original data values ​​to a unified numerical interval [0, 1] or [-1, 1] to eliminate the differences in magnitude and dimension between different modalities and improve the comparability and processing efficiency of the data in the subsequent modeling process. Assume that the original data value is x i , its normalized result x i ′ can be expressed as:

[0069]

[0070] Among them, x min Indicates the minimum value of the current modal data, x max Indicates the maximum value of the current modal data, x i is any data point in the original data. This formula ensures that the normalized data has a uniform scale, which facilitates data fusion and weight calculation in the multimodal tensor modeling process.

[0071] After normalization, the data will be stored in a unified structure and input into the subsequent tensor modeling and optimization modules.

[0072] A tensor modeling and optimization module is used to construct the multimodal preprocessed data into a tensor expression of uniform dimension, and perform tensor feature optimization based on the importance of each modal data to obtain a sparse feature structure;

[0073] In this embodiment, the tensor modeling and optimization module receives multimodal data processed by the data acquisition and preprocessing module. After time alignment, noise processing, missing value filling and normalization, the multimodal data already has a unified time scale and numerical range. On this basis, the module analyzes each modal data and constructs a tensor expression structure of unified dimension based on its structural characteristics in time, space and semantic dimensions. Specifically, each dimension of the tensor represents a different data feature, such as time, space or specific parameters of each modality. The construction of the tensor follows the following principles:

[0074] Time dimension: Data from all modalities will be aligned according to a unified timestamp to ensure temporal consistency of tensors.

[0075] Spatial dimension: The spatial features of each modality will be mapped to the same spatial scale, thus ensuring the spatial consistency of data from different modalities.

[0076] Semantic dimension: The semantic information contained in data of different modalities, such as color data, image data, and environmental parameters, will be converted into the same representation so that they can be reasonably integrated in the tensor structure.

[0077] After completing the tensor structure construction, the module enters the stage of assigning modal importance weights. The degree of influence of each modality in color deviation modeling is different. Therefore, the module will analyze the response intensity and sensitivity of each modal data in the target prediction to determine its importance in the tensor. Preferably, this process uses an attention allocation mechanism to dynamically adjust the weight of each modality. The attention allocation mechanism automatically assigns corresponding weights to each modality based on the data characteristics of each modality and its contribution to the color prediction results. The core idea of ​​this mechanism is to dynamically adjust the importance of each modality by calculating its performance on historical data and its impact on the current task, so that the model pays more attention to the modal data that has a greater impact on the prediction results. Specifically, the following attention weight calculation formula can be used:

[0078]

[0079] Among them, w i represents the weight of the i-th mode; α i is the attention score calculated based on the modality's performance in historical data, and j is the index of all modalities. This formula can convert the contribution of each modality into its weight value, thereby optimizing information fusion in tensor representation.

[0080] After the weight distribution is completed, the tensor modeling and optimization module further performs feature compression processing to extract a sparse feature structure with low data redundancy and strong representation ability. The goal of feature compression is to reduce the dimension of the data and remove unnecessary redundant information, thereby improving computational efficiency and reducing the complexity of the model. This process preferably uses sparse matrix decomposition, principal component analysis (PCA) or other feature selection methods based on constrained optimization. Through these methods, it is possible to eliminate irrelevant features that contribute little to the prediction results while maintaining the important features of the data. Specifically, the compression process can be expressed by the following mathematical formula:

[0081] X′=Φ(X);

[0082] Here, X represents the original tensor data, Φ represents the compression operation, and X′ represents the compressed sparse feature tensor. The compressed tensor retains the most representative features and removes redundant features, thereby optimizing the subsequent model training and inference process.

[0083] A causal reasoning module constructs a causal graph model based on the causal relationship between the variables contained in the sparse feature structure, identifies the key factors affecting the color deviation, and outputs the reasoning results of the deviation cause;

[0084] In this embodiment, the causal reasoning module is used to identify and analyze the key factors that cause color deviation based on the sparse feature structure output by the tensor modeling and optimization module, forming an explanatory data-driven causal model, thereby providing mechanism-level support and guidance for the color dynamic calibration process.

[0085] The causal inference module takes as input a sparse feature tensor that has undergone feature compression and modal fusion. This sparse feature structure fully preserves the representative parameters of each key modality while removing redundant or noisy features. Based on these feature variables, the module constructs a causal graph model. The core of this model is to identify causal dependencies between variables, rather than simply modeling correlations.

[0086] To achieve this, the module first performs causal structure learning based on the joint probability distribution characteristics of each variable in the sparse feature structure. This structure learning uses a data-driven search strategy, combining a scoring function with constraints to jointly determine the topology of the causal graph. Specifically, the module uses a structure learning algorithm based on the score-search paradigm, such as Greedy-Equivalence-Search (GES) or the neighborhood-restricted Bayesian network structure learning method. The structure learning process can be abstracted into the following mathematical modeling problem:

[0087] Let the variable set be Its joint distribution is P(v1,v2,…,v n), the goal is to find a directed acyclic graph (DAG) Each edge e ij ∈ε represents the variable v i v j The structure learning process optimizes the graph structure by minimizing the following objective function:

[0088]

[0089] in, is a scoring function that measures the value of each variable in its parent node set Pa(v i ; G) The predictive ability under the conditions, the scoring function preferably uses the Bayesian Information Criterion (BIC) or the Akaike Information Criterion (AIC) and the like.

[0090] In the process of constructing the causal graph model, each node represents a modal variable in the sparse feature structure, such as color parameters, ambient humidity, inkjet pressure, etc.; each directed edge represents the direct causal dependency between two variables. For example, if the edge v i →v j If exists, it means variable v i is the variable v j The causal graph model can not only be used to model static causal relationships, but can also be expanded into a temporal-Bayesian-network to adapt to non-stationary characteristics when the system operating environment is subject to temporal dynamic changes.

[0091] After the causal graph is constructed, the module further performs inference operations based on the observed data to estimate the cause of the target variable (i.e., the color deviation indicator). The inference process is completed based on the conditional probability propagation mechanism in the Bayesian network. Let the color deviation result be the target variable y, its prior distribution is P(y), and the set of observed variables is O = {o1,o2,…,o k}, then the posterior inference goal is:

[0092]

[0093] Among them, the joint probability P(y,O) can be expanded by the chain rule based on the parent-child relationship of the nodes in the causal graph:

[0094]

[0095] During the inference process, the variable elimination or forward-backward propagation algorithm is used to quantitatively evaluate the possible sources of target deviation based on the current values ​​of the observed variables, and the contribution of each candidate dependent variable to the color deviation is ranked.

[0096] To further enhance the interpretability and practicality of the inference results, the module also introduces a causal path analysis mechanism to analyze the change path of the target variable. This path analysis is based on the set of paths from specific input variables to the output variable y in the causal graph, and evaluates the cumulative impact of each path. The path impact is calculated as follows:

[0097]

[0098] Among them, δ p Indicates the cumulative impact of path p on y, θ ij For edge v i →v j The corresponding edge weight is obtained through learning from historical data and reflects the marginal impact of the dependent variable on the outcome variable.

[0099] Ultimately, the module outputs causal reasoning results for color deviations, presented as a causal graph structure, impact paths, and variable contribution rankings. This provides a theoretical basis for subsequent dynamic color control strategies. These outputs can also be used in system-level feedback control mechanisms to achieve online correction of color states.

[0100] A feedback control module combines the inference result with the current color state and the preset color target to generate corresponding equipment adjustment parameters for dynamically adjusting the operating state of the digital printing equipment;

[0101] In this embodiment, the feedback control module is used to generate device control parameters for dynamically adjusting the operating state of the digital printing device based on the color deviation inference results output by the causal reasoning module, the current color state, and the difference between the preset target color state, thereby achieving accurate correction of color deviation.

[0102] The feedback control module first receives the color deviation inference results from the causal inference module. These results include the primary causal factors influencing color deviation and their respective contributions. Furthermore, the module obtains the current color status of the digital printing device. This information is typically acquired through sensors and provides real-time feedback on the device's actual output color. The current color status typically refers to the color characteristic parameters currently output by the printing device, such as hue, saturation, and brightness. These parameters accurately reflect the color characteristics of the device in actual operation.

[0103] Next, the feedback control module compares the current color state with a preset target color state. The target color state is the desired ideal color state, based on user requirements or device settings. This state is typically defined by a set color standard (such as an international color standard). The module calculates the difference between the current color state and the target color state to determine the amount of color adjustment required.

[0104] After calculating the color difference, the module further calculates the adjustment amount based on this difference. The adjustment amount reflects the correction operation required to achieve the target color state. Specifically, the calculation of the adjustment amount can be expressed by the following formula:

[0105] ΔC=C target -C current ;

[0106] Among them, ΔC is the required adjustment amount, C target is the target color state, C current The current color state. This difference provides the feedback control module with the direction and amplitude that need to be adjusted.

[0107] After calculating the adjustment amount, the module inputs it into an adjustment coefficient matrix for weighted adjustment. The adjustment coefficient matrix is ​​constructed based on the impact of various color adjustment parameters on the output color. This matrix typically involves multiple device parameters that affect color, such as inkjet control, ink flow rate, voltage signal, and temperature. These device parameters have a complex causal relationship with the final output color, and different device parameters may contribute differently to the color adjustment effect.

[0108] To ensure that each device parameter can be accurately adjusted according to the deviation of the current color state, the adjustment coefficient matrix assigns different weights to each adjustment parameter. Specifically, the calculation process of the matrix can be expressed as follows:

[0109] P = W·ΔC;

[0110] Where P is the device control parameter vector, W is the adjustment coefficient matrix, and ΔC is the required adjustment amount. The result of matrix multiplication is the final device control parameter. Through this process, the module can assign appropriate adjustment ranges to each device adjustment parameter based on its importance.

[0111] Finally, the generated device control parameters are fed into the digital printing equipment's control system, which dynamically adjusts the equipment's operating state. These adjustments include inkjet control (such as droplet size and jetting frequency), ink flow rate, voltage signal, temperature parameters, and more. By precisely adjusting these parameters, the device's color output gradually approaches the preset target color state, thereby correcting color deviations.

[0112] This feedback control module calibrates and optimizes the system in real time during operation based on inference results and current device status, ensuring consistent and stable color output. Furthermore, because this process is based on causal reasoning and a data-driven approach, it effectively identifies the primary causes of color deviation and provides a more scientific and rational basis for color adjustment decisions.

[0113] A cross-device mapping module is used to establish a modal structure mapping relationship between different types of digital printing devices and migrate the tensor expression structure and causal model on the existing device to the target device, so that the feedback control module maintains the continuity of the calibration function in heterogeneous devices;

[0114] In this embodiment, the cross-device mapping module is intended to migrate the tensor expression structure and causal model on the source device to the target device to ensure that the color dynamic calibration function can be implemented continuously and consistently between different types of digital printing devices.

[0115] The cross-device mapping module's workflow begins by receiving a tensor representation structure and causal model from the source device. The source device's tensor structure, derived from a previous model training process, incorporates the multimodal data features generated by the device during color calibration. The causal model describes the causal dependencies between various device parameters and color output. This information is then transferred and adapted to the target device, enabling it to inherit and perform the same calibration task.

[0116] The module first extracts modal structure information related to the target device based on its operating environment. The modal structure of the target device may differ from that of the source device. These differences may include differences in device hardware, operating parameters, or specific modal variables related to the device's operating environment (such as ambient temperature and humidity). This modal structure information will be used to guide the subsequent mapping process, ensuring that the migrated model is adapted to the characteristics of the target device.

[0117] After extracting the modal information related to the target device, the cross-device mapping module constructs a mapping function. This function adjusts the tensor structure in the source device in terms of dimensions and semantics to adapt it to the modal characteristics of the target device. Specifically, the mapping function must not only consider the conversion of data dimensions but also handle device-specific semantic differences. For example, the target device may use different parameter units, data ranges, or color spaces. Therefore, the mapping function transforms the tensor structure in the source device to facilitate its use on the target device.

[0118] The specific expression of the mapping function can be formalized in the following way:

[0119]

[0120] Among them, T target Represents the tensor structure on the target device, T source is the tensor structure in the source device, is the modal feature of the target device, and f(·) is the mapping function. The output of this function is the adjusted tensor structure of the target device, ensuring that the target device can perform effective color calibration based on the empirical data of the source device.

[0121] The application of mapping functions is not limited to adjusting data structures; it also involves converting the modeling parameters and control logic of the source device to adapt to the control system of the target device. These modeling parameters typically include device control parameters that affect color output, such as inkjet control, ink flow rate, and voltage regulation. The control logic, derived from the causal model of the source device, includes a strategy for adjusting device states based on color deviations. By applying the mapping function, these parameters and logic of the source device are converted into a form suitable for the target device.

[0122] After completing the mapping between devices, the cross-device mapping module further uses a small amount of sample data collected by the target device to adaptively adjust the migrated model. This process is based on the error feedback mechanism. By calibrating the actual output of the target device and dynamically correcting the control parameters, the migrated model can meet the color calibration accuracy requirements of the target device and remain consistent with the operating data of the target device. Specifically, the target device collects actual operating data (such as color characteristics such as output chromaticity and brightness), compares it with the target color state, and calculates the error between the current model and the target state. The module then adjusts the control parameters based on the error to optimize the color output of the device.

[0123] During the model adjustment process, the feedback mechanism can be quantified by the following formula:

[0124] ΔP adjust =g(ε);

[0125] Where ΔP adjust represents the adjustment amount of the control parameter, ε is the current error, and g(·) is the error-based adjustment function. This adjustment function dynamically adjusts the control parameter based on the error to minimize the difference between the output color of the target device and the preset target color state.

[0126] Ultimately, the cross-device mapping module achieves model migration and adaptation between the source and target devices through these steps. This mapped and adaptively adjusted model allows for continuous color calibration on the target device, ensuring consistent and accurate color output, with a high degree of continuity across devices.

[0127] A self-diagnosis module is used to collect various operating status data and environmental variables during operation, generate equipment health assessment results based on multi-dimensional indicators, and trigger the feedback control module to perform parameter correction and control strategy update when abnormal equipment operating status is detected;

[0128] In this embodiment, the self-diagnosis module is configured to collect multi-dimensional operating status data and environmental variables in real time during the operation of the digital printing device, evaluate the current operating health of the device through multi-indicator fusion calculation, and automatically link the feedback control module to perform parameter correction and control strategy update when potential abnormalities are detected, thereby ensuring the continuous stability of the color calibration process.

[0129] During system operation, the self-diagnostic module first collects various operating status data and external environmental parameters from the device. This data preferably includes, but is not limited to, internal device temperature, ambient humidity, device load, ink viscosity, inkjet head operating status, power supply voltage stability, system operating frequency, and device usage time. This data is acquired in real time via a multimodal sensing unit located internally or externally on the device and fed into the subsequent evaluation and calculation process in a unified data format.

[0130] To quantitatively assess the device's operating status, the self-diagnosis module integrates and analyzes multiple collected parameters based on a pre-defined health assessment function. This evaluation function considers the weight of each parameter in the device's health assessment and its historical evolution. By constructing a multidimensional indicator system, it derives the current device's operational health score. The basic expression of the evaluation function is as follows:

[0131]

[0132] Where H represents the health assessment score of the device, w i is the weight factor of the i-th type state variable, satisfying ∑w i =1,x i represents the current collected value of the i-th state parameter, f i(x i ) is a conversion function that maps the parameter value to a normalized health score. This function can be linear, nonlinear, or based on empirical rules, and is set based on the operating characteristics of the parameter in the device.

[0133] Weight factors w of various state parameters i The evaluation process is preferably based on historical data statistics, empirical knowledge bases, or machine learning algorithm training results, making the overall evaluation results dynamically adaptable. To further enhance the accuracy of the evaluation, historical trend comparison, time series sliding window processing, and anomaly detection algorithms can also be introduced into the evaluation process to improve the ability to respond to sudden situations.

[0134] After the evaluation is completed, the module compares the currently calculated health assessment score H with the set threshold H th Compare. If the following conditions exist:

[0135] H <H th ;

[0136] The system determines that the current device has a potential risk of abnormal operation and needs to execute a response mechanism. th It can be dynamically set according to the device model, frequency of use and operation strategy, and has a certain flexible adjustment range.

[0137] When the device is determined to be at risk of an abnormality, the self-diagnosis module automatically generates an abnormality flag and initiates an adaptive correction process with the feedback control module. This correction process involves real-time adjustment of the device's operating parameters and updating of the control strategy. Specifically, the module calculates the required control parameter adjustment based on the difference between the current state parameters and the target health state. The correction process can be expressed as follows:

[0138] ΔP=φ(H th -H);

[0139] Where ΔP is the correction amount for the control parameter, and φ(·) represents the control parameter mapping function based on the health assessment difference. This function maps changes in health status to changes in device control layer commands, such as adjusting inkjet pressure, changing the ink heating strategy, or activating the cooling unit.

[0140] This correction process directly impacts the parameter system relied upon by the feedback control module and influences the input distribution of its control decision model. Through dynamic adjustments, the device can promptly respond to potential operational anomalies, preventing further deterioration of the device's condition and effectively preventing the accumulation of color deviations and the resulting degradation of output quality.

[0141] In addition, to enhance the stability of the system during long-term operation, the self-diagnosis module supports periodic health assessment and data recording functions, which can store historical health status data and input it into the data-driven model for continuous training and updating of the equipment status recognition logic, thereby achieving trend monitoring and predictive maintenance of equipment operating conditions in the long term.

[0142] See also Figure 2 The present invention also provides a method for dynamic color calibration of a digital printing device using multimodal data fusion, the method comprising the following steps:

[0143] S1. Acquire multimodal original data of a digital printing device, wherein the multimodal original data includes color data, image data, environmental parameter data, and device operation parameter data;

[0144] The multimodal raw data covers four aspects: color data, image data, environmental parameter data, and equipment operation parameter data. Color data is usually collected in real time by a built-in or external color sensor of the device to reflect the color performance of the actual printed output; image data is collected by a high-speed camera device to collect the structural information and spatial distribution of the device output image, which is used to assist in analyzing the stability of color in different areas; environmental parameter data includes but is not limited to external variables that affect color performance, such as temperature, humidity, and air quality, and is collected by environmental sensors distributed around the device; equipment operation parameter data refers to the system operation status information recorded by the device during operation, such as inkjet frequency, voltage, current, heating unit power, etc. These data are collected uniformly through a distributed acquisition system and provide a basis for subsequent processing.

[0145] S2. performing time alignment, noise removal, missing complementation, and normalization processing on the multimodal raw data to generate preprocessed data with a unified structure;

[0146] First, the timestamp alignment operation is used to align different modal data in the same time dimension, so that they have a time series correspondence; then, noise removal is performed based on statistical methods and filtering algorithms to filter out outliers introduced by equipment vibration, electromagnetic interference or acquisition errors; then, the missing values ​​in the data are filled using methods such as neighbor interpolation and model inference to avoid the impact of incomplete sample information on the modeling process; finally, the numerical scale of various types of data is unified through normalization processing to ensure that data of different physical quantity units can be integrated and used in the same structure, which is convenient for subsequent tensor structure construction and feature analysis.

[0147] S3. Constructing a multimodal tensor structure based on the preprocessed data, and performing sparse feature compression after setting weights for each mode;

[0148] The tensor structure is a mathematical structure suitable for expressing high-dimensional multimodal data. It can simultaneously characterize the intrinsic relationships of data in multiple modal dimensions. During the construction process, corresponding weight coefficients are set for different modalities to reflect their relative importance in the overall modeling. Subsequently, the system performs sparse feature compression operations on the constructed tensor structure, retaining key features through methods such as principal component extraction, tensor decomposition, or dictionary learning, while eliminating redundant information and irrelevant variables, reducing model complexity and improving the interpretability of subsequent causal modeling.

[0149] S4. Causal structure learning is performed based on sparse feature structure, a causal graph model is constructed, and color deviation is inferred and estimated in combination with observed variables;

[0150] The goal is to explore the causal dependencies between different modal features and construct a causal graph model. Causal graphs express the direct and indirect influence paths between variables through graph theory, enabling the system to not only identify correlations between variables but also understand their mechanisms. During model construction, the system incorporates joint distribution analysis and conditional independence testing strategies based on observed variables to identify key paths and directions of influence, ultimately forming a causal graph structure for color deviation reasoning. This structure provides the foundation for subsequent color state prediction and deviation source analysis.

[0151] S5. Compare the inference result with the target color state to generate the adjustment amount required for color correction;

[0152] During the inference process, the system leverages established causal pathways, combines the current device state with collected data, and calculates the deviation between the current color output and the expected state. This deviation information is then transformed and mapped into parameter adjustments that can be directly applied to the control layer, including fine-tuning recommendations for multiple control variables (such as inkjet volume and heating temperature).

[0153] S6. Adjusting the equipment operating parameters according to the adjustment amount, dynamically correcting the control logic of the digital printing equipment for closed-loop control of the color state;

[0154] Through closed-loop feedback control, the inference results are converted into specific parameter setting commands, which are then applied to the color calibration-related control units. During operation, the control logic continuously responds to external input adjustment information, maintaining consistency between color output and target state. This enables dynamic closed-loop control of color state, ensuring stable printing quality over long periods of operation.

[0155] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A digital printing equipment color dynamic calibration system based on multimodal data fusion, characterized in that: The system includes the following modules: The data acquisition and preprocessing module is used to collect multimodal raw data from digital printing equipment to perform denoising, missing value filling, numerical normalization and time alignment operations to generate multimodal preprocessed data; A tensor modeling and optimization module is used to construct the multimodal preprocessed data into a tensor expression of uniform dimension, and perform tensor feature optimization based on the importance of each modal data to obtain a sparse feature structure; A causal reasoning module constructs a causal graph model based on the causal relationship between the variables contained in the sparse feature structure, identifies the key factors affecting the color deviation, and outputs the reasoning results of the deviation cause; A feedback control module combines the inference result with the current color state and the preset color target to generate corresponding equipment adjustment parameters for dynamically adjusting the operating state of the digital printing equipment; A cross-device mapping module is used to establish a modal structure mapping relationship between different types of digital printing devices and migrate the tensor expression structure and causal model on the existing device to the target device, so that the feedback control module maintains the continuity of the calibration function in heterogeneous devices; The self-diagnosis module is used to collect various operating status data and environmental variables during operation, generate equipment health assessment results based on multi-dimensional indicators, and trigger the feedback control module to perform parameter correction and control strategy update when an abnormal equipment operating status is detected.

2. The digital printing equipment color dynamic calibration system with multimodal data fusion according to claim 1 is characterized in that: The data acquisition and preprocessing module includes the following steps: Acquiring multimodal original data from a digital printing device, wherein the multimodal original data includes color data, image data, environmental parameter data, and device operation parameter data; Performing a time alignment operation on the multimodal raw data based on a unified time reference benchmark; performing noise processing on the time-aligned data, wherein the noise processing includes a filtering operation based on a local statistical method; In the case of missing data, the missing values ​​are filled using a numerical completion method based on interpolation functions; The processed multimodal data is normalized according to a unified numerical range, and the normalization is a linear normalization mapping.

3. The digital printing equipment color dynamic calibration system with multimodal data fusion according to claim 1, characterized in that: The tensor modeling and optimization module includes the following steps: Receive pre-processed multimodal data and construct a tensor representation structure of unified dimensions based on the structural characteristics of each modal data in time, space and semantic dimensions; After the tensor is constructed, the influence of each mode in color deviation modeling is analyzed, and the corresponding importance weight is assigned to each mode accordingly; The tensor structure is subjected to feature compression processing according to the weights of each modality to extract a sparse feature structure with strong representation ability and low data redundancy.

4. The digital printing equipment color dynamic calibration system with multimodal data fusion according to claim 3 is characterized in that: The importance weight is determined based on an attention allocation mechanism, which dynamically adjusts the weight of each modality according to the response strength and sensitivity of different modalities in target prediction.

5. The digital printing equipment color dynamic calibration system with multimodal data fusion according to claim 1, characterized in that: The causal reasoning module includes the following steps: It receives modal variables in the sparse feature structure as inference input, learns causal structures based on the joint statistical properties of the variables, and constructs a causal graph model containing multiple nodes and directed connections. In the causal graph, each node represents a modal variable, and each edge represents a causal dependency relationship between two variables. The causal graph structure is determined by a data-driven structure search algorithm. After the causal graph is constructed, the target color deviation result is inferred and estimated in combination with the observed variables, and the corresponding color deviation cause information is output through the conditional probability propagation mechanism.

6. The digital printing equipment color dynamic calibration system with multimodal data fusion according to claim 1, characterized in that: The feedback control module includes the following steps: Receive the color deviation inference result and the current color state provided by the causal reasoning module, and compare them with the preset target color state; Calculating an adjustment amount to perform color correction based on a difference between a current color state and a target color state; Inputting the adjustment amount into the adjustment coefficient matrix for weighted adjustment to generate device control parameters; The generated device control parameters are used to dynamically adjust the operating status of the digital printing device. The adjustments include inkjet control, ink flow rate, voltage signal and temperature parameters, which are used to adjust the output color of the device to a preset target value.

7. The digital printing equipment color dynamic calibration system with multimodal data fusion according to claim 1, characterized in that: The cross-device mapping module includes the following steps: Receive the tensor representation structure and causal model from the source device and extract the modal structure information related to the target device's operating environment; Construct a mapping function for inter-device conversion, adjusting the tensor structure constructed in the source device in terms of dimension and semantics to adapt to the modal characteristics of the target device; The modeling parameters and control logic in the source device are converted into a structural representation suitable for the target device according to the mapping function, so as to be used continuously after the model migration.

8. The digital printing equipment color dynamic calibration system with multimodal data fusion according to claim 7, characterized in that: The cross-device mapping module further comprises the following steps: After completing the model structure mapping, the migrated model is adaptively adjusted using a small amount of sample data collected from the target device; The adaptive adjustment dynamically corrects control parameters based on an error feedback mechanism, so that the migrated model can meet the color calibration accuracy requirements on the target device and remain consistent with the operating data of the target device.

9. The digital printing equipment color dynamic calibration system with multimodal data fusion according to claim 1, characterized in that: The self-diagnosis module comprises the following steps: During the operation of digital printing equipment, multiple operating status data and environmental parameters including temperature, humidity, equipment load and ink viscosity are collected in real time; Based on the preset health assessment function, the collected data is calculated in multiple dimensions to generate the current health assessment results of the device; The health assessment result is compared with the set threshold. If the assessment result is lower than the threshold, it is determined that the device has a potential abnormal risk, and the feedback control module is automatically triggered to dynamically correct the current device operating parameters and control strategy to ensure that the color calibration process is continuous and stable.

10. A method for dynamic color calibration of digital printing equipment using multimodal data fusion, applied to the system according to any one of claims 1 to 9, characterized in that: The method comprises the following steps: S1. Acquire multimodal original data of a digital printing device, wherein the multimodal original data includes color data, image data, environmental parameter data, and device operation parameter data; S2. performing time alignment, noise removal, missing complementation, and normalization processing on the multimodal raw data to generate preprocessed data with a unified structure; S3. Constructing a multimodal tensor structure based on the preprocessed data, and performing sparse feature compression after setting weights for each mode; S4. Causal structure learning is performed based on sparse feature structure, a causal graph model is constructed, and color deviation is inferred and estimated in combination with observed variables; S5. Compare the inference result with the target color state to generate the adjustment amount required for color correction; S6. Adjusting the equipment operating parameters according to the adjustment amount, dynamically correcting the control logic of the digital printing equipment for closed-loop control of the color state.

Images