Self-evolutionary variable data printing method based on latent space embedding and complexity grading

Abstract

This invention relates to the field of variable data printing technology. Specifically, it is a self-evolving variable data printing method based on latent space embedding and complexity grading. By pre-rendering and caching static regions, latent space embedding parameters and feature encoding networks are introduced to compress and map high-dimensional original dynamic data into low-dimensional latent space parameters. These parameters are then decomposed to extract font style features, color distribution features, geometric layout features, and local texture features, achieving decoupled expression of different visual attributes of dynamic regions. This allows for independent adjustment and optimization of different attributes such as font, color, layout, and texture during subsequent generation. Simultaneously, a comprehensive complexity score is calculated by considering the region area ratio, edge variation degree, number of graphic structures, and historical generation time, and compared with a preset threshold. An adaptive path selection mechanism is used to dynamically allocate computing resources according to the actual needs of different types of dynamic regions.

Description

Technical Field

[0001] This invention relates to the field of variable data printing technology, specifically to a self-evolving variable data printing method based on latent space embedding and complexity hierarchy. Background Technology

[0002] Variable data printing, as a core branch of digital printing, is based on digital processes and database technology. It allows for real-time changes to text, images, barcodes, and other content during continuous printing, enabling personalized customization of mass-produced printed materials. It is widely used in scenarios such as commercial invoices, direct mail marketing, smart packaging, and card production. Existing technology CN202310210980.7 discloses a method that involves pre-converting static template data, determining the current dynamic area occupied by the current dynamic data in the current printing drawing, extracting a predetermined area larger than the current dynamic area, converting and processing the predetermined area to obtain converted data, and extracting the current dynamic converted data from this data. This current dynamic converted data replaces the corresponding data in the static template data, resulting in new printing data composed of the static template data and the current dynamic data. Compared to directly converting the current printing drawing, this significantly reduces conversion time, thereby reducing loading time and meeting the needs of continuous production; it also reduces the size of the converted file, thus reducing storage requirements.

[0003] Existing technologies require truncating a predetermined region larger than the dynamic region each time the current dynamic region is processed. This region is obtained by extending the dynamic region outward by a predetermined length, and all data within this predetermined region is re-converted, such as in raster image processing. This has certain drawbacks: First, it requires truncating the predetermined region extending beyond the dynamic region and re-converting the data as a whole, repeatedly processing the pre-processed static data, resulting in computational redundancy. Second, all dynamic data uses the same processing flow, lacking the ability to adaptively balance generation efficiency and quality. Summary of the Invention

[0004] The purpose of this invention is to provide a self-evolving variable data printing method based on latent space embedding and complexity hierarchy to solve the problems mentioned in the background art.

[0005] To address the aforementioned technical problems, this invention provides the following technical solution: a self-evolving variable data printing method based on latent space embedding and complexity grading, which obtains an initial template for variable data printing containing fixed and variable content regions, and performs layout parsing and region division processing into static and dynamic regions, and pre-renders and caches the static regions; wherein the static regions are the content that remains unchanged in different printing tasks, and the dynamic regions are the content that changes with the task data. Record the corresponding spatial coordinates, hierarchical relationships, output resolution requirements and associated data field identifiers for each dynamic area. At the same time, perform pre-rendering and caching processing on the static area so that the static area can be directly called in multiple subsequent printing tasks, thereby establishing a template structure in which the static area and the dynamic area are independent of each other. Based on the data types corresponding to the dynamic regions, a differentiated feature coding model is established. Structural feature extraction and coding processing are performed on dynamic data such as text, barcode, QR code, image, graphics and combined layout. The original data of the dynamic regions is compressed into low-dimensional latent space parameters and split into four independent features. The complexity of the dynamic regions is evaluated based on multi-dimensional indicators, and two path types are combined: threshold adaptive switching rule rendering and latent space parameter driven. Based on low-dimensional latent space parameters and path types, a dynamic region template node, coordinate mapping matrix, and layout constraint rule set bound to multiple types of information are constructed. A generation parameter stability evaluation mechanism is added based on the output quality statistics of historical printing tasks. The optimal parameter combination is selected according to the stability evaluation mechanism to complete the standardized construction of the dynamic region generation template. The multiple types of information include spatial coordinate information, dynamic region identifier, low-dimensional latent space parameters, and path type. When receiving a new printing task, the target dynamic area is accurately located by generating a template based on the dynamic area. Only the dynamic content is updated locally, while the static cache and layout constraints are preserved. The page is partially replaced and the path type priority is dynamically optimized based on the dynamic area replacement running statistics. After the dynamic area update is completed, the dynamic area generation result and the static area cache result are combined and the final printed page is output based on the coordinate mapping matrix and the set of page layout constraint rules.

[0006] Furthermore, based on the data type corresponding to the dynamic region, a differentiated feature coding model is established. The specific method for compressing the original data of the dynamic region into latent space parameters and splitting it into four independent features is as follows: Feature encoding is performed on the original data of each dynamic region separately, and the feature encoder is used to encode the original data of a single region into latent space parameters; the encoding mapping relationship is Zi=E θ (Xi), where Xi represents the original input data of the i-th dynamic region, E θLet θ represent the feature encoder, and Zi represent the latent space embedding parameters corresponding to the dynamic region. Encoding achieves compressed mapping from the original high-dimensional dynamic region data to low-dimensional latent space parameters that characterize the main changes in the region, reducing the computational overhead caused by repeatedly processing complete image data during subsequent page updates. The generated latent space parameters undergo multi-feature-dimensional structured decomposition. The decomposed latent space parameters are composed of four independent feature types: font style feature parameters, color distribution feature parameters, geometric layout feature parameters, and local texture feature parameters. The vector structure of the latent space parameters is as follows: ,in Corresponding font style feature parameters, Corresponding color distribution characteristic parameters, Corresponding geometric layout feature parameters, Corresponding to local texture feature parameters; the above-mentioned structured decomposition achieves decoupled representation of different visual attributes of dynamic regions, facilitating individual parameter adjustment for the four types of visual feature attributes in subsequent content generation stages. Furthermore, the specific method for evaluating the complexity of dynamic regions based on multi-dimensional indicators is as follows: A dynamic region complexity evaluation model is constructed, quantifying the dynamic region complexity from four dimensions: region area ratio, edge change degree, number of graphic and textual structures, and historical average generation time. A weighted calculation is used to obtain the comprehensive complexity score Ci for the i-th dynamic region. The comprehensive complexity calculation formula is: Ci = αAi / A page +βEi+γFi+δTi; where Ai is the area of ​​a single dynamic region, A page Let Ei be the total area of ​​the entire page, Fi be the region edge complexity index, Fi be the number of graphic and textual structure objects within the region, Ti be the historical average generation time of the dynamic region, and α, β, γ, and δ be the weighting coefficients corresponding to the area proportion, edge complexity, number of graphic and textual objects, and historical generation time, respectively. The multi-dimensional index includes the region area proportion, edge change degree, number of graphic and textual structures, and historical average generation time. Furthermore, the specific method for adaptively switching path types based on thresholds is as follows: A complexity switching threshold Cth is pre-configured, and the magnitude of the comprehensive complexity score and the switching threshold is used as the path switching criterion. When the comprehensive complexity of the dynamic region satisfies Ci < Cth, the system uses the rule rendering path to directly generate the dynamic region. When the comprehensive complexity of the dynamic region satisfies Ci ≥ Cth, the system switches to the latent space parameter-driven generation path to complete the generation of the dynamic region. The threshold-based hierarchical path selection mechanism avoids the unnecessary computational overhead caused by using latent space inference for simple dynamic regions, while ensuring that high-complexity dynamic regions maintain the flexibility of layout generation and the visual consistency of the screen by relying on parameter-driven paths.

[0007] Furthermore, the specific methods for constructing dynamic region template nodes, coordinate mapping matrices, and layout constraint rule sets that bind multiple types of information are as follows: For each dynamic region, a dedicated template description structure is built, binding and summarizing the dynamic region identifier, latent space parameters, generation path type, and spatial coordinate information to form a dynamic region template node. The template node is represented by a set Ti={IDi,Zi,Pi,Ri}; where Ti is the template node corresponding to the i-th dynamic region, IDi is the dynamic region identifier number, Zi is the previously generated latent space embedding parameter, Pi is the generation path type matched by the dynamic region, and Ri is the set of layout constraint rules corresponding to the dynamic region. By establishing a unified mapping relationship between dynamic region parameters and template structure, subsequent printing tasks can directly call the corresponding template node to complete the dynamic region update. Based on the recorded spatial coordinate information of the dynamic region, a dynamic region coordinate mapping matrix constraint is constructed. The coordinate mapping matrix is ​​used to constrain the position of the content generated by the dynamic region in the template. The coordinate mapping parameters are represented as Mi={xi,yi,wi,hi}; where Mi represents the coordinate mapping parameters of the i-th dynamic region, xi and yi are the horizontal and vertical coordinates of the upper left corner of the dynamic region in the page, wi is the width of the dynamic region, and hi is the height of the dynamic region. The coordinate mapping matrix of a single dynamic area consists of four quantifiable parameters: the horizontal coordinate is the distance from the top left corner of the area to the left edge of the page, the vertical coordinate is the distance from the top left corner of the area to the top edge of the page, and the area width and height correspond to the horizontal and vertical dimensions of the dynamic area, respectively; by using the coordinate mapping matrix to lock the generation position, the defects of dynamic area offset and page misalignment are avoided; The set of layout constraint rules corresponding to the dynamic region includes color constraint rules and resolution constraint rules. The degree of color matching between the generated dynamic region and the original template layout is calculated using a color consistency evaluation index. The formula for calculating the color consistency index is as follows: ,in Let Ni be the color consistency score for the i-th dynamic region, and Ni be the number of color sampling points in the i-th dynamic region. Generate the color value of the image at the k-th sampling point for the i-th dynamic region. The reference color value corresponding to the i-th dynamic region template and the k-th sampling point is used to control the color consistency of the dynamic region generated image. The resolution standard of the dynamic region output image is limited by adding dynamic region resolution constraint rules. The resolution constraint rule is that the actual output resolution of the dynamic region generation result is greater than or equal to the minimum output resolution threshold preset by the template. The color constraint rule and the resolution constraint rule are combined to limit the visual consistency of the dynamic region and the static region of the page.

[0008] Furthermore, the specific method for adding a generation parameter stability evaluation mechanism based on the output quality statistics of historical printing tasks is as follows: Based on the output quality statistics of historical printing tasks, and by integrating data from four dimensions—edge consistency, color consistency, page offset, and generation time—a weighted stability score is calculated for the j-th combination of generation parameters. The stability score calculation formula is as follows: Where Sj is the stability score of the j-th group of generated parameters, Bj is the edge blending quality score, Qj is the color consistency score, Oj is the normalized result of page offset, Tj is the normalized result of generation time, and λ1, λ2, λ3, and λ4 are the weight coefficients corresponding to the four indicators of edge, color, offset, and time, respectively. The above formula is used to quantify the comprehensive stability level of each group of generated parameters in historical printing tasks. The generated parameter combinations are sorted according to their stability scores, and the parameter combination with the highest stability score is prioritized when performing dynamic region generation tasks. An offline statistical optimization parameter selection mechanism is adopted to improve the layout consistency and output stability of multiple printing batches while eliminating parameter drift problems caused by online continuous learning.

[0009] Furthermore, the specific method for completing partial page replacement is as follows: After the dynamic region is generated, the new dynamic region result is written to the corresponding dynamic region cache, and a partial page replacement relationship is established: P t+1 =P t -D t +D t+1 , where P t D represents the page result before the update. t D represents the dynamic region content before the update. t+1 P represents the new dynamic region content generated after the current task is created. t+1 This represents the updated page result. Since the system only performs partial replacement on the dynamic areas of the page, while the static and unchanged areas remain unchanged, it avoids the redundant calculation overhead caused by regenerating the entire page. At the same time, the system statistically analyzes the processing time, cache call count, and partial replacement success rate during the dynamic area update process, and dynamically adjusts the path scheduling priority corresponding to different dynamic areas based on the statistical results. This reduces unnecessary calls to the implicit space parameter generation path in low-complexity areas, thereby improving the page update efficiency and system resource utilization in batch variable data printing tasks. For text numbers, ordinary barcodes, QR codes, or low-complexity graphic areas, the corresponding dynamic area content is directly generated by calling the rule rendering path; for dynamic areas containing complex graphic structures, high-frequency style changes, or multi-element combination layouts, the implicit space parameters are called to drive the generation path to complete the dynamic area generation. During the dynamic area generation process, the system only allows updating the template node data corresponding to the current dynamic area. The static area cache data, layout structure relationship, layer hierarchy relationship, spatial coordinate mapping relationship, and color constraint rules and resolution constraint rules established in step three remain unchanged, thereby ensuring that the layout stability of other areas of the page is not affected during the dynamic area replacement process.

[0010] Furthermore, the specific method for outputting the final printed page is as follows: After the dynamic area update process is completed, the coordinate mapping matrix constraint and the set of page constraint rules are called. Based on the above constraint rules, the dynamic area generation result and the static area cache result are combined into a page layout and the final printed page is output. During the page layout compositing process, the edge transitions, resolution matching, and color matching at the junctions of dynamic and static areas are corrected to ensure that the output appearance of the dynamic areas and the original template page remains consistent. At the same time, the static area reuse mechanism is used to eliminate the processing delay caused by repeated rendering of the entire page, thereby improving the page output efficiency of batch variable data printing tasks.

[0011] Compared with existing technologies, the beneficial effects achieved by this invention are as follows: By clearly dividing the variable data printing template into static and dynamic regions and performing pre-rendering and caching processing on the static regions, the static content can be directly reused in subsequent large-scale printing tasks without repeatedly performing page layout parsing, rasterization conversion, or full-page layer reconstruction operations. Simultaneously, the solution introduces latent space embedding parameters for the dynamic regions, compressing and mapping high-dimensional text, barcodes, images, and other raw data into low-dimensional latent space parameters through a feature encoding network, thereby transforming the changing patterns of the dynamic regions into a compact parameter representation. This mechanism fundamentally avoids the huge computational overhead of reprocessing complete image data for each printing task. Especially in large-scale batch printing scenarios, the combined effect of multiple reuses of static region rendering and the low-dimensional parameterized representation of dynamic regions significantly reduces the average processing latency and system resource consumption per page. This invention establishes dynamic region template nodes, uniformly binding dynamic region identifiers, latent space parameters, generation path types, and page layout constraint rules, and further constructs coordinate mapping matrices, color constraint rules, and resolution constraint rules to ensure that the dynamic region generation results can be stably mapped to fixed page layout positions. Building upon this foundation, the solution uses historical printing task output quality statistics to comprehensively score edge consistency, color consistency, layout offset, and generation time, generating a parameter stability index. Based on this score, the generated parameter combinations are ranked, prioritizing combinations with higher stability scores for subsequent dynamic area generation. This offline statistical optimization mechanism avoids parameter drift issues that can arise from continuous online learning, ensuring highly consistent layout output across different batches and time periods of variable data printing tasks. This effectively addresses the pain point of significant layout differences between multiple print runs in traditional methods.

[0012] When a new variable data printing task is received, the system only identifies the dynamic areas involved in the current task and locates the corresponding template nodes. Updates are performed only on these areas, while static area cache data, page layout relationships, layer hierarchy relationships, spatial coordinate mapping relationships, and pre-established color and resolution constraints remain unchanged. After the dynamic areas are generated, the system writes the new dynamic area content to the corresponding cache area using a partial page replacement formula, replacing the original content. Static areas and unchanged areas are left unprocessed. Simultaneously, the system tracks the processing time, cache call count, and partial replacement success rate during the dynamic area update process and dynamically adjusts the path scheduling priority for different dynamic areas based on the statistical results. This partial update mechanism ensures that each new printing task only processes the truly changed parts, avoiding redundant calculations caused by regenerating the entire page. Attached Figure Description

[0013] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of the steps of the self-evolving variable data printing method based on latent space embedding and complexity hierarchy of the present invention. Figure 2 This is a flowchart of the template node and coordinate mapping construction process of the self-evolving variable data printing method based on latent space embedding and complexity hierarchy of the present invention. Figure 3 This is a flowchart of the constraint rules and parameter stability evaluation of the self-evolving variable data printing method based on latent space embedding and complexity hierarchy in this invention. Figure 4 This is a flowchart of the dynamic region local update process of the self-evolving variable data printing method based on latent space embedding and complexity hierarchy of the present invention. Detailed Implementation

[0014] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0015] Please see Figure 1-4 The present invention provides a technical solution: a self-evolving variable data printing method based on latent space embedding and complexity gradation, which obtains an initial template for variable data printing containing fixed content regions and variable content regions, and performs layout parsing and region division processing into static regions and dynamic regions, and performs pre-rendering and caching processing on the static regions; wherein the static regions are the content that remains unchanged in different printing tasks, and the dynamic regions are the content that changes with the task data. Record the corresponding spatial coordinates, hierarchical relationships, output resolution requirements and associated data field identifiers for each dynamic area. At the same time, perform pre-rendering and caching processing on the static area so that the static area can be directly called in multiple subsequent printing tasks, thereby establishing a template structure in which the static area and the dynamic area are independent of each other. Based on the data types corresponding to the dynamic regions, a differentiated feature coding model is established. Structural feature extraction and coding processing are performed on dynamic data such as text, barcode, QR code, image, graphics and combined layout. The original data of the dynamic regions is compressed into low-dimensional latent space parameters and split into four independent features. The complexity of the dynamic regions is evaluated based on multi-dimensional indicators, and two path types are combined: threshold adaptive switching rule rendering and latent space parameter driven. Based on low-dimensional latent space parameters and path types, a dynamic region template node, coordinate mapping matrix, and layout constraint rule set bound to multiple types of information are constructed. A generation parameter stability evaluation mechanism is added based on the output quality statistics of historical printing tasks. The optimal parameter combination is selected according to the stability evaluation mechanism to complete the standardized construction of the dynamic region generation template. The multiple types of information include spatial coordinate information, dynamic region identifier, low-dimensional latent space parameters, and path type. When receiving a new printing task, the target dynamic area is accurately located by generating a template based on the dynamic area. Only the dynamic content is updated locally, while the static cache and layout constraints are preserved. The page is partially replaced and the path type priority is dynamically optimized based on the dynamic area replacement running statistics. After the dynamic area update is completed, the dynamic area generation result and the static area cache result are combined and the final printed page is output based on the coordinate mapping matrix and the set of page layout constraint rules.

[0016] Furthermore, based on the data type corresponding to the dynamic region, a differentiated feature coding model is established. The specific method for compressing the original data of the dynamic region into latent space parameters and splitting it into four independent features is as follows: Feature encoding is performed on the original data of each dynamic region separately, and the feature encoder is used to encode the original data of a single region into latent space parameters; the encoding mapping relationship is Zi=E θ (Xi), where Xi represents the original input data of the i-th dynamic region, E θ Let θ represent the feature encoder, and Zi represent the latent space embedding parameters corresponding to the dynamic region. Encoding achieves compressed mapping from the original high-dimensional dynamic region data to low-dimensional latent space parameters that characterize the main changes in the region, reducing the computational overhead caused by repeatedly processing complete image data during subsequent page updates. The generated latent space parameters undergo multi-feature-dimensional structured decomposition. The decomposed latent space parameters are composed of four independent feature types: font style feature parameters, color distribution feature parameters, geometric layout feature parameters, and local texture feature parameters. The vector structure of the latent space parameters is as follows: ,in Corresponding font style feature parameters, Corresponding color distribution characteristic parameters, Corresponding geometric layout feature parameters, Corresponding local texture feature parameters; through the above structured decomposition, the different visual attributes of dynamic regions are decoupled and represented, which facilitates the subsequent content generation process to adjust the parameters of the four types of visual feature attributes separately.

[0017] Specifically, the raw data of dynamic regions in variable printing scenarios are input into a multi-type dynamic data feature encoding network. This network can selectively extract core visual features from dynamic data such as text, barcodes, and images. The encoding network sequentially passes through three encoding hidden layers to output corresponding low-dimensional latent space parameters. The first encoding hidden layer has an output dimension of 128 dimensions and uses the ReLU activation function to complete non-linear feature mapping. The second encoding hidden layer has an output dimension of 64 dimensions, uses the ReLU activation function, and leverages L2 regularization to constrain network weights, suppressing sample overfitting. The third encoding hidden layer has an output dimension of 32 dimensions and uses the ReLU activation function to complete feature condensation. Finally, the low-dimensional latent space parameters of the corresponding dynamic regions are obtained through mapping via the encoding output layer, completing the compressed mapping of multiple types of dynamic data (text, barcodes, images, etc.) from high-dimensional raw data to low-dimensional latent space. Preferably, the training dataset for the multi-type dynamic data feature encoding network uses various dynamic region samples from real-world variable data printing applications, covering real printing data of varying complexity, such as text, barcodes, and complex graphics. Offline supervised training is employed, and the loss function is a weighted sum of mean squared error loss and feature decoupling loss. The mean squared error loss constrains the accuracy of latent space parameter restoration, while the feature decoupling loss ensures the independence of the four feature classes. The optimizer uses the Adam optimizer with an initial learning rate of 0.001 and 1000 training iterations. The encoder parameters are iteratively optimized based on the generation quality and processing latency statistics of historical printing tasks, with the optimization objectives being the dynamic region feature decoupling effect and the accuracy of latent space parameter restoration. Simultaneously, continuous tuning is performed using a generation parameter stability evaluation mechanism to avoid parameter drift caused by online continuous learning, ensuring that the encoder stably completes the compression mapping from high-dimensional dynamic data to low-dimensional latent space parameters.

[0018] Furthermore, the specific method for evaluating the complexity of dynamic regions based on multi-dimensional indicators is as follows: A dynamic region complexity evaluation model is constructed, quantifying the dynamic region complexity from four dimensions: region area ratio, edge change degree, number of graphic and textual structures, and historical average generation time. A weighted calculation is used to obtain the comprehensive complexity score Ci for the i-th dynamic region. The comprehensive complexity calculation formula is: Ci = αAi / A page +βEi+γFi+δTi; where Ai is the area of ​​a single dynamic region, A pageEi represents the total area of ​​the entire page, and Ei is the region edge complexity index, calculated using the Canny edge detection algorithm. Specifically, it is the ratio of the number of edge pixels within the region to the total number of pixels in the region. Fi represents the number of graphic and textual structural objects within the region, Ti represents the historical average generation time of this dynamic region, and α, β, γ, and δ are the weight coefficients corresponding to the area proportion, edge complexity, number of graphic and textual objects, and historical generation time, respectively. The multi-dimensional index includes the region area proportion, the degree of edge change, the number of graphic and textual structural objects, and the historical average generation time.

[0019] Furthermore, the specific method for adaptively switching path types based on thresholds is as follows: A complexity switching threshold Cth is pre-configured, and the magnitude of the comprehensive complexity score and the switching threshold is used as the path switching criterion. When the comprehensive complexity of the dynamic region satisfies Ci < Cth, the system uses the rule rendering path to directly generate the dynamic region. When the comprehensive complexity of the dynamic region satisfies Ci ≥ Cth, the system switches to the latent space parameter-driven generation path to complete the generation of the dynamic region. The threshold-based hierarchical path selection mechanism avoids the unnecessary computational overhead caused by using latent space inference for simple dynamic regions, while ensuring that high-complexity dynamic regions maintain the flexibility of layout generation and the visual consistency of the screen by relying on parameter-driven paths.

[0020] Furthermore, the specific methods for constructing dynamic region template nodes, coordinate mapping matrices, and layout constraint rule sets that bind multiple types of information are as follows: For each dynamic region, a dedicated template description structure is built, binding and summarizing the dynamic region identifier, latent space parameters, generation path type, and spatial coordinate information to form a dynamic region template node. The template node is represented by a set Ti={IDi,Zi,Pi,Ri}; where Ti is the template node corresponding to the i-th dynamic region, IDi is the dynamic region identifier number, Zi is the previously generated latent space embedding parameter, Pi is the generation path type matched by the dynamic region, and Ri is the set of layout constraint rules corresponding to the dynamic region. By establishing a unified mapping relationship between dynamic region parameters and template structure, subsequent printing tasks can directly call the corresponding template node to complete the dynamic region update. Based on the recorded spatial coordinate information of the dynamic region, a dynamic region coordinate mapping matrix constraint is constructed. The coordinate mapping matrix is ​​used to constrain the position of the content generated by the dynamic region in the template. The coordinate mapping parameters are represented as Mi={xi,yi,wi,hi}; where Mi represents the coordinate mapping parameters of the i-th dynamic region, xi and yi are the horizontal and vertical coordinates of the upper left corner of the dynamic region in the page, wi is the width of the dynamic region, and hi is the height of the dynamic region. The coordinate mapping matrix of a single dynamic area consists of four quantifiable parameters: the horizontal coordinate is the distance from the top left corner of the area to the left edge of the page, the vertical coordinate is the distance from the top left corner of the area to the top edge of the page, and the area width and height correspond to the horizontal and vertical dimensions of the dynamic area, respectively; by using the coordinate mapping matrix to lock the generation position, the defects of dynamic area offset and page misalignment are avoided; Specifically, dynamic area 1: horizontal 50, vertical 80, width 200, height 40; Dynamic area 2: 300mm horizontally, 80mm vertically, 150mm wide, and 150mm high; The position and size of the dynamic region are fixed by four quantization parameters to ensure that the generated result does not shift.

[0021] The set of layout constraint rules corresponding to the dynamic region includes at least color constraint rules and resolution constraint rules. The degree of color matching between the generated dynamic region and the original template layout is calculated using a color consistency evaluation index. The formula for calculating the color consistency index is as follows: ,in Let Ni be the color consistency score for the i-th dynamic region, and Ni be the number of color sampling points in the i-th dynamic region. Generate the color value of the image at the k-th sampling point for the i-th dynamic region. The reference color value corresponding to the i-th dynamic region template and the k-th sampling point is used to control the color consistency of the generated image of the dynamic region. By adding dynamic region resolution constraint rules to limit the resolution standard of the output image of the dynamic region, the resolution constraint rule is that the actual output resolution of the dynamic region generation result is greater than or equal to the minimum output resolution threshold preset by the template. Combined with color constraint rules and resolution constraint rules, the visual consistency between the dynamic region and the static region of the page is limited.

[0022] Furthermore, the specific method for adding a generation parameter stability evaluation mechanism based on the output quality statistics of historical printing tasks is as follows: Based on the output quality statistics of historical printing tasks, and by integrating data from four dimensions—edge consistency, color consistency, page offset, and generation time—a weighted stability score is calculated for the j-th combination of generation parameters. The stability score calculation formula is as follows: Where Sj is the stability score of the j-th group of generated parameters, Bj is the edge blending quality score, Qj is the color consistency score, Oj is the normalized result of page offset, Tj is the normalized result of generation time, and λ1, λ2, λ3, and λ4 are the weight coefficients corresponding to the four indicators of edge, color, offset, and time, respectively. The above formula is used to quantify the comprehensive stability level of each group of generated parameters in historical printing tasks. The generated parameter combinations are sorted according to their stability scores, and the parameter combination with the highest stability score is prioritized when performing dynamic region generation tasks. An offline statistical optimization parameter selection mechanism is adopted to improve the layout consistency and output stability of multiple printing batches while eliminating parameter drift problems caused by online continuous learning.

[0023] Specifically, taking a variable data printing page as an example, the page contains a static area, dynamic area 1, and dynamic area 2, where dynamic area 1 is a simple text area and dynamic area 2 is a complex graphic area; first, a template node is constructed for each dynamic area; the identifier number of dynamic area 1 is... ; Latent space embedding parameters: ; Generate path type Rendering paths for rules, set of layout constraint rules Includes center alignment and font style constraints, corresponding to the template node. ; Identification number of dynamic area 2 ; Latent space embedding parameters:

[0024] Generate path type Generate paths and layout constraint rule sets for latent space parameters. Includes edge blending and resolution constraints, corresponding to template nodes. ; Establish a coordinate mapping matrix, with the coordinate mapping parameters for dynamic region 1 as follows:

[0025] Coordinate mapping parameters for dynamic region 2:

[0026] Set consistency constraints and apply color consistency evaluation metrics. The matching threshold is set to 0.92, and the resolution constraint threshold is set to... Finally, a stability evaluation of the generated parameters is conducted, with weighting coefficients set. , Stability indices were calculated for the three sets of historical generation parameter combinations. ,get: The parameter combination with the highest stability score is selected first to complete the dynamic area generation, ensuring the consistency of the layout and output stability of multiple printing batches. By using the template construction method described above, dynamic area parameters and page constraints can be uniformly bound, ensuring that the dynamic area generation results accurately match the preset coordinate positions. At the same time, the generation parameters are optimized based on the stability evaluation mechanism, effectively reducing the risk of page offset and edge misalignment. This allows the dynamic and static areas to blend naturally and have a unified visual effect in multiple printing tasks. Furthermore, subsequent printing tasks can directly call the constructed template nodes without repeating page parsing and parameter configuration, significantly improving the task processing efficiency and output stability of batch variable data printing.

[0027] Furthermore, the specific method for completing partial page replacement is as follows: After the dynamic region is generated, the new dynamic region result is written to the corresponding dynamic region cache, and a partial page replacement relationship is established: P t+1 =P t -D t +D t+1 , where P t D represents the page result before the update. t D represents the dynamic region content before the update. t+1 P represents the new dynamic region content generated after the current task is created. t+1 This represents the updated page result. Since the system only performs partial replacement on the dynamic areas of the page, while the static and unchanged areas remain unchanged, it avoids the redundant calculation overhead caused by regenerating the entire page. At the same time, the system statistically analyzes the processing time, cache call count, and partial replacement success rate during the dynamic area update process, and dynamically adjusts the path scheduling priority corresponding to different dynamic areas based on the statistical results. This reduces unnecessary calls to the implicit space parameter generation path in low-complexity areas, thereby improving the page update efficiency and system resource utilization in batch variable data printing tasks. For text numbers, ordinary barcodes, QR codes, or low-complexity graphic areas, the corresponding dynamic area content is directly generated by calling the rule rendering path. For dynamic areas containing complex graphic structures, high-frequency style changes, or multi-element combination layouts, the implicit space parameters are called to drive the generation path to complete the dynamic area generation. During the dynamic area generation process, the system only allows the template node data corresponding to the current dynamic area to be updated. The static area cache data, layout structure relationships, layer hierarchy relationships, spatial coordinate mapping relationships, and the color constraint rules and resolution constraint rules established in step three remain unchanged, thereby ensuring that the layout stability of other areas of the page is not affected during the dynamic area replacement process.

[0028] Furthermore, the specific method for outputting the final printed page is as follows: After the dynamic area update process is completed, the coordinate mapping matrix constraint and the set of page constraint rules are called. Based on the above constraint rules, the dynamic area generation result and the static area cache result are combined into a page layout and the final printed page is output. During the page layout compositing process, the edge transitions, resolution matching, and color matching at the junctions of dynamic and static areas are corrected to ensure that the output appearance of the dynamic areas and the original template page remains consistent. At the same time, the static area reuse mechanism is used to eliminate the processing delay caused by repeated rendering of the entire page, thereby improving the page output efficiency of batch variable data printing tasks.

[0029] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0030] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A self-evolving variable data printing method based on latent space embedding and complexity hierarchy, characterized in that: Obtain the initial template for variable data printing, and perform layout parsing and area division into static and dynamic areas. Perform pre-rendering and caching processing on the static areas. A differential feature coding model is established based on the data type corresponding to the dynamic region. The original data of the dynamic region is compressed into latent space parameters and split into four independent features. The complexity of dynamic regions is evaluated based on multi-dimensional indicators, and the path type is adaptively switched by combining thresholds. Based on latent space parameters and path types, a dynamic region template node, coordinate mapping matrix, and layout constraint rule set are constructed and bound to multiple types of information. A generation parameter stability evaluation mechanism is added based on the output quality statistics of historical printing tasks. The parameter combination is selected and called according to the stability evaluation mechanism to complete the standardized construction of dynamic region generation template. When a new printing task is received, a template is generated based on the dynamic area to locate the target dynamic area and complete the partial replacement of the page. After the dynamic area update is completed, the dynamic area generation result and the static area cache result are combined and the final printed page is output based on the coordinate mapping matrix and the set of page layout constraint rules.

2. The self-evolving variable data printing method based on latent space embedding and complexity hierarchy according to claim 1, characterized in that: Based on the data type corresponding to the dynamic region, a differential feature coding model is established. The specific method for compressing the original data of the dynamic region into latent space parameters and splitting it into four independent features is as follows: The original data of each dynamic region are respectively encoded with features. The feature encoder is used to encode the original data of a single region into latent space parameters. The generated latent space parameters are subjected to multi-feature dimension structured decomposition. The decomposed latent space parameters are composed of four independent features: font style feature parameters, color distribution feature parameters, geometric layout feature parameters, and local texture feature parameters.

3. A self-evolutionary variable data printing method based on latent space embedding and complexity grading as claimed in claim 2, wherein: The specific method for evaluating the complexity of dynamic regions based on multi-dimensional indicators is as follows: A dynamic region complexity evaluation model is constructed, quantifying the dynamic region complexity from four dimensions: region area ratio, edge change degree, number of graphic and textual structures, and historical average generation time. A weighted calculation is used to obtain the comprehensive complexity score Ci for the i-th dynamic region. The comprehensive complexity calculation formula is: Ci = αAi / A page +βEi+γFi+δTi; where Ai is the area of ​​a single dynamic region, A page Ei represents the total area of ​​the entire page, Fi represents the number of graphic and textual structural objects within the area, Ti represents the historical average generation time of this dynamic area, and α, β, γ, and δ represent the weighting coefficients corresponding to the area percentage, edge complexity, number of graphic and textual objects, and historical generation time, respectively.

4. A self-evolutionary variable data printing method based on latent space embedding and complexity grading according to claim 3, characterized in that: The specific method for adaptively switching path types based on thresholds is as follows: A complexity switching threshold Cth is pre-configured, and the magnitude of the comprehensive complexity score and the switching threshold is used as the path switching criterion; when the comprehensive complexity of the dynamic region satisfies Ci < Cth, the system enables rule-based rendering path to directly generate the dynamic region; When the overall complexity of the dynamic region satisfies Ci≥Cth, the system switches to the latent space parameter-driven generation path to complete the dynamic region generation.

5. A self-evolutionary variable data printing method based on latent space embedding and complexity grading as claimed in claim 4, wherein: The specific method for constructing dynamic region template nodes bound to multiple types of information, coordinate mapping matrices, and a set of layout constraint rules is as follows: For each dynamic region, a dedicated template description structure is built, binding and summarizing the dynamic region identifier, latent space parameters, generated path type, and spatial coordinate information to form a dynamic region template node. The template node is represented by a set Ti={IDi,Zi,Pi,Ri}; where Ti is the template node corresponding to the i-th dynamic region, IDi is the dynamic region identifier number, Zi is the latent space embedding parameter generated beforehand, Pi is the generated path type matched by the dynamic region, and Ri is the set of layout constraint rules corresponding to the dynamic region. Based on the recorded spatial coordinate information of the dynamic region, a dynamic region coordinate mapping matrix constraint is constructed. The coordinate mapping matrix is ​​used to constrain the position of the content generated by the dynamic region in the template. The coordinate mapping parameters are represented as Mi={xi,yi,wi,hi}; where Mi represents the coordinate mapping parameters of the i-th dynamic region, xi and yi are the horizontal and vertical coordinates of the upper left corner of the dynamic region in the page, wi is the width of the dynamic region, and hi is the height of the dynamic region. The set of layout constraint rules corresponding to the dynamic region includes color constraint rules and resolution constraint rules. The degree of color matching between the generated dynamic region and the original template layout is calculated using a color consistency evaluation index. The formula for calculating the color consistency index is as follows: ,in Let Ni be the color consistency score for the i-th dynamic region, and Ni be the number of color sampling points in the i-th dynamic region. Generate the color value of the image at the k-th sampling point for the i-th dynamic region. The reference color value corresponding to the i-th dynamic region template and the k-th sampling point; By adding dynamic region resolution constraint rules to limit the resolution standard of the output image of the dynamic region, the resolution constraint rule is that the actual output resolution of the dynamic region generation result is greater than or equal to the minimum output resolution threshold preset by the template. Combined with color constraint rules and resolution constraint rules, the visual consistency between the dynamic region and the static region of the page is limited.

6. A self-evolutionary variable data printing method based on latent space embedding and complexity grading as claimed in claim 5, wherein: The specific method for adding a generation parameter stability evaluation mechanism based on the output quality statistics of historical printing tasks is as follows: Based on the output quality statistics of historical printing tasks, and by integrating data from four dimensions—edge consistency, color consistency, page offset, and generation time—a weighted stability score is calculated for the j-th combination of generation parameters. The stability score calculation formula is as follows: Where Sj is the stability score of the j-th generation parameter combination, Bj is the edge blending quality score, Qj is the color consistency score, Oj is the normalized result of the page offset, Tj is the normalized result of the generation time, and λ1, λ2, λ3, and λ4 are the weight coefficients corresponding to the four indicators of edge, color, offset, and time, respectively. The generated parameter combinations are sorted according to their stability scores, and the parameter combination with the highest stability score is selected first when performing the dynamic region generation task.

7. The self-evolving variable data printing method based on latent space embedding and complexity hierarchy according to claim 6, characterized in that: The specific method for completing partial page replacement is as follows: After the dynamic region is generated, the new dynamic region result is written to the corresponding dynamic region cache, and a partial page replacement relationship is established: P t+1 =P t -D t +D t+1 , where P t D represents the page result before the update. t D represents the dynamic region content before the update. t+1 P represents the new dynamic region content generated after the current task is created. t+1 This indicates the updated page result.

8. A self-evolutionary variable data printing method based on latent space embedding and complexity grading according to claim 7, characterized in that: The specific method for outputting the final printed page is as follows: After the dynamic area update process is completed, the coordinate mapping matrix constraint and the set of layout constraint rules are called. Based on the above constraint rules, the dynamic area generation result and the static area cache result are combined into a layout and the final printed page is output.

9. A computer device, comprising: It includes a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform the steps of the method as described in any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that: The device stores a computer program that, when executed by a processor, causes the processor to perform the steps of the method as described in any one of claims 1 to 8.

Images