Package design language model training method

Abstract

The invention relates to the technical field of package design language data processing, in particular to a training method of a package design language model. The method comprises the following steps: acquiring a package design original data set; performing data preprocessing on the package design original data set to obtain a standard package design data set; performing package design information extraction on the standard package design data set to obtain package design information; and carrying out package design term analysis on the package design information to obtain package design term information. A data processing technology, a mode recognition technology, a semantic analysis technology and a machine learning technology are adopted; according to the method, packaging structure component integration evaluation is carried out on the packaging design language graphic and text information, and correlation mapping is carried out on the packaging design function and the packaging application scene, so that the identification evaluation accuracy of the packaging design language model is improved, and the packaging design function and the application scene are presented.

Description

Technical Field

[0001] The present invention relates to the technical field of packaging design language data processing, and in particular to a training method for a packaging design language model. Background Art

[0002] The packaging design language includes visual element design features, structural morphological features, and design-specific terms. These feature information presents the functionality and applicable scenarios of the packaging design. In the past, packaging design was mainly constructed by designers through subjective design, lacking objective analysis of the packaging design language features. With the application of semantic analysis technology in the field of packaging design, it is possible to obtain packaging design terminology data for design language training, and perform word segmentation on the packaging design terminology data to extract keywords, phrases, or industry terms, and finally construct a packaging design language model for packaging design language recognition. However, the existing training methods for packaging design language models fail to evaluate the integrity of packaging structural components based on the graphic information of the packaging design language, and fail to associate and map the packaging design function with the packaging application scenario, resulting in a low recognition and evaluation accuracy of the packaging design language model, making it difficult to present the functions and application scenarios of the packaging design. Summary of the invention

[0003] Based on this, it is necessary to provide a training method for a packaging design language model to solve at least one of the above technical problems.

[0004] To achieve the above object, a method for training a packaging design language model is provided, the method comprising the following steps:

[0005] Step S1: obtaining an original packaging design data set; performing data preprocessing on the original packaging design data set to obtain a standard packaging design data set; extracting packaging design information from the standard packaging design data set to obtain packaging design information; performing packaging design terminology analysis on the packaging design information to obtain packaging design terminology information;

[0006] Step S2: performing design element recognition on the packaging design term information to obtain packaging design element data; performing packaging graphic information recognition on the packaging design element data to generate packaging design graphic information; performing visual feature scanning based on the packaging design graphic information to generate packaging design visual features;

[0007] Step S3: dividing the packaging structure components according to the packaging design visual features to obtain packaging structure design components; classifying the packaging structure design components into design structure types to obtain packaging design structure types; evaluating the packaging structure integrity of the packaging design structure types to obtain structure integrity evaluation data; mapping the packaging design structure types to packaging design functions based on the structure integrity evaluation data to generate packaging design function data; matching the packaging structure design components to packaging application scenarios using the packaging design function data to obtain packaging application scenario data;

[0008] Step S4: construct a packaging design language model based on the structural integrity evaluation data, packaging design function data and packaging application scenario data to obtain a packaging design language model; evaluate the packaging design language features of the original packaging design data set through the packaging design language model to generate packaging design language features; visualize the packaging design language features to obtain a packaging design language feature report.

[0009] The present invention can provide an accurate and consistent data basis for subsequent analysis by obtaining the original data set of packaging design and performing data preprocessing; the extraction of packaging design information enables key design elements to be directly obtained from the data set, and the packaging design term analysis further refines this information and clarifies the professional terms used in the design; this series of operations ensures the accuracy and professionalism of the design information, and lays a solid foundation for the identification and further analysis of design elements. The design element identification of the packaging design term information provides a clear definition and classification for the specific components of the design; the identification of packaging graphic information and the scanning of visual features enable the effective identification and analysis of graphic and text elements in the design, which helps to deeply understand the visual effects and information conveyed by the design, thereby enhancing the understanding and analysis capabilities of the visual elements of the packaging design. The packaging structure component division based on the visual features of the packaging design clarifies the structural composition in the design, and the design structure type classification systematically classifies these components; the acquisition of structural integrity evaluation data provides quantitative indicators for evaluating the integrity and coordination of the packaging design; the packaging design function mapping and application scenario matching further link the structure type with the actual function and usage scenario to ensure the practicality and adaptability of the design. Through the comprehensive analysis of structural integrity evaluation data, packaging design function data and packaging application scenario data, the constructed packaging design language model can fully reflect the internal logic and external performance of the design. Through the language feature evaluation of the original data set of packaging design by the model, the generated packaging design language features can accurately describe the core attributes of the design; feature visualization further converts these abstract features into an intuitive report form, making the language features of the design easy to understand. Therefore, the present invention uses data processing technology, pattern recognition technology, semantic analysis technology and machine learning technology to realize the evaluation of the integrity of packaging structure components of packaging design language graphic information, and realizes the association mapping of packaging design functions with packaging application scenarios, thereby improving the recognition and evaluation accuracy of the packaging design language model, so as to present the functions and application scenarios of packaging design.

[0010] Preferably, step S1 comprises the following steps:

[0011] Step S11: Obtaining the original packaging design data set;

[0012] Step S12: performing data cleaning on the original packaging design data set to obtain a packaging design cleaned data set; performing data standardization on the packaging design cleaned data set to obtain a standard packaging design data set;

[0013] Step S13: marking the standard packaging design data set with packaging design features to obtain packaging design feature information; extracting packaging design information from the standard packaging design data set based on the packaging design feature information to obtain packaging design information;

[0014] Step S14: performing professional terminology recognition on the packaging design information to obtain a set of professional terminology for packaging design; performing packaging design industry terminology matching on the set of professional terminology for packaging design to generate industry terminology matching data; performing industry semantic association on the industry terminology matching data to obtain packaging design terminology information.

[0015] The present invention obtains the original data set of packaging design, which is the starting point of the entire analysis process, provides a basis for subsequent data cleaning and standardization, ensures that the original data for analysis is complete, and thus ensures the purity of the data in the subsequent steps and the reliability of the analysis results; performs data cleaning on the original data set of packaging design, removes invalid or erroneous data, and improves the quality of the data; subsequently, the data standardization process unifies the data format, provides consistency and comparability for further feature marking and information extraction, and improves the accuracy and efficiency of subsequent analysis; performs packaging design feature marking on the standard packaging design data set, annotates the key design features for the data set, and can identify and distinguish different design elements; extracts packaging design information from the data set based on these feature information, directly reflects the core elements of the design, and provides accurate data support for in-depth analysis of design elements and understanding of design intent; performs professional terminology recognition on packaging design information, and the collection concentrates professional vocabulary in the design field, which can professionally describe and understand design features; performs packaging design industry terminology matching on these professional terminology sets, and ensures the accuracy of terminology use and industry versatility; performs industry semantic association on industry terminology matching data, clarifies the specific meaning of the terminology, and reveals the relevance and application background in the industry.

[0016] Preferably, step S2 comprises the following steps:

[0017] Step S21: performing design term semantic decomposition on the packaging design term information to obtain design term decomposition data; performing design element term recognition on the design term decomposition data to generate design element term data; performing term feature mapping on the design element term data to obtain packaging design element data;

[0018] Step S22: performing design element structured coding on the packaging design element data to obtain design element coded data; performing attribute extraction on the design element coded data to obtain design element attribute information;

[0019] Step S23: performing image segmentation on the design element attribute information to generate packaging image and text separation data; performing design graphic recognition on the packaging image and text separation data to obtain packaging design graphic features; performing design text recognition on the packaging image and text separation data to obtain packaging design text features;

[0020] Step S24: integrating the packaging design graphic features and the packaging design text features into packaging graphic information to generate packaging design graphic information;

[0021] Step S25: Scan the visual features based on the packaging design graphic information to generate the packaging design visual features.

[0022] The present invention performs semantic decomposition of design terminology on packaging design terminology information, deepens the understanding of the meaning of terms, and provides a basis for further term identification; generates design element terminology data based on the decomposed data, which makes the terms of design elements clearer, and the term feature mapping further associates these terms with specific design elements; performs structured coding of design elements on packaging design element data, so that design elements are recorded and classified in a standardized form for easy management and retrieval; extracts design element attribute information from the coded data, provides a detailed description for understanding the specific characteristics and functions of design elements, and enhances the recognition and understanding of design element attributes; performs image segmentation on design element attribute information, which provides a clear description for the recognition of graphics and text. boundaries; identifying the graphic features of packaging design from the separated data can clarify the graphic elements in the design; identifying the text features of packaging design from the separated data makes it possible to understand the text elements in the design; the execution of these two steps enables the graphic and text features to be independently and accurately identified and analyzed; integrating the graphic features of packaging design and the text features of packaging design into the packaging graphic information can fully understand the visual elements of packaging design, ensuring that the visual information of the design remains complete and coordinated during analysis and application; performing visual feature scanning based on the packaging design graphic information and a detailed description of the visual elements of the design, so that the visual features of the design can be quantified and evaluated, providing important visual data support for subsequent design analysis and optimization.

[0023] Preferably, step S25 comprises the following steps:

[0024] Step S251: Scan the packaging design graphic information with high resolution to obtain graphic image scan information; enhance the graphic details of the graphic image scan information to generate a graphic detail enhanced image;

[0025] Step S252: performing color correction on the image with enhanced detail to generate an image with color correction; performing contrast enhancement on the image with color correction to generate an image with contrast enhancement;

[0026] Step S253: performing packaging visual design edge detection on the image with enhanced contrast between text and images to obtain packaging edge visual information; performing texture feature recognition on the packaging edge visual information to generate packaging texture visual information;

[0027] Step S254: performing spatial shape recognition on the packaging texture visual information to obtain spatial shape visual information; performing visual hierarchy analysis on the spatial shape visual information to obtain visual hierarchy information; and merging packaging design visual features on the packaging design graphic information based on the visual hierarchy information to generate packaging design visual features.

[0028] The present invention performs high-resolution scanning on packaging design graphic information, improves the clarity of the image, and provides high-quality basic data for subsequent detail enhancement; the generation of graphic detail enhanced images makes the details in the image more obvious, enhances the readability of the image and the accuracy of analysis; color correction is performed on the graphic detail enhanced image, and the generated graphic color corrected image ensures the authenticity and consistency of color; contrast enhancement further optimizes the visual effect of the image, making the graphic contrast enhanced image more prominent in visual recognition; packaging visual design edge detection is performed on the graphic contrast enhanced image, providing accurate data for identifying the boundaries of design elements; texture feature recognition is performed on the packaging edge visual information to clarify the packaging texture visual information; spatial shape recognition is performed on the packaging texture visual information, providing important data for understanding the three-dimensional structure of the packaging; visual hierarchy analysis is performed on the spatial shape visual information to reveal the visual layout information of the design; visual feature merging of the packaging design graphic information based on the visual hierarchy information provides comprehensive data support for comprehensively evaluating the visual appeal and functionality of the packaging design.

[0029] Preferably, step S3 comprises the following steps:

[0030] Step S31: performing packaging structure feature recognition on the packaging design visual features to obtain packaging structure feature information; performing structural contour detection on the packaging structure feature information to generate packaging structure contour data;

[0031] Step S32: mapping the packaging structure outline data to structural components to obtain structural component mapping data; dividing the packaging structure components based on the structural component mapping data to obtain packaging structure design components;

[0032] Step S33: performing component structure similarity measurement on the packaging structure design components to generate component structure similarity measurement values; performing design structure clustering on the packaging structure design components according to the component structure similarity measurement values ​​to obtain design structure clustering data; performing design structure type classification based on the design structure clustering data to obtain the packaging design structure type;

[0033] Step S34: performing packaging structural integrity evaluation on the packaging design structure type to obtain structural integrity evaluation data;

[0034] Step S35: mapping the packaging design structure type to packaging design function based on the structural integrity evaluation data to generate packaging design function data;

[0035] Step S36: Match the packaging structure design components with the packaging application scenarios using the packaging design function data to obtain packaging application scenario data.

[0036] The present invention identifies packaging structural features based on packaging design visual features, providing key data for analyzing the physical structure of the packaging; performs structural contour detection based on these feature information, which can accurately capture the structural details of the packaging; performs structural component mapping on the packaging structural contour data, clarifying the relationship between the various structural parts; the packaging structural component division based on this data provides specific component units for further analysis and design optimization; performs component structural similarity measurement on the packaging structural design components, providing quantitative indicators for identifying similar or repeated components; clusters the design structures based on these measurement values, and classifies similar components, which can identify and standardize design patterns; performs packaging structural integrity evaluation on the packaging design structure type, providing a basis for evaluating the overall stability and coordination of the packaging; performs packaging design function mapping on the packaging design structure type based on the structural integrity evaluation data, providing clear guidance for the functional realization of the design; performs packaging application scenario matching on the packaging structural design components through the packaging design function data, ensuring that the design components can adapt to specific usage environments, thereby improving the practicality of the packaging.

[0037] Preferably, step S34 includes the following steps:

[0038] Step S341: performing topological relationship mapping on the packaging design structure type data to obtain structure topological relationship data; performing topological path continuity verification on the structure topological relationship data to obtain topological path verification data;

[0039] Step S342: performing structural consistency detection on the packaging design structure type data through the topological path verification data to generate a structural consistency detection result; performing consistency degree measurement on the structural consistency detection result to obtain structural consistency degree data;

[0040] Step S343: performing structural space alignment accuracy identification on the topological path verification data to obtain the structural space alignment accuracy; performing structural geometric compatibility detection on the packaging design structure type data based on the structural space alignment accuracy to generate geometric compatibility data;

[0041] Step S344: performing a structural component connection point strength test on the topology path verification data to obtain a component connection point strength value; performing a structural component coordination test on the packaging design structure type data according to the component connection point strength value to generate structural component coordination data;

[0042] Step S345: Integrate the structural coherence degree data, geometric compatibility data and structural component coordination data to obtain structural integrity evaluation data.

[0043] The present invention maps the topological relationship of packaging design structure type data, which can reveal the spatial and logical connections between various parts in the packaging design; the topological path continuity verification further ensures the integrity and continuity of these connections, and the generated topological path verification data provides accurate path information for subsequent structural consistency detection; the packaging design structure type data is tested for structural consistency through the topological path verification data, which directly reflects the connectivity state of the packaging design; the consistency degree of these results is calculated, which can quantitatively evaluate the connectivity and consistency of the packaging design; the topological path verification data is identified for the structural space alignment accuracy, which provides an accurate measurement for the spatial alignment of the packaging design; the packaging design structure type data is tested for structural geometric compatibility based on this accuracy, which ensures the compatibility and coordination of the packaging design in the geometric structure; the topological path verification data is tested for the strength of the structural component connection points, which provides key data for evaluating the stability and reliability of the connection points in the packaging design; the packaging design structure type data is tested for structural component coordination according to these strength values, which ensures the coordination between the components in the packaging design; the structural consistency degree data, the geometric compatibility data and the structural component coordination data are integrated to obtain the structural integrity evaluation data.

[0044] Preferably, step S35 includes the following steps:

[0045] Step S351: extracting packaging design functional requirements from the packaging design structure type to obtain packaging functional requirement data; matching the packaging functional requirement data with design parameters to generate functional requirement design parameters;

[0046] Step S352: Score the structural integrity evaluation data to obtain an integrity score value; set mapping rules for the integrity score value according to the functional requirement design parameters to obtain a packaging function mapping rule set;

[0047] Step S353: performing a structural function matching degree detection on the packaging design structure type based on the packaging function mapping rule set to generate a structural function matching degree; performing a packaging design function integrity check on the structural function matching degree to obtain packaging function integrity check data;

[0048] Step S354: Based on the packaging function integrity verification data, the packaging design structure type is presented with packaging structure functions to generate packaging structure function data; the packaging structure function data is mapped with packaging design functions to generate packaging design function data.

[0049] The present invention extracts packaging design functional requirements for packaging design structure types, and clarifies the specific functions that the packaging design needs to meet; the design parameters correspond to the generated functional requirement design parameters, and clarify the functional standards of the design; the structural integrity evaluation data is scored for the degree of integrity, and a quantitative evaluation is provided for the integrity of the packaging design; the mapping rules are set for the degree of integrity score values ​​according to the functional requirement design parameters, and the obtained packaging function mapping rule set provides a standardized evaluation basis for the matching of structure and function; based on the packaging function mapping rule set, the packaging design structure type is tested for the structural function matching degree, and direct data is provided for evaluating the fit between the structure and the functional requirements; the packaging design function integrity check is performed on the structural function matching degree, and the integrity and effectiveness of the design in functional realization are ensured; based on the packaging function integrity check data, the packaging design structure type is presented with the packaging structure function, and a specific structured description is provided for the functional realization of the design; the packaging structure function mapping is performed on the packaging structure function data, and the structural characteristics are closely associated with the functional requirements, and accurate data support is provided for design optimization and functional enhancement.

[0050] Preferably, step S36 includes the following steps:

[0051] Step S361: performing functional pattern recognition on the packaging design functional data to obtain the packaging design functional pattern; performing clustering processing on the packaging design functional pattern to obtain packaging design functional clustering data; performing packaging design functional type classification on the packaging design functional clustering data to generate packaging design functional types;

[0052] Step S362: according to the packaging design function type, the packaging structure design components are associated with functional scenarios to obtain functional application scenario associated data; the functional scenario feature recognition is performed on the functional application scenario associated data to generate packaging functional scenario features;

[0053] Step S363: Based on the packaging functional scenario characteristics, the functional application scenario associated data is evaluated for application scenario adaptability to obtain a functional scenario adaptability score; the packaging structure design components are ranked by packaging component functional scenario scores according to the functional scenario adaptability score to generate component scenario score ranking information;

[0054] Step S364: Screen application scenario components according to component scenario score sorting information to obtain an application scenario component list; match packaging structure design components with packaging application scenarios through the application scenario component list to obtain packaging application scenario data.

[0055] The present invention performs functional pattern recognition on packaging design functional data, providing a patterned understanding for understanding the distribution of functions in packaging design; the packaging design function clustering data obtained after clustering processing classifies similar functions for easy management and analysis; these clustering data are classified to further clarify the characteristics and uses of different functions; functional scenarios are associated with packaging structure design components according to the packaging design function type, so that the functional requirements are matched with specific structural components; functional scenario feature recognition extracts packaging function scenario features from the associated data, and can understand the performance of different functions in actual applications; based on the packaging function scenario features, the application scenario adaptability of the function application scenario associated data is evaluated, and the fitness of each function in a specific scenario is quantified; the packaging structure design components are sorted according to these scores, providing a basis for selecting the most suitable components; application scenario components are screened according to the component scenario score sorting information, and a group of selected component options are provided for specific scenarios; the packaging structure design components are matched through this list, ensuring that the design components can meet the needs of specific application scenarios, thereby improving the applicability and efficiency of the design.

[0056] Preferably, step S4 comprises the following steps:

[0057] Step S41: constructing a packaging design language model according to the structural integrity evaluation data, the packaging design function data and the packaging application scenario data to obtain a packaging design language pre-model; training the packaging design language pre-model based on the structural integrity evaluation data, the packaging design function data and the packaging application scenario data to obtain a packaging design language training model;

[0058] Step S42: performing model cross-validation evaluation on the packaging design language training model to obtain model evaluation data; adjusting model parameters of the packaging design language training model using the model evaluation data to obtain a packaging design language model;

[0059] Step S43: evaluating the packaging design language features of the original packaging design data set through the packaging design language model to generate packaging design language features; and visualizing the packaging design language features to obtain a packaging design language feature report.

[0060] The present invention combines structural integrity evaluation data, packaging design function data and packaging application scenario data to construct a packaging design language pre-model. This step integrates the key elements of packaging design and provides comprehensive data support for the model. Subsequently, the pre-model is trained, and the obtained packaging design language training model can learn and understand the key features and patterns in packaging design, providing a basis for subsequent evaluation and optimization. The packaging design language training model is subjected to model cross-validation evaluation, and the obtained model evaluation data provides quantitative feedback on the model performance. Based on these evaluation data, the model parameters are adjusted, and the optimized packaging design language model can more accurately reflect the actual needs and characteristics of packaging design, thereby improving the prediction ability and reliability of the model. The original packaging design data set is evaluated through the packaging design language model, so that the core attributes of the design can be accurately captured. These features are visualized, and the obtained packaging design language feature report displays the key information of the design in an intuitive way, which is convenient for designers and decision makers to understand and apply, thereby improving the efficiency and quality of design decisions.

[0061] Preferably, step S43 includes the following steps:

[0062] Step S431: inputting the original packaging design data set into the packaging design language model to perform packaging design language recognition to obtain packaging design language recognition data;

[0063] Step S432: Based on the packaging design language model, the packaging design language recognition data is evaluated for structural integrity to obtain a packaging design integrity evaluation result; based on the packaging design language model, the packaging design function is evaluated for the packaging design language recognition data to obtain a packaging design function feature result; based on the packaging design language model, the packaging design language recognition data is evaluated for packaging application scenarios to obtain packaging application scenario feature results;

[0064] Step S433: integrating the packaging design integrity evaluation results, the packaging design functional feature results and the packaging application scenario feature results into packaging design language features to generate packaging design language features;

[0065] Step S434: graphically process the packaging design language features to obtain a packaging design language graph; and compile a visual report on the packaging design language graph to obtain a packaging design language feature report.

[0066] The present invention inputs the original data set of packaging design into the packaging design language model to perform packaging design language recognition. The obtained packaging design language recognition data directly reflects the language characteristics of the design elements, so that the language attributes of the design can be clarified, and an accurate data basis is provided for subsequent evaluation and analysis; based on the packaging design language model, the structural integrity, function and application scenario of the recognition data are evaluated, and packaging design integrity evaluation results, packaging design functional feature results and packaging application scenario feature results are obtained respectively, and these results provide quantitative evaluation for the three key dimensions of packaging design, ensuring the comprehensiveness and practicality of the design; the packaging design integrity evaluation results, functional feature results and application scenario feature results are integrated, and the generated packaging design language features comprehensively describe the comprehensive characteristics of the packaging design; the integration effect provides a multi-dimensional feature view for the design; the packaging design language features are processed in a graphical manner, and the obtained packaging design language chart intuitively displays the key features of the design in a graphical manner, which is convenient for comparison and analysis; further, these charts are compiled into a visual report, and the obtained packaging design language feature report provides decision makers with easy-to-understand visual materials, thereby improving the efficiency of information transmission and the quality of decision-making.

[0067] The present invention also provides a method for constructing a packaging design language model, the method comprising the following steps:

[0068] Step S1: obtaining an original packaging design data set; performing data preprocessing on the original packaging design data set to obtain a standard packaging design data set; extracting packaging design information from the standard packaging design data set to obtain packaging design information; performing packaging design terminology analysis on the packaging design information to obtain packaging design terminology information;

[0069] Step S2: performing design element recognition on the packaging design term information to obtain packaging design element data; performing packaging graphic information recognition on the packaging design element data to generate packaging design graphic information; performing visual feature scanning based on the packaging design graphic information to generate packaging design visual features;

[0070] Step S3: dividing the packaging structure components according to the packaging design visual features to obtain packaging structure design components; classifying the packaging structure design components into design structure types to obtain packaging design structure types; evaluating the packaging structure integrity of the packaging design structure types to obtain structure integrity evaluation data; mapping the packaging design structure types to packaging design functions based on the structure integrity evaluation data to generate packaging design function data; matching the packaging structure design components to packaging application scenarios using the packaging design function data to obtain packaging application scenario data;

[0071] Step S4: construct a packaging design language model based on the structural integrity evaluation data, packaging design function data and packaging application scenario data to obtain a packaging design language model; evaluate the packaging design language features of the original packaging design data set through the packaging design language model to generate packaging design language features; visualize the packaging design language features to obtain a packaging design language feature report. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Figure 1 A flowchart of the steps of a training method for a packaging design language model;

[0073] Figure 2 for Figure 1 Detailed implementation steps of step S4 in FIG.

[0074] Figure 3 for Figure 2 Detailed implementation steps of step S43;

[0075] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0076] The technical method of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are 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 technicians in this field without creative work are within the scope of protection of the present invention.

[0077] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0078] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0079] To achieve this, please refer to Figures 1 to 3 , a method for training a packaging design language model, the method comprising the following steps:

[0080] Step S1: obtaining an original packaging design data set; performing data preprocessing on the original packaging design data set to obtain a standard packaging design data set; extracting packaging design information from the standard packaging design data set to obtain packaging design information; performing packaging design terminology analysis on the packaging design information to obtain packaging design terminology information;

[0081] Step S2: performing design element recognition on the packaging design term information to obtain packaging design element data; performing packaging graphic information recognition on the packaging design element data to generate packaging design graphic information; performing visual feature scanning based on the packaging design graphic information to generate packaging design visual features;

[0082] Step S3: dividing the packaging structure components according to the packaging design visual features to obtain packaging structure design components; classifying the packaging structure design components into design structure types to obtain packaging design structure types; evaluating the packaging structure integrity of the packaging design structure types to obtain structure integrity evaluation data; mapping the packaging design structure types to packaging design functions based on the structure integrity evaluation data to generate packaging design function data; matching the packaging structure design components to packaging application scenarios using the packaging design function data to obtain packaging application scenario data;

[0083] Step S4: construct a packaging design language model based on the structural integrity evaluation data, packaging design function data and packaging application scenario data to obtain a packaging design language model; evaluate the packaging design language features of the original packaging design data set through the packaging design language model to generate packaging design language features; visualize the packaging design language features to obtain a packaging design language feature report.

[0084] The present invention can provide an accurate and consistent data basis for subsequent analysis by obtaining the original data set of packaging design and performing data preprocessing; the extraction of packaging design information enables key design elements to be directly obtained from the data set, and the packaging design term analysis further refines this information and clarifies the professional terms used in the design; this series of operations ensures the accuracy and professionalism of the design information, and lays a solid foundation for the identification and further analysis of design elements. The design element identification of the packaging design term information provides a clear definition and classification for the specific components of the design; the identification of packaging graphic information and the scanning of visual features enable the effective identification and analysis of graphic and text elements in the design, which helps to deeply understand the visual effects and information conveyed by the design, thereby enhancing the understanding and analysis capabilities of the visual elements of the packaging design. The packaging structure component division based on the visual features of the packaging design clarifies the structural composition in the design, and the design structure type classification systematically classifies these components; the acquisition of structural integrity evaluation data provides quantitative indicators for evaluating the integrity and coordination of the packaging design; the packaging design function mapping and application scenario matching further link the structure type with the actual function and usage scenario to ensure the practicality and adaptability of the design. Through the comprehensive analysis of structural integrity evaluation data, packaging design function data and packaging application scenario data, the constructed packaging design language model can fully reflect the internal logic and external performance of the design. Through the language feature evaluation of the original data set of packaging design by the model, the generated packaging design language features can accurately describe the core attributes of the design; feature visualization further converts these abstract features into an intuitive report form, making the language features of the design easy to understand. Therefore, the present invention uses data processing technology, pattern recognition technology, semantic analysis technology and machine learning technology to realize the evaluation of the integrity of packaging structure components of packaging design language graphic information, and realizes the association mapping of packaging design functions with packaging application scenarios, thereby improving the recognition and evaluation accuracy of the packaging design language model, so as to present the functions and application scenarios of packaging design.

[0085] In the embodiment of the present invention, reference Figure 1 FIG. 1 is a schematic diagram of a step flow chart of a method for training a packaging design language model of the present invention. In this example, the method for training a packaging design language model includes the following steps:

[0086] Step S1: obtaining an original packaging design data set; performing data preprocessing on the original packaging design data set to obtain a standard packaging design data set; extracting packaging design information from the standard packaging design data set to obtain packaging design information; performing packaging design terminology analysis on the packaging design information to obtain packaging design terminology information;

[0087] In the embodiment of the present invention, a large amount of packaging images and corresponding label data are obtained through open source data sets or self-collection, and these data will serve as the basis of the original data set of packaging design; for example, image data related to packaging design is downloaded from a public data set platform to finally obtain the original data set of packaging design; data preprocessing is performed on the original data set of packaging design to obtain a standard packaging design data set; data preprocessing includes operations such as image reading, resizing, and normalization, and specifically uses Python language in combination with OpenCV library to read and resize images, and uniformly adjusts the images to a fixed size, such as 128x128 pixels; at the same time, the Sklearn library is used to split the data set, and the data set is divided into a training set and a test set, usually 80% training set and 20% test set. The data set is divided according to the proportion of the data set; packaging design information is extracted from the standard packaging design data set to obtain packaging design information; information extraction is achieved through a deep learning model, such as building a convolutional neural network (CNN) model, using TensorFlow and Keras frameworks to build the model, the model contains multiple convolutional layers, pooling layers and fully connected layers to extract key features in packaging design; packaging design terminology analysis is performed on the packaging design information to obtain packaging design terminology information; terminology analysis is achieved through natural language processing technology, such as using the NELL system to automatically extract structured knowledge from the descriptive text of packaging design, including professional terms such as packaging materials, structure, and printing; in addition, machine learning algorithms are used to classify and cluster the terms to further understand various aspects of packaging design.

[0088] Step S2: performing design element recognition on the packaging design term information to obtain packaging design element data; performing packaging graphic information recognition on the packaging design element data to generate packaging design graphic information; performing visual feature scanning based on the packaging design graphic information to generate packaging design visual features;

[0089] In an embodiment of the present invention, design elements are identified on packaging design terminology information to obtain packaging design element data; computer vision technology, especially deep learning frameworks TensorFlow and Keras, is used to build a convolutional neural network (CNN) model to identify elements in packaging design; for example, a CNN model named Packaging Element Recognition is created, which extracts features of design elements through multiple convolutional layers and pooling layers, and outputs recognition results through a fully connected layer. The specific implementation operation is as follows: first, Python language and TensorFlow library are used to build the model, and a CNN structure containing multiple convolutional layers and pooling layers is defined; the input of the model is the image data corresponding to the packaging design term information, and the output is the identified design elements; the model contains a tf.keras.layers.Conv2D convolutional layer for extracting local features in the image, and then the feature dimension is reduced through the tf.keras.layers.MaxPooling2D pooling layer, and finally the features are mapped to the category of the design element through the tf.keras.layers.Flatten and tf.keras.layers.Dense layers; the packaging design element data is used to identify the packaging graphic information and generate the packaging design graphic information; the natural language processing (NLP) technology, especially the NELL system, is used to extract text information from the packaging design element data, where the NELL system It can automatically identify and extract structured knowledge, including text content and related attributes in packaging design; specifically, through the NELL system, text analysis is performed on each identified design element to extract text information around the element, such as brand name, product description, etc., which is organized into a structured graphic information data set; visual feature scanning is performed based on the packaging design graphic information. Specifically, the SIFT (scale-invariant feature transform) feature detection algorithm in the OpenCV library is used to detect and describe feature points of the image part in the packaging design graphic information, where these feature points and descriptors can represent the visual features of the image and are used for subsequent image matching and recognition tasks; such as setting the parameters of the SIFT algorithm, such as nfeatures, nOctaveLayers, contrastThreshold, etc., to adapt to the specific needs of packaging design, and extract key visual features from the packaging design graphic information to generate packaging design visual features.

[0090] Step S3: dividing the packaging structure components according to the packaging design visual features to obtain packaging structure design components; classifying the packaging structure design components into design structure types to obtain packaging design structure types; evaluating the packaging structure integrity of the packaging design structure types to obtain structure integrity evaluation data; mapping the packaging design structure types to packaging design functions based on the structure integrity evaluation data to generate packaging design function data; matching the packaging structure design components to packaging application scenarios using the packaging design function data to obtain packaging application scenario data;

[0091] In the embodiment of the present invention, the packaging structure components are divided according to the visual features of the packaging design to obtain the packaging structure design components; the visual features of the packaging design are analyzed by image processing technology, the packaging image is input into the image segmentation algorithm, and the different structural components (such as the box body, the cover, the bottom, etc.) in the image are separated by using a threshold-based segmentation method or an edge detection algorithm (such as Canny edge detection); after processing, the generated structural components are marked as "box body component", "cover component", etc., for subsequent analysis and classification; the packaging structure design components are classified into design structure types, and the divided structural components are analyzed by feature extraction technology; shape descriptors (such as Hu moments) are specifically used to extract the geometric features of each component, and these features are input into a classification algorithm, such as K nearest neighbor (KNN) or support vector machine (SVM), to classify the components; by training the model, different structural components are classified into types such as "folding carton", "corrugated carton", etc.; the packaging structure integrity of the packaging design structure type is evaluated; specifically, the finite element analysis (FEA) technology is used to evaluate the mechanical properties of different structural types. The design geometry model is evaluated and imported into finite element analysis software (such as ANSYS), material properties and boundary conditions are set, and static analysis is performed; the integrity of the structure is evaluated through the calculated stress, deformation and other data, and the structural integrity evaluation data is generated; the packaging design structure type is mapped to the packaging design function based on the structural integrity evaluation data, and specifically, the structural integrity evaluation results are associated with the functional requirements using design rules; for example, a mapping table is created to associate the "corrugated box" structure with its function (such as "strong load-bearing capacity" and "shockproof"); the specific function of each structure type is determined by analyzing the evaluation data, and the packaging design function data is generated; the packaging structure design components are matched with the packaging application scenario through the packaging design function data, and the packaging design function data is matched with the actual application scenario using a rule-based matching algorithm, and an application scenario database is established to record the packaging requirements of different commodities and the corresponding packaging design functions; the most suitable packaging application scenario is quickly matched by querying the function data, such as matching the "strong shockproof function" packaging design with the transportation application scenario of fragile commodities, thereby obtaining the packaging application scenario data.

[0092] Step S4: construct a packaging design language model based on the structural integrity evaluation data, packaging design function data and packaging application scenario data to obtain a packaging design language model; evaluate the packaging design language features of the original packaging design data set through the packaging design language model to generate packaging design language features; visualize the packaging design language features to obtain a packaging design language feature report.

[0093] In an embodiment of the present invention, a packaging design language model is constructed based on structural integrity evaluation data, packaging design function data and packaging application scenario data to obtain a packaging design language model; specifically, a machine learning technology is used to combine the structure, function and application scenario data of the packaging design to construct a comprehensive language model; the specific operation is to integrate the structural integrity evaluation data, packaging design function data and packaging application scenario data into a feature data set, and then use the data set to train a classification model, such as a random forest or a gradient boosting machine (GBM), to construct a packaging design language model; the structural integrity evaluation data, packaging design function data and packaging application scenario data are merged into a data set, which contains multiple features, such as structural stability, load-bearing capacity, environmental protection, etc.; then, a random forest model is trained using a Python machine learning library, such as scikit-learn, with the input being the feature data set and the output being the classification label of the packaging design; through model training, a packaging design language model that can predict the classification of the packaging design according to the input features is obtained. The packaging design language features of the original packaging design data set are evaluated through the packaging design language model, and each packaging design in the original data set is evaluated using the trained packaging design language model; the model will output the language features of each design based on the input design image and related data, such as feature labels such as "environmentally friendly", "portable", and "strongly protective", to form a packaging design language feature data set; the packaging design language features are visualized, and the feature labels in the packaging design language feature data set are statistically analyzed, and the distribution of different features in all designs is displayed using bar charts or heat maps; a packaging design language feature report is generated through the visualization results, which details the application frequency and importance of each feature in the design.

[0094] Preferably, step S1 comprises the following steps:

[0095] Step S11: Obtaining the original packaging design data set;

[0096] Step S12: performing data cleaning on the original packaging design data set to obtain a packaging design cleaned data set; performing data standardization on the packaging design cleaned data set to obtain a standard packaging design data set;

[0097] Step S13: marking the standard packaging design data set with packaging design features to obtain packaging design feature information; extracting packaging design information from the standard packaging design data set based on the packaging design feature information to obtain packaging design information;

[0098] Step S14: performing professional terminology recognition on the packaging design information to obtain a set of professional terminology for packaging design; performing packaging design industry terminology matching on the set of professional terminology for packaging design to generate industry terminology matching data; performing industry semantic association on the industry terminology matching data to obtain packaging design terminology information.

[0099] In an embodiment of the present invention, an original data set of packaging design is obtained, and data related to packaging design is collected from different sources, including images, text descriptions, design parameters, etc.; such as obtaining a data set from an online database, and finally obtaining an original data set of packaging design; data cleaning is performed on the original data set of packaging design, and data cleaning includes processing missing values, outliers, duplicate records, etc. Using Python's Pandas library, the dropna() function is used to delete missing values, the DataFrame.info() and describe() methods are used to identify outliers, and the duplicated() method is used to remove duplicate records to obtain a packaging design cleaning data set; the packaging design cleaning data set is data standardized to obtain a standard packaging design data set. Data standardization refers to scaling the data so that it falls into a small specific interval. For example, the Z-score normalization method is used to standardize numerical features to obtain a standard packaging design data set; the standard packaging design data set is marked with packaging design features, and the specific feature marking involves extracting key information from the data and labeling it; for example, image processing technology is used to identify specific elements in packaging design, and annotation tools such as LabelImg are used to mark them to form packaging design feature information; based on the packaging design feature information, packaging design information is extracted from the standard packaging design data set, and information extraction is achieved through natural language processing technology, such as using the jieba library for Chinese word segmentation to extract keywords in the design description to obtain packaging design information; professional terminology is identified for the packaging design information to obtain a set of packaging design professional terminology, and professional terminology is identified by constructing a professional terminology dictionary and using regular expression matching technology to identify professional terminology in the packaging design information; packaging design industry terminology is matched for the set of packaging design professional terminology, and industry terminology matching is performed by comparing the professional terminology set with the industry standard terminology library, and Python's difflib library is used for similarity matching to generate industry terminology matching data.

[0100] Preferably, step S2 comprises the following steps:

[0101] Step S21: performing design term semantic decomposition on the packaging design term information to obtain design term decomposition data; performing design element term recognition on the design term decomposition data to generate design element term data; performing term feature mapping on the design element term data to obtain packaging design element data;

[0102] Step S22: performing design element structured coding on the packaging design element data to obtain design element coded data; performing attribute extraction on the design element coded data to obtain design element attribute information;

[0103] Step S23: performing image segmentation on the design element attribute information to generate packaging image and text separation data; performing design graphic recognition on the packaging image and text separation data to obtain packaging design graphic features; performing design text recognition on the packaging image and text separation data to obtain packaging design text features;

[0104] Step S24: integrating the packaging design graphic features and the packaging design text features into packaging graphic information to generate packaging design graphic information;

[0105] Step S25: Scan the visual features based on the packaging design graphic information to generate the packaging design visual features.

[0106] In an embodiment of the present invention, the packaging design term information is subjected to semantic decomposition of design terms, and specifically, natural language processing technology, especially semantic analysis method, is used to decompose the term information. The specific operations include using jieba to perform Chinese word segmentation, and then using word meaning association tools such as WordNet to perform semantic decomposition of terms to form design term decomposition data; performing design element term recognition on the design term decomposition data, and identifying terms related to design elements from the decomposed data through pattern matching and regular expression technology to generate design element term data; performing term feature mapping on the design element term data, and using feature extraction technology, such as TF-IDF or Word2Vec, to map the term data to the feature vector space to obtain packaging design element data; performing design element structured coding on the packaging design element data, and using structured coding technology, such as JSON or XML, to encode the packaging design element data into a structured format to obtain design element coded data; performing attribute extraction on the design element coded data, and extracting key attributes, such as color, shape, material, etc., from the coded data through an attribute extraction algorithm to obtain design element attribute information; Perform image segmentation based on the attribute information of design elements, and use image segmentation technology, such as watershed algorithm or GrabCut algorithm, to separate the graphics and text in the image of packaging design, and generate packaging graphics and text separation data; perform design graphics recognition on the packaging graphics and text separation data, and use graphic recognition technology, such as image recognition algorithm based on deep learning, to identify the graphic features in the separation data, and obtain the packaging design graphic features; perform design text recognition on the packaging graphics and text separation data, and use OCR technology, such as CRNN network, to identify the text information in the separation data, and obtain the packaging design text features; integrate the packaging design graphic features and packaging design text features into the packaging graphics and text information, and integrate the graphic features and text features together through data fusion technology to generate packaging design graphics and text information; perform visual feature scanning based on the packaging design graphics and text information, and use computer vision technology, such as feature extraction and matching technology, to perform visual feature analysis on the packaging design graphics and text information, and generate packaging design visual features.

[0107] Preferably, step S25 comprises the following steps:

[0108] Step S251: Scan the packaging design graphic information with high resolution to obtain graphic image scan information; enhance the graphic details of the graphic image scan information to generate a graphic detail enhanced image;

[0109] Step S252: performing color correction on the image with enhanced detail to generate an image with color correction; performing contrast enhancement on the image with color correction to generate an image with contrast enhancement;

[0110] Step S253: performing packaging visual design edge detection on the image with enhanced contrast between text and images to obtain packaging edge visual information; performing texture feature recognition on the packaging edge visual information to generate packaging texture visual information;

[0111] Step S254: performing spatial shape recognition on the packaging texture visual information to obtain spatial shape visual information; performing visual hierarchy analysis on the spatial shape visual information to obtain visual hierarchy information; and merging packaging design visual features on the packaging design graphic information based on the visual hierarchy information to generate packaging design visual features.

[0112] In the embodiment of the present invention, the packaging design graphic information is scanned at high resolution. In this step, ultra-high resolution laser scanning technology is used to scan the image to obtain clearer image details and obtain graphic image scanning information; the graphic image scanning information is enhanced in graphic details, and image sharpening technology, such as Laplace operator and Sobel operator, is used to enhance the edge and details of the image; the specific operation is to apply Laplace filtering and Sobel edge detection to the "graphic image scanning information", implement sharpening processing through code, and generate a graphic detail enhanced image; color correction is performed on the graphic detail enhanced image, and color correction technology, such as white balance, is used. Adjust and color interpolate to correct the color deviation of the image; the specific operation is to analyze the color distribution of the "image with enhanced image details", use the color correction algorithm to adjust the RGB channels of the image to make its color more accurate, and generate a color-corrected image with image details; perform contrast enhancement on the color-corrected image with image details, and use image enhancement techniques such as histogram equalization to improve the contrast of the image; the specific operation is to apply the histogram equalization algorithm to the "image with enhanced image details", adjust the pixel value distribution, enhance the contrast of the image, and obtain a contrast-enhanced image with image details; perform edge detection on the contrast-enhanced image with image details, and use edge detection algorithms such as CR Net model to detect edge information in the image; the specific operation is to input the "image with enhanced contrast" into the CRNet model, and the model outputs the edge detection result to obtain the visual information of the packaging edge; the texture feature recognition is performed on the visual information of the packaging edge, and texture recognition technology, such as texture analysis based on local binary pattern (LBP), is used to identify the texture features in the image; the specific operation is to apply the LBP algorithm to the "visual information of the packaging edge" to extract texture features; the spatial shape recognition is performed on the visual information of the packaging texture, and the spatial shape recognition technology, such as the convolutional neural network (CNN) in deep learning, is used to recognize the spatial shape in the image. shape; the specific operation is to input the "packaging texture visual information" into the CNN model, and the model outputs the spatial shape recognition result; the spatial shape visual information is subjected to visual hierarchy analysis, and the hierarchy analysis technology, such as image segmentation algorithm, is used to analyze the visual hierarchy in the image; the specific operation is to apply the image segmentation algorithm to the "spatial shape visual information" to separate the image features at different levels; the feature fusion technology is used to combine the "visual hierarchy information" with the original image and text information to extract the comprehensive visual features; the specific operation is to fuse the "visual hierarchy information" with the "packaging design image and text information" to extract the key visual features.

[0113] Preferably, step S3 comprises the following steps:

[0114] Step S31: performing packaging structure feature recognition on the packaging design visual features to obtain packaging structure feature information; performing structural contour detection on the packaging structure feature information to generate packaging structure contour data;

[0115] Step S32: mapping the packaging structure outline data to structural components to obtain structural component mapping data; dividing the packaging structure components based on the structural component mapping data to obtain packaging structure design components;

[0116] Step S33: performing component structure similarity measurement on the packaging structure design components to generate component structure similarity measurement values; performing design structure clustering on the packaging structure design components according to the component structure similarity measurement values ​​to obtain design structure clustering data; performing design structure type classification based on the design structure clustering data to obtain the packaging design structure type;

[0117] Step S34: performing packaging structural integrity evaluation on the packaging design structure type to obtain structural integrity evaluation data;

[0118] Step S35: mapping the packaging design structure type to packaging design function based on the structural integrity evaluation data to generate packaging design function data;

[0119] Step S36: Match the packaging structure design components with the packaging application scenarios using the packaging design function data to obtain packaging application scenario data.

[0120] In the embodiment of the present invention, the packaging structure feature recognition is performed on the packaging design visual features, specifically using line structured light imaging technology to build a visual imaging model of customized product packaging design, and realizing packaging visual image fusion through light stripe parameter analysis, so as to identify the packaging structure features; structural contour detection is performed on the packaging structure feature information, specifically using edge detection algorithms in image processing technology, such as Canny edge detector; structural component mapping is performed on the packaging structure contour data, and image segmentation technology, such as GraphCut algorithm, is used to map the contour data to specific structural components; packaging structure components are divided based on the structural component mapping data, and the mapping data is divided into different components through clustering algorithms such as K-means or spectral clustering; component structure analysis is performed on the packaging structure design components Similarity measurement, using feature matching technology, such as SIFT algorithm, to calculate the similarity measurement value between components, and using clustering algorithm to cluster the design structure according to the component structure similarity measurement value; classifying the design structure type based on the design structure clustering data, using random forest classification algorithm to obtain the packaging design structure type; evaluating the packaging structure integrity of the packaging design structure type, using machine learning methods of structural state recognition and evaluation to evaluate the integrity of the structure type; mapping the packaging design function of the packaging design structure type based on the structural integrity evaluation data, and associating the structural integrity evaluation data with the design function; matching the packaging structure design components with the packaging application scenarios through the packaging design function data, using matching algorithms, such as nearest neighbor search, to match the function data with the actual application scenarios.

[0121] Preferably, step S34 includes the following steps:

[0122] Step S341: performing topological relationship mapping on the packaging design structure type data to obtain structure topological relationship data; performing topological path continuity verification on the structure topological relationship data to obtain topological path verification data;

[0123] Step S342: performing structural consistency detection on the packaging design structure type data through the topological path verification data to generate a structural consistency detection result; performing consistency degree measurement on the structural consistency detection result to obtain structural consistency degree data;

[0124] Step S343: performing structural space alignment accuracy identification on the topological path verification data to obtain the structural space alignment accuracy; performing structural geometric compatibility detection on the packaging design structure type data based on the structural space alignment accuracy to generate geometric compatibility data;

[0125] Step S344: performing a structural component connection point strength test on the topology path verification data to obtain a component connection point strength value; performing a structural component coordination test on the packaging design structure type data according to the component connection point strength value to generate structural component coordination data;

[0126] Step S345: Integrate the structural coherence degree data, geometric compatibility data and structural component coordination data to obtain structural integrity evaluation data.

[0127] In the embodiment of the present invention, the packaging design structure type data is topologically mapped, and the topological analysis technology in graph theory is used to construct a graph model to represent the various structural components of the packaging design and their mutual relationships. The specific operation is to use a graph theory library such as NetworkX, take the components in the packaging design structure type data as nodes, and the connection relationships between the components as edges, to construct a graph model, and perform topological relationship mapping to obtain structural topological relationship data; perform topological path continuity verification on the structural topological relationship data, and use a path search algorithm such as the Dijkstra algorithm to perform path analysis on the graph model in the "structural topological relationship data", specifically, perform path search from the source node to the target node. Search, check whether the path exists and the integrity of the path, and obtain the topological path verification data; use the topological path verification data to perform structural consistency detection on the packaging design structure type data to determine whether the connection between the structural components meets the design requirements; according to the path information in the topological path verification data, evaluate the connection order and logical relationship of each structural component; measure the consistency of the structural consistency detection results, specifically calculate the path length, node connectivity and other indicators, and evaluate the consistency of the structure; identify the structural space alignment accuracy of the topological path verification data, specifically use the PASTE algorithm to calculate the spatial similarity and alignment accuracy between the structural components, and form the structural space alignment accuracy. Based on the structural space alignment accuracy, the structural geometric compatibility test is performed on the packaging design structure type data, and the geometric shapes and sizes of the structural components are specifically analyzed to evaluate the compatibility between them and form geometric compatibility data; the strength of the structural component connection points is tested on the topological path verification data, and the finite element analysis software is used to perform stress analysis on the connection points, evaluate the strength of the connection points, and form the component connection point strength value; the structural component coordination test is performed on the packaging design structure type data according to the component connection point strength value, and the coordination test algorithm is used to specifically analyze whether the strength of the connection points meets the structural stability requirements; the data fusion technology is used to integrate the structural coherence degree data, geometric compatibility data and structural component coordination data to obtain the structural integrity evaluation data.

[0128] Preferably, step S35 includes the following steps:

[0129] Step S351: extracting packaging design functional requirements from the packaging design structure type to obtain packaging functional requirement data; matching the packaging functional requirement data with design parameters to generate functional requirement design parameters;

[0130] Step S352: Score the structural integrity evaluation data to obtain an integrity score value; set mapping rules for the integrity score value according to the functional requirement design parameters to obtain a packaging function mapping rule set;

[0131] Step S353: performing a structural function matching degree detection on the packaging design structure type based on the packaging function mapping rule set to generate a structural function matching degree; performing a packaging design function integrity check on the structural function matching degree to obtain packaging function integrity check data;

[0132] Step S354: Based on the packaging function integrity verification data, the packaging design structure type is presented with packaging structure functions to generate packaging structure function data; the packaging structure function data is mapped with packaging design functions to generate packaging design function data.

[0133] In the embodiment of the present invention, packaging design functional requirements are extracted from packaging design structure types, and the structural types of packaging design are converted into corresponding functional requirements through the function-behavior-structure (FBS) model method; the specific operations are: analyzing the packaging design structure types, determining the functions they should have, such as protection, portability, etc., to form packaging functional requirement data; matching the packaging functional requirement data with design parameters, using the axiomatic design function-structure mapping method to convert the functional requirements into design parameters, such as material selection, structural stability, etc., to form functional requirement design parameters; and scoring the degree of integrity of the structural integrity assessment data, specifically using data fusion technology, combining information from multiple sensors or data sources, and comprehensively scoring the structural integrity assessment data to form an integrity score value. The mapping rules for the integrity score are set according to the functional requirement design parameters. Specifically, according to the axiomatic design function-structure mapping method, a mapping model from the functional domain to the structural domain is established; this model converts the functional requirements of packaging design (such as protection and portability) into specific structural design parameters (such as material type and structural stability); according to the corrugated box packaging design requirement analysis mapping method, the technical requirement characteristics are gradually decomposed into the key attributes of the product; in this process, the quality function deployment (QFD) method is used to map the technical attributes to the key attributes, such as box type selection, internal size, cardboard category, etc.; thereby determining the relationship between the technical requirements, and based on this, determining the importance of the technical requirements, and finally setting the mapping rules. By using data mapping technology, a corresponding relationship is established between the functional requirement design parameters and the integrity degree score value to form a packaging function mapping rule set; based on the packaging function mapping rule set, the packaging design structure type is tested for structure-function matching, and the function-behavior-structure (FBS) model method is applied to evaluate the functional matching of the packaging design structure type according to the mapping rule set to form a structural function matching degree; the packaging design function integrity is checked for the structural function matching degree, and the structural function matching degree is checked for integrity through the axiomatic design function-structure mapping method to ensure that all functional requirements are met and form packaging function integrity verification data; based on the packaging function integrity verification data, the packaging design structure type is presented for packaging structure function, and machine vision technology is used; according to the functional integrity verification data, the packaging design structure type is presented for function to form packaging structure function data; the packaging structure function data is mapped for packaging design function to generate packaging design function data; the data mapping technology is applied again to map the packaging structure function data back to the design function requirements to form packaging design function data.

[0134] Preferably, step S36 includes the following steps:

[0135] Step S361: performing functional pattern recognition on the packaging design functional data to obtain the packaging design functional pattern; performing clustering processing on the packaging design functional pattern to obtain packaging design functional clustering data; performing packaging design functional type classification on the packaging design functional clustering data to generate packaging design functional types;

[0136] Step S362: according to the packaging design function type, the packaging structure design components are associated with functional scenarios to obtain functional application scenario associated data; the functional scenario feature recognition is performed on the functional application scenario associated data to generate packaging functional scenario features;

[0137] Step S363: Based on the packaging functional scenario characteristics, the functional application scenario associated data is evaluated for application scenario adaptability to obtain a functional scenario adaptability score; the packaging structure design components are ranked by packaging component functional scenario scores according to the functional scenario adaptability score to generate component scenario score ranking information;

[0138] Step S364: Screen application scenario components according to component scenario score sorting information to obtain an application scenario component list; match packaging structure design components with packaging application scenarios through the application scenario component list to obtain packaging application scenario data.

[0139] In the embodiment of the present invention, functional pattern recognition is performed on the packaging design functional data. Specifically, the K-means clustering algorithm is used to perform pattern recognition on the functional data. By calculating the Euclidean distance between data points, similar functional data are clustered together to form a packaging design functional pattern. The packaging design functional pattern is clustered according to the similarity measure between the patterns, such as the Hamming distance. distance), build a clustering tree, obtain clustering results of different levels by pruning, and form packaging design function clustering data; classify the packaging design function types for the packaging design function clustering data, and classify the clustering data based on the clustering results using a decision tree classification algorithm to determine the function type corresponding to each clustering category to form the packaging design function type; associate the packaging structure design components with the functional scenarios according to the packaging design function type, and specifically simulate the interaction between the packaging design function type and the packaging structure design components through the digital twin model, record the matching results, and form functional application scenario association data; identify the functional scenario features of the functional application scenario association data, and use feature extraction technology in machine learning, such as principal component analysis (PCA), to identify the features of the association data; specifically extract key features through PCA dimensionality reduction to form the packaging Functional scenario characteristics; based on the packaging functional scenario characteristics, the functional application scenario associated data is evaluated for application scenario adaptability, specifically, a random forest model is trained, the input is the packaging functional scenario characteristics, and the output is the adaptability score, so as to form a functional scenario adaptability score; the packaging structure design components are ranked according to the functional scenario scores according to the functional scenario adaptability scores, and the packaging structure design components are ranked according to the functional scenario adaptability scores to obtain the score ranking information; the application scenario components are screened according to the component scenario score ranking information, specifically, the screening control technology is used to screen out components suitable for specific application scenarios according to the score ranking information; the packaging structure design components are matched with the packaging application scenarios through the application scenario component list, specifically, a matching algorithm is used to match the packaging structure design components according to the application scenario component list to form packaging application scenario data.

[0140] As an example of the present invention, refer to Figure 2 As shown, in this example, step S4 includes:

[0141] Step S41: constructing a packaging design language model according to the structural integrity evaluation data, the packaging design function data and the packaging application scenario data to obtain a packaging design language pre-model; training the packaging design language pre-model based on the structural integrity evaluation data, the packaging design function data and the packaging application scenario data to obtain a packaging design language training model;

[0142] Step S42: performing model cross-validation evaluation on the packaging design language training model to obtain model evaluation data; adjusting model parameters of the packaging design language training model using the model evaluation data to obtain a packaging design language model;

[0143] Step S43: evaluating the packaging design language features of the original packaging design data set through the packaging design language model to generate packaging design language features; and visualizing the packaging design language features to obtain a packaging design language feature report.

[0144] In an embodiment of the present invention, a packaging design language model is constructed according to structural integrity evaluation data, packaging design function data and packaging application scenario data to obtain a packaging design language pre-model. Using deep learning technology, especially a neural network model, a comprehensive packaging design language pre-model is constructed in combination with structural integrity evaluation data, packaging design function data and packaging application scenario data. A specific embodiment is to use the TensorFlow framework to construct a neural network model including a convolutional layer and a fully connected layer, input the feature vector of structural integrity evaluation data, packaging design function data and packaging application scenario data, and output the language features of the packaging design to form a packaging design language pre-model; based on the structural integrity evaluation data, packaging design function data and packaging application scenario data, the packaging design language pre-model is trained to obtain a packaging design language training model. During the training process, the model is evaluated for feature importance using a random forest algorithm to determine which features contribute the most to the model's predictive ability, and the model structure is adjusted accordingly; the specific operation is to use the RandomForestClassifier class in the scikit-learn library, input the feature vector, and the label is the language feature of the packaging design, and the packaging design language training model is obtained through training. The packaging design language training model was cross-validated and evaluated by using the K-fold cross-validation method. The data set was divided into K subsets, one of which was used as the test set and the rest as the training set in turn, and the performance indicators of the model on each subset were calculated. The specific implementation example is: using the cross_val_score function in the scikit-learn library, with the input being the packaging design language training model and the feature vector, and the output being the model evaluation data; adjusting the model parameters of the packaging design language training model through the model evaluation data, and adjusting the model hyperparameters, such as the learning rate, the regularization coefficient, etc., according to the results of the cross validation, to improve the generalization ability of the model; specifically using GridSearchCV or RandomizedSearchCV for hyperparameter tuning, with the input being the model evaluation data, and the output being the adjusted packaging design language model; evaluating the packaging design language features of the original data set of packaging design through the packaging design language model, applying the trained model to the original data set, and obtaining the design language features of each sample; specifically using the trained packaging design language model to identify the original data set, and outputting the packaging design language features; visualizing the packaging design language features, and using data visualization technology, such as Matplotlib and Seaborn libraries, to display the packaging design language features in the form of charts, and specifically using the Matplotlib library to draw bar charts and line charts to display the distribution of different design features, and forming a packaging design language feature report.

[0145] As an example of the present invention, refer to Figure 3As shown, in this example, step S43 includes:

[0146] Step S431: inputting the original packaging design data set into the packaging design language model to perform packaging design language recognition to obtain packaging design language recognition data;

[0147] Step S432: Based on the packaging design language model, the packaging design language recognition data is evaluated for structural integrity to obtain a packaging design integrity evaluation result; based on the packaging design language model, the packaging design function is evaluated for the packaging design language recognition data to obtain a packaging design function feature result; based on the packaging design language model, the packaging design language recognition data is evaluated for packaging application scenarios to obtain packaging application scenario feature results;

[0148] Step S433: integrating the packaging design integrity evaluation results, the packaging design functional feature results and the packaging application scenario feature results into packaging design language features to generate packaging design language features;

[0149] Step S434: graphically process the packaging design language features to obtain a packaging design language graph; and compile a visual report on the packaging design language graph to obtain a packaging design language feature report.

[0150] In an embodiment of the present invention, an original data set of packaging design is input into a packaging design language model to perform packaging design language recognition, specifically, the original data set is forward propagated through the model to obtain packaging design language recognition data; a structural integrity assessment is performed on the packaging design language recognition data based on the packaging design language model to obtain a packaging design integrity assessment result; a structural integrity assessment algorithm, such as a convolutional neural network (CNN) based on deep learning, is used to evaluate the recognition data; specifically, the packaging design language recognition data is input into the CNN model, and the model outputs a structural integrity assessment result to form a packaging design integrity assessment result; based on the packaging design language model, a packaging design function assessment is performed on the packaging design language recognition data, and a functional assessment algorithm, such as a random forest algorithm, is used to perform functional feature assessment on the recognition data; specifically, the model is used to output a functional feature assessment result to form a packaging design functional feature result; based on the packaging design language model, the packaging design language recognition data is subjected to packaging application scenario assessment. Evaluation, use application scenario evaluation algorithms, such as support vector machine (SVM), to evaluate the application scenarios of the identified data; specifically use the model to output the application scenario evaluation results to form the packaging application scenario feature results; use feature fusion technology, such as principal component analysis (PCA), to form the packaging design integrity evaluation results, packaging design functional feature results and packaging application scenario feature results into packaging design language features; graphically process the packaging design language features, and use data visualization libraries, such as Matplotlib and Seaborn, to graphically display the packaging design language features; specifically use Matplotlib to draw scatter plots and Seaborn to draw box plots to display the distribution of packaging design language features and form packaging design language charts; compile a visual report for the packaging design language chart to obtain a packaging design language feature report; use report compilation tools to integrate the packaging design language chart into a visual report; specifically write code in Jupyter Notebook to integrate charts and analysis results to form a packaging design language feature report.

[0151] The present invention also provides a method for constructing a packaging design language model, the method comprising the following steps:

[0152] Step S1: obtaining an original packaging design data set; performing data preprocessing on the original packaging design data set to obtain a standard packaging design data set; extracting packaging design information from the standard packaging design data set to obtain packaging design information; performing packaging design terminology analysis on the packaging design information to obtain packaging design terminology information;

[0153] Step S2: performing design element recognition on the packaging design term information to obtain packaging design element data; performing packaging graphic information recognition on the packaging design element data to generate packaging design graphic information; performing visual feature scanning based on the packaging design graphic information to generate packaging design visual features;

[0154] Step S3: dividing the packaging structure components according to the packaging design visual features to obtain packaging structure design components; classifying the packaging structure design components into design structure types to obtain packaging design structure types; evaluating the packaging structure integrity of the packaging design structure types to obtain structure integrity evaluation data; mapping the packaging design structure types to packaging design functions based on the structure integrity evaluation data to generate packaging design function data; matching the packaging structure design components to packaging application scenarios using the packaging design function data to obtain packaging application scenario data;

[0155] Step S4: construct a packaging design language model based on the structural integrity evaluation data, packaging design function data and packaging application scenario data to obtain a packaging design language model; evaluate the packaging design language features of the original packaging design data set through the packaging design language model to generate packaging design language features; visualize the packaging design language features to obtain a packaging design language feature report.

[0156] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is therefore intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0157] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A method for training a packaging design language model, characterized in that: The following steps are involved: Step S1: Obtain the original packaging design data set; Preprocess the original packaging design data set to obtain a standard packaging design data set; Extracting packaging design information from a standard packaging design data set to obtain packaging design information; Performing packaging design term analysis on packaging design information to obtain packaging design term information; Step S2: performing design element recognition on the packaging design term information to obtain packaging design element data; Recognize packaging graphic information on packaging design element data to generate packaging design graphic information; Scan visual features based on packaging design graphic information to generate packaging design visual features; Step S3: Divide the packaging structure components according to the packaging design visual features to obtain packaging structure design components; Classifying the design structure types of packaging structure design components to obtain packaging design structure types; Conducting packaging structural integrity evaluation on packaging design structural types to obtain structural integrity evaluation data; Mapping the packaging design function to the packaging design structure type based on the structural integrity evaluation data to generate packaging design function data; Matching packaging application scenarios with packaging structure design components through packaging design function data to obtain packaging application scenario data; Step S4: constructing a packaging design language model based on the structural integrity evaluation data, the packaging design function data, and the packaging application scenario data to obtain a packaging design language model; evaluating the packaging design language features of the original packaging design data set through the packaging design language model to generate packaging design language features; The packaging design language features are visualized to obtain a packaging design language feature report.

2. The method for training a packaging design language model according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: Obtaining the original packaging design data set; Step S12: performing data cleaning on the original packaging design data set to obtain a packaging design cleaned data set; performing data standardization on the packaging design cleaned data set to obtain a standard packaging design data set; Step S13: marking the standard packaging design data set with packaging design features to obtain packaging design feature information; extracting packaging design information from the standard packaging design data set based on the packaging design feature information to obtain packaging design information; Step S14: performing professional terminology recognition on the packaging design information to obtain a set of professional terminology for packaging design; performing packaging design industry terminology matching on the set of professional terminology for packaging design to generate industry terminology matching data; performing industry semantic association on the industry terminology matching data to obtain packaging design terminology information.

3. The method for training a packaging design language model according to claim 1, characterized in that: Step S2 includes the following steps: Step S21: performing design term semantic decomposition on the packaging design term information to obtain design term decomposition data; performing design element term recognition on the design term decomposition data to generate design element term data; performing term feature mapping on the design element term data to obtain packaging design element data; Step S22: performing design element structured coding on the packaging design element data to obtain design element coded data; performing attribute extraction on the design element coded data to obtain design element attribute information; Step S23: performing image segmentation on the design element attribute information to generate packaging image and text separation data; performing design graphic recognition on the packaging image and text separation data to obtain packaging design graphic features; performing design text recognition on the packaging image and text separation data to obtain packaging design text features; Step S24: integrating the packaging design graphic features and the packaging design text features into packaging graphic information to generate packaging design graphic information; Step S25: Scan the visual features based on the packaging design graphic information to generate the packaging design visual features.

4. The method for training a packaging design language model according to claim 3, characterized in that: Step S25 includes the following steps: Step S251: Scan the packaging design graphic information with high resolution to obtain graphic image scan information; enhance the graphic details of the graphic image scan information to generate a graphic detail enhanced image; Step S252: performing color correction on the image with enhanced detail to generate an image with color correction; performing contrast enhancement on the image with color correction to generate an image with contrast enhancement; Step S253: performing packaging visual design edge detection on the image with enhanced contrast between text and images to obtain packaging edge visual information; performing texture feature recognition on the packaging edge visual information to generate packaging texture visual information; Step S254: performing spatial shape recognition on the packaging texture visual information to obtain spatial shape visual information; performing visual hierarchy analysis on the spatial shape visual information to obtain visual hierarchy information; and merging packaging design visual features on the packaging design graphic information based on the visual hierarchy information to generate packaging design visual features.

5. The method for training a packaging design language model according to claim 1, characterized in that: Step S3 includes the following steps: Step S31: performing packaging structure feature recognition on the packaging design visual features to obtain packaging structure feature information; performing structural contour detection on the packaging structure feature information to generate packaging structure contour data; Step S32: mapping the packaging structure outline data to structural components to obtain structural component mapping data; dividing the packaging structure components based on the structural component mapping data to obtain packaging structure design components; Step S33: performing component structure similarity measurement on the packaging structure design components to generate component structure similarity measurement values; performing design structure clustering on the packaging structure design components according to the component structure similarity measurement values ​​to obtain design structure clustering data; performing design structure type classification based on the design structure clustering data to obtain the packaging design structure type; Step S34: performing packaging structural integrity evaluation on the packaging design structure type to obtain structural integrity evaluation data; Step S35: mapping the packaging design structure type to packaging design function based on the structural integrity evaluation data to generate packaging design function data; Step S36: Match the packaging structure design components with the packaging application scenarios using the packaging design function data to obtain packaging application scenario data.

6. The method for training a packaging design language model according to claim 5, characterized in that: Step S34 includes the following steps: Step S341: performing topological relationship mapping on the packaging design structure type data to obtain structure topological relationship data; performing topological path continuity verification on the structure topological relationship data to obtain topological path verification data; Step S342: performing structural consistency detection on the packaging design structure type data through the topological path verification data to generate a structural consistency detection result; performing consistency degree measurement on the structural consistency detection result to obtain structural consistency degree data; Step S343: performing structural space alignment accuracy identification on the topological path verification data to obtain the structural space alignment accuracy; performing structural geometric compatibility detection on the packaging design structure type data based on the structural space alignment accuracy to generate geometric compatibility data; Step S344: performing a structural component connection point strength test on the topology path verification data to obtain a component connection point strength value; performing a structural component coordination test on the packaging design structure type data according to the component connection point strength value to generate structural component coordination data; Step S345: Integrate the structural coherence degree data, geometric compatibility data and structural component coordination data to obtain structural integrity evaluation data.

7. The method for training a packaging design language model according to claim 5, characterized in that: Step S35 includes the following steps: Step S351: extracting packaging design functional requirements from the packaging design structure type to obtain packaging functional requirement data; matching the packaging functional requirement data with design parameters to generate functional requirement design parameters; Step S352: Score the structural integrity evaluation data to obtain an integrity score value; set mapping rules for the integrity score value according to the functional requirement design parameters to obtain a packaging function mapping rule set; Step S353: performing a structural function matching degree detection on the packaging design structure type based on the packaging function mapping rule set to generate a structural function matching degree; performing a packaging design function integrity check on the structural function matching degree to obtain packaging function integrity check data; Step S354: Based on the packaging function integrity verification data, the packaging design structure type is presented with packaging structure functions to generate packaging structure function data; the packaging structure function data is mapped with packaging design functions to generate packaging design function data.

8. The method for training a packaging design language model according to claim 5, characterized in that: Step S36 includes the following steps: Step S361: performing functional pattern recognition on the packaging design functional data to obtain the packaging design functional pattern; performing clustering processing on the packaging design functional pattern to obtain packaging design functional clustering data; performing packaging design functional type classification on the packaging design functional clustering data to generate packaging design functional types; Step S362: according to the packaging design function type, the packaging structure design components are associated with functional scenarios to obtain functional application scenario associated data; the functional scenario feature recognition is performed on the functional application scenario associated data to generate packaging functional scenario features; Step S363: Based on the packaging functional scenario characteristics, the functional application scenario associated data is evaluated for application scenario adaptability to obtain a functional scenario adaptability score; the packaging structure design components are ranked by packaging component functional scenario scores according to the functional scenario adaptability score to generate component scenario score ranking information; Step S364: Screen application scenario components according to component scenario score sorting information to obtain an application scenario component list; match packaging structure design components with packaging application scenarios through the application scenario component list to obtain packaging application scenario data.

9. The method for training a packaging design language model according to claim 1, characterized in that: Step S4 includes the following steps: Step S41: constructing a packaging design language model according to the structural integrity evaluation data, the packaging design function data and the packaging application scenario data to obtain a packaging design language pre-model; training the packaging design language pre-model based on the structural integrity evaluation data, the packaging design function data and the packaging application scenario data to obtain a packaging design language training model; Step S42: performing model cross-validation evaluation on the packaging design language training model to obtain model evaluation data; adjusting model parameters of the packaging design language training model using the model evaluation data to obtain a packaging design language model; Step S43: evaluating the packaging design language features of the original packaging design data set through the packaging design language model to generate packaging design language features; and visualizing the packaging design language features to obtain a packaging design language feature report.

10. The method for training a packaging design language model according to claim 9, characterized in that: Step S43 includes the following steps: Step S431: inputting the original packaging design data set into the packaging design language model to perform packaging design language recognition to obtain packaging design language recognition data; Step S432: Based on the packaging design language model, the packaging design language recognition data is evaluated for structural integrity to obtain a packaging design integrity evaluation result; based on the packaging design language model, the packaging design function is evaluated for the packaging design language recognition data to obtain a packaging design function feature result; based on the packaging design language model, the packaging design language recognition data is evaluated for packaging application scenarios to obtain packaging application scenario feature results; Step S433: integrating the packaging design integrity evaluation results, the packaging design functional feature results and the packaging application scenario feature results into packaging design language features to generate packaging design language features; Step S434: graphically process the packaging design language features to obtain a packaging design language graph; and compile a visual report on the packaging design language graph to obtain a packaging design language feature report.

Images