An agricultural product packaging box type and appearance design generation method based on artificial intelligence

Abstract

The application provides a kind of agricultural product packaging box box type and appearance design generation method based on artificial intelligence, it is related to intelligent design technology.The agricultural product packaging box box type and appearance design generation method based on artificial intelligence, the specific steps of the design generation method are step 1: design requirement and basic data acquisition preprocessing, step 2: multi-dimensional feature extraction and semantic coding, step 3: box type structure intelligent generation and topology optimization, step 4: appearance design intelligent generation.The application overcomes the defects of low artificial efficiency, AI design heavy appearance, poor box type and appearance coordination and insufficient agricultural product specific adaptability in the prior art, provides an agricultural product packaging box box type and appearance design generation method based on artificial intelligence, realizes the integrated intelligent generation of box type structure and appearance design.

Description

Technical Field

[0001] This invention relates to the field of intelligent design technology, specifically to a method for generating the box shape and appearance design of agricultural product packaging boxes based on artificial intelligence. Background Technology

[0002] Agricultural product packaging boxes are the core carriers for ensuring the safety of agricultural product storage and transportation, enhancing product added value, and conveying brand and regional culture. my country has a rich variety of agricultural products, with significant differences in size, fragility, preservation requirements, and storage and transportation environments among different categories. At the same time, agricultural product packaging needs to meet specific requirements such as food contact compliance, geographical indication presentation, and embedded traceability information. This places extremely high demands on the personalized adaptation of the box structure and appearance design. Currently, agricultural product packaging design mainly relies on manual work. Designers need to simultaneously consider multiple dimensions such as structural engineering, preservation adaptation, visual design, production process, and compliance requirements. The design cycle can take several days to several weeks, resulting in high design costs and making it difficult to adapt to the small-batch, personalized design needs of small and medium-sized farmers and regional specialty agricultural products.

[0003] Therefore, this invention proposes a method for generating the shape and appearance design of agricultural product packaging boxes based on artificial intelligence. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides an AI-based method for generating the box shape and appearance design of agricultural product packaging boxes. The aim is to overcome the deficiencies in existing agricultural product packaging design, such as low manual efficiency, AI design prioritizing appearance over practical application, poor coordination between box shape and appearance, and insufficient adaptability to specific agricultural products. This AI-based method achieves integrated intelligent generation of box structure and appearance design, simultaneously meeting the comprehensive needs of agricultural products in terms of storage, transportation, preservation, production implementation, brand culture, and compliance.

[0005] To achieve the above objectives, the present invention provides the following technical solution: A method for generating the box type and appearance design of agricultural product packaging boxes based on artificial intelligence, wherein the specific steps of the design generation method are as follows: Step 1: Design requirements and basic data collection and preprocessing; Collect packaging design requirements data for target agricultural products, and construct and preprocess a dedicated dataset for agricultural product packaging; Step 2: Multi-dimensional feature extraction and semantic encoding; Multi-dimensional feature extraction and multi-modal semantic encoding are performed on the collected requirement data to generate a joint design feature vector that is fused across modalities; Step 3: Intelligent generation and topology optimization of box-shaped structures; Using the joint design feature vector as input, multiple candidate box schemes are generated through a pre-trained box structure generation model. The candidate box schemes are then topologically optimized by combining multi-physics constraints and multi-objective optimization algorithms to select the optimal box structure scheme. Step 4: Intelligent generation of appearance design; Based on the 3D model and UV layout information of the optimal box structure, and guided by the joint design feature vector, a pre-trained appearance generation model is used to generate an appearance design scheme that is accurately adapted to the box structure, and the adaptability and compliance optimization are completed. Step 5: Multi-objective comprehensive verification and feasibility assessment; A multi-dimensional evaluation system was constructed to comprehensively verify and quantitatively score the design scheme that integrates box shape and appearance, and to select the final candidate design scheme. Step 6: Collaborative Iterative Optimization; Receive user feedback on modifications, perform semantic parsing and feature updates on the feedback, feed it back to the corresponding generation model to complete local iterative optimization of the design scheme, repeat the verification process of steps S3-S5, and obtain the final design scheme confirmed by the user. Step 7: Output production-level files; For the final design scheme, output box structure files, appearance printing files and supporting evaluation files that can be directly used for production.

[0006] Furthermore, step 1, which involves designing requirements and preprocessing basic data collection, specifically includes: Step 101: Demand Data Collection; The collected demand data includes four main categories: basic attributes of the target agricultural products, storage and transportation environment parameters, brand culture requirements, and production process constraints. Step 102: Constructing a dedicated dataset for agricultural product packaging; The dataset consists of three sub-databases: box structure dataset, appearance design dataset, and packaging engineering knowledge base; Step 103: Data preprocessing; The dataset is cleaned, deduplicated, and standardized. The box structure data is labeled with structural type, load-bearing threshold, material compatibility, and molding difficulty. The appearance design data is labeled with element semantics, applicable categories, color system, and compliance. The packaging engineering knowledge base is processed into a knowledge graph and a constraint rule base is constructed. Through the above technical solutions, during the demand data collection process, the basic attributes of the target agricultural products include category, single fruit / individual size, total box weight, fragility level, freshness preservation and ventilation / sealing requirements, and shelf life requirements; the storage and transportation environment parameters include transportation temperature and humidity range, stacking load-bearing requirements, logistics turnover method, and warehousing and stacking cycle; the brand culture requirements include brand logo, geographical indication elements, origin culture characteristics, visual style preferences, product information and traceability labeling requirements; and the production process constraints include the types of packaging materials available, the process limits of production equipment, die-cutting and folding requirements, and cost control thresholds. The agricultural product packaging-specific dataset comprises three sub-databases: the box structure dataset includes 3D structural parameters, planar unfolded diagrams, die-cutting line data, mechanical performance test data, and molding process parameters for standard packaging boxes and customized agricultural product boxes; the appearance design dataset includes visual elements related to agricultural product categories, geographical indication patterns, compliant color matching systems for agricultural product packaging, a library of food packaging labeling standards, and a library of origin culture materials; and the packaging engineering knowledge base includes performance parameters of packaging materials, design specifications for agricultural product preservation structures, design standards for cushioning and anti-collision structures, food packaging compliance regulations, and printing and die-cutting production process specifications.

[0007] Furthermore, the specific steps of step 2, multi-dimensional feature extraction and semantic encoding, are as follows: Step 201: Constraint Feature Extraction; A pre-trained CNN feature extraction network is used to extract numerical features from the basic attributes, storage and transportation environment parameters, and production process constraints of the target agricultural products, generating a box-shaped design constraint feature vector. Step 202: Visual semantic encoding; The pre-trained multimodal large model CLIP is used to perform cross-modal semantic encoding of text and visual elements to meet brand culture requirements, visual style preferences, and compliance labeling requirements, thereby aligning textual requirements with visual features and generating semantic feature vectors for appearance design. Step 203: Cross-modal feature fusion; A feature fusion module based on an attention mechanism is constructed to perform cross-dimensional association and fusion of box-shaped design constraint feature vectors and appearance design semantic feature vectors to generate joint design feature vectors, which serve as the unified input for subsequent generation models.

[0008] Furthermore, the specific steps of step 3, intelligent generation and topology optimization of the box structure, are as follows: Step 301: Initial generation of candidate box types; Using the joint design feature vector as input, the box structure generation model based on the Transformer architecture and fine-tuned by the agricultural product box structure dataset is called to generate multiple sets of parametric 3D structure models and corresponding planar unfolded diagrams of candidate box types. The candidate box types include three core components: the box body, the inner lining buffer structure, and the functional structure. Step 302: Topology optimization under multiphysics constraints; By connecting to the finite element analysis interface and using constraint rules from the packaging engineering knowledge base, multi-dimensional topology optimization is performed on candidate box types, specifically including: Mechanical performance optimization: Simulation calculations were performed on the stacking load-bearing capacity and drop impact resistance of the box type, and the parameters of the box body reinforcing ribs, corner impact protection structure and inner lining buffer structure were optimized to ensure that the storage and transportation load-bearing and damage prevention requirements are met; Preservation function optimization: Optimize the pore size, layout, and number of ventilation holes in the box, or the design of sealing structure and partition layer, to meet the breathing characteristics and preservation requirements of agricultural products in response to their preservation needs. Production process optimization: Optimize the die-cutting path, number of folded edges, overlapping structure, and interlocking design of the box shape, eliminate structures that exceed the process limits of the production equipment, and reduce material waste and production difficulty; Step 303: Multi-objective optimization and screening; The NSGA-II multi-objective optimization algorithm is adopted, with the optimization objectives of maximizing structural strength, minimizing material consumption, and minimizing production difficulty. The optimal solutions of all candidate box types are sorted, and the box structure scheme with the best overall performance is selected by combining the user cost threshold.

[0009] Furthermore, the specific steps of step 4, intelligent generation of the appearance design, are as follows: Step 401: Pre-fitting of the box-shaped structure; Extract the 3D model of the optimal box structure, the UV layout information of the planar unfolded diagram, and the area division information of the box's display surface / folded surface / overlapping surface. Input these into the appearance generation model to establish the coordinate mapping relationship between the appearance pattern and the box structure. Step 402: Generation of the initial design draft; Guided by the semantic features of appearance in the joint design feature vector, the diffusion generation model based on the Stable Diffusion architecture and fine-tuned by the agricultural product appearance design dataset is called to generate an initial draft of appearance design adapted to the box-shaped UV layout. The initial draft includes the main visual pattern, color scheme, product information layout area, brand logo reserved space, and traceability code embedding space. Step 403: Adaptability and compliance optimization; This includes structural adaptation optimization, automatic compliance optimization, and brand consistency optimization; The specific optimizations to adaptability and compliance achieved through the above technical solution include: Structural adaptation optimization: For the folding edges, overlapping areas, and locking structure of the box, the patterns are spliced ​​and seamlessly connected to ensure that the box is visually complete from all angles after folding, without any key patterns being obscured or misaligned; For the main display surface, sides, and bottom of the box, the hierarchy and layout of visual elements are optimized to ensure that core brand information and product information are clearly visible in warehousing and display scenarios. Automatic compliance optimization: Based on regulatory requirements in the packaging engineering knowledge base, it automatically embeds compliance labels for food contact materials, agricultural product origin information, shelf life, and product implementation standard number in designated areas, and automatically generates and embeds product traceability QR codes; Brand consistency optimization: Standardize and optimize the size, position, and color of the brand logo, standard colors, and standard fonts to ensure the consistency of the brand visual system.

[0010] Furthermore, the multi-objective comprehensive verification and feasibility assessment in step 5 includes four verification dimensions: structural feasibility verification, production feasibility verification, compliance verification, and visual effect verification. Each dimension has a set quantitative scoring threshold, specifically: Step 501: Structural feasibility verification; Check the mechanical properties, molding stability, preservation suitability, and material compatibility of the box shape. If there are no structural defects, it is judged as passing and a structural performance score is generated. Step 502: Production feasibility verification; Verify the feasibility of the die-cutting, printing, and folding processes for the box, calculate the material utilization rate and the production cost per box, and if they meet the production equipment requirements and cost thresholds, they are deemed to have passed. Step 503: Compliance verification; Verify compliance with relevant regulations on food packaging, advertising law, and regulations on agricultural product labeling. If there are no compliance risks, the application is deemed passed, and a compliance score is generated. Step 504: Visual effect verification; A 360° full-view rendering of the box after folding is generated through 3D rendering. The integrity of the pattern, the presentation of brand elements, and the matching degree of visual style are checked to generate a visual effect score. Step 505: Comprehensive scoring and screening; For schemes that pass the threshold verification in all dimensions, a weighted comprehensive score is calculated based on structural performance, production cost, visual effect, and agricultural product suitability. The top 1-2 schemes with the highest comprehensive scores are selected as the final candidate design schemes. Schemes that fail the threshold verification are returned to the corresponding steps for re-optimization.

[0011] Furthermore, the specific steps of step 6, collaborative iterative optimization, include: Step 601: Solution presentation and feedback reception; The final candidate design scheme's 3D rendering model, plan development diagram, process parameters, and comprehensive evaluation report are presented to the user in a visual manner, and the user's global or partial modification feedback is received. Step 602: Feedback semantic parsing and feature update; A large language model is used to perform semantic parsing on user modification feedback, distinguish between box structure modification requirements and appearance design modification requirements, extract modification features, and update the joint design feature vector. Step 603: Local iterative optimization; The updated feature vectors are fed back to the corresponding box structure generation model or appearance generation model for local parameter adjustment and scheme iteration. There is no need to regenerate the entire scheme. After iteration, the verification process of steps 3-5 is repeated until the user confirms the final design scheme.

[0012] Furthermore, the production-level files output in step 7 specifically include: Box-type structure production documents: vector format plan view, die-cutting line drawing, folding process instructions, packaging material specifications, and inner lining structure processing drawings; Production documents for industrial design: printing source files after CMYK color separation, UV alignment files, marking specifications, and pattern splicing alignment instructions; Supporting evaluation documents: box-type structure mechanical simulation report, production feasibility analysis report, compliance verification report, 3D rendering; An artificial intelligence-based system for designing the shape and appearance of agricultural product packaging boxes is provided to implement the artificial intelligence-based method for generating the shape and appearance of agricultural product packaging boxes as described in any one of claims 1-8. The system includes a data acquisition and preprocessing module, a multimodal feature extraction and encoding module, a box structure generation and optimization module, an appearance design generation and optimization module, a multi-objective comprehensive verification module, a human-machine collaborative iteration module, and a production-level file output module. The data acquisition and preprocessing module is used to collect target agricultural product packaging design requirement data, and to construct and preprocess a dedicated dataset for agricultural product packaging. The multimodal feature extraction and encoding module is used to extract multi-dimensional features and encode multimodal semantics from the collected requirement data to generate a joint design feature vector that integrates cross-modal features. The box structure generation and optimization module is used to generate multiple sets of candidate box structure schemes and perform topology optimization, and select the optimal box structure scheme, taking the joint design feature vector as input. The appearance design generation and optimization module is used to generate an appearance design scheme that is precisely adapted to the box shape based on the optimal box shape structure information and to complete the adaptation and compliance optimization. The multi-objective comprehensive verification module is used to construct a multi-dimensional evaluation system to comprehensively verify and quantitatively score the integrated design scheme of box type and appearance, and to screen the final candidate design scheme. The human-machine collaborative iterative module is used to receive user modification feedback, parse the feedback content and update the feature vector to complete the local iterative optimization of the design scheme. The production-level document output module is used to output box structure documents, appearance printing documents, and supporting evaluation documents that can be directly used for production based on the final design scheme.

[0013] This invention provides a method for generating the box shape and appearance design of agricultural product packaging boxes based on artificial intelligence. It has the following beneficial effects: 1. This invention provides an artificial intelligence-based method for generating the box shape and appearance design of agricultural product packaging boxes. This design generation method systematically constructs a dataset and AI generation model specifically for agricultural product packaging, realizing the integrated intelligent generation of box structure and appearance design. It shortens the traditional manual design cycle of several days to hours, significantly reducing design costs and adapting to the personalized design needs of small and medium-sized farmers and regional specialty agricultural products. Furthermore, the box shape generation stage incorporates multi-physics field constraints on mechanical performance, preservation function, and production process, simultaneously optimizing core structures such as buffering and anti-collision, breathable preservation, and stacking load-bearing capacity. This solves the pain point of existing AI designs that prioritize appearance over structure and cannot meet the storage and preservation needs of agricultural products.

[0014] 2. This invention provides an artificial intelligence-based method for generating the box shape and appearance design of agricultural product packaging boxes. The design system supports local iterative optimization through human-machine collaboration. It can make local modifications to the box shape or appearance based on user feedback without regenerating the entire solution. It takes into account both the generation efficiency of AI and the personalized needs of human. At the same time, it has a built-in automatic compliance verification module to ensure that the solution complies with the relevant regulations for agricultural product packaging and avoids compliance risks. Attached Figure Description

[0015] Figure 1 This diagram illustrates the implementation steps of the artificial intelligence-based method for generating the shape and appearance design of agricultural product packaging boxes according to the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Example 1:

[0018] like Figure 1 As shown, an artificial intelligence-based method for generating the box type and appearance design of agricultural product packaging boxes is characterized by the following specific steps: Step 1: Design requirements and basic data collection and preprocessing; Collect packaging design requirements data for target agricultural products, and construct and preprocess a dedicated dataset for agricultural product packaging; Step 1, which involves designing requirements and preprocessing basic data collection, specifically includes: Step 101: Demand Data Collection; The collected demand data includes four main categories: basic attributes of the target agricultural products, storage and transportation environment parameters, brand culture requirements, and production process constraints. Step 102: Constructing a dedicated dataset for agricultural product packaging; The dataset consists of three sub-databases: box structure dataset, appearance design dataset, and packaging engineering knowledge base; Step 103: Data preprocessing; The dataset is cleaned, deduplicated, and standardized. The box structure data is labeled with structural type, load-bearing threshold, material compatibility, and molding difficulty. The appearance design data is labeled with element semantics, applicable categories, color system, and compliance. The packaging engineering knowledge base is processed into a knowledge graph and a constraint rule base is constructed. Step 2: Multi-dimensional feature extraction and semantic encoding; Multi-dimensional feature extraction and multi-modal semantic encoding are performed on the collected requirement data to generate a joint design feature vector that is fused across modalities; The specific steps of step 2, multi-dimensional feature extraction and semantic encoding, are as follows: Step 201: Constraint Feature Extraction; A pre-trained CNN feature extraction network is used to extract numerical features from the basic attributes, storage and transportation environment parameters, and production process constraints of the target agricultural products, generating a box-shaped design constraint feature vector. Step 202: Visual semantic encoding; The pre-trained multimodal large model CLIP is used to perform cross-modal semantic encoding of text and visual elements to meet brand culture requirements, visual style preferences, and compliance labeling requirements. This aligns textual requirements with visual features and generates semantic feature vectors for the product design. Step 203: Cross-modal feature fusion; A feature fusion module based on an attention mechanism is constructed to perform cross-dimensional association and fusion of box-shaped design constraint feature vectors and appearance design semantic feature vectors to generate joint design feature vectors, which serve as the unified input for subsequent generation models. Step 3: Intelligent generation and topology optimization of box-shaped structures; Using the joint design feature vector as input, multiple candidate box schemes are generated through a pre-trained box structure generation model. The candidate box schemes are then topologically optimized by combining multi-physics constraints and multi-objective optimization algorithms to select the optimal box structure scheme. The specific steps of step 3, intelligent generation and topology optimization of the box structure, are as follows: Step 301: Initial generation of candidate box types; Using the joint design feature vector as input, the box structure generation model based on the Transformer architecture and fine-tuned by the agricultural product box structure dataset is called to generate multiple sets of parametric 3D structure models and corresponding planar unfolded diagrams of candidate box types. The candidate box types include three core components: the box body, the inner lining buffer structure, and the functional structure. Step 302: Topology optimization under multiphysics constraints; By connecting to the finite element analysis interface and using constraint rules from the packaging engineering knowledge base, multi-dimensional topology optimization is performed on candidate box types, specifically including: Mechanical performance optimization: Simulation calculations were performed on the stacking load-bearing capacity and drop impact resistance of the box type, and the parameters of the box body reinforcing ribs, corner impact protection structure and inner lining buffer structure were optimized to ensure that the storage and transportation load-bearing and damage prevention requirements are met; Preservation function optimization: Optimize the pore size, layout, and number of ventilation holes in the box, or the design of sealing structure and partition layer, to meet the breathing characteristics and preservation requirements of agricultural products in response to their preservation needs. Production process optimization: Optimize the die-cutting path, number of folded edges, overlapping structure, and interlocking design of the box shape, eliminate structures that exceed the process limits of the production equipment, and reduce material waste and production difficulty; Step 303: Multi-objective optimization and screening; The NSGA-II multi-objective optimization algorithm is employed, with the optimization objectives of maximizing structural strength, minimizing material consumption, and minimizing production difficulty. All candidate box types are ranked for optimal solutions, and the optimal box structure scheme with the best overall performance is selected by combining this with a user cost threshold. Step 4: Intelligent generation of appearance design; Based on the 3D model and UV layout information of the optimal box structure, and guided by the joint design feature vector, a pre-trained appearance generation model is used to generate an appearance design scheme that is accurately adapted to the box structure, and the adaptability and compliance optimization are completed. Step 401: Pre-fitting of the box-shaped structure; Extract the 3D model of the optimal box structure, the UV layout information of the planar unfolded diagram, and the area division information of the box's display surface / folded surface / overlapping surface. Input these into the appearance generation model to establish the coordinate mapping relationship between the appearance pattern and the box structure. Step 402: Generation of the initial design draft; Guided by the semantic features of appearance in the joint design feature vector, the diffusion generation model based on the Stable Diffusion architecture and fine-tuned by the agricultural product appearance design dataset is called to generate an initial draft of appearance design adapted to the box-shaped UV layout. The initial draft includes the main visual pattern, color scheme, product information layout area, brand logo reserved space, and traceability code embedding space. Step 403: Adaptability and compliance optimization; This includes structural adaptation optimization, automatic compliance optimization, and brand consistency optimization; Step 5: Multi-objective comprehensive verification and feasibility assessment; A multi-dimensional evaluation system was constructed to comprehensively verify and quantitatively score the design scheme that integrates box shape and appearance, and to select the final candidate design scheme. The multi-objective comprehensive verification and feasibility assessment in step 5 includes four verification dimensions: structural feasibility verification, production feasibility verification, compliance verification, and visual effect verification. Each dimension has a quantitative scoring threshold, specifically: Step 501: Structural feasibility verification; Check the mechanical properties, molding stability, preservation suitability, and material compatibility of the box shape. If there are no structural defects, it is judged as passing and a structural performance score is generated. Step 502: Production feasibility verification; Verify the feasibility of the die-cutting, printing, and folding processes for the box, calculate the material utilization rate and the production cost per box, and if they meet the production equipment requirements and cost thresholds, they are deemed to have passed. Step 503: Compliance verification; Verify compliance with relevant regulations on food packaging, advertising law, and regulations on agricultural product labeling. If there are no compliance risks, the application is deemed passed, and a compliance score is generated. Step 504: Visual Effect Verification A 360° full-view rendering of the box after folding is generated through 3D rendering. The integrity of the pattern, the presentation of brand elements, and the matching degree of visual style are checked to generate a visual effect score. Step 505: Comprehensive scoring and screening; For schemes that pass the threshold verification in all dimensions, a weighted comprehensive score is calculated based on structural performance, production cost, visual effect, and agricultural product suitability. The top 1-2 schemes with the highest comprehensive scores are selected as the final candidate design schemes. Schemes that fail the threshold verification are returned to the corresponding steps for re-optimization. Step 6: Collaborative Iterative Optimization; Receive user feedback on modifications, perform semantic parsing and feature updates on the feedback, feed it back to the corresponding generation model to complete local iterative optimization of the design scheme, repeat the verification process of steps S3-S5, and obtain the final design scheme confirmed by the user. The specific steps of step 6, collaborative iterative optimization, include: Step 601: Solution presentation and feedback reception; The final candidate design scheme's 3D rendering model, plan development diagram, process parameters, and comprehensive evaluation report are presented to the user in a visual manner, and the user's global or partial modification feedback is received. Step 602: Feedback semantic parsing and feature update; A large language model is used to perform semantic parsing on user modification feedback, distinguish between box structure modification requirements and appearance design modification requirements, extract modification features, and update the joint design feature vector. Step 603: Local iterative optimization; The updated feature vectors are fed back to the corresponding box structure generation model or appearance generation model for local parameter adjustment and scheme iteration. There is no need to regenerate the entire scheme. After iteration, the verification process of steps 3-5 is repeated until the user confirms the final design scheme. Step 7: Output production-level files; For the final design scheme, output box structure files, appearance printing files and supporting evaluation files that can be directly used for production; The production-level files output in step 7, production-level file output, specifically include: Box-type structure production documents: vector format plan view, die-cutting line drawing, folding process instructions, packaging material specifications, and inner lining structure processing drawings; Production documents for industrial design: printing source files after CMYK color separation, UV alignment files, marking specifications, and pattern splicing alignment instructions; Supporting evaluation documents: box-type structure mechanical simulation report, production feasibility analysis report, compliance verification report, and 3D rendering.

[0019] Example 2:

[0020] An AI-based system for designing the shape and appearance of agricultural product packaging boxes is provided. This system is used to generate the shape and appearance of agricultural product packaging boxes based on AI. The system includes a data acquisition and preprocessing module, a multimodal feature extraction and encoding module, a box structure generation and optimization module, an appearance design generation and optimization module, a multi-objective comprehensive verification module, a human-machine collaborative iteration module, and a production-level file output module. The data acquisition and preprocessing module is used to collect target agricultural product packaging design requirement data, and to construct and preprocess a dedicated dataset for agricultural product packaging. The multimodal feature extraction and encoding module is used to extract multi-dimensional features and encode multimodal semantics from the collected requirement data to generate a joint design feature vector that integrates cross-modal features. The box structure generation and optimization module is used to generate multiple sets of candidate box structure schemes and perform topology optimization, and select the optimal box structure scheme, taking the joint design feature vector as input. The appearance design generation and optimization module is used to generate an appearance design scheme that is precisely adapted to the box shape based on the optimal box shape structure information and to complete the adaptation and compliance optimization. The multi-objective comprehensive verification module is used to construct a multi-dimensional evaluation system to comprehensively verify and quantitatively score the integrated design scheme of box type and appearance, and to screen the final candidate design scheme. The human-machine collaborative iterative module is used to receive user modification feedback, parse the feedback content and update the feature vector to complete the local iterative optimization of the design scheme. The production-level document output module is used to output box structure documents, appearance printing documents, and supporting evaluation documents that can be directly used for production based on the final design scheme.

[0021] The following points should be noted in this article: 1. The accompanying drawings of the embodiments disclosed herein only relate to the structures involved in the embodiments disclosed herein; other structures can be referred to in general design.

[0022] 2. Where there is no conflict, the embodiments of this disclosure and the features in the embodiments can be combined with each other to obtain new embodiments.

[0023] Although embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

Claims

1. A method for generating the box shape and appearance design of agricultural product packaging boxes based on artificial intelligence, characterized in that: The specific steps of the design generation method are as follows: Step 1: Design requirements and basic data collection and preprocessing; Collect packaging design requirements data for target agricultural products, and construct and preprocess a dedicated dataset for agricultural product packaging; Step 2: Multi-dimensional feature extraction and semantic encoding; Multi-dimensional feature extraction and multi-modal semantic encoding are performed on the collected requirement data to generate a joint design feature vector that is fused across modalities; Step 3: Intelligent generation and topology optimization of box-shaped structures; Using the joint design feature vector as input, multiple candidate box schemes are generated through a pre-trained box structure generation model. The candidate box schemes are then topologically optimized by combining multi-physics constraints and multi-objective optimization algorithms to select the optimal box structure scheme. Step 4: Intelligent generation of appearance design; Based on the 3D model and UV layout information of the optimal box structure, and guided by the joint design feature vector, a pre-trained appearance generation model is used to generate an appearance design scheme that is accurately adapted to the box structure, and the adaptability and compliance optimization are completed. Step 5: Multi-objective comprehensive verification and feasibility assessment; A multi-dimensional evaluation system was constructed to comprehensively verify and quantitatively score the design scheme that integrates box shape and appearance, and to select the final candidate design scheme. Step 6: Collaborative Iterative Optimization; Receive user feedback on modifications, perform semantic parsing and feature updates on the feedback, feed it back to the corresponding generation model to complete local iterative optimization of the design scheme, repeat the verification process of steps S3-S5, and obtain the final design scheme confirmed by the user. Step 7: Output production-level files; For the final design scheme, output box structure files, appearance printing files and supporting evaluation files that can be directly used for production.

2. The method for generating the shape and appearance design of agricultural product packaging boxes based on artificial intelligence according to claim 1, characterized in that: Step 1, which involves designing requirements and preprocessing basic data collection, specifically includes: Step 101: Demand Data Collection; The collected demand data includes four main categories: basic attributes of the target agricultural products, storage and transportation environment parameters, brand culture requirements, and production process constraints. Step 102: Constructing a dedicated dataset for agricultural product packaging; The dataset consists of three sub-databases: box structure dataset, appearance design dataset, and packaging engineering knowledge base; Step 103: Data preprocessing; The dataset is cleaned, deduplicated, and standardized. The box structure data is labeled with structural type, load-bearing threshold, material compatibility, and molding difficulty. The appearance design data is labeled with element semantics, applicable categories, color scheme, and compliance. The packaging engineering knowledge base is processed into a knowledge graph and a constraint rule base is constructed.

3. The method for generating the shape and appearance design of agricultural product packaging boxes based on artificial intelligence according to claim 1, characterized in that: The specific steps of step 2, multi-dimensional feature extraction and semantic encoding, are as follows: Step 201: Constraint Feature Extraction; A pre-trained CNN feature extraction network is used to extract numerical features from the basic attributes, storage and transportation environment parameters, and production process constraints of the target agricultural products, generating a box-shaped design constraint feature vector. Step 202: Visual semantic encoding; The pre-trained multimodal large model CLIP is used to perform cross-modal semantic encoding of text and visual elements to meet brand culture requirements, visual style preferences, and compliance labeling requirements, thereby aligning textual requirements with visual features and generating semantic feature vectors for appearance design. Step 203: Cross-modal feature fusion; A feature fusion module based on an attention mechanism is constructed to perform cross-dimensional association and fusion of box-shaped design constraint feature vectors and appearance design semantic feature vectors to generate joint design feature vectors, which serve as the unified input for subsequent generation models.

4. The method for generating the shape and appearance design of agricultural product packaging boxes based on artificial intelligence according to claim 1, characterized in that: The specific steps of step 3, intelligent generation and topology optimization of the box structure, are as follows: Step 301: Initial generation of candidate box types; Using the joint design feature vector as input, the box structure generation model based on the Transformer architecture and fine-tuned by the agricultural product box structure dataset is called to generate multiple sets of parametric 3D structure models and corresponding planar unfolded diagrams of candidate box types. The candidate box types include three core components: the box body, the inner lining buffer structure, and the functional structure. Step 302: Topology optimization under multiphysics constraints; By connecting to the finite element analysis interface and using constraint rules from the packaging engineering knowledge base, multi-dimensional topology optimization is performed on candidate box types, specifically including: Mechanical performance optimization: Simulation calculations were performed on the stacking load-bearing capacity and drop impact resistance of the box type, and the parameters of the box body reinforcing ribs, corner impact protection structure and inner lining buffer structure were optimized to ensure that the storage and transportation load-bearing and damage prevention requirements are met; Preservation function optimization: Optimize the pore size, layout, and number of ventilation holes in the box, or the design of sealing structure and partition layer, to meet the breathing characteristics and preservation requirements of agricultural products in response to their preservation needs. Production process optimization: Optimize the die-cutting path, number of folded edges, overlapping structure, and interlocking design of the box shape, eliminate structures that exceed the process limits of the production equipment, and reduce material waste and production difficulty; Step 303: Multi-objective optimization and screening; The NSGA-II multi-objective optimization algorithm is adopted, with the optimization objectives of maximizing structural strength, minimizing material consumption, and minimizing production difficulty. The optimal solutions of all candidate box types are sorted, and the box structure scheme with the best overall performance is selected by combining the user cost threshold.

5. The method for generating the shape and appearance design of agricultural product packaging boxes based on artificial intelligence according to claim 1, characterized in that: The specific steps for intelligent generation of the appearance design in step 4 are as follows: Step 401: Pre-fitting of the box-shaped structure; Extract the 3D model of the optimal box structure, the UV layout information of the planar unfolded diagram, and the area division information of the box's display surface / folded surface / overlapping surface. Input these into the appearance generation model to establish the coordinate mapping relationship between the appearance pattern and the box structure. Step 402: Generation of the initial design draft; Guided by the semantic features of appearance in the joint design feature vector, the diffusion generation model based on the Stable Diffusion architecture and fine-tuned by the agricultural product appearance design dataset is called to generate an initial draft of appearance design adapted to the box-shaped UV layout. The initial draft includes the main visual pattern, color scheme, product information layout area, brand logo reserved space, and traceability code embedding space. Step 403: Adaptability and compliance optimization; This includes structural adaptation optimization, automatic compliance optimization, and brand consistency optimization.

6. The method for generating the box type and appearance design of agricultural product packaging boxes based on artificial intelligence according to claim 1, characterized in that: The multi-objective comprehensive verification and feasibility assessment in step 5 includes four verification dimensions: structural feasibility verification, production feasibility verification, compliance verification, and visual effect verification. Each dimension has a quantitative scoring threshold, specifically: Step 501: Structural feasibility verification; Check the mechanical properties, molding stability, preservation suitability, and material compatibility of the box shape. If there are no structural defects, it is judged as passing and a structural performance score is generated. Step 502: Production feasibility verification; Verify the feasibility of the die-cutting, printing, and folding processes for the box, calculate the material utilization rate and the production cost per box, and if they meet the production equipment requirements and cost thresholds, they are deemed to have passed. Step 503: Compliance verification; The system checks for compliance with relevant regulations on food packaging, advertising laws, and regulations on the management of agricultural product labeling. If there are no compliance risks, the system is deemed to have passed and a compliance score is generated. Step 504: Visual effect verification; A 360° full-view rendering of the box after folding is generated through 3D rendering. The integrity of the pattern, the presentation of brand elements, and the matching degree of visual style are checked to generate a visual effect score. Step 505: Comprehensive scoring and screening; For schemes that pass the threshold verification in all dimensions, a weighted comprehensive score is calculated based on structural performance, production cost, visual effect, and agricultural product suitability. The top 1-2 schemes with the highest comprehensive scores are selected as the final candidate design schemes. Schemes that fail the threshold verification are returned to the corresponding steps for re-optimization.

7. The method for generating the shape and appearance design of agricultural product packaging boxes based on artificial intelligence according to claim 1, characterized in that: The specific steps of step 6, collaborative iterative optimization, include: Step 601: Solution presentation and feedback reception; The final candidate design scheme's 3D rendering model, plan development diagram, process parameters, and comprehensive evaluation report are presented to the user in a visual manner, and the user's global or partial modification feedback is received. Step 602: Feedback semantic parsing and feature update; A large language model is used to perform semantic parsing on user modification feedback, distinguish between box structure modification requirements and appearance design modification requirements, extract modification features, and update the joint design feature vector. Step 603: Local iterative optimization; The updated feature vectors are fed back to the corresponding box structure generation model or appearance generation model for local parameter adjustment and scheme iteration. There is no need to regenerate the entire scheme. After iteration, the verification process of steps 3-5 is repeated until the user confirms the final design scheme.

8. The method for generating the shape and appearance design of agricultural product packaging boxes based on artificial intelligence according to claim 1, characterized in that: The production-level files output in step 7, production-level file output, specifically include: Box-type structure production documents: vector format plan view, die-cutting line drawing, folding process instructions, packaging material specifications, and inner lining structure processing drawings; Production documents for industrial design: printing source files after CMYK color separation, UV alignment files, marking specifications, and pattern splicing alignment instructions; Supporting evaluation documents: box-type structure mechanical simulation report, production feasibility analysis report, compliance verification report, and 3D rendering.

9. A system for designing the shape and appearance of agricultural product packaging boxes based on artificial intelligence, characterized in that: The system is used to implement the artificial intelligence-based method for generating the box type and appearance design of agricultural product packaging boxes as described in any one of claims 1-8. The system includes a data acquisition and preprocessing module, a multimodal feature extraction and encoding module, a box type structure generation and optimization module, an appearance design generation and optimization module, a multi-objective comprehensive verification module, a human-machine collaborative iteration module, and a production-level file output module. The data acquisition and preprocessing module is used to collect target agricultural product packaging design requirement data, and to construct and preprocess a dedicated dataset for agricultural product packaging. The multimodal feature extraction and encoding module is used to extract multi-dimensional features and encode multimodal semantics from the collected requirement data to generate a joint design feature vector that integrates cross-modal features. The box structure generation and optimization module is used to generate multiple sets of candidate box structure schemes and perform topology optimization, and select the optimal box structure scheme, taking the joint design feature vector as input. The appearance design generation and optimization module is used to generate an appearance design scheme that is precisely adapted to the box shape based on the optimal box shape structure information and to complete the adaptation and compliance optimization. The multi-objective comprehensive verification module is used to construct a multi-dimensional evaluation system to comprehensively verify and quantitatively score the integrated design scheme of box type and appearance, and to screen the final candidate design scheme. The human-machine collaborative iterative module is used to receive user modification feedback, parse the feedback content and update the feature vector to complete the local iterative optimization of the design scheme. The production-level document output module is used to output box structure documents, appearance printing documents, and supporting evaluation documents that can be directly used for production based on the final design scheme.

Images