Intelligent pattern design generation method

Abstract

The invention discloses an intelligent pattern design generation method, and belongs to the technical field of pattern design. The method comprises the following steps: S1, acquiring demand data and style feature data generated by clothing patterns in intelligent pattern design, and performing normalization summarization on the demand data and the style feature data to generate a user demand feature data set; s2, acquiring basic element data of the clothing pattern, analyzing the basic element data based on image segmentation, and extracting characteristic parameters which are frequently retrieved from the basic element data; according to the method, a closed loop is formed through style consistency parameters and design insight data, adaptive screening is produced based on the step S4, and it is ensured that the feasibility of single-piece pattern design is improved; based on double-track optimization in the step S5, the design consistency of the series of patterns is improved; based on the insight iteration reverse training model in the step S6, such as supplementing the national wind element library and adjusting the weight of the generator, the series design development cycle is compressed, the market matching degree is continuously improved, and the pattern design generation efficiency is greatly improved through cooperation of the insight iteration reverse training model, the national wind element library and the generator weight.

Description

Technical Field

[0001] This invention relates to the field of pattern design technology, specifically to an intelligent pattern design generation method. Background Technology

[0002] Pattern design generation is the process of creating graphic works with specific aesthetic value and application functions through the creative combination of visual elements (such as lines, colors, and shapes). As an important carrier of visual communication, patterns are widely used in textiles and clothing, packaging and decoration, print advertising, digital media and other fields. The design quality directly affects the visual appeal and market competitiveness of products.

[0003] Traditional pattern design relies on the professional skills and creative experience of designers, and requires multiple stages such as inspiration, sketching, detail adjustment, and effect rendering to finally form a design solution that meets the requirements. With the development of technologies such as artificial intelligence, computer vision, and deep learning, intelligent methods have gradually penetrated into the field of pattern design, driving the design mode to shift from "human-led" to "human-machine collaboration".

[0004] At the same time, relying on purely aesthetic generation models ignores physical constraints such as clothing pattern parameters, fabric color rendering characteristics, and production process costs, resulting in a large number of designs needing manual rework and adjustments. It lacks visual aesthetics as an optimization goal and fails to integrate production feasibility analysis and user preference feedback, causing the designs to be unacceptable and unused.

[0005] To address the aforementioned technical shortcomings, a solution is proposed. Summary of the Invention

[0006] The purpose of this invention is to provide an intelligent pattern design and generation method to solve the problems mentioned above.

[0007] To achieve the above objectives, the present invention provides the following technical solution: an intelligent pattern design and generation method, comprising the following steps:

[0008] S1. Obtain the demand data and style feature data for clothing pattern generation in intelligent pattern design, and normalize and summarize them to generate a user demand feature dataset.

[0009] S2. Obtain basic element data of clothing patterns, analyze and process the basic element data based on image segmentation, extract frequently accessed feature parameters, and generate a clothing pattern element feature database.

[0010] S3. The acquired demand data, style feature data and element feature database are correlated and trained to build a service design generation model. Based on the preset style constraint threshold and element combination weight of the service design generation model, a model training dataset is generated.

[0011] S4. Generate an initial set of clothing patterns based on the pattern generation model and demand data. Through pattern-clothing category compatibility analysis, select patterns with a matching degree higher than the preset compatibility threshold with clothing style and fabric characteristics, and generate candidate clothing pattern data.

[0012] S5. Obtain user feedback on candidate garment pattern data, construct user feedback data, combine with garment production process feasibility analysis, optimize candidate patterns in detail, and generate optimized garment pattern data; based on the optimized garment pattern data, perform batch generation of garment patterns in the same series, and ensure element correspondence and style unity of the series patterns through style consistency verification, and generate complete series garment pattern data.

[0013] Furthermore, the intelligent pattern design in S1 includes the following processing steps:

[0014] Acquire user aggregation information, which includes users' style preferences, theme requirements, and target group characteristics for clothing patterns. Normalize and summarize the style preferences, theme requirements, and target group characteristics to generate a user demand feature dataset.

[0015] Acquire basic data for clothing categories, including garment pattern parameters, fabric characteristics, and common processes for different categories, and normalize and summarize them to generate a clothing category feature dataset.

[0016] Obtain market distribution information, which includes the popular colors, high-frequency elements and style evolution patterns of clothing patterns over the past N years. N represents a natural number greater than zero, and can take the value N=3. Normalize and summarize the popular colors, popular colors and style evolution patterns to generate a popular style feature dataset.

[0017] The user demand feature dataset, clothing category feature dataset, and popular style feature dataset generated by the aggregation are cleaned and integrated to jointly construct comprehensive demand feature data.

[0018] Furthermore, the process of constructing the clothing pattern element feature database in S2 is as follows:

[0019] We obtained basic element data for clothing patterns, which included traditional clothing patterns, natural elements, modern abstract elements, and popular culture elements. After classifying and summarizing these elements, we constructed an initial element library.

[0020] Based on image segmentation model The patterns in the initial element library are deconstructed and analyzed to separate several independent visual units, and the feature parameters of each independent visual unit are extracted: line features, color features, texture features and composition features.

[0021] The extracted feature parameters are standardized, and then... Clustering algorithms group similar elements into the same feature category, generating a labeled database of clothing pattern element features. The labels contain information such as element style, applicable scenarios, and matching processes, making it easy for the model to call them accurately.

[0022] Furthermore, the process of obtaining the model training dataset in S3 is as follows:

[0023] The system retrieves a pre-stored garment pattern generation model from the intelligent pattern design process. The garment pattern generation model consists of a generator and a discriminator. The generator is responsible for generating garment patterns from input features, and the discriminator is responsible for judging the similarity between the generated pattern and a real high-quality pattern.

[0024] The pattern generation model associates and maps the comprehensive demand feature data in S1 with the clothing pattern element feature database in S2 to construct training samples, taking the comprehensive demand feature data as input and the matched high-quality patterns as output labels.

[0025] Retrieve the pre-stored style constraint threshold, element combination weights, and iteration count from the pattern generation model, and input them into the pattern generation model using the backpropagation algorithm. Calculate the style similarity between the generated pattern and the label pattern every K iterations, where K is a natural number greater than zero, and can be K=100. Retrieve the preset similarity threshold from the pattern generation model and compare it with the style similarity. If the style similarity is greater than the preset similarity threshold, mark the data used in training as the model training dataset, and save the pattern generation model when training is complete.

[0026] Furthermore, the process of generating the S4 candidate clothing pattern data through analysis is as follows:

[0027] Input the user demand feature dataset S1 into the trained clothing pattern generation model to generate P initial clothing patterns. P represents a combination of natural numbers greater than zero, which can take values ​​from 10 to 20. It covers different combinations of elements and composition schemes, and these are marked as the initial pattern set.

[0028] The demand data in S1 is retrieved and compared with the clothing pattern element feature database in S2. The pattern size, color, and detail complexity in the demand data are compared with the clothing pattern fit, fabric color fit, and process fit in the clothing pattern element feature database to construct a pattern-clothing category fit evaluation index.

[0029] Each pattern in the initial pattern set is compared with the pattern-clothing category compatibility evaluation index. At the same time, the compatibility threshold Q, which is a natural number greater than zero and can be Q=70, is retrieved from the intelligent pattern design. Patterns that are higher than the compatibility threshold Q are selected in the comparison process, candidate clothing pattern data is generated, and the compatibility advantages of each pattern are marked.

[0030] Furthermore, the processing procedure for user feedback data in S5 is as follows:

[0031] The system obtains user feedback on candidate garment pattern data, including evaluation dimensions such as aesthetics, uniqueness, and thematic relevance, generating user feedback data. It then collaborates with garment manufacturers to conduct a feasibility analysis of the candidate pattern data, retrieving actual production costs, efficiency, and quality stability data to construct a process feasibility report. Combining user feedback data, the process feasibility report, and pre-stored adjustable percentage thresholds in the intelligent pattern design, the system performs detailed optimization on the candidate garment pattern data, generating optimized garment pattern data: reducing the complexity of details in high-cost patterns, adjusting color schemes with low ratings in user feedback data, and strengthening the proportion of elements with insufficient thematic relevance.

[0032] Furthermore, based on the optimized clothing pattern data, the main color, core elements, and composition logic are extracted from the S1 style feature data and labeled as series pattern style consistency parameters. For different clothing categories in the same series, the pattern size and layout are adjusted according to the pattern characteristics of each category, and style consistency verification is performed accordingly. If there are patterns that do not meet the consistency parameters, their colors, elements, and layouts are adjusted according to the parameter differences until all meet the consistency parameter requirements, thus generating complete series clothing pattern data.

[0033] The beneficial effects of this invention are:

[0034] 1. This invention achieves precise matching between generated patterns and garment production characteristics through a quantitative matching mechanism of pattern-clothing category compatibility evaluation index and preset compatibility threshold Q. This solves the problem of pattern and garment physical properties being disconnected in traditional solutions, significantly improves the first-pass yield of design drafts, significantly reduces sampling and rework costs, and constructs intelligent screening for production compatibility.

[0035] 2. This invention breaks through the limitations of existing technologies by combining user feedback data and process feasibility reports with adjustable proportion thresholds to dynamically adjust design details. It automatically replaces high-cost embroidery elements with equivalent digital printing solutions, significantly reducing the design modification cycle while ensuring market acceptance, and achieving dual-track feedback joint optimization. Attached Figure Description

[0036] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0037] Figure 1 This is a schematic diagram of the method flow of the present invention;

[0038] Figure 2 This is a schematic diagram of the closed-loop process iteration of the present invention. Detailed Implementation

[0039] 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.

[0040] Example 1: Please refer to Figure 1 - Figure 2 As shown, this embodiment is an intelligent pattern design and generation method, which includes the following steps;

[0041] S1. Obtain the demand data and style feature data for clothing pattern generation in intelligent pattern design. The demand data includes user style preferences, clothing category characteristics, and target group characteristics. The style feature data includes popular color trends, classic pattern style characteristics, and fabric compatibility characteristics. These data are obtained by pattern demanders through market research and are then summarized and constructed. During the intelligent pattern design process, these data are normalized and summarized to generate a user demand feature dataset. The intelligent pattern design in S1 includes the following processing steps.

[0042] S11. Obtain user distribution information. User distribution information is obtained through user questionnaires, e-commerce platform review crawling, designer interviews, etc. User distribution information includes obtaining users' style preferences, theme needs, and target group characteristics for clothing patterns. It should be noted that: style preferences include retro, minimalist, Chinese style, etc.; theme needs include festivals, seasons, scenes, etc.; and target group characteristics include age, gender, consumption scenarios, etc. The style preferences, theme needs, and target group characteristics are normalized and summarized for the purpose of eliminating or unifying the dimensions of the obtained data information to generate a user demand feature dataset.

[0043] S12. Obtain basic data for clothing categories. Basic data for clothing categories includes clothing pattern parameters, fabric characteristics, and common processes for different categories. It should be noted that different categories of clothing include T-shirts, dresses, coats, etc. Clothing pattern parameters include garment length, bust, sleeve type, etc. Fabric characteristics include texture / stretchability of cotton, silk, knit, etc. Common processes include digital printing, embroidery, jacquard, etc. Then, normalize and summarize them to generate a clothing category feature dataset.

[0044] S13. Obtain market distribution information. Market distribution information is obtained through fashion trend websites, industry reports, runway analysis, and other channels. Market distribution information includes obtaining the popular colors, high-frequency elements, and style evolution patterns of clothing patterns over the past N years. It should be noted that: popular colors include Pantone's color of the year and seasonal colors; high-frequency elements include flowers, geometric shapes, and cartoon characters; style evolution patterns are influenced by factors such as historical culture, technological development, social concepts, and aesthetic changes. They show the characteristics of pattern inheritance and innovation, from single patterns to multiple patterns, and dynamic adjustment of patterns according to the needs of the times. It is not limited to these. Popular colors, popular color and style evolution patterns are normalized and summarized to generate a popular style feature dataset.

[0045] S14. Clean and integrate the generated user demand feature dataset, clothing category feature dataset, and popular style feature dataset to jointly construct comprehensive demand feature data.

[0046] S15. It should be noted that: the cleaning and integration process involves preprocessing the three types of datasets separately. The user demand feature dataset needs to remove miscellaneous or vague descriptions, such as feedback without clear direction like "okay" or "average". Style preferences such as "retro" and "traditional Chinese style" should be standardized into the standard terminology used in intelligent pattern design. The clothing category feature dataset needs to unify the units of pattern parameters, such as converting "centimeter" and "inch" to a unified unit. Fabric characteristic description deviations, such as "good breathability", should be corrected and quantified into a breathability coefficient range to match the standard terminology used in intelligent pattern design. The popular style feature dataset needs to remove duplicate trend records, such as different sources of descriptions for the same color trend, and unify the time dimension to a "quarterly" granularity to ensure data timeliness.

[0047] S16. Jointly construct a mapping relationship between three types of data, with the clothing category feature dataset as the core association dimension. For example, bind the dress length and collar type of the dress category to the user's style demand for French retro dresses, and then match the popular style features of the category in recent seasons, such as retro floral elements and Morandi colors. Through feature weighted fusion, the weight ratio is pre-stored in the intelligent pattern design, and the possible value ratios are 40% for the user demand feature dataset, 30% for the clothing category feature dataset, and 30% for the popular style feature dataset. Transform the scattered features into a unified vector containing four-dimensional constraints of category, style, element, and process, such as dress - Chinese style - peony element - digital printing. Finally, generate comprehensive demand feature data to provide clear basic constraint boundaries for subsequent pattern generation.

[0048] S2. Obtain basic element data of clothing patterns, parse and process the basic element data based on image segmentation, extract frequently accessed feature parameters, and generate a clothing pattern element feature database; the construction process of the clothing pattern element feature database in S2 is as follows;

[0049] S21. Obtain basic element data for clothing patterns, which includes traditional clothing patterns, natural elements, modern abstract elements, and popular culture elements. It should be noted that: traditional clothing patterns include Miao silver ornament patterns and blue and white porcelain patterns; natural elements include plants and flowers, animal shapes, and landscape textures; modern abstract elements include geometric shapes, gradient colors, and textures; and popular culture elements include film and television IPs and art movement symbols. After classifying and summarizing these elements, construct an initial element library.

[0050] S22, Based on image segmentation model The patterns in the initial element library are deconstructed and analyzed to separate several independent visual units. The feature parameters of each independent visual unit are extracted: line features, color features, texture features, and composition features. It should be noted that these independent visual units may include pattern petal visual units, line visual units, and color block visual units. Line features include thickness, straightness, and continuity; color features include RGB values, brightness, and saturation; texture features include roughness, regularity, and layering; and composition features include symmetry, repetition, and gradient layout. (Image segmentation model) It is an image segmentation model based on convolutional neural networks (CNNs). The patterns in the initial element library are segmented at the pixel level to accurately separate several groups of independent visual units. Then, based on the segmentation results, feature parameters such as line features, color features, texture features, and composition features of each unit are extracted. Finally, the several groups of independent visual units obtained from the segmentation and the extracted feature parameters are combined to form a tagged clothing pattern element feature database, which provides basic element data support that can be accurately called for subsequent pattern generation.

[0051] S23. Standardize the extracted feature parameters, through... Clustering algorithms group similar elements into the same feature category, generating a labeled database of clothing pattern element features. It should be noted that for the same feature category, such as retro floral or geometric abstract, the labels include information such as element style, applicable scenarios, and suitable processes, which facilitates accurate model calls.

[0052] S3. The acquired demand data, style feature data and element feature database are correlated and trained to build a service design generation model. Based on the preset style constraint threshold and element combination weight of the service design generation model, the model training dataset is generated. The process of obtaining the model training dataset in S3 is as follows.

[0053] S31. Retrieve the pre-stored garment pattern generation model in the intelligent pattern design. The garment pattern generation model consists of a generator and a discriminator. The generator is responsible for generating garment patterns from input features, and the discriminator is responsible for judging the similarity between the generated pattern and the real high-quality pattern.

[0054] S32. The clothing pattern generation model associates and maps the comprehensive demand feature data in S1 with the clothing pattern element feature database in S2 to construct training samples. It takes the comprehensive demand feature data as input and the matched high-quality patterns as output labels. It should be noted that the comprehensive demand feature data can extract summer dresses + fresh style, but is not limited to this.

[0055] S33. Retrieve the pre-stored style constraint threshold, element combination weights, and iteration count from the garment pattern generation model. Combine this with the backpropagation algorithm and input it into the garment pattern generation model. Calculate the style similarity between the generated pattern and the label pattern every K iterations. Compare the style similarity with the preset similarity threshold retrieved from the garment pattern generation model. If the style similarity is greater than the preset similarity threshold, mark the data used in training as the model training dataset and save the garment pattern generation model at the end of training. It should be noted that: the style constraint threshold represents the deviation between the generated pattern and the target style, and the deviation can be ≤15%; the element combination weight can be represented as the proportion of floral elements + the proportion of geometric elements, and the proportions can be 30% and 20% respectively, but are not limited to these values; the initial iteration count can be set to 5000.

[0056] S34. An example of using the backpropagation algorithm combined with a garment pattern generation model to optimize the training process of an improved generative adversarial network (GAN). The training sample pairs are input with the demand features of summer dresses and a fresh, clean style, and the matched high-quality patterns are input as output labels. These are then fed into the generator, which generates garment patterns based on the current parameters. The discriminator judges the similarity between the generated pattern and the real high-quality pattern, retrieving a preset probability threshold from the intelligent pattern design database. This threshold can range from 0 to 1; the closer the output probability value is to 1, the closer it is to the real pattern. The classification loss of the discriminator and the generation loss of the generator are calculated. The classification loss can be represented as the loss of the generated pattern if it is mistakenly generated. The error in classifying a pattern as real can be represented by the generation loss, which is the error in generating a pattern that does not meet the features of a real pattern. The loss value is passed from the output layer to the input layer through the backpropagation algorithm, adjusting the weight parameters of each layer of the generator and discriminator, such as convolution kernel parameters and bias terms. The specific values ​​are adjusted according to the actual situation and are not limited to these. In each iteration, the algorithm first optimizes the discriminator to improve its discrimination ability, and then optimizes the generator to deceive the discriminator. This cycle continues until the loss value stabilizes at the probability threshold, such as generation loss < 0.05, to ensure that the matching degree between the pattern generated by the model and the required features and style constraints continues to improve. Finally, a well-trained clothing pattern generation model is output.

[0057] S4. Based on the garment pattern generation model and demand data, an initial set of garment patterns is generated. Through pattern-garment category compatibility analysis, patterns with a matching degree higher than the preset compatibility threshold for garment style and fabric characteristics are selected to generate candidate garment pattern data. The process of generating candidate garment pattern data through analysis in S4 is as follows.

[0058] S41. Input the user demand feature dataset of S1 into the trained clothing pattern generation model to generate P initial clothing patterns, which cover different element combinations and composition schemes, and mark them as the initial pattern set.

[0059] S42. Retrieve the demand data from S1 and compare it with the clothing pattern element feature database in S2. Compare the pattern size, color, and detail complexity in the demand data with the clothing pattern element feature database for clothing fit, fabric color matching, and process matching. Construct a pattern-clothing category compatibility evaluation index. It should be noted that: clothing fit can be expressed as large-area patterns matching loose T-shirts, small-area patterns matching tight-fitting knitwear, etc.; fabric color matching can be expressed as high-saturation patterns matching cotton and linen, low-saturation patterns matching silk, etc.; process matching can be expressed as fine line patterns matching digital printing, rough texture patterns matching embroidery, etc. In summary, compare the size with clothing fit, color with fabric color matching, and detail complexity with process matching.

[0060] S43. Compare each pattern in the initial pattern set with the pattern-clothing category compatibility evaluation index. At the same time, retrieve the pre-stored compatibility threshold Q in the intelligent pattern design, filter out the patterns that are higher than the compatibility threshold Q in the comparison process, generate candidate clothing pattern data, and mark the compatibility advantages of each pattern. The compatibility advantages can be represented as excellent color-fabric compatibility, excellent process-detail compatibility, etc.

[0061] Example 2

[0062] S5. Obtain user feedback on candidate clothing pattern data, construct user feedback data, combine with clothing production process feasibility analysis, optimize candidate patterns in detail, and generate optimized clothing pattern data; based on optimized clothing pattern data, perform batch generation of clothing patterns in the same series, ensure element correspondence and style unity of the series patterns through style consistency verification, and generate complete series clothing pattern data. The process of processing user feedback data in S5 is as follows.

[0063] S51. Obtain user evaluations of candidate garment pattern data, including aesthetics, uniqueness, and thematic relevance, and generate user feedback data; collaborate with garment manufacturers to conduct feasibility analysis on candidate pattern data, assess the actual production cost, efficiency, and quality stability of the patterns, and construct a process feasibility report. It should be noted that user evaluations are obtained through online voting, focus group interviews, etc.; the actual production cost of a pattern can be expressed as the relationship between the printed area and the cost—the larger the printed area, the higher the cost, and vice versa; efficiency can be expressed as the relationship between embroidery complexity and production time—the more complex the embroidery pattern, the longer the production time; quality stability can be expressed as the difference in color reproduction between the pattern and the original pattern reference.

[0064] S52. Combining user feedback data and process feasibility reports, as well as the adjustable percentage thresholds pre-stored in the intelligent pattern design, perform detailed optimization on the candidate garment pattern data to generate optimized garment pattern data. It should be noted that the adjustable percentage thresholds are pre-stored in the intelligent pattern design and can be adjusted as needed. Here, the values ​​can be set to 60% for user feedback data and 40% for process feasibility reports. Detail optimization can be expressed as reducing the complexity of details in high-cost patterns, adjusting color matching with low evaluation in user feedback data, and strengthening the proportion of elements with insufficient theme fit.

[0065] S53. Based on the optimized garment pattern data, extract the main color tone, core elements, and composition logic from the style feature data in S1, and label them as series pattern style consistency parameters. It should be noted that the main color tone, core elements, and composition logic are all taken from a portion of the style feature data, extracted and constructed separately as needed. The deviation of the main color tone can be ≤10%, the repetition rate of the core elements can be ≥60%, and the composition rhythm represents the density distribution, referring to the spatial layout density difference of garment pattern elements on the garment layout. It is a key feature for measuring the composition rhythm of the pattern and needs to be maintained. For example, the patterns on T-shirts and jackets in the same series should maintain a consistent layout logic between densely packed areas (main pattern on the chest) and sparsely spaced areas (embellishments on the cuffs). This avoids one item having an overly crowded pattern while another is too loose, thus achieving visual harmony and stylistic unity in the series of garment patterns. For different garment categories within the same series, the pattern size and layout are adjusted according to the characteristics of each category's cut. Stylistic consistency is then verified. If any patterns do not meet the consistency parameters, their colors, elements, and layout are adjusted based on the parameter differences until all meet the consistency parameter requirements, generating complete series garment pattern data.

[0066] After generating and constructing the complete series of garment pattern data, iterative closed-loop processing can be performed according to the needs of intelligent pattern design generation, thereby constructing step S6, the specific steps of which are as follows;

[0067] S6. Integrate the demand data, clothing pattern element feature database, model training dataset, candidate clothing pattern data, user feedback data, and complete series clothing pattern data to generate clothing pattern design insight data, providing an iterative basis for subsequent design. The data integration and iterative data processing process in S6 is as follows.

[0068] S61. Collect the user demand feature dataset in S1, the clothing pattern element feature database in S2, the model training dataset in S3, the candidate clothing pattern data in S4, the user feedback data in S5, and the complete series of clothing pattern data, and store them in association according to the logical chain of demand-element-generation-feedback-output.

[0069] S62. Perform statistical analysis on the collected data to generate clothing pattern design insight data. This data includes high-frequency demand styles, highly adaptable elements, and model optimization directions. It should be noted that: high-frequency demand styles, such as the proportion of "Chinese style" demand in the past six months, can be taken as 35%; highly adaptable elements, such as the average adaptation score of floral elements on dresses, can be taken as 82; and model optimization directions, such as improving the accuracy of "minimalist style" generation, can be taken as 10%.

[0070] S63. Feedback design insight data to previous steps: update the clothing pattern element feature database to supplement high-demand elements, adjust the model training dataset to optimize the training weights of styles with low accuracy, and refine the user demand feature dataset to add sustainable design-related demand items, forming a design iteration closed loop.

[0071] Combining Examples 1 and 2, a closed loop is formed by style consistency parameters and design insight data. Based on the production adaptation screening in step S4, the feasibility of individual pattern designs is improved. Based on the dual-track optimization in step S5, the consistency of series pattern designs is improved. Based on the insight iteration reverse training model in step S6, such as supplementing the Chinese style element library and adjusting the generator weights, the three work together to compress the series design and development cycle, continuously improve market matching, and significantly improve the efficiency of pattern design generation.

[0072] The above description is merely an example and illustration of the structure of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the structure of the invention or exceed the scope defined in the claims, all of which should fall within the protection scope of the present invention.

[0073] In the description of this specification, the references to terms such as "an embodiment," "example," and "specific example" indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described can be combined in any suitable manner in one or more embodiments or examples. The system performs dimensional unification processing on the collected / acquired data, retrieved historical / related data, and preset thresholds, or converts them into dimensionless values ​​to ensure consistency in calculation, comparison, and decision-making (such as standardization, normalization, proportionalization, etc.), eliminating the influence of dimensional differences on the analysis results, but is not limited to this.

[0074] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. An intelligent pattern design and generation method, characterized in that, Includes the following steps: S1. Obtain the demand data and style feature data for clothing pattern generation in intelligent pattern design, and normalize and summarize them to generate a user demand feature dataset. S2. Obtain basic element data of clothing patterns, analyze and process the basic element data based on image segmentation, extract frequently accessed feature parameters, and generate a clothing pattern element feature database. S3. The acquired demand data, style feature data and element feature database are correlated and trained to build a service design generation model. Based on the preset style constraint threshold and element combination weight of the service design generation model, a model training dataset is generated. S4. Generate an initial set of clothing patterns based on the pattern generation model and demand data. Through pattern-clothing category compatibility analysis, select patterns with a matching degree higher than the preset compatibility threshold with clothing style and fabric characteristics, and generate candidate clothing pattern data. S5. Obtain user feedback on candidate clothing pattern data, construct user feedback data, combine with clothing production process feasibility analysis, optimize candidate patterns in detail, and generate optimized clothing pattern data. Based on the optimized clothing pattern data, batch generation of clothing patterns for the same series is performed. Style consistency verification is used to ensure that the elements of the series patterns echo each other and the style is unified, thus generating complete series clothing pattern data.

2. The intelligent pattern design and generation method according to claim 1, characterized in that, The S1 type of intelligent pattern design includes the following processing steps; Acquire user aggregation information, which includes users' style preferences, theme requirements, and target group characteristics for clothing patterns. Normalize and summarize the style preferences, theme requirements, and target group characteristics to generate a user demand feature dataset. Obtain basic data for clothing categories, including garment pattern parameters, fabric characteristics, and common processes for different categories, and normalize and summarize them to generate a clothing category feature dataset.

3. The intelligent pattern design and generation method according to claim 2, characterized in that, Acquire market distribution information, which includes the popular colors, high-frequency elements and style evolution patterns of clothing patterns over the past N years. Normalize and summarize the popular colors, popular colors and style evolution patterns to generate a popular style feature dataset. The user demand feature dataset, clothing category feature dataset, and popular style feature dataset generated by the aggregation are cleaned and integrated to jointly construct comprehensive demand feature data.

4. The intelligent pattern design and generation method according to claim 1, characterized in that, The process of constructing the clothing pattern element feature database in S2 is as follows: We obtained basic element data for clothing patterns, which included traditional clothing patterns, natural elements, modern abstract elements, and popular culture elements. After classifying and summarizing these elements, we constructed an initial element library. Based on image segmentation model The patterns in the initial element library are deconstructed and analyzed to separate several independent visual units, and the feature parameters of each independent visual unit are extracted: line features, color features, texture features and composition features. The extracted feature parameters are standardized, and then... Clustering algorithms group similar elements into the same feature category, generating a labeled database of clothing pattern element features.

5. The intelligent pattern design and generation method according to claim 1, characterized in that, The process of obtaining the model training dataset in S3 is as follows; The system retrieves the pre-stored garment pattern generation model from the intelligent pattern design. The garment pattern generation model consists of a generator and a discriminator. The generator is responsible for generating garment patterns from input features, and the discriminator is responsible for judging the similarity between the generated pattern and the real high-quality pattern. The pattern generation model associates and maps the comprehensive demand feature data in S1 with the clothing pattern element feature database in S2 to construct training samples. It takes the comprehensive demand feature data as input and the matched high-quality patterns as output labels.

6. The intelligent pattern design and generation method according to claim 5, characterized in that, Retrieve the pre-stored style constraint threshold, element combination weights, and iteration count from the pattern generation model, and input them into the pattern generation model using the backpropagation algorithm. Calculate the style similarity between the generated pattern and the label pattern every K iterations. Retrieve the preset similarity threshold from the pattern generation model and compare it with the style similarity. If the style similarity is greater than the preset similarity threshold, mark the data used in training as the model training dataset, and save the pattern generation model when training is complete.

7. The intelligent pattern design and generation method according to claim 1, characterized in that, The process of generating the S4 candidate clothing pattern data through analysis is as follows; Input the user demand feature dataset of S1 into the trained garment pattern generation model to generate P initial garment patterns, which cover different element combinations and composition schemes, and mark them as the initial pattern set. The demand data in S1 is retrieved and compared with the clothing pattern element feature database in S2. The pattern size, color, and detail complexity in the demand data are compared with the clothing pattern fit, fabric color fit, and process fit in the clothing pattern element feature database to construct a pattern-clothing category fit evaluation index.

8. The intelligent pattern design and generation method according to claim 7, characterized in that, Each pattern in the initial pattern set is compared with the pattern-clothing category compatibility evaluation index. At the same time, the compatibility threshold Q pre-stored in the intelligent pattern design is retrieved, and patterns that are higher than the compatibility threshold Q in the comparison process are selected to generate candidate clothing pattern data and mark the compatibility advantages of each pattern.

9. The intelligent pattern design and generation method according to claim 1, characterized in that, The process by which S5 processes user feedback data is as follows: We obtain user evaluations of candidate clothing patterns, including aesthetics, uniqueness, and thematic relevance, and generate user feedback data. We also collaborate with clothing manufacturers to conduct feasibility analysis on the candidate pattern data, retrieve actual production costs, efficiency, and quality stability of the patterns, and construct a process feasibility report. By combining user feedback data and process feasibility reports, as well as the adjustable proportion thresholds pre-stored in intelligent pattern design, the candidate garment pattern data is optimized in detail to generate optimized garment pattern data.

10. The intelligent pattern design and generation method according to claim 9, characterized in that, Based on the optimized clothing pattern data, the main color, core elements, and composition logic are extracted from the S1 style feature data and labeled as series pattern style consistency parameters. For different clothing categories in the same series, the pattern size and layout are adjusted according to the pattern characteristics of each category. The style consistency is verified accordingly. If there are patterns that do not meet the consistency parameters, their colors, elements, and layouts are adjusted according to the parameter differences until all meet the consistency parameter requirements, thus generating complete series clothing pattern data.

Images