Intelligent patchwork pattern design optimization system and method

Abstract

The invention relates to the technical field of intelligent patchwork pattern design, and discloses an intelligent patchwork pattern design optimization system and method.The intelligent patchwork pattern design optimization system is characterized in that (1) a pattern input module is used for receiving initial patchwork element patterns and style requirements input by a user and supporting image uploading, graph drawing or pattern model selection; (2) a pattern feature extraction module for extracting shape features, color vectors, texture distribution and symmetry indexes of an input pattern based on a convolutional neural network (CNN), (3) a patchwork template generation module for segmenting the pattern into basic unit blocks by using a segmentation algorithm and constructing a patchwork template capable of being spliced; the pattern optimization module is used for optimizing the combination mode of the cloth splicing units through a genetic algorithm or a differential evolution algorithm; (5) a pattern evaluation module: scoring the visual aesthetic feeling of the optimized pattern based on an Aestpic Score model, and feeding back to an optimization module; and (6) a pattern output module for outputting the optimized patchwork pattern and supporting a digital embroidery machine or printer interface.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent patchwork pattern design, and in particular to an intelligent patchwork pattern design optimization system and method. Background Art

[0002] Patchwork is an ancient pattern design and fabric processing method widely used in clothing, home textiles, handicrafts, and other fields. It creates aesthetically pleasing, functional, and personalized patterns by cutting and reassembling pieces of fabric in various shapes, colors, and materials. Traditional patchwork design relies heavily on manual experience. Designers typically create patterns and combine pieces based on intuition, aesthetic preferences, and hand-drawing. This is not only time-consuming and labor-intensive, but also difficult to achieve optimal pattern complexity, aesthetic harmony, and fabric efficiency.

[0003] In recent years, although some computer-aided design (CAD) software has been applied to patchwork pattern design, these systems are mostly limited to graphic drawing and style editing, and lack the functions of intelligent analysis, automatic optimization and aesthetic evaluation of pattern structure. In particular, there is still a large technical gap in terms of meeting the user's personalized aesthetic while taking into account the coordination of patchwork piece shapes, color matching and adaptability of weaving equipment.

[0004] Furthermore, existing pattern design tools often lack sufficient integration of computer vision and intelligent optimization algorithms, making it difficult to quantitatively identify pattern style features and automatically optimize their combination methods. On the one hand, factors such as color matching and structural symmetry within a pattern significantly impact the aesthetics of the finished product, but traditional design tools are unable to effectively measure and optimize them. On the other hand, in practical patchwork applications, such as industrial embroidery or automated sewing, pattern output must be converted into a control instruction format that can be recognized by the device, placing higher demands on the scalability of the design system.

[0005] In summary, a patchwork pattern design system that combines image recognition, deep learning, and intelligent optimization algorithms is needed. It can automatically extract pattern features, perform combination optimization based on user input preferences, and output control data adapted to weaving equipment in digital form, thereby greatly improving the intelligence level and production efficiency of patchwork pattern design. Summary of the Invention

[0006] The purpose of the present invention is to address the shortcomings of the prior art and provide an intelligent patchwork pattern design optimization system and method. To achieve the above purpose, the present invention provides the following technical solutions:

[0007] An intelligent patchwork pattern design optimization system, comprising:

[0008] (S1) Pattern input module: used to receive the initial patchwork element pattern and style requirements input by the user, supporting image uploading, graphic drawing or pattern template selection;

[0009] (S2) Pattern feature extraction module: Based on the convolutional neural network (CNN), the shape features, color vectors, texture distribution and symmetry index of the input pattern are extracted and the output feature vector F = [f1, f2, ..., f n ];

[0010] (S3) Patchwork template generation module: using a segmentation algorithm (such as K-means image segmentation or Canny algorithm based on edge detection) to divide the pattern into basic unit blocks and construct a patchwork template that can be spliced;

[0011] (S4) Pattern Optimization Module: Optimize the patchwork unit combination mode through genetic algorithm or differential evolution algorithm to minimize the following objective function:

[0012] L=αD c +βD s +γC f

[0013] in:

[0014] D c is the sum of squares of color differences between adjacent blocks of color;

[0015] D s is the structural symmetry loss function;

[0016] C f is the cosine distance between the user preference and the current pattern style;

[0017] (S5) Pattern evaluation module: scores the visual aesthetics of the optimized pattern based on the Aesthetic Score model (e.g., NIMA) and feeds back to the optimization module;

[0018] (S6) Pattern output module: outputs the optimized patchwork pattern in SVG or high-resolution PNG format and supports digital embroidery machine or printer interface.

[0019] Preferably, the pattern feature extraction module is trained using a ResNet-50 network structure with an output channel number of 2048, and features are reduced to 128-dimensional vectors through PCA (principal component analysis) for subsequent matching and style clustering.

[0020] Preferably, the patchwork template generation module includes a pattern segmentation method based on boundary curvature, and uses the following formula for corner detection:

[0021]

[0022] where θ i is the direction angle of the i-th edge point, l i The unit side length is used to locate the area with sudden shape changes and perform boundary cutting.

[0023] Preferably, the genetic algorithm in the pattern optimization module comprises the following steps:

[0024] (1) Initialize the population P = {x1, x2, ..., x n}, where each individual represents a patchwork combination;

[0025] (2) Individual selection using roulette wheel selection method;

[0026] (3) Use the two-point crossover operator to combine individuals;

[0027] (4) A mutation operator based on local perturbation is used to enhance diversity, and the mutation probability is set to 0.1;

[0028] (5) In each iteration, the fitness function L is calculated for the individuals, and the top 10% of the best individuals are retained for elite retention.

[0029] Preferably, the square sum of color differences between adjacent blocks of color in the objective function D c The calculation is as follows:

[0030]

[0031] Among them C i represents the RGB color vector of the i-th block, and N(i) is the set of blocks adjacent to the i-th block.

[0032] Preferably, the pattern optimization module further includes a style constraint mechanism, using clustering style centers S k The cosine distance between the style vector F and the current pattern is used as the style loss:

[0033]

[0034] And set the threshold θ f =0.85, triggering soft constraint adjustment when the style deviates from the target cluster center.

[0035] Preferably, the pattern evaluation module adopts an image aesthetic evaluation network fine-tuned by the Inception-v3 model, with an output score range of 0-10, and the condition for the pattern to be fed back into the optimization module is that the aesthetic score is lower than 6.0 points.

[0036] Preferably, the system supports user-defined patchwork fabric texture library and introduces texture repetition rate penalty term T during optimization. r , the modified objective function is as follows:

[0037]

[0038] Where δ(t i =t j )=1 means that adjacent blocks use the same texture cloth, λ is the penalty coefficient, and the default value is 0.3.

[0039] Preferably, the pattern output module includes a G-code conversion engine for weaving printing or embroidery equipment, which can convert the optimized pattern into G-code control instructions and support command set formats of industrial embroidery equipment models including Brother, Bernina, etc.

[0040] Preferably, the method comprises the following steps:

[0041] (1) The user inputs the initial pattern and style requirements;

[0042] (2) extracting pattern feature vector F;

[0043] (3) Generate basic patchwork templates and quantify segmentation units;

[0044] (4) Optimize the patchwork combination through genetic algorithm to minimize the objective function L = αD c +βD s +γC f ;

[0045] (5) Iterative optimization based on the evaluation module scoring feedback;

[0046] (6) Output optimized pattern file or G-code control instruction;

[0047] If the user feedback is not satisfactory, the system records the preference vector and updates the style database for secondary optimization.

[0048] Compared with the prior art, the intelligent patchwork pattern design optimization system and method provided by the present invention has the following beneficial effects:

[0049] (1) Significantly improve design efficiency

[0050] Image recognition and deep convolutional neural networks (CNNs) are used to automatically extract style features (such as color distribution, geometric shape, texture, etc.) of user-input patterns or samples, and quickly reconstruct patterns based on the constructed style vector space, significantly reducing manual drawing and typesetting time.

[0051] (2) Enhance pattern aesthetic optimization capabilities

[0052] The present invention introduces a multi-objective optimization algorithm (such as the improved NSGA-II) to convert color coordination, symmetry, aesthetic score, pattern complexity, fabric utilization, etc. into quantifiable indicators. The optimal pattern design scheme is screened through the Pareto frontier solution set to ensure that the design results are both aesthetically pleasing and practical.

[0053] (3) Automatically generate personalized patterns

[0054] The system supports users to input preferred parameters (such as favorite color style, pattern symmetry requirements, complexity level, etc.), and combines variational autoencoder (VAE) or generative adversarial network (GAN) models to achieve personalized generation of pattern styles to meet users' diverse customization needs.

[0055] (4) Improve fabric utilization and process adaptability

[0056] The system effectively reduces fabric waste by introducing a two-dimensional fabric arrangement optimization algorithm (such as an improved ant colony algorithm), and automatically generates cutting paths and stitching sequences based on the output pattern structure, providing precise control instructions for subsequent automated sewing equipment.

[0057] (5) Ability to reuse patterns and innovate combinations

[0058] The present invention supports extracting sub-blocks from existing patterns, constructing a pattern library and a block template database, and can generate new patterns through reorganization and rearrangement, thereby improving the creative scalability and reusability of the design and promoting modular innovation of patchwork patterns.

[0059] (6) Support multi-platform collaborative output

[0060] The generated patterns support multiple output formats, such as vector graphics (SVG), CAD drawings (.dxf), and machine embroidery control codes (such as .dst, .exp, etc.). They can be directly adapted to embroidery machines, laser cutting machines, sewing machines and other equipment, opening up the digital process from design to manufacturing.

[0061] (7) Human-computer interaction and iterative optimization can be realized

[0062] The system is equipped with a visual interactive interface that allows users to fine-tune, score and provide feedback based on the generated patterns. The system automatically records user behavior and uses it to optimize algorithm updates, realizing a "human-machine co-creation" design model and continuously improving the quality of pattern recommendations.

[0063] (8) Support the integration of multi-style and cross-cultural patchwork art

[0064] The system uses style transfer technology (such as StyleGAN) to support the fusion and re-creation of different Chinese and Western style patterns, expand the cultural expressiveness of patchwork design, and promote the diversified development of patchwork art. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 The figure is a schematic diagram of the process structure of an intelligent patchwork pattern design optimization system and method of the present invention. DETAILED DESCRIPTION

[0066] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments described in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0067] In order to enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings.

[0068] like Figure 1 As shown, an embodiment of the present invention provides an intelligent patchwork pattern design optimization system and method, an intelligent patchwork pattern design optimization system, characterized in that the system includes:

[0069] (S1) Pattern input module: used to receive the initial patchwork element pattern and style requirements input by the user, supporting image uploading, graphic drawing or pattern template selection;

[0070] (S2) Pattern feature extraction module: Based on the convolutional neural network (CNN), the shape features, color vectors, texture distribution and symmetry index of the input pattern are extracted and the output feature vector F = [f1, f2, ..., f n ];

[0071] (S3) Patchwork template generation module: using a segmentation algorithm (such as K-means image segmentation or Canny algorithm based on edge detection) to divide the pattern into basic unit blocks and construct a patchwork template that can be spliced;

[0072] (S4) Pattern Optimization Module: Optimize the patchwork unit combination mode through genetic algorithm or differential evolution algorithm to minimize the following objective function:

[0073] L=αD c +βD s +γC f

[0074] in:

[0075] D c is the sum of squares of color differences between adjacent blocks of color;

[0076] D s is the structural symmetry loss function;

[0077] Cf is the cosine distance between the user preference and the current pattern style;

[0078] (S5) Pattern evaluation module: scores the visual aesthetics of the optimized pattern based on the Aesthetic Score model (e.g., NIMA) and feeds back to the optimization module;

[0079] (S6) Pattern output module: outputs the optimized patchwork pattern in SVG or high-resolution PNG format and supports digital embroidery machine or printer interface.

[0080] Furthermore, the pattern feature extraction module is trained using the ResNet-50 network structure with an output channel number of 2048, and PCA (principal component analysis) is used to reduce the feature dimension to a 128-dimensional vector for subsequent matching and style clustering.

[0081] The patchwork template generation module includes a pattern segmentation method based on boundary curvature and uses the following formula for corner detection:

[0082]

[0083] where θ i is the direction angle of the i-th edge point, l i The unit side length is used to locate the area with sudden shape changes and perform boundary cutting.

[0084] In one embodiment of the present invention, the genetic algorithm in the pattern optimization module includes the following steps:

[0085] (1) Initialize the population P = {x1, x2, ..., x n}, where each individual represents a patchwork combination;

[0086] (2) Individual selection using roulette wheel selection method;

[0087] (3) Use the two-point crossover operator to combine individuals;

[0088] (4) A mutation operator based on local perturbation is used to enhance diversity, and the mutation probability is set to 0.1;

[0089] (5) In each iteration, the fitness function L is calculated for the individuals, and the top 10% of the best individuals are retained for elite retention.

[0090] The sum of squares of color differences between adjacent color blocks in the objective function D c The calculation is as follows:

[0091]

[0092] Among them C irepresents the RGB color vector of the i-th block, and N(i) is the set of blocks adjacent to the i-th block.

[0093] The pattern optimization module further includes a style constraint mechanism, which uses clustering style centers S k The cosine distance between the style vector F and the current pattern is used as the style loss:

[0094]

[0095] And set the threshold θ f =0.85, triggering soft constraint adjustment when the style deviates from the target cluster center.

[0096] The pattern evaluation module uses an image aesthetic evaluation network fine-tuned by the Inception-v3 model, with an output score range of 0-10. The condition for the pattern to be fed back into the optimization module is that the aesthetic score is lower than 6.0 points.

[0097] The system supports user-defined patchwork fabric texture library and introduces texture repetition rate penalty term T during optimization. r , the modified objective function is as follows:

[0098]

[0099] Where δ(t i =t j )=1 means that adjacent blocks use the same texture cloth, λ is the penalty coefficient, and the default value is 0.3.

[0100] The pattern output module includes a G-code conversion engine for textile printing or embroidery equipment, which can convert optimized patterns into G-code control instructions and support command set formats of industrial embroidery equipment models including Brother and Bernina.

[0101] Another embodiment of the present invention is: a pattern design optimization method of the intelligent patchwork pattern design optimization system, comprising the following steps:

[0102] (1) The user inputs the initial pattern and style requirements;

[0103] (2) extracting pattern feature vector F;

[0104] (3) Generate basic patchwork templates and quantify segmentation units;

[0105] (4) Optimize the patchwork combination through genetic algorithm to minimize the objective function L = αD c +βD s +γC f ;

[0106] (5) Iterative optimization based on the evaluation module scoring feedback;

[0107] (6) Output optimized pattern file or G-code control instruction;

[0108] If the user feedback is not satisfactory, the system records the preference vector and updates the style database for secondary optimization.

[0109] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent patchwork pattern design optimization system, characterized in that: The system includes: (1) Pattern input module: used to receive the initial patchwork element pattern and style requirements input by the user, supporting image uploading, graphic drawing or pattern template selection; (2) Pattern feature extraction module: Based on the convolutional neural network (CNN), the shape features, color vectors, texture distribution and symmetry index of the input pattern are extracted and the output feature vector F = [f1, f2, ..., f n ]; (3) Patchwork template generation module: Use segmentation algorithms (such as K-means image segmentation or Canny algorithm based on edge detection) to divide the pattern into basic unit blocks and construct a patchwork template that can be spliced; (4) Pattern optimization module: Optimize the patchwork unit combination mode through genetic algorithm or differential evolution algorithm to minimize the following objective function: L=αD c +βD s +γC f in: D c is the sum of squares of color differences between adjacent blocks of color; D s is the structural symmetry loss function; C f is the cosine distance between the user preference and the current pattern style; (5) Pattern evaluation module: scores the visual aesthetics of the optimized pattern based on the Aesthetic Score model (such as NIMA) and feeds it back to the optimization module; (6) Pattern output module: Output the optimized patchwork pattern in SVG or high-resolution PNG format and support digital embroidery machine or printer interface.

2. The intelligent patchwork pattern design optimization system according to claim 1, characterized in that: The pattern feature extraction module is trained using a ResNet-50 network structure with 2048 output channels, and PCA (principal component analysis) is used to reduce the feature dimension to a 128-dimensional vector for subsequent matching and style clustering.

3. The intelligent patchwork pattern design optimization system according to claim 1, characterized in that: The patchwork template generation module includes a pattern segmentation method based on boundary curvature, and uses the following formula for corner detection: where θ i is the direction angle of the i-th edge point, l i The unit side length is used to locate the area with sudden shape changes and perform boundary cutting.

4. The intelligent patchwork pattern design optimization system according to claim 1, characterized in that: The genetic algorithm in the pattern optimization module comprises the following steps: (1) Initialize the population P = {x1, x2, ..., x n }, where each individual represents a patchwork combination; (2) Individual selection using roulette wheel selection method; (3) Use the two-point crossover operator to combine individuals; (4) A mutation operator based on local perturbation is used to enhance diversity, and the mutation probability is set to 0.1; (5) In each iteration, the fitness function L is calculated for the individuals, and the top 10% of the best individuals are retained for elite retention.

5. The intelligent patchwork pattern design optimization system according to claim 1, characterized in that: The objective function is the sum of squares of color differences between adjacent blocks of color D c The calculation is as follows: Among them C i represents the RGB color vector of the i-th block, and N(i) is the set of blocks adjacent to the i-th block.

6. The intelligent patchwork pattern design optimization system according to claim 1, characterized in that: The pattern optimization module further includes a style constraint mechanism, using clustering style centers S k The cosine distance between the style vector F and the current pattern is used as the style loss: And set the threshold θ f =0.85, triggering soft constraint adjustment when the style deviates from the target cluster center.

7. The intelligent patchwork pattern design optimization system according to claim 1, characterized in that: The pattern evaluation module uses an image aesthetic evaluation network fine-tuned by the Inception-v3 model, with an output score range of 0-10. The condition for a pattern to be fed back into the optimization module is that the aesthetic score is lower than 6.0 points.

8. The intelligent patchwork pattern design optimization system according to claim 1, characterized in that: The system supports user-defined patchwork fabric texture library and introduces texture repetition rate penalty term T during optimization r , the modified objective function is as follows: Where δ(t i =t j )=1 means that adjacent blocks use the same texture cloth, λ is the penalty coefficient, and the default value is 0.

3.

9. The intelligent patchwork pattern design optimization system according to claim 1, characterized in that: The pattern output module includes a G-code conversion engine for weaving printing or embroidery equipment, which can convert optimized patterns into G-code control instructions and support command set formats of industrial embroidery equipment models including Brother and Bernina.

10. A pattern design optimization method using the above-mentioned intelligent patchwork pattern design optimization system, characterized in that: The following steps are involved: (1) The user inputs the initial pattern and style requirements; (2) extracting pattern feature vector F; (3) Generate basic patchwork templates and quantify segmentation units; (4) Optimize the patchwork combination through genetic algorithm to minimize the objective function L = αD c +βD s +γC f ; (5) Iterative optimization based on the evaluation module scoring feedback; (6) Output optimized pattern file or G-code control instruction; If the user feedback is not satisfactory, the system records the preference vector and updates the style database for secondary optimization.

Images