A method and system for generating a braided product, a computer device and a storage medium

Abstract

This invention provides a method, system, computer equipment, and storage medium for generating Qiang woven products, belonging to the field of digital protection and innovative design of intangible cultural heritage. The method includes: collecting historical records of sheep totemism and physical examples of handmade Qiang woven crafts; extracting vectorized feature curves and visual features of the craftsmanship using an improved Mask R-CNN model; constructing a Qiang cultural gene library containing visual features, semantic features, and translation rules based on grounded theory; transforming the gene library into a cue word matrix, design intention graphics, and creative sketches using shape grammar rules, and inputting them into an AIGC model to generate a candidate solution set; dynamically adjusting the AIGC model parameters through a multi-strategy enhanced escape algorithm, and finally outputting the optimal design elements and generating Qiang woven products. This achieves accurate extraction and intelligent translation of Qiang cultural genes, balancing the accuracy of cultural inheritance with the efficiency of design innovation, and promoting the intelligent transformation of intangible cultural heritage from protection to innovation.

Description

Technical Field

[0001] This invention belongs to the field of digital protection and innovative design of intangible cultural heritage, specifically involving a method, system, computer equipment and storage medium for producing Qiang weaving products. Background Technology

[0002] The sheep totem symbolizes "good fortune." The Qiang people's sheep totem can be traced back to the image of the argali sheep 6,000 years ago. The Qiang people have worshipped the sheep totem for thousands of years, forming its inherent symbolic meaning. Qiang weaving, as one of China's important intangible cultural heritages, possesses profound cultural connotations and unique craftsmanship. In recent years, with the development of digital technology and artificial intelligence, the protection, inheritance, and innovation of intangible cultural heritage have gradually moved towards intelligent and data-driven approaches.

[0003] Currently, traditional design methods related to the innovative design of Qiang weaving products rely on the experience and craftsmanship of inheritors, using methods such as hand-drawn patterns and physical reproduction for product design. This method emphasizes the inheritance of skills, but suffers from low innovation efficiency and difficulty in scaling up. Another approach utilizes CAD software for pattern design and modeling, improving efficiency and accuracy, but still requires manual extraction and translation of cultural symbols, heavily relying on the designer's cultural understanding. Existing technologies also construct mapping models between cultural symbols and design elements through schema theory and grounded theory to translate cultural connotations into product design; however, research has found that model construction is highly dependent on subjective experience, resulting in poor reproducibility and objectivity.

[0004] While generative artificial intelligence (AIGC), as a mature existing technology, has been widely applied in general image and text generation and has demonstrated strong creative potential, its application in the innovative design of intangible cultural heritage such as Qiang weaving, which highly relies on specific cultural contexts, still faces fundamental challenges. The root cause is that the excellent performance of AIGC models depends on large-scale, high-quality training data, while for specialized fields like Qiang culture, there is a lack of systematically organized and labeled datasets specifically designed for it. This directly results in existing general-purpose AIGC models failing to deeply understand the unique symbolic system and deep semantics of Qiang culture, often leading to significant deviations in cultural connotation and accuracy in their generated results. Therefore, to effectively apply AIGC technology in the innovative design of Qiang weaving products, the primary prerequisite is to solve the problem of constructing a dataset specifically for intangible cultural heritage.

[0005] Therefore, in the innovative design of Qiang weaving products, AIGC technology faces the problem of difficulty in constructing a dedicated dataset for intangible cultural heritage, which urgently needs to be addressed. Summary of the Invention

[0006] To address the aforementioned problems, this invention provides a method, system, computer equipment, and storage medium for producing Qiang woven products.

[0007] To achieve the above objectives, the present invention provides a method for producing Qiang woven products, comprising:

[0008] Collect documents and images of sheep totem records from Qiang culture and handicrafts made of Qiang cloth.

[0009] The process involves extracting the spiral curves and contour features from sheep totem records, locating key inflection points of these features, and obtaining vectorized feature curves representing cultural semantics. Visual features of weaving lines and pattern units from images of handcrafted Qiang handicrafts are extracted, and the direction of these weaving lines and the product outline are located, outputting the craft's visual features. Based on grounded theory, the vectorized feature curves and craft visual features are encoded to generate a Qiang cultural gene library containing visual features, semantic features, and translation rules. Combining shape grammar rules, the Qiang cultural gene library is transformed into a Prompt word matrix, design intention graphics, and creative sketches. Independent Qiang weaving design elements are extracted from the sheep totem records and images of handcrafted Qiang handicrafts to form a design element image library.

[0010] Using the Prompt word matrix, design intention graphics, and creative sketches as input, and the design element image library as ground truth labels, an AIGC model is trained; an AIGC model that can output the optimal Qiang weaving product design elements is obtained, and Qiang weaving products are generated using the Qiang weaving product design elements.

[0011] Preferably, the improved Mask R-CNN model is used to extract vectorized feature curves and craft visual features from the sheep totem record documents and images of handmade Qiang handicrafts, respectively; the improved Mask R-CNN model inserts an SE attention module after the ReLU activation function of the ResNet backbone network of the traditional Mask R-CNN model and before the residual connection fusion; edge detection is added after the multi-scale feature output layer of the ResNet backbone network.

[0012] Preferably, the improved Mask R-CNN model is used to extract vectorized feature curves and craft visual features from sheep totem records and images of hand-woven Qiang handicrafts, respectively. Specifically, this includes: filtering background interference information from the sheep totem records using the SE attention module, extracting the spiral curves and contour features of the filtered sheep totem records, and then locating the key inflection points of the spiral curves and contour features using the edge detection module to obtain vectorized feature curves representing cultural semantics; filtering background noise from the images of hand-woven Qiang handicrafts using the SE attention module, extracting the visual features of the weaving lines and pattern units of the filtered hand-woven Qiang handicrafts, and then locating the direction of the weaving lines and the product contour using the edge detection module to output the craft visual features.

[0013] Preferably, the training of the AIGC model further includes: using the global convergence of cultural gene saliency, user cognitive preference, and styling factors as optimization objectives, and employing a multi-strategy enhanced escape algorithm to dynamically adjust the parameters of the AIGC model, thereby obtaining an AIGC model that can output the optimal Qiang weaving product design elements.

[0014] Preferably, the encoding includes open encoding, axial encoding, and selective encoding; open encoding decomposes the vectorized feature curves and process visual features into initial concepts, establishes the correlation between the initial concepts through axial encoding, and summarizes them into main categories; selective encoding determines the core categories from the main categories and systematically integrates all categories to form the Qiang ethnic culture gene pool.

[0015] Preferably, the step of combining shape grammar rules to transform the Qiang cultural gene pool into a Prompt word matrix, design intention graphics, and creative sketches specifically includes: parsing basic visual symbol units carrying cultural semantics from the Qiang cultural gene pool; using the basic visual symbol units as input, automatically combining and transforming them using shape grammar rules to generate multiple derivative symbol combination schemes; automatically generating text descriptions describing the cultural connotations and formal characteristics of each derivative symbol combination scheme, with the text descriptions forming the Prompt word matrix; rendering the vector data of each derivative symbol combination scheme into design intention graphics expressing style intentions and creative sketches expressing clear outlines, respectively; the shape grammar rules include copying, scaling, body rotation, point rotation, vertical mirroring, horizontal mirroring, shearing, Bézier curve transformation, and adding / deleting generative rules.

[0016] Preferably, the multi-strategy enhanced escape algorithm employs multi-strategy optimization, which includes an elite pool selection strategy, an adaptive perturbation factor adjustment strategy, and a dynamic centroid reverse learning strategy; the elite pool selection strategy constructs an elite pool by selecting design elements.

[0017] The present invention also provides a system for producing Qiang woven products, comprising:

[0018] The data acquisition module is used to collect documents recording sheep totems in Qiang culture and images of handmade Qiang handicrafts.

[0019] The sample construction module is used to extract the spiral curves and contour morphological features of the sheep totem records, locate the key inflection points of the spiral curves and contour morphological features, and obtain vectorized feature curves representing cultural semantics; extract the visual features of the weaving lines and pattern units of the images of handmade Qiang handicrafts, locate the direction of the weaving lines of the visual features and the product outline, and output the visual features of the craftsmanship; based on grounded theory, the vectorized feature curves and visual features of the craftsmanship are encoded to generate a Qiang cultural gene library containing visual features, semantic features and translation rules; combined with shape grammar rules, the Qiang cultural gene library is transformed into a Prompt word matrix, design intention graphics and creative sketches; from the sheep totem records and images of handmade Qiang handicrafts, independent Qiang weaving design elements are extracted to form a design element image library.

[0020] The training module is used to train the AIGC model with the Prompt word matrix, design intention graphics and creative sketches as input, and the design element image library as ground truth labels; to obtain the AIGC model that can output the optimal Qiang weaving product design elements, and to generate Qiang weaving products using the Qiang weaving product design elements.

[0021] The present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement any of the steps in the method for generating the Qiang woven product.

[0022] The present invention also provides a computer-readable storage medium storing a computer program that, when loaded by a processor, can execute any of the steps in the method for generating the Qiang-woven product.

[0023] The method for generating Qiang woven products provided by this invention has the following beneficial effects: Machine-readable cultural genes are extracted from a small number of core sheep totem records and representative images of handcrafted Qiang woven artifacts. These genes are then encoded into a structured gene bank using grounded theory, ensuring cultural accuracy and logic. Furthermore, based on shape grammar rules, the abstract genes are automatically transformed into a Prompt matrix, design intention graphics, and creative sketches suitable for AIGC model training, achieving the construction of a high-quality, large-scale training dataset. The resulting AIGC model can deeply understand the semantics of intangible cultural heritage and generate design elements that conform to the characteristics of the craft, achieving a leap from imitation and replication to culturally insightful innovation. The AIGC model rapidly generates diverse design schemes that fit the intangible cultural heritage of the Qiang people, balancing the accuracy of cultural transmission with the efficiency of innovative design, and achieving efficient synergy between intangible cultural heritage and modern design needs. Attached Figure Description

[0024] To more clearly illustrate the embodiments and design schemes of the present invention, the accompanying drawings required for this embodiment will be briefly described below. 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.

[0025] Figure 1 This is a flowchart illustrating a method for producing Qiang woven products according to an embodiment of the present invention;

[0026] Figure 2 This is the set of shape grammar-based transformation rules for the Qiang cultural symbol primitives in this embodiment of the invention;

[0027] Figure 3 This is a demonstration of the evolution steps of the Qiang weaving style of the Qiang cultural symbol primitives in this embodiment of the invention;

[0028] Figure 4 This is the framework for the synesthetic translation mode of Qiang culture in this embodiment of the invention. Detailed Implementation

[0029] To enable those skilled in the art to better understand and implement the technical solutions of the present invention, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and should not be construed as limiting the scope of protection of the present invention.

[0030] The "sheep" totem cultural symbol was extracted, and a semantic database of cultural symbol sets was constructed based on AIGC. Open coding was then applied to the corresponding semantic database. The Qiang cultural symbols in the open system were associated in linguistic form, and the interactive relationships between various types of Qiang cultural symbols were studied.

[0031] This invention aims to construct a complete digital narrative and product generation system for Qiang culture by integrating artificial intelligence technology and qualitative research methods, thereby solving the coordination problem between cultural protection and innovation, and realizing innovation in the form of intangible cultural heritage protection and the methods of dissemination.

[0032] Based on this, the present invention provides a method for producing Qiang woven products, specifically as follows: Figure 1 As shown, it includes:

[0033] S1. Collect documents and images of Qiang ethnic culture related to sheep totems and handicrafts made of Qiang cloth.

[0034] S2. Extract the spiral curves and contour features from the sheep totem records, locate the key turning points of the spiral curves and contour features, and obtain vectorized feature curves representing cultural semantics; extract the visual features of the weaving lines and pattern units from the images of handmade Qiang handicrafts, locate the direction of the weaving lines and the product outline of the visual features, and output the craft visual features; based on grounded theory, encode the vectorized feature curves and craft visual features to generate a Qiang cultural gene library containing visual features, semantic features, and translation rules; combine shape grammar rules to transform the Qiang cultural gene library into a Prompt word matrix, design intention graphics, and creative sketches; extract independent Qiang weaving design elements from the sheep totem records and images of handmade Qiang handicrafts to form a design element image library.

[0035] Images of handcrafted Qiang weaving artifacts were collected, and image segmentation techniques or manual annotation were used to extract independent Qiang weaving design elements (such as sheep totem pattern units, characteristic contour curves, typical color combinations, etc.) to form a design element image library as ground truth labels.

[0036] To extract the feature lines of the sheep totem, a Mask R-CNN model was introduced, integrating SE attention mechanism and edge detection algorithm to improve the accuracy and efficiency of capturing the feature curves of the sheep totem. Simultaneously, data augmentation techniques were employed to expand the diversity of the dataset and enhance the model's generalization ability. The AI ​​was trained to generate sheep totem feature genes that conform to cognitive orientations. Based on the input requirements of the Midjourney AI model, these were transformed into prompts that reasonably describe design preferences, graphics and images of design intentions, and creative sketches. Inputting these into the Midjourney model, it can generate a large number of design scheme sets at a speed far exceeding that of designers generating creative solutions. Then, based on cognitive tasks such as shape, outline, and color, the generated design schemes are encoded and classified to form open-ended codes, training the AIGC model to generate accurate codes.

[0037] Maintaining consistency and unity in human intelligence collaboration. Compared to the methods mentioned above, eye-tracking experiments combined with psychological measurements during interviews allow for further optimization of decisions regarding sheep totem characteristic genes, yielding relatively reasonable cognitive data.

[0038] The Qiang cultural genes are organized using selective coding to realize the translation and mapping logic of the design factors of Qiang weaving products. An escape algorithm is then used to realize the logical reasoning of converting the main axis coding into selective coding.

[0039] The translation process implements a mapping logic between feature curves and product shape curves based on the homeomorphism principle. A translation path is constructed using the translation interaction model, comprising "symbolic genes, translation methods, and carrier forms." Translation methods utilize shape grammar to perform copying (R1), scaling (R2), ontological rotation (R3), point rotation (R4), vertical mirroring (R5), horizontal mirroring (R6), and Bézier curve transformation (R7) on cultural gene elements. Furthermore, generative rule addition and deletion (R8) will be employed, such as... Figure 2 and Figure 3 As shown.

[0040] Figure 2 This is a "basic toolbox" for the evolution of Qiang weaving designs. It breaks down the transformation logic of Qiang cultural symbols (such as the core primitives of traditional Qiang weaving patterns, like the curling patterns in the image) into six standardized shape grammar rules. Simultaneously, it uses "legends and parametric vector graphics" to map "cultural patterns into digital geometric factors," facilitating subsequent algorithmic applications. The rule column (R1-R6) defines six basic transformation methods (copying, scaling, rotation, etc.) for Qiang cultural symbol primitives, serving as the "basic operation unit" of the shape grammar. The legend column demonstrates the visual transformation effect of each rule on the Qiang cultural pattern primitives (the curling patterns in the image, corresponding to traditional decorative elements in Qiang weaving). The parametric vector graphics column transforms "visual transformation" into geometric vectors (changes in the position, size, and angle of triangles), achieving a digital and calculable expression of cultural patterns. This is crucial for the subsequent "translation of Qiang cultural genes into Qiang weaving design factors." The algorithm can precisely control the design details of Qiang weaving products through these vector parameters.

[0041] Figure 3 yes Figure 2 The "Combined Application Cases" of the basic rules demonstrate how to gradually evolve individual Qiang cultural symbols into complex shapes of Qiang woven products through the combination of multiple rules. The evolutionary steps begin with a "single Qiang cultural symbol" and are then layered sequentially. Figure 2 The basic rules (such as R1 copy and R6 horizontal mirror) enable iterative generation from "simple primitives" to "complex Qiang weaving patterns". Figure 1 / Figure 2 show two different combined evolution paths (corresponding to different styles of Qiang weaving products): Figure 1 generates a dense Qiang weaving pattern by "R1 (copy) to R1 and R6 (copy + horizontal mirror) to multiple rule superposition"; Figure 2 generates a symmetrical radial Qiang weaving pattern by "R6 (horizontal mirror) to R1 (copy) to R4 (rotate with points)". Parametric vector graphics synchronously record the geometric vector changes of each evolution step to ensure that the whole process is "quantifiable and reproducible", which makes the design of Qiang weaving products change from "experience-based creation" to "digitalized and standardized generation".

[0042] Figure 2 and 3Its core function is to transform the "artistic variations of Qiang cultural patterns" into "digital rules that can be controlled by algorithms," providing an operable path from "cultural genes to design factors" for subsequent AIGC to generate Qiang weaving design schemes and mESC algorithm to optimize modeling factors.

[0043] By combining shape grammar rules, the Qiang cultural gene pool is transformed into multimodal design input. Specifically, the following steps are included: First, basic visual symbol units carrying cultural semantics and their associated rules are parsed from the Qiang cultural gene pool. A shape grammar rule library is defined, including copying, scaling, rotation, mirroring, shearing, and Bézier curve transformations. Using the basic visual symbol units as input, the shape grammar rules are applied for automatic combination and transformation, generating a series of derived symbol combination schemes. Based on the basic symbol semantics and shape grammar rules invoked by each derived symbol combination scheme, a text description describing its cultural connotations and formal characteristics is automatically generated, forming a Prompt word matrix. Finally, the vector data of each derived symbol combination scheme is rendered into a design intention graphic expressing stylistic intent and a creative sketch expressing a clear outline, respectively.

[0044] The mapping logic between feature curves and product shape curves is formed based on the homeomorphism principle. A translation path is constructed using the translation interaction model of "symbolic genes, translation methods, and carrier forms": First, a multi-strategy enhanced escape algorithm (mESC) is applied, employing an adaptive perturbation factor strategy to address the diversity of Qiang cultural genes; second, a restart mechanism is introduced to enhance mESC's ability to explore variations in cultural genes; finally, a dynamic centroid reverse learning strategy is designed to handle the iteration of the basic and growth libraries of cultural genes. Furthermore, to accelerate global convergence, a boundary adjustment strategy based on an elite pool is used to replace inferior genes by selecting superior ones.

[0045] y i,d =L d +Rand i,d ×(U d -L d );

[0046] In the formula, L d and U d Rand represents the lower and upper bounds of the d-th dimension, respectively. i，d The values ​​of follow a uniform distribution between 0 and 1. The fitness values ​​of the population are then calculated and sorted in ascending order, with the best individual stored in the elite pool E. pool (ElitePool), these elites are the best possible solutions discovered. The specific expression is as follows:

[0047] E pool ={y (1) ,y (2) ,...,y (ex)};

[0048] Here, ex represents the approximate optimal solution for the knitting case in the elite pool.

[0049] The elite pool initially uses a case study library of weaving inheritors as its elements. These elements are extracted in digital form and used to construct an input feature prompt word matrix for the AIGC model.

[0050] The adaptive perturbation factor can dynamically adjust the probability of generating non-standard cases that promote design optimization progress as the iteration process progresses, while also enhancing the diversity of usability-intended innovations. The expression is:

[0051] ;

[0052] in RAND It is a random number in the equation. As the adaptive perturbation factor (ADF) increases, the system can enhance its global search capability and avoid getting trapped in local optima.

[0053] The combination of Qiang cultural elements and the elite pool forms woven products with significant cultural characteristics. Qiang cultural elements are characterized by their dispersed nature; to enhance the salience of these characteristics, it is necessary to improve local search accuracy and accelerate the convergence process, thus requiring a reduction in the ADF value. In the game between the optimal and near-optimal solutions, by adjusting individual extrema, global extrema, and position terms, the adaptive perturbation factor can effectively increase the probability of finding the global optimum.

[0054] The adaptive perturbation factor construction mechanism for Qiang weaving culture is based on shape grammar for the evolution of cultural genes. The evolution mechanism is that the smallest gene unit is its basic symbol, which is the cultural cognitive recognition symbol. The acquisition of cognitive symbols is achieved by collecting the most representative image forms of cultural carriers, such as the sheep totem in Qiang culture, and refining and modifying them to obtain a library of recognition symbol elements. The process involves copying (R1), scaling (R2), body rotation (R3), point rotation (R4), vertical mirroring (R5), horizontal mirroring (R6), and Bézier curve transformation (R7). In addition, generative rules for addition and deletion (R8) will be used. This yields symbol combinations suitable for the product's shape. The optimal solution is obtained by generating different combinations, and the extreme values ​​and distance scales are adjusted through a game between the near-optimal solution and the optimal solution to optimize the design effect.

[0055] Based on schema theory and using cognitive feature data, this study investigates the digital narrative structure of cultural evolution in Qiang weaving products, namely, the translation pattern of Qiang cultural genes into Qiang weaving products. Furthermore, it optimizes the design factors of Qiang weaving products through decision-making, revising the reconstruction and redesign methods of multi-dimensional modeling factors such as curves, patterns, colors, and techniques.

[0056] The narrative structure is based on the logical deduction of the mapping relationship between the A0 cultural layer, the A1 cognitive link layer, and the A2 element layer. The A0 layer establishes a pattern library and semantic library of open-source encoding of Qiang culture, which maps to the A1 layer as the user's emotional decision library for A0, serving as the main axis encoding. Decisions are made through visual perception, depth perception, and patterns that the user understands and accepts, respectively, through eye-tracking experiments and psychological scale measurements of visual stimulus responses. The A2 layer is a library mapped to product design factors.

[0057] S3. Using the Prompt word matrix, design intention graphics, and creative sketches as input, and the design element image library as ground truth labels, train the AIGC model; obtain the AIGC model that can output the optimal Qiang weaving product design elements, and generate Qiang weaving products using the Qiang weaving product design elements.

[0058] For the design factors of Qiang-style textile products, the enhanced escape algorithm is used for classification optimization to accurately capture the design factor elements of the A2 layer for decision optimization. According to previous experimental verification, when the dimension of the feature modality element is set to 10 or 20, the convergence degree can reach the optimal state, indicating that the accuracy of the feature is relatively optimal. Figure 4 This is the visualized implementation process of the "Qiang Culture Synesthetic Translation Model Framework," which corely carries the translation logic of "from abstract Qiang cultural synesthetic imagery to concrete Qiang woven product design elements and then to intrinsic design language," corresponding to the core path of the "Synesthetic Translation Model Framework." The input synesthetic imagery (C design coding) corresponds to the multi-sensory cognitive elements of Qiang culture (natural patterns, woven patterns, cultural identity, etc.), serving as the "cultural material source" for synesthetic translation. It covers synesthetic dimensions such as visual, auditory, and perceptual senses, satisfying the input requirement of "capturing the synesthetic elements of Qiang culture." The "synesthetic bridge" is the core carrier of translation: through "feature mapping, translation, and association," abstract synesthetic images (such as the cultural cognition of "Three Sheep Bring Prosperity") are transformed into concrete "innate perceptions" (sheep horn flowers, argali patterns, etc. in the elite pool of sheep totem symbols), matching the logic of "synesthetic mapping relationship deduction"; the "elite pool of sheep totem cultural symbols" is a standardized library of Qiang cultural genes, and at the same time, it continuously updates symbol materials through "iterative optimization" to provide accurate cultural carriers for translation; the output-end styling design factors are transformed into innate design language. From the innate perceptions, the shape, texture, color, and other Qiang weaving styling factors are "mapped and extracted," and finally, multiple sets of innate design language schemes for Qiang weaving products are generated, realizing the closed-loop translation from "cultural synesthetic images to product design schemes," which is a visual presentation of the process of "constructing innate product design language."

[0059] The distillation of sheep totem characteristic genes: This involves integrating schema theory and grounded theory to summarize and organize cognitive data, combining AIGC and Mask R-CNN models, and incorporating SE attention mechanisms and edge detection algorithms to verify the consistency of human-intellectual interaction in capturing the inheritors' understanding of Qiang culture characteristics. An enhanced escape algorithm optimizes decisions based on multi-dimensional modeling factors to form a digital narrative structure.

[0060] The original Mask R-CNN's ResNet backbone extracts multi-scale features through residual block stacking, but all feature channels are treated equally, failing to specifically enhance the channel responses corresponding to sheep totem features. This invention modifies the process by inserting an SE attention module into each residual block of the backbone network. Specifically, it is positioned after the convolutional layers, batch normalization (BN), and activation function (ReLU), and before residual connection fusion, forming a new residual block structure of "Conv, BN, ReLU, SE, and Add (residual connection)". The SE attention mechanism uses a "compression-activation" logic to calibrate the weights of feature channels, strengthening effective feature channels related to the sheep totem and suppressing irrelevant background channels. The learned channel weights are multiplied element-wise with each channel of the original feature map, allowing the model to focus on the key feature channels of the sheep totem and weaken the influence of background channels (such as irrelevant textures or noise). By dynamically assigning higher weights to the sheep totem feature channels through the SE module, the "targeting" of feature extraction is improved, providing clearer channel-level feature support for subsequent edge detection.

[0061] In the multi-scale feature output layers of the Mask R-CNN backbone network (such as layers C2, C3, C4, and C5, corresponding to features of different resolutions), edge detection is performed separately on the feature map of each scale (generating an edge feature map at that scale); the semantic feature map of that scale is weighted and fused with the edge feature map to enhance the spatial expression of the sheep totem line features at multiple scales, avoid the dilution of fine line features in deep networks, and improve the extraction accuracy of complex pattern lines (such as sheep totem derived patterns).

[0062] The SE attention mechanism (channel level) strengthens effective channels by filtering "which channels contain sheep totem features"; the edge detection algorithm (spatial level) strengthens spatial signals by locating "the spatial position of sheep totem feature lines"; the two work together to enable the model to simultaneously possess the ability of "channel-level feature filtering" and "spatial-level line localization", accurately capturing the feature curves of sheep totems (such as the curvature of sheep horns, the outline of sheep's body, and the continuous lines of patterns).

[0063] Specifically, the collected sheep totem pattern samples are input into the AIGC model for feature derivation, simulating the generation of a pattern library from its features. Global average pooling is then performed, compressing the two-dimensional feature (H*W) into a single real number, and transforming the feature map from [h,w,c]==>[1,1,c]. A weight value is generated for each channel of the totem contour feature, with the number of output weight values ​​being the same as the number of channels in the input feature map: [1,1,c]==>[1,1,c]. The normalized weights are then applied to the totem contour feature: [h,w,c]*[1,1,c]==>[h,w,c].

[0064] Based on the improved Mask R-CNN model, an edge detection algorithm is introduced to extract the feature lines of the sheep's head. The improved model first accurately segments the sheep's head region, and then uses edge detection algorithms such as Canny to quickly and accurately extract the feature lines of the sheep's head from the segmented image. Compared with traditional manual extraction methods, this method reduces the feature contour curve extraction time to 5-10% of the original time, and the extracted lines are more complete and accurate, unaffected by subjective factors.

[0065] The innovation of this invention is mainly reflected in the improvement of the Mask R-CNN model. By introducing the SE attention mechanism, the model can more intelligently focus on the key features of the sheep's head. The application of data augmentation technology effectively expands the dataset and improves the model's generalization ability. Combining the improved model with edge detection and AI creation models opens up a new path from cultural relic feature extraction to artistic creation.

[0066] This invention focuses on the mapping method between cultural symbols and forms. It integrates schema theory and grounded theory to deduce a logical mapping model of Qiang culture in the translation process of Qiang woven products. This invention combines human and generative artificial intelligence (AIGC) models to collaboratively conduct divergent thinking to acquire data on the collective cognition of Qiang culture. It then uses graph segmentation and deep learning clustering algorithms for convergence to extract the graphic features of Qiang cultural elements and constructs an AI-assisted design agent for the creative design of Qiang woven products. Based on the homeomorphism principle in topology, it forms a mapping logic for the product's form curve based on curve fitting, thus creating a digital narrative structure for the translation of Qiang culture into Qiang woven product design.

[0067] This invention, in the process of exploring users' genetic cognitive preferences for Qiang culture, employs visual and perceptual signal detection to compensate for the deficiency of excessive sensory data in synesthetic translation. Furthermore, it uses an enhanced escape algorithm to optimize the design factors of multi-dimensional Qiang woven products. This invention is based on a framework for the formation of synesthetic translation patterns within Qiang culture. It applies design thinking theories and methods for solving complex design problems proposed by domestic scholars, and uses a multi-dimensional cognitive model that links key concepts, intent capture, and evolutionary path analysis based on knowledge. It captures synesthetic intentional elements of Qiang culture as a foundation, deduces synesthetic mapping relationships, translates them into design factors, constructs the design language of intrinsically perceptual products, and thus forms an evolutionary path.

[0068] Based on the same inventive concept, this invention also provides a system for producing Qiang woven products, comprising:

[0069] The data acquisition module is used to collect documents recording sheep totems in Qiang culture and images of handmade Qiang handicrafts.

[0070] The sample construction module is used to extract the spiral curves and contour morphological features of the sheep totem records, locate the key inflection points of the spiral curves and contour morphological features, and obtain vectorized feature curves representing cultural semantics; extract the visual features of the weaving lines and pattern units of the images of handmade Qiang handicrafts, locate the direction of the weaving lines of the visual features and the product outline, and output the visual features of the craftsmanship; based on grounded theory, the vectorized feature curves and visual features of the craftsmanship are encoded to generate a Qiang cultural gene library containing visual features, semantic features and translation rules; combined with shape grammar rules, the Qiang cultural gene library is transformed into a Prompt word matrix, design intention graphics and creative sketches; from the sheep totem records and images of handmade Qiang handicrafts, independent Qiang weaving design elements are extracted to form a design element image library.

[0071] The training module is used to train the AIGC model with the Prompt word matrix, design intention graphics and creative sketches as input, and the design element image library as ground truth labels; to obtain the AIGC model that can output the optimal Qiang weaving product design elements, and to generate Qiang weaving products using the Qiang weaving product design elements.

[0072] This invention also provides a computer device. At the hardware level, the computer device includes a processor, an internal bus, a network interface, memory, and non-volatile storage, and may also include other hardware required for business operations. The processor reads the corresponding computer program from the non-volatile storage into memory and then runs it to implement the above-described method for generating the Qiang-style product.

[0073] The present invention also provides a computer-readable storage medium storing a computer program that can be used to execute the above-described method for generating Qiang-style woven products.

[0074] Specific limitations regarding the computational system for the generation method of Qiang woven products can be found in the limitations of the generation method of Qiang woven products mentioned above, and will not be repeated here. Each module in the aforementioned generation system of Qiang woven products can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0075] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification. Furthermore, the above embodiments only illustrate several implementation methods of this application, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method for producing Qiang woven products, characterized in that, include: Collect documents documenting sheep totems and images of handcrafted Qiang handicrafts from Qiang culture; The process involves extracting the spiral curves and contour features from sheep totem records, locating key inflection points of these features, and obtaining vectorized feature curves representing cultural semantics. Visual features of weaving lines and pattern units from images of handcrafted Qiang woven artifacts are extracted, and the direction of these weaving lines and the product outline are located, outputting the visual features of the craft. Based on grounded theory, the vectorized feature curves and visual features of the craft are encoded to generate a Qiang cultural gene library containing visual features, semantic features, and translation rules. Combining shape grammar rules, the Qiang cultural gene library is transformed into a Prompt word matrix, design intention graphics, and creative sketches. Independent Qiang woven design elements are extracted from the sheep totem records and images of handcrafted Qiang woven artifacts to form a design element image library. The AIGC model is trained using the Prompt word matrix, design intention graphics, and creative sketches as input, and the design element image library as ground truth labels. Obtain an AIGC model that can output the optimal Qiang weaving product design elements, and use the Qiang weaving product design elements to generate Qiang weaving products; Using an improved Mask R-CNN model, vectorized feature curves and craft visual features are extracted from sheep totem records and images of hand-woven Qiang handicrafts. Specifically, this includes: filtering background interference information from sheep totem records using an SE attention module, extracting the spiral curves and contour features of the filtered sheep totem records, and then locating the key inflection points of the spiral curves and contour features using an edge detection module to obtain vectorized feature curves representing cultural semantics; filtering background noise from images of hand-woven Qiang handicrafts using an SE attention module, extracting the visual features of the weaving lines and pattern units of the filtered hand-woven Qiang handicrafts, and then locating the direction of the weaving lines and the product contour using an edge detection module to output the craft visual features. The improved Mask R-CNN model inserts an SE attention module after the ReLU activation function of the ResNet backbone network and before residual connection fusion in the traditional Mask R-CNN model; and adds edge detection after the multi-scale feature output layer of the ResNet backbone network.

2. The method for producing Qiang woven products according to claim 1, characterized in that, The training of the AIGC model also includes: using the global convergence of cultural gene saliency, user cognitive preference, and styling factors as optimization objectives, and employing a multi-strategy enhanced escape algorithm to dynamically adjust the parameters of the AIGC model, thereby obtaining an AIGC model that can output the optimal Qiang weaving product design elements.

3. The method for producing Qiang woven products according to claim 1, characterized in that, The encoding includes open encoding, axial encoding, and selective encoding. Open encoding decomposes the vectorized feature curves and process visual features into initial concepts, establishes the relationship between the initial concepts through axial encoding, and summarizes them into main categories. Selective encoding determines the core categories from the main categories and systematically integrates all categories to form the Qiang culture gene pool.

4. The method for producing Qiang woven products according to claim 1, characterized in that, The method of combining shape grammar rules to transform the Qiang cultural gene pool into a Prompt word matrix, design intention graphics, and creative sketches specifically includes: parsing basic visual symbol units carrying cultural semantics from the Qiang cultural gene pool; using the basic visual symbol units as input, automatically combining and transforming them using shape grammar rules to generate multiple derivative symbol combination schemes; automatically generating text descriptions describing the cultural connotations and formal characteristics of each derivative symbol combination scheme, with the text descriptions forming the Prompt word matrix; and rendering the vector data of each derivative symbol combination scheme into design intention graphics expressing stylistic intentions and creative sketches expressing clear outlines, respectively. The shape grammar rules include copying, scaling, body rotation, point rotation, vertical mirroring, horizontal mirroring, shearing, Bézier curve transformation, and adding and deleting generative rules.

5. The method for producing Qiang woven products according to claim 2, characterized in that, The multi-strategy enhanced escape algorithm employs multi-strategy optimization, which includes an elite pool selection strategy, an adaptive perturbation factor adjustment strategy, and a dynamic centroid reverse learning strategy. The elite pool selection strategy constructs an elite pool by selecting design elements.

6. A system for implementing the method for producing Qiang woven products according to claim 1, characterized in that, include: The data acquisition module is used to collect documents recording sheep totems in Qiang culture and images of handmade Qiang handicrafts. The sample construction module is used to extract the spiral curves and contour features of the sheep totem records, locate the key inflection points of the spiral curves and contour features, and obtain vectorized feature curves representing cultural semantics; extract the visual features of the weaving lines and pattern units of the images of handmade Qiang woven handicrafts, locate the direction of the weaving lines of the visual features and the product outline, and output the visual features of the craft; based on grounded theory, the vectorized feature curves and visual features of the craft are encoded to generate a Qiang cultural gene library containing visual features, semantic features and translation rules; combined with shape grammar rules, the Qiang cultural gene library is transformed into a Prompt word matrix, design intention graphics and creative sketches; from the sheep totem records and images of handmade Qiang woven handicrafts, independent Qiang weaving design elements are extracted to form a design element image library; The training module is used to train the AIGC model by taking the Prompt prompt word matrix, design intention graphics and creative sketches as inputs and the design element image library as ground value labels. An AIGC model is obtained that can output the optimal design elements for Qiang woven products, and Qiang woven products are generated using these design elements.

7. A computer device, comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is loaded by the processor, it is able to perform the steps of the method according to any one of claims 1 to 5.

Images