Method and system for constructing non-perpetual inheritance atlas

Abstract

The invention discloses a construction method and system of a non-abandoned inheritance atlas, and relates to the technical field of atlas construction. The construction method of the non-abandoned inheritance atlas comprises the following steps: acquiring work image data, identity information and skill interview voice information of a plurality of embroidery inheritors, inputting the work image data, the identity information and the skill interview voice information into a style clustering model to extract an attribute triple, and constructing an initial non-abandoned inheritance atlas; the method comprises the following steps: acquiring propagation behavior data of each embroidery inheritor, performing feature analysis on the propagation behavior data and an attribute triple to obtain a propagation traction index and a structure process stability index, and performing comprehensive analysis to obtain a skill inheritance index. According to the method, an embroidery non-perpetual inheritance atlas is generated, so that a structural connection relationship between an inheritor and a style category is established, an atlas node system with expression levels and comparability is constructed, and the precision and practicability of the atlas in structural expression are enhanced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of atlas construction, in particular to a non-heritage inheritance atlas construction method and system. BACKGROUND

[0002] With the continuous promotion of non-material cultural heritage protection policy, the digital preservation and inheritance of traditional skill type non-heritage resources have gradually become the core task. Among them, traditional skills represented by embroidery have strong regional characteristics, diverse styles, and oral and mental transmission as the main inheritance characteristics. For a long time, there has been a lack of unified standard structured expression mechanism and quantifiable evaluation system, making it difficult to realize the overall modeling and visual display of non-heritage skill inheritance status. In the prior art, the non-heritage embroidery skills are recorded and managed by adopting the way of archiving based on text files or work pictures, lacking systematic modeling and analysis of the skill structure behind the embroidery works, and failing to establish a skill style classification and evolution relationship network with clear structure and centered on inheritors. In addition, for the propagation behavior in the non-heritage inheritance process, the existing methods mostly stop at statistical publication quantity or participation record, and fail to effectively combine the positive traction effect of the propagation behavior on skill inheritance, resulting in the problems of data disconnection and management fragmentation in skill protection and propagation promotion.

[0003] The prior art such as the patent application with the announcement number CN117312578B discloses a non-heritage inheritance atlas construction method and system, which includes the following steps: based on multiple types of public databases and geographic information systems, data mining technology is adopted to collect and clean the original non-heritage element space-time data, and standardized processing is performed to generate the purified non-heritage space-time data set. In the present application, the long short-term memory network reveals the non-heritage time law, optimizes the space-time evolution model, the convolutional neural network and the Transformer algorithm deep learning non-heritage audio-visual information, enriches the multi-modal semantic features, the graph database and the ontology construct the systematic non-heritage knowledge graph, provides structured knowledge support for inheritance, strengthens the learning, community detection and key node recognition algorithm to optimize the inheritance path prediction, provides scientific decision support, the generative adversarial network deeply digs the deep semantic relationship of non-heritage elements, expands the non-heritage deep semantic association network, and provides strong data and theoretical support for inheritance and protection.

[0004] Based on the above scheme discovery, the limitations of the prior art at least include the following problems: the prior art ignores the importance and structural difference of the inheritor as the core practice subject in the non-heritage skill inheritance process, fails to realize the identification and quantitative modeling of the real bearer behavior, ability and stability in the inheritance path, and lacks structured modeling centered on the inheritor, which means that the system cannot evaluate whether a certain skill has expression deviation, transmission distortion, structure damage or cognitive misplacement, and also cannot identify the role intensity of which type of inheritor in actual transmission or the depth of inheritance, thereby directly leading to the fact that the graph is only on the static information display level, and cannot support the actual needs of classification management, risk assessment and precise intervention in non-heritage policy making. SUMMARY

[0005] In view of the shortcomings of the prior art, the present application provides a non-heritage inheritance graph construction method and system, which solves the problem that the prior art ignores the modeling of the inheritor, thereby causing the graph to lack the expression and evaluation ability of the behavior and stability.

[0006] To achieve the above purpose, the present application is realized by the following technical scheme: a non-heritage inheritance graph construction method, comprising the following steps: obtaining work image data, identity information and skill interview voice information of a plurality of embroidery inheritors, and inputting them into a style clustering model for identification analysis to obtain attribute triples of each embroidery inheritor; constructing an initial non-heritage inheritance graph based on the attribute triples of each embroidery inheritor; obtaining the propagation behavior data of each embroidery inheritor, and respectively performing feature analysis on the attribute triples to obtain the propagation traction index and the structure process stability index of each embroidery inheritor; comprehensively analyzing the propagation traction index and the structure process stability index of each embroidery inheritor to obtain the skill inheritance index of each embroidery inheritor; and performing label expansion processing on the initial non-heritage inheritance graph based on the skill inheritance index of each embroidery inheritor to generate an embroidery non-heritage inheritance graph.

[0007] Further, the specific formula for calculating the skill inheritance index of a certain embroidery inheritor is as follows: ; wherein, is the skill inheritance index of a certain embroidery inheritor, is the structure process stability index of a certain embroidery inheritor, is the process adjustment coefficient stored in the database, is the propagation traction index of a certain embroidery inheritor, is the propagation adjustment coefficient stored in the database, is the matching deviation adjustment coefficient stored in the database, is the difference adjustment coefficient stored in the database.

[0008] Further, the work image data is specifically pixel values and two-dimensional coordinates of each pixel point in the work image, and the style clustering model is specifically a joint self-supervised clustering network, which includes an input layer, a feature extraction layer, a fusion representation layer, and an output layer.

[0009] Further, the specific steps of obtaining the attribute triplets of each embroidery inheritor are as follows: in the input layer of the joint self-supervised clustering network, receiving the work image data, identity information, and skill interview voice information of each embroidery inheritor and performing preprocessing; in the feature extraction layer of the joint self-supervised clustering network, performing feature encoding processing on the work image data, identity information, and skill interview voice information of each embroidery inheritor after preprocessing to obtain a work image vector and a semantic embedding vector of each embroidery inheritor; in the fusion representation layer of the joint self-supervised clustering network, performing fusion processing on the work image vector and the semantic embedding vector of each embroidery inheritor to obtain a joint feature representation vector of each embroidery inheritor; and in the output layer of the joint self-supervised clustering network, performing attribute extraction processing on the joint feature representation vector of each embroidery inheritor to obtain the attribute triplets of each embroidery inheritor, including an inheritor ID, a style label, a belonging confidence, and a process structural evaluation index set; the process structural evaluation index set includes a process restoration index, a main needle method focusing index, and a process logic deflection index.

[0010] Further, the specific steps of constructing the initial intangible cultural heritage inheritance graph based on the attribute triplets of each embroidery inheritor are as follows: reading the style label of each embroidery inheritor and constructing a style category main node set, i.e., a plurality of style category main nodes; reading the inheritor ID of each embroidery inheritor and constructing an inheritor node set; reading the belonging confidence of each embroidery inheritor and establishing a graph connection edge between the inheritor node set and the corresponding style category main node to generate the initial intangible cultural heritage inheritance graph.

[0011] Further, the propagation behavior data includes a video upload quantity value, an account fan value, an exhibition participation frequency value, a propagation interaction guide index, a propagation trajectory index, and a platform co-occurrence coverage index, and the specific steps of obtaining the propagation traction index of each embroidery inheritor are as follows: based on a particle swarm optimization algorithm, comprehensively analyzing the video upload quantity value, the account fan value, and the exhibition participation frequency value of each embroidery inheritor to obtain a propagation diffusion driving index of each embroidery inheritor; based on the particle swarm optimization algorithm, comprehensively analyzing the propagation interaction guide index, the propagation trajectory index, and the platform co-occurrence coverage index of each embroidery inheritor to obtain a content touch coverage index of each embroidery inheritor; and comprehensively analyzing the propagation diffusion driving index and the content touch coverage index of each embroidery inheritor to obtain a propagation traction index of each embroidery inheritor.

[0012] Further, the specific formula of the propagation traction index of a certain embroidery inheritor is as follows: ; wherein, is the propagation traction index of a certain embroidery inheritor, is the propagation diffusion driving index of a certain embroidery inheritor, is the diffusion driving adjustment coefficient stored in the database, is the content touch coverage index of a certain embroidery inheritor, is the touch coverage adjustment coefficient stored in the database, is the balance adjustment coefficient stored in the database.

[0013] Further, the specific steps of obtaining the structural process stability index of each embroidery inheritor are as follows: reading the process structural evaluation index set in the attribute triple of each embroidery inheritor, that is, the process reduction index, the main needle method focusing index and the process logic deflection index; comprehensively analyzing the process reduction index, the main needle method focusing index and the process logic deflection index of each embroidery inheritor to obtain the structural process stability index of each embroidery inheritor.

[0014] Further, the specific steps of generating the embroidery intangible cultural heritage inheritance graph are as follows: connecting the skill inheritance index of each embroidery inheritor to the inheritor node of the corresponding style category main node in the initial intangible cultural heritage inheritance graph, and performing clustering mean analysis to obtain the style skill inheritance index of each style category main node in the initial intangible cultural heritage inheritance graph.

[0015] Color marking processing is performed on the initial intangible cultural heritage inheritance graph based on the style skill inheritance index of each style category main node to generate the embroidery intangible cultural heritage inheritance graph.

[0016] An intangible cultural heritage inheritance graph construction system comprises: a data acquisition and identification module configured to acquire work image data, identity information and skill interview voice information of a plurality of embroidery inheritors, and input the data into a style clustering model for identification analysis to obtain an attribute triple of each embroidery inheritor; an initial graph construction module configured to construct an initial intangible cultural heritage inheritance graph based on the attribute triple of each embroidery inheritor; a feature analysis module configured to acquire propagation behavior data of each embroidery inheritor, and perform feature analysis on the attribute triple to obtain a propagation traction index and a structural process stability index of each embroidery inheritor; a comprehensive analysis module configured to comprehensively analyze the propagation traction index and the structural process stability index of each embroidery inheritor to obtain a skill inheritance index of each embroidery inheritor; and an intangible cultural heritage graph generation module configured to perform marking and expansion processing on the initial intangible cultural heritage inheritance graph based on the skill inheritance index of each embroidery inheritor to generate an embroidery intangible cultural heritage inheritance graph.

[0017] The present application has the following beneficial effects:

[0018] (1) The non-heritage map construction method, by collecting the image data, identity and skill voice information of the embroidery heritage, and combining the joint self-supervised clustering network, extracting the style label, attribution confidence and process structural evaluation index, then combining the propagation behavior data to further generate the propagation traction index and skill inheritance index, then hanging these quantitative indexes as node attributes to the non-heritage map, thereby establishing the structural connection relationship between the heritage and the style category, and embedding the numerical label with behavior characteristic significance in the map, thereby constructing the map node system with expression level, comparability and dynamic expansion capability, and then making each node can present its characteristic difference in process expression, propagation participation and style attribution dimensions, thereby enhancing the precision and practicality of the map in structural expression.

[0019] (2) The non-heritage map construction method, by constructing the structural process stability index and the propagation traction index double-channel evaluation mechanism, thereby realizing the quantitative expression of the skill fidelity and propagation ability of the heritage in the heritage dimension, and extracting the process restoration index, main needle method focusing index and process logic deflection index reflected in the heritage works image and voice semantics by the joint self-supervised clustering network, combining the propagation interaction behavior data to calculate its diffusion driving and reach ability, and using genetic algorithm and particle swarm optimization algorithm for weight fitting and index balance processing, the finally output skill inheritance index can represent the comprehensive performance of the heritage in preserving the traditional process specification and promoting the skill propagation, thereby enabling the map to support the ability sorting and structure comparison of the heritage under multi-dimensional indexes, and thereby enhancing the expression ability of the map to the real heritage efficiency.

[0020] (3) The non-heritage map construction method, by binding the skill inheritance index of each heritage to the style main node in the embroidery non-heritage map during generation, and introducing a color marking mechanism to intuitively distinguish the inheritance state, thereby realizing the semantic enhancement of the map in the graphical visualization level, giving the map nodes red, yellow and green color state marks, corresponding to the three inheritance states of skill recession, maintenance and health, thereby enabling the user to quickly identify the skill risk level and development trend of each style category in the map structure, at the same time, this method also has good scalability, can dynamically refresh the map annotation combined with subsequent inheritance index update, thereby forming an intelligent visual structure with self-evolution and risk warning capability, thereby facilitating data tracking and analysis response under complex inheritance system.

[0021] (4) The system for constructing the intangible cultural heritage inheritance map has a modular design that divides the data collection, attribute identification, feature analysis, index calculation and map generation process into structured tasks, thereby realizing the full-process automated construction capability from raw data input to map generation. The data acquisition and identification module set in the system can automatically access multi-source image and voice information to complete the extraction of individual characteristics of inheritors; while the feature analysis module and the comprehensive analysis module work together to complete bidirectional index modeling from behavioral and process dimensions; finally, the intangible cultural heritage map generation module embeds multi-dimensional indicators into the map structure, thereby effectively improving the systematization level of the intangible cultural heritage inheritance modeling process and avoiding the problems of information loss, indicator isolation and update lag in manual map construction, thus making the entire map construction process highly reusable, scalable and automatically updated.

[0022] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0023] Figure 1 This is a flowchart of a method for constructing a non-genetic inheritance map according to the present invention.

[0024] Figure 2 This is a flowchart illustrating the specific steps involved in obtaining the attribute triplets for each embroidery inheritor in the construction method of a non-heritage inheritance map according to the present invention.

[0025] Figure 3 This is a schematic diagram of the embroidery inheritor sequence process structure evaluation index set in the non-heritage inheritance map construction method of the present invention.

[0026] Figure 4 This is a block diagram of a non-genetic inheritance map construction system according to the present invention. Detailed Implementation

[0027] Please see Figure 1This invention provides a technical solution: a method for constructing an intangible cultural heritage (ICH) transmission map, comprising the following steps: acquiring image data of works by several ICH inheritors, identity information (including but not limited to inheritor number, level identifier, such as national level, provincial level, region, and corresponding work number, used as unique identifiers for individual inheritors), and voice information from skill interviews, and inputting them into a style clustering model for identification and analysis to obtain attribute triples for each ICH inheritor; constructing an initial ICH transmission map based on the attribute triples of each ICH inheritor; acquiring dissemination behavior data of each ICH inheritor and performing feature analysis with the attribute triples to obtain a dissemination traction index and a structural process stability index (evaluating the integrity and standardization of the ICH work at the skill structure level, and reflecting the stability and orthodoxy of the ICH inheritor's skill transmission); comprehensively analyzing the dissemination traction index and structural process stability index of each ICH inheritor to obtain a skill transmission index for each ICH inheritor; and performing labeling and expansion processing on the initial ICH transmission map based on the skill transmission index of each ICH inheritor to generate an ICH transmission map for embroidery.

[0028] The specific formula for calculating the skill inheritance index of a certain embroidery inheritor is as follows: ;in, This refers to the skill inheritance index of a certain embroidery inheritor. This refers to the structural and technical stability index of a certain embroidery inheritor. These are process adjustment coefficients stored in the database. The dissemination index of a certain embroidery inheritor. The propagation modulation coefficients are stored in the database. The matching deviation adjustment coefficient is stored in the database. These are the difference adjustment coefficients stored in the database.

[0029] It needs to be explained that the specific form of the tanh function is as follows: ,in, It is a natural constant, and in this example it can be taken as 2.71, with a domain of (−∞, +∞) and a range of (−1, +1).

[0030] in the formula This item is used to adjust the numerical matching degree between the structural craftsmanship stability index and the dissemination traction index of individual inheritors of intangible cultural heritage embroidery, so as to avoid the craftsmanship inheritance index being too high or too low.

[0031] , , , The following steps can be taken to obtain the following: Utilize historical data, combined with the structural process stability index and the propagation traction index, to conduct statistical regression analysis, quantify the specific impact of each factor on the skill inheritance index, and thus fit the initial weight values. Secondly, use sensitivity analysis to adjust the value range of each coefficient and observe its impact on the skill inheritance assessment results, ensuring the stability and rationality of the model.

[0032] The image data of the artwork specifically includes the pixel value and two-dimensional coordinates of each pixel in the artwork image. The style clustering model is specifically a joint self-supervised clustering network, which includes an input layer, a feature extraction layer, a fusion representation layer, and an output layer.

[0033] The input layer is used to receive image data of the inheritor's works, identity information, and audio information from interviews with the inheritor, and to complete data preprocessing.

[0034] The feature extraction layer is used to extract texture features from images and semantic features from spoken text.

[0035] The fusion representation layer is used to align and fuse image and semantic features to generate a unified joint feature representation;

[0036] The output layer is used to output style tags, attribution confidence scores, and structural evaluation indices, and to generate attribute triples.

[0037] Specifically, such as Figure 2As shown, the specific steps to obtain the attribute triplet for each embroidery inheritor are as follows: In the input layer of the joint self-supervised clustering network, the image data of each embroidery inheritor's work, identity information, and audio information from skill interviews are received and preprocessed (for image data, size normalization, illumination compensation, and format standardization are performed; for audio information, endpoint detection and audio enhancement are performed before being sent to the speech-to-text module; and the identity information is directly bound to the input sample as a structured label, i.e., the inheritor ID, which is carried as an implicit identifier field throughout the process to ensure that it can be accurately associated with the individual inheritor in the final output, forming a traceable output structure); in the feature extraction layer of the joint self-supervised clustering network, the preprocessed data of each embroidery inheritor is processed... The image data of the embroidery inheritors' works, identity information, and audio information from skill interviews are processed by feature encoding to obtain the image vector and semantic embedding vector of each embroidery inheritor's works. In the fusion representation layer of the joint self-supervised clustering network, the image vector and semantic embedding vector of each embroidery inheritor's works are fused to obtain the joint feature representation vector of each embroidery inheritor. In the output layer of the joint self-supervised clustering network, the joint feature representation vector of each embroidery inheritor is processed by attribute extraction to obtain the attribute triples of each embroidery inheritor, including inheritor ID, style tag, attribution confidence, and process structure evaluation index set. The process structure evaluation index set includes process restoration index, main needle method focus index, and process logic deviation index.

[0038] The specific steps of feature encoding processing are as follows: The input image data is fed into a convolutional coding network with ResNet-101 as the backbone. Basic edge information extraction and low-level texture perception are completed through its initial layer. Then, it passes through multiple residual blocks to achieve deep extraction of multi-level texture features, obtaining an intermediate image feature tensor with rich detail representation capabilities. This tensor is then fed into a ViT (Vision Transformer)-based visual modeling module. This module first divides the image features into several fixed-size image patches. The system first embeds the image data and then constructs its spatial location awareness through a positional encoding mechanism. Subsequently, a multi-layer Transformer encoder structure is used to model the global semantic relationships between image patches, extracting cross-regional contextual information to form a high-dimensional image feature representation with global texture awareness and spatial structure expression capabilities. Finally, the convolutional features output by ResNet and the global semantic features output by ViT are concatenated and fused, and mapped to a fixed-length work image vector through linear transformation and a normalization layer (LayerNorm) to express the main texture form, color distribution, pattern construction features, and spatial combination structure of the embroidery work. The input interview speech data is fed into an automatic speech recognition module based on Wav2Vec 2.0. A convolutional front-end and a time-contrast pre-trained network complete self-supervised modeling and language unit extraction of the speech frames, ultimately outputting the corresponding speech-to-text sequence. This text sequence is then input into a semantic encoding module based on BERT, where a multi-layer bidirectional Transformer encoder is used to model its context embedding, combined with a language masking mechanism. LanguageModeling automatically captures the semantic relationships of each word in the context. At the same time, through named entity recognition (NER) and keyword annotation mechanisms, it extracts representative style words (such as Suzhou embroidery, random stitch), needlework terms (such as satin stitch, flat stitch), and typical operation action words (such as starting stitch, finishing stitch) from the embedded sequence. Finally, the semantic embedded sequence corresponding to the text is mapped into a semantic embedded vector after mean pooling and fully connected transformation. This vector is used to express the inheritor's structured understanding of the technique content, operation process, and style self-description at the linguistic level.

[0039] The specific steps of the fusion processing are as follows: Linear transformation and normalization are performed on the image vector and semantic embedding vector of the artwork, respectively, to ensure that the two types of features maintain dimensional consistency and are comparable in the feature space, eliminating the inconsistency in feature expression caused by modal differences. The processed image features and semantic features are then concatenated along the feature dimension to form a single fusion feature representation, which simultaneously covers key information such as image texture morphology, pattern structure information, and style category, technique terminology, and operation path expressed in the language. After concatenation, the fusion feature undergoes further nonlinear mapping processing. A multi-layer neural network structure is used to compress and transform the features, and the weights of different modal information in the fusion expression are dynamically adjusted based on the actual distribution of the input content. This makes the fusion expression more adaptable and discriminative among different types of inheritor samples. Finally, the fusion representation layer outputs a joint feature representation vector for each embroidery inheritor, used to uniformly represent the style attribution and craft structure characteristics of their representative works in the visual feature domain and the linguistic semantic domain.

[0040] The specific steps of attribute extraction processing are as follows: Collect the joint feature representation vectors of all inheritor samples, and use cluster analysis methods to perform online clustering processing on the batch of features. For example, the KMeans algorithm can be used to divide the samples into several style subclasses, and generate multiple temporary cluster centers from the clustering results. Each cluster center can be regarded as a style aggregation trend formed in the sample space, which is used to represent a specific style expression structure. In the process of attribution judgment, the joint feature representation vector of the inheritor to be identified is compared with the above multiple local cluster centers one by one. The cosine similarity method is used to measure the relative closeness between the input vector and each center. Based on the calculated similarity results, the style direction represented by the center with the highest matching degree is further compared, and the attribution confidence is used as the attribution confidence, and the style attribution label of the inheritor is output.

[0041] Based on the texture region vectors contained in the image channels in the joint feature representation, texture fragment regions with typical stitch patterns in the artwork are identified. The number, distribution area, and position density of stitch fragments are counted, and each is matched with a predefined set of standard stitch templates. For each standard stitch category, its matching coverage, missing rate, and morphological similarity in the current artwork are calculated to form a stitch structure matching matrix. Combining the overall similarity score of this matrix, the Preservation Degree (PRD) is output to measure the degree to which the artwork retains traditional embroidery techniques at the stitch structure level.

[0042] Secondly, by utilizing the heat distribution characteristics of the image attention mechanism in the joint vector, the main texture area focused by the model is located, and spatial overlap analysis is performed on it with the key stitch areas (such as the overlock area and the filling area) defined in the standard embroidery technique. The offset distance between the attention focus center and the center of the key area is calculated, and the heat weight ratio of the area is evaluated. Thus, the Key-focus Index (KFI) is calculated. This index reflects whether the representative stitch area in the work has become the main compositional element of visual expression.

[0043] Next, from the keyword paths extracted from the semantic channels in the joint representation vector, the sequence of operational actions representing the technique process is identified, including descriptions such as starting the stitch, moving the stitch, turning the stitch, and finishing the stitch, and arranged in order to form an operational path sequence. Subsequently, the sequence is compared with the preset standard process flow in the skill database. The Levenshtein Distance or Dynamic Time Warping (DTW) algorithm is used to quantify the degree of deviation between the current semantic path and the standard path, and output the Step Consistency Deviation (SCD) index, which is used to measure the logical consistency and standardization of the operational steps in the inheritor's narration.

[0044] Finally, based on the style attribution label and the craft structure evaluation index extracted from the joint feature representation vector of the inheritor sample, a composite attribute triplet containing identity identifier, work information and multi-dimensional structural attributes is constructed. The triplet fields include: inheritor ID, style category, attribution confidence, craft restoration index, main needlework focus index, and process logic deviation index.

[0045] Furthermore, the pre-training process of the joint self-supervised clustering network is as follows:

[0046] A dataset of intangible cultural heritage embroidery styles containing labeled images and audio information was obtained. The dataset was provided by several embroidery inheritors with clear identities. The sample content includes multi-angle high-resolution image data of their representative works, audio content describing the techniques (including work descriptions, operation steps, etc.), and their corresponding style category labels and key technique area annotation information. The intangible cultural heritage embroidery style dataset was divided into training set and validation set.

[0047] The Joint Self-Supervised Clustering Network (IMTC-Net) is initialized and configured. The image encoding branch uses ResNet or ViT as the feature backbone, the semantic branch loads the BERT encoder as the language representation basis, the modality alignment module sets up a dual mechanism of splicing and attention, the clustering discriminator connects to the DeepCluster or SwAV scheme, and the structural index output head sets up three branches to output three structural indicators: process restoration index, main needle method focusing index, and process logic deflection index. All network layer weight parameters are initialized using Kaiming, and optional feature extraction modules pre-trained on general image semantic datasets such as COCO+ImageNet+BERT can be loaded to enhance representation transfer capability and pre-training stability.

[0048] Training is performed on the training set, with a set number of training rounds (e.g., 100 rounds). Each round of training is executed in the following order: In the forward propagation phase, input image and speech data are processed through a feature extraction layer and a fusion representation layer to generate a joint feature representation vector. The style attribution probability distribution and attribution confidence are output in the clustering discrimination path, while the process restoration index, main needle focusing index, and process logic deflection index are output in the structural index path. In the loss function calculation phase, the style clustering task is used as the self-supervised clustering objective. A clustering consistency loss based on the original features and cluster pseudo-labels is adopted. Simultaneously, a multi-task loss function for three structural indices is constructed, including structural fragment matching error, attention hotspot shift loss, and process path deviation loss. These are weighted and summed using preset task weights to form the total loss function. In the backpropagation and weight update phase, gradient backpropagation is performed. The Adam optimizer is used to iteratively optimize the network parameters, and strategies such as gradient clipping, parameter normalization, and learning rate cosine annealing are introduced to improve model training stability and structural clustering generalization ability.

[0049] After each training round, a complete inference process evaluation is performed based on the validation set. The style category distribution map and the predicted value of the structure index are output respectively. The consistency score of the style clustering results (such as NMI, ARI), the mean squared error (MSE) of the structure output, the structure evaluation accuracy, and the trend stability index of the three indices are calculated. The loss function decline curve, the structure evaluation curve and the clustering consistency curve are dynamically plotted according to the performance change trend of the validation set. If the validation index does not improve for several consecutive rounds, the Early Stopping strategy is triggered to terminate the training process in advance.

[0050] After training is complete, the system saves the network weight parameter file in the final converged state and supports exporting it as a standard model format for deployment and use.

[0051] The specific steps for constructing an initial intangible cultural heritage transmission graph based on the attribute triplet of each embroidery inheritor are as follows: Read the style tag of each embroidery inheritor and construct a set of style category master nodes, i.e., several style category master nodes (each style tag corresponds to a unique style category node, and then statistics are performed to cluster nodes belonging to the same style tag into a unified style category node); read the inheritor ID of each embroidery inheritor and construct an inheritor node set (including the inheritor's basic information field, which serves as an individual node in the graph); read the attribution confidence of each embroidery inheritor and establish graph connection edges between the inheritor node set and the corresponding style category master nodes (i.e., the connection edge represents the attribution relationship between the inheritor and the style, and the attribution confidence is attached as edge weight information), generating the initial intangible cultural heritage transmission graph.

[0052] This implementation plan introduces a joint self-supervised clustering network, integrating image data of embroidery inheritors' works, identity information, and technical voice information to comprehensively extract style tags, attribution confidence, and structural evaluation indicators of craftsmanship. This constructs attribute triples with semantic attribution and craft structure expression capabilities. Furthermore, inheritor IDs are used as unique identifiers to construct individual node sets, style tags are aggregated to construct style category master node sets, and graph connection edges are established through attribution confidence. This achieves a hierarchical connection between styles and inheritors, thereby improving the structural clarity and organizational rationality of the intangible cultural heritage graph. This ensures the accurate attribution and positioning of individual inheritors in the graph, ultimately giving the graph traceable, scalable, and computable graph structure expression capabilities.

[0053] Specifically, the dissemination behavior data includes video upload count, account follower count, exhibition participation count, dissemination interaction guidance index, dissemination trajectory index, and platform co-occurrence coverage index. The specific steps to obtain the dissemination traction index for each embroidery inheritor are as follows: A comprehensive analysis of the video upload count, account follower count, and exhibition participation count for each embroidery inheritor is performed based on the particle swarm optimization algorithm (first, standardization is performed, and then weighting is applied based on the standardization results. The weighting coefficients corresponding to the standardized video upload count, account follower count, and exhibition participation count are adaptively obtained through the particle swarm optimization algorithm. Specifically, after standardization, the three standardized values ​​are...). The system uses the number of video uploads, account followers, and exhibition participations as input vectors to construct a data structure for weighted fusion. Each weight coefficient combination is used as a particle, and each particle represents a candidate solution for weighted fusion of three standardized propagation parameters. The particle's position is the current weight combination, and its velocity represents the direction and magnitude of its search adjustment in the solution space. Multiple particles are initialized to form an initial particle swarm, and each particle is given a random initial position and search velocity. During the search process, a global optimal position and an individual optimal position are defined to guide the particle's position update. In each iteration, the system evaluates the propagation parameters and preset evaluation objectives, such as exponential distribution balance and difference recognition. Alternatively, the fitness value of each particle can be calculated based on its fit with artificial labels, and its position and velocity can be updated accordingly. After multiple rounds of search, the particles continuously approach the globally optimal weighted combination. When the particle swarm evolution reaches a preset stopping condition, such as the maximum number of iterations or the stable number of rounds of the globally optimal solution, it is used as the globally optimal weighted combination, i.e., the weighting coefficients corresponding to the standardized video upload count, account follower count, and exhibition participation count. This yields the dissemination and diffusion driving index for each embroidery inheritor (representing the active diffusion ability of intangible cultural heritage inheritors in the process of multi-channel dissemination). Based on the particle swarm optimization algorithm, the dissemination interaction guidance index, dissemination trajectory index, and platform co-occurrence coverage of each embroidery inheritor are calculated. The indices are comprehensively analyzed (firstly, they are standardized, and then weighted based on the standardized results. The weighting coefficients corresponding to the standardized dissemination interaction guidance index, dissemination trajectory index, and platform co-occurrence coverage index are adaptively obtained through particle swarm optimization algorithm. This is consistent with the logic of the weighting coefficients corresponding to the standardized video upload quantity, account follower count, and exhibition participation count, which are also adaptively obtained through particle swarm optimization algorithm). This yields the content reach coverage index for each embroidery inheritor (representing the breadth of user perception and the extensibility of the dissemination path of intangible cultural heritage embroidery works in the actual dissemination process, and measuring their dissemination coverage capabilities in spatial, platform, and interactive dimensions).Furthermore, a comprehensive analysis was conducted on the dissemination and diffusion driving index and content reach and coverage index of each embroidery inheritor to obtain the dissemination traction index for each inheritor (representing the degree of comprehensive response of the inheritor's proactive driving ability and actual dissemination response ability during the dissemination of the intangible cultural heritage embroidery techniques they represent).

[0054] The video upload count is the total number of videos related to the embroidery works uploaded on various social media platforms (such as Douyin, Kuaishou, Bilibili, etc.). Data is retrieved from the content data interfaces of open platforms such as Douyin, Kuaishou, and Bilibili, combined with the pre-entered real-name account information of the inheritors, and the video posting records are filtered. The content hash fingerprint matching (i.e., embedding a unique hash fingerprint into each work image and then matching it) and title semantic keyword retrieval are used to determine whether the content is the original work of the inheritor. The number of successfully matched videos is counted to obtain the video upload count of the inheritor.

[0055] The account follower count is the cumulative number of followers of the inheritor's social media account across major communication platforms. It is used to assess the public attention base and follower influence. By calling the platform's open API (such as Douyin Developer Platform, Weibo data interface, etc.), the current follower count is obtained based on the inheritor's account ID field. If the inheritor has accounts on multiple platforms, the weighted average is taken as the account follower count.

[0056] The exhibition participation count is the cumulative number of times an embroidery inheritor has participated in various offline intangible cultural heritage exhibitions, handicraft displays, special lectures, or local cultural exchange activities in the past 5 years. It reflects the level of activity in the industry or public communication channels. By connecting with the data interfaces of cultural departments at all levels, exhibition organizers, and intangible cultural heritage platforms, activity registration information and participant lists are obtained. Based on the joint index fields such as the inheritor's name, ID number, and registration number, precise matching is performed, and historical participation records are deduplicated to obtain the exhibition participation count.

[0057] The Dissemination Interaction Guidance Index reflects the degree of user interaction triggered by proactive content explanation during the dissemination process. It demonstrates the actual guiding ability of the disseminated content on audience behavior. The index is generated by acquiring several explanatory videos, performing automatic speech recognition (ASR), extracting the transcribed text, identifying phrases containing proactive speaking intent such as "Listen to me," "Pay attention here," and "You can watch" through keyword filtering rules, locating the explanation segments in the video timeline, and calling platform interfaces to obtain the number of likes and comments generated within the explanation segment. These data are then weighted, and the weighted results for each video are averaged to obtain the Dissemination Interaction Guidance Index.

[0058] The propagation trajectory index is the maximum geographical radius of a work's spread after it has been published on various platforms and is used to measure the spatial breadth of the work's spread. The specific steps to obtain it are as follows: when uploading video or image content, a unique hash fingerprint is embedded in each work image. When the work is forwarded or referenced on other user devices or platform pages, the content reuse behavior is detected by fingerprint recognition, and the IP address of each occurrence is recorded. Then, the spatial straight-line distance between the farthest forwarding node and the first upload node is calculated. This maximum geographical distance is the propagation trajectory index.

[0059] The platform co-occurrence coverage index measures the degree to which the same embroidery work is simultaneously published and appears on different platforms. It is used to measure the multi-platform distribution capability of content dissemination. It captures content indexes from mainstream platforms such as Douyin, Kuaishou, Xiaohongshu, and Bilibili, and uses image similarity recognition (such as Perceptual Hash) to compare with title keywords to determine whether they are the same content. It counts the number of different platforms on which the work appears within a preset time window (such as 72 hours), and then outputs the platform co-occurrence coverage index as the ratio of the number of co-occurring platforms to the total number of platforms.

[0060] The specific formula for calculating the dissemination influence index of a certain embroidery inheritor is as follows: ;in, The dissemination index of a certain embroidery inheritor. The dissemination and diffusion driving index for a certain embroidery inheritor. The diffusion-driven adjustment coefficients are stored in the database. The content reach and coverage index for a certain embroidery inheritor The reach coverage adjustment coefficients are stored in the database. These are the balance adjustment coefficients stored in the database.

[0061] It needs to be explained that in the formula This item is used to dynamically regulate the degree of matching between the inheritor's proactive dissemination behavior (i.e., the dissemination and diffusion driving index) and the actual dissemination feedback of their works (i.e., the content reach and coverage index), thereby enhancing the behavioral consistency constraint ability of the dissemination traction index.

[0062] , , The following steps can be used to obtain the initial influence weights of each variable (diffusion-driving index and content reach-coverage index) on the dissemination traction index based on historical data through statistical regression analysis. Then, the range of values ​​of the coefficients is adjusted using sensitivity analysis to assess the stability and applicability of these parameters to the formula output. Next, the weights are further fitted through model optimization (such as machine learning algorithms or multi-objective optimization) to ensure that the formula can accurately reflect the actual dissemination status.

[0063] This implementation plan constructs a dissemination traction index system composed of a dissemination and diffusion driving index and a content reach and coverage index. This system systematically evaluates the proactive driving ability and actual content feedback ability of embroidery inheritors in the process of disseminating intangible cultural heritage. Secondly, it uses a particle swarm optimization algorithm to standardize and adaptively weight and fuse multiple dissemination-related parameters, effectively solving the problems of heterogeneous dissemination data, inconsistent dimensions, and subjective weight settings. Next, a balance adjustment term is introduced into the formula to achieve dynamic matching constraints between dissemination behavior and content feedback, thereby avoiding unbalanced evaluation results dominated by a single indicator. This improves the accuracy and consistency of the dissemination traction index. Finally, this dissemination traction index not only truly reflects the overall performance of inheritors in the multi-platform dissemination system, but also provides a high-quality behavioral quantification basis for subsequent skill inheritance index calculation and dynamic updating of the intangible cultural heritage map.

[0064] Specifically, the steps to obtain the structural process stability index of each embroidery inheritor are as follows: Read the set of process structure evaluation indices in the attribute triplets of each embroidery inheritor, namely, process restoration index, main needlework focus index, and process logic deviation index; conduct a comprehensive analysis of the process restoration index, main needlework focus index, and process logic deviation index of each embroidery inheritor to obtain the structural process stability index of each embroidery inheritor.

[0065] The specific formula for calculating the structural craftsmanship stability index of a certain embroidery inheritor is as follows: ;in, This refers to the structural and technical stability index of a certain embroidery inheritor. The craftsmanship reproduction index of a certain embroidery inheritor. These are the process restoration adjustment coefficients stored in the database. Focusing on the master stitch technique of a certain embroidery inheritor. The focus adjustment coefficients are stored in the database. This refers to the process logic deviation index of a certain embroidery inheritor. The logical deflection adjustment coefficients are stored in the database. These are the interaction adjustment coefficients stored in the database.

[0066] It needs to be explained that the specific form of the tanh function is as follows: ,in, It is a natural constant, and in this example it can be taken as 2.71, with a domain of (−∞, +∞) and a range of (−1, +1).

[0067] , , , The following steps can be used to obtain the initial influence weights of each variable (process restoration index, main needle method focusing index, and process logic deflection index) on the structural process stability index through statistical regression analysis based on historical data. Then, the range of values ​​of the coefficients is adjusted using sensitivity analysis to evaluate the stability and applicability of these parameters to the formula output. Next, the weights are further fitted through model optimization (such as machine learning algorithms or multi-objective optimization) to ensure that the formula can accurately reflect the actual process inheritance status.

[0068] The following is a specific example of calculating the structural process stability index of a certain embroidery inheritor. The available data includes the process restoration index, main stitch focus index, and process logic deflection index of five randomly selected embroidery inheritors, as detailed in Table 1 and... Figure 3 As shown:

[0069] Table 1. Example of data from the structural assessment index set of embroidery inheritor sequence craftsmanship.

[0070] Process reduction index Main needle method focusing index Process logic deflection index Embroidery heir 1 0.924 0.887 0.152 Embroidery heir 2 0.873 0.794 0.114 Embroidery heir 3 0.912 0.837 0.134 Embroidery heir 4 0.954 0.912 0.098 Embroidery heir 5 0.849 0.763 0.217

[0071] Process reduction adjustment coefficients stored in the database Approximately 0.348;

[0072] Focusing adjustment coefficients stored in the database Approximately 0.246;

[0073] Logical deflection adjustment coefficients stored in the database Approximately 0.297;

[0074] Interaction adjustment coefficients stored in the database Approximately 0.584;

[0075] Substituting the data from Table 1 and the aforementioned adjustment coefficients into the specific formula for calculating the structural craftsmanship stability index of a certain embroidery inheritor, we obtain:

[0076] The structural craftsmanship stability index of the first-ranked embroidery inheritor is calculated as: tanh((ln(1+0.924)) 0.348 ×0.887 0.246 × (0.297 / 0.152) × 0.584) ≈ 0.545;

[0077] The structural craftsmanship stability index of the second-ranked embroidery inheritor is calculated as: tanh((ln(1+0.873)) 0.348 ×0.794 0.246 × (0.297 / 0.114) × 0.584) ≈ 0.607;

[0078] The structural craftsmanship stability index of the third-ranked embroidery inheritor is calculated as: tanh((ln(1+0.912)) 0.348 ×0.837 0.246 × (0.297 / 0.134) × 0.584) ≈ 0.573;

[0079] The structural craftsmanship stability index of the fourth embroidery inheritor = tanh((ln(1+0.954)) 0.348 ×0.912 0.246 × (0.297 / 0.098) × 0.584) ≈ 0.662;

[0080] The structural craftsmanship stability index of the fifth embroidery inheritor = tanh((ln(1+0.849)) 0.348 ×0.763 0.246 ×(0.297 / 0.217)))×0.584)≈0.432.

[0081] This implementation plan comprehensively considers three dimensions: the process restoration index, the main needlework focus index, and the process logic deflection index. This allows for an accurate depiction of the degree to which the embroidery work retains its traditional process structure, the visual focus of key needlework techniques, and the logical consistency of the operational path. Furthermore, personalized weighted adjustments are made using adjustment coefficients stored in the database, enabling adaptive matching and dynamic adjustment among the indicators. In addition, the structural process stability index incorporates a tanh nonlinear function to normalize the overall evaluation results, exhibiting good numerical stability, strong boundary controllability, and high outlier suppression capabilities. This effectively enhances the model's robustness in distinguishing different types of heritage samples and its structural representation ability. It provides a repeatable, traceable, and logically transparent quantitative calculation path, thereby supporting the explicit expression and hierarchical display of structural stability in the graph.

[0082] Specifically, the steps for generating the embroidery intangible cultural heritage inheritance map are as follows: The skill inheritance index of each embroidery inheritor is linked to the inheritor node of the corresponding style category main node in the initial intangible cultural heritage inheritance map, and cluster mean analysis is performed to obtain the style skill inheritance index of each style category main node in the initial intangible cultural heritage inheritance map; based on the style skill inheritance index of each style category main node, the initial intangible cultural heritage inheritance map is color-coded to generate the embroidery intangible cultural heritage inheritance map.

[0083] The process involves comparing the style and skill inheritance index of each style category main node in the initial intangible cultural heritage inheritance map with a preset threshold range. If the style and skill inheritance index of each style category main node in the initial intangible cultural heritage inheritance map is lower than the lower limit of the threshold range, the main node of that style category is marked in red (indicating that the style faces a risk of skill discontinuity or overall skill decline within the current inheritor system). If the style and skill inheritance index of each style category main node in the initial intangible cultural heritage inheritance map is within the threshold range, the main node of that style category is marked in yellow (indicating that the style's skills are basically stable within the current inheritor system, but have not yet reached a high-quality inheritance level). If the style and skill inheritance index of each style category main node in the initial intangible cultural heritage inheritance map is higher than the upper limit of the threshold range, the main node of that style category is marked in green (indicating that the style has a good foundation for inheritance and a high level of skill within the current inheritor system, and its inheritance status is healthy and stable).

[0084] In this implementation plan, by introducing the cluster mean of the individual skill status of inheritors as a structural indicator of style hierarchy, style category nodes are no longer static labels, but possess dynamic and quantitative inheritance expression capabilities. This enhances the evaluation dimensions and hierarchical clarity of the graph nodes. Secondly, by matching the style skill inheritance index with multiple threshold ranges and using red, yellow, and green color markings, intuitive identification of various style inheritance statuses is achieved. This effectively distinguishes between skill discontinuity risk, stable but needing improvement status, and high-level healthy inheritance status, and gives the graph the function of identifying the strength of inheritance. Finally, this visualization process not only simplifies the cognitive threshold of the graph, but also provides a quantitative basis for subsequent tracking analysis, key support, or intervention strategies, thereby giving the graph structure the ability to dynamically analyze and update, thus improving its application breadth and practical decision-making reference value.

[0085] Please see Figure 4This invention provides a technical solution: a system for constructing an intangible cultural heritage (ICH) transmission map, comprising: a data acquisition and identification module for acquiring image data of works, identity information, and audio information from skill interviews of several ICH inheritors, and inputting them into a style clustering model for identification and analysis to obtain attribute triples for each ICH inheritor; an initial map construction module for constructing an initial ICH transmission map based on the attribute triples of each ICH inheritor; a feature analysis module for acquiring dissemination behavior data of each ICH inheritor and performing feature analysis with the attribute triples to obtain the dissemination traction index and structural process stability index of each ICH inheritor; a comprehensive analysis module for comprehensively analyzing the dissemination traction index and structural process stability index of each ICH inheritor to obtain the skill transmission index of each ICH inheritor; and an ICH transmission map generation module for labeling and expanding the initial ICH transmission map based on the skill transmission index of each ICH inheritor to generate an ICH transmission map for embroidery.

[0086] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0087] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for constructing a non-genetic inheritance map, characterized in that, Includes the following steps: The image data of the works, identity information and audio information of the skill interviews of several embroidery inheritors were obtained and input into the style clustering model for recognition and analysis to obtain the attribute triplet of each embroidery inheritor. An initial intangible cultural heritage transmission map is constructed based on the attribute ternary set of each embroidery inheritor; Data on the dissemination behavior of each embroidery inheritor was obtained and feature analysis was performed on each of the attribute triples to obtain the dissemination traction index and structural process stability index of each embroidery inheritor. A comprehensive analysis of the dissemination traction index and structural process stability index of each embroidery inheritor was conducted to obtain the skill inheritance index of each embroidery inheritor. Based on the skill inheritance index of each embroidery inheritor, the initial intangible cultural heritage inheritance map is marked and expanded to generate an embroidery intangible cultural heritage inheritance map.

2. The method for constructing a non-genetic inheritance map according to claim 1, characterized in that, The specific formula for calculating the skill inheritance index of a certain embroidery inheritor is as follows: ； in, , , The indices are, in order, the skill inheritance index, the structural and technological stability index, and the dissemination and influence index of a certain embroidery inheritor. , , , The parameters are, in order, the process adjustment coefficient, the propagation adjustment coefficient, the matching deviation adjustment coefficient, and the difference adjustment coefficient stored in the database.

3. The method for constructing a non-genetic inheritance map according to claim 1, characterized in that, The image data of the artwork specifically refers to the pixel value and two-dimensional coordinates of each pixel in the artwork image. The style clustering model is specifically a joint self-supervised clustering network, which includes an input layer, a feature extraction layer, a fusion representation layer, and an output layer.

4. The method for constructing a non-genetic inheritance map according to claim 3, characterized in that, The specific steps to obtain the attribute triad for each embroidery inheritor are as follows: In the input layer of the joint self-supervised clustering network, image data of each embroidery inheritor's work, identity information, and audio information from skill interviews are received and preprocessed. In the feature extraction layer of the joint self-supervised clustering network, the preprocessed image data of each embroidery inheritor's works, identity information, and voice information of the skill interview are respectively processed by feature encoding to obtain the image vector and semantic embedding vector of each embroidery inheritor's works. In the fusion representation layer of the joint self-supervised clustering network, the image vector and semantic embedding vector of each embroidery inheritor's work are fused to obtain the joint feature representation vector of each embroidery inheritor. In the output layer of the joint self-supervised clustering network, attribute extraction processing is performed on the joint feature representation vector of each embroidery inheritor to obtain the attribute triples of each embroidery inheritor, including inheritor ID, style tag, attribution confidence, and set of craft structure evaluation indexes; The set of process structural evaluation indices includes the process reduction index, the main needle method focusing index, and the process logic deflection index.

5. The method for constructing a non-genetic inheritance map according to claim 4, characterized in that, The specific steps for constructing an initial intangible cultural heritage transmission map based on the attribute triplets of each embroidery inheritor are as follows: Read the style tags of each embroidery inheritor and construct a set of style category master nodes, that is, several style category master nodes; Read the inheritor ID of each embroidery inheritor and construct an inheritor node set; Read the attribution confidence of each embroidery inheritor, and establish graph connection edges between the inheritor node set and the corresponding style category master node to generate an initial intangible cultural heritage inheritance graph.

6. The method for constructing a non-genetic inheritance map according to claim 1, characterized in that, The dissemination behavior data includes the number of videos uploaded, the number of account followers, the number of times participants attended exhibitions, the dissemination interaction guidance index, the dissemination trajectory index, and the platform co-occurrence coverage index. The specific steps to obtain the dissemination traction index for each embroidery inheritor are as follows: Based on the particle swarm optimization algorithm, the number of videos uploaded, the number of account followers, and the number of times the embroidery was participated in exhibitions for each embroidery inheritor were comprehensively analyzed to obtain the dissemination and diffusion driving index of each embroidery inheritor. Based on the particle swarm optimization algorithm, a comprehensive analysis was conducted on the dissemination interaction guidance index, dissemination trajectory index, and platform co-occurrence coverage index of each embroidery inheritor to obtain the content reach coverage index of each embroidery inheritor; Furthermore, a comprehensive analysis was conducted on the dissemination and diffusion driving index and content reach and coverage index of each embroidery inheritor to obtain the dissemination traction index of each embroidery inheritor.

7. The method for constructing a non-genetic inheritance map according to claim 6, characterized in that, The specific formula for calculating the dissemination influence index of a certain embroidery inheritor is as follows: ； in, , , The three indices, in order, are the dissemination pull index, dissemination and diffusion drive index, and content reach and coverage index of a certain embroidery inheritor. The dissemination and diffusion driving index for a certain embroidery inheritor. , , The coefficients are, in order, the diffusion-driven adjustment coefficient, the reach-coverage adjustment coefficient, and the balance adjustment coefficient stored in the database.

8. The method for constructing a non-genetic inheritance map according to claim 4, characterized in that, The specific steps to obtain the structural craftsmanship stability index for each embroidery inheritor are as follows: Read the set of process structure evaluation indices from the attribute triplets of each embroidery inheritor, namely process restoration index, main needlework focus index, and process logic deviation index; By comprehensively analyzing the craft restoration index, main needlework focus index, and process logic deviation index of each embroidery inheritor, the structural craft stability index of each embroidery inheritor is obtained.

9. The method for constructing a non-genetic inheritance map according to claim 1, characterized in that, The specific steps for generating an intangible cultural heritage map of embroidery are as follows: The skill inheritance index of each embroidery inheritor is attached to the inheritor node of the corresponding style category main node in the initial intangible cultural heritage inheritance map, and cluster mean analysis is performed to obtain the style skill inheritance index of each style category main node in the initial intangible cultural heritage inheritance map. Based on the style and skill inheritance index of each style category master node, the initial intangible cultural heritage map is color-coded to generate an embroidery intangible cultural heritage map.

10. A system for constructing a non-heritage inheritance map, using the method for constructing a non-heritage inheritance map according to any one of claims 1-9, characterized in that, include: The data acquisition and recognition module is used to acquire image data of works, identity information and audio information of skill interviews of several embroidery inheritors, and input them into the style clustering model for recognition and analysis to obtain the attribute triplet of each embroidery inheritor; The initial atlas construction module is used to construct an initial intangible cultural heritage atlas based on the attribute triplets of each embroidery inheritor; The feature analysis module is used to acquire the dissemination behavior data of each embroidery inheritor and perform feature analysis with the attribute triples to obtain the dissemination traction index and structural process stability index of each embroidery inheritor. The comprehensive analysis module is used to conduct a comprehensive analysis of the dissemination traction index and structural process stability index of each embroidery inheritor, so as to obtain the skill inheritance index of each embroidery inheritor. The Intangible Cultural Heritage Atlas Generation Module is used to mark and expand the initial Intangible Cultural Heritage Atlas based on the skill inheritance index of each embroidery inheritor, and generate an embroidery Intangible Cultural Heritage Atlas.

Images