Non-perpetual numeral intelligent village construction method based on digital twinborn and generative artificial intelligence
By using digital twin technology and generative artificial intelligence technology, digital twin villages of intangible cultural heritage in rural areas are constructed. This enables the accurate reproduction of the dynamic process flow and artisan interaction of intangible cultural heritage skills, generates derivative content with cultural accuracy, and establishes real-time linkage between digital intangible cultural heritage content and physical villages, forming a self-evolving closed loop and improving the protection effect of intangible cultural heritage resources.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-23
- Publication Date
- 2026-03-27
AI Technical Summary
Existing digital museums cannot accurately reproduce the dynamic process of rural intangible cultural heritage skills and the interaction methods of artisans. Moreover, the collection and production of rural intangible cultural heritage digital assets are costly and time-consuming. They lack a deep semantic understanding of the historical texts and folk symbols behind intangible cultural heritage, and cannot automatically generate derivative content with cultural accuracy. It is difficult to achieve real-time linkage between intangible cultural heritage knowledge and offline rural entities.
By employing digital twin technology and generative artificial intelligence technology, data on intangible cultural heritage projects are collected through multimodal sensing devices to construct high-fidelity digital twins, train generative artificial intelligence models, and fine-tune them using LoRA in conjunction with knowledge graphs. A collaborative operation mechanism of "cloud-edge-device" is established to achieve high-fidelity mapping and immersive interaction of intangible cultural heritage resources.
It has achieved accurate restoration of the dynamic process flow and artisan interaction of rural intangible cultural heritage skills, generated derivative content with cultural accuracy, established real-time linkage between digital intangible cultural heritage content and physical villages, formed a self-evolving closed loop, and improved the protection effect of intangible cultural heritage resources.
Smart Images
Figure CN121744906A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The application relates to a non-heritage digital smart village construction method based on digital twinning and generative artificial intelligence. BACKGROUND
[0002] Most existing digital museums adopt the form of displaying static pictures and playing recorded videos, and cannot restore the dynamic process of rural non-heritage skills, the interaction mode of artisans and the cultural context of rural areas.
[0003] Further, the collection and production of rural non-heritage digital assets are highly dependent on professional teams, and the cycle is too long and the cost is too high, which is difficult to cover a large number of rural non-heritage projects, and at the same time lacks deep semantic understanding of historical texts, folk symbols and oral history behind non-heritage, and cannot automatically generate derivative content with cultural accuracy, so as to form a self-evolution construction system around the real-time linkage of artificial intelligence content generation of non-heritage knowledge and offline rural entities. SUMMARY
[0004] The embodiment of the application provides a non-heritage digital smart village construction method based on digital twinning and generative artificial intelligence, which is reasonable in design, and through the cooperation of digital twinning technology and generative artificial intelligence technology, can accurately restore the dynamic process of rural non-heritage skills, the interaction mode of artisans and the cultural context of rural areas, automatically generate derivative content with cultural accuracy, and at the same time establish real-time linkage between rural non-heritage digital content and physical villages, workshops and inheritors, so as to form a self-evolution closed loop mechanism around the real-time linkage of artificial intelligence content generation of non-heritage knowledge and offline rural entities, realize high-fidelity mapping, intelligent generation and immersive interactive experience of rural non-heritage resources, construct a self-evolution non-heritage digital smart twin village, and improve the effective protection of rural non-heritage content, thereby solving the problems in the prior art.
[0005] The technical scheme adopted by the application to solve the above technical problems is: The non-heritage digital smart village construction method based on digital twinning and generative artificial intelligence comprises the following steps: S1, constructing a rural non-heritage digital twinning base, collecting rural non-heritage project static asset parameters and dynamic adjustment processes through multi-modal perception devices, and fusing village geographical space data to construct a high-fidelity digital twinning body; S2, constructing a rural non-heritage cultural knowledge graph, extracting rural non-heritage entities and their semantic relationships from historical documents, oral history and folk records to form a structured knowledge base; S3, training a rural non-heritage generative artificial intelligence large model, using the knowledge graph as a constraint condition, using a multi-modal data set for LoRA fine-tuning, so that the rural non-heritage generative artificial intelligence large model can generate culture-compliant content; S4, interact with the non-heritage digital intelligence twin village via natural language or gestures, and generate personalized narratives, virtual heritage person dialogues, and derivative digital collections in real time through generative artificial intelligence, to realize immersive interaction and content generation; S5, establish a "cloud-edge-end" collaborative operation mechanism, and return user interaction data and offline workshop sales and reservation data to the construction system for knowledge graph updating and artificial intelligence model iteration, to form a self-evolution closed loop mechanism, realize high-fidelity mapping, intelligent generation and immersive interaction experience of rural non-heritage resources, and build a self-evolution non-heritage digital intelligence twin village, and improve the effective protection of rural non-heritage content.
[0006] The multi-modal perception device comprises: A static acquisition component acquires a millimeter-level 3D model of the rural non-heritage artifact by using a three-dimensional laser scanner or photogrammetry; A dynamic capture component records joint angles and tool trajectories during the artisan's making process by using an inertial motion capture suit, and synchronously acquires environmental sound and commentary voice; A space mapping component constructs a three-dimensional model of the village scene by using a UAV oblique photography, and fuses with BIM data or GIS data.
[0007] Construct a rural non-heritage cultural knowledge graph, extract rural non-heritage entities and their semantic relationships from historical documents, oral history, and folklore records, and form a structured knowledge base comprising the following steps: S2.1, recognize non-heritage entities from unstructured text by using a BERT-BiLSTM-CRF model; S2.2, extract triple relationships based on a predefined ontology; S2.3, form a directed graph with non-heritage entities as nodes and semantic relationships as edges.
[0008] Build a total loss function corresponding to the rural non-heritage generative artificial intelligence large model, wherein the total loss function is: Ltotal = Lcrossentropy + Lmultimodalalignment + Luserpreference + Lgraphembeddingconstraint CE Lcrossentropy is a cross-entropy loss, Lmultimodalalignment is a multi-modal alignment loss, Luserpreference is a user preference loss, and Lgraphembeddingconstraint is a knowledge graph embedding constraint loss. align preference culture =||h+r-t|| 2 h, r, and t are TransE embedding vectors of the head entity, the relationship, and the tail entity respectively; λ1, λ2, and λ3 are dynamic weights for balancing cultural consistency and other objectives, and the value range is [0.1, 1.0].
[0009] The immersive interaction comprises: Virtual inheritor dialogue, combined with knowledge graph and action data to generate voice answers and drive virtual artisans to demonstrate actions; Generate AI digital collectibles, output 3D models and NFT metadata that meet cultural symbol specifications.
[0010] The self-evolution closed loop mechanism includes: Offline workshop, every time a physical product is sold, the system automatically triggers the airdrop of online digital collectibles; When the user's virtual experience score is lower than the preset threshold, trigger knowledge graph review and AI model incremental training.
[0011] The construction system includes: Digital twin construction module, used to construct a rural intangible cultural heritage digital twin base, collect static asset parameters and dynamic adjustment processes of rural intangible cultural heritage projects through multi-modal perception devices, and fuse village geographic space data to construct a high-fidelity digital twin; Knowledge graph module, used to construct a rural intangible cultural heritage knowledge graph, extract rural intangible cultural heritage entities and their semantic relationships from historical documents, oral history, and folklore records to form a structured knowledge base; Artificial intelligence module, used to train a rural intangible cultural heritage generative artificial intelligence large model, use the knowledge graph as a constraint condition, and use multi-modal data sets for LoRA fine-tuning, so that the rural intangible cultural heritage generative artificial intelligence large model can generate culturally compliant content; Application interaction module, used to interact with the intangible cultural heritage digital twin village through natural language or gestures, generate personalized narratives, virtual inheritor dialogues, and derivative digital collectibles in real time, achieve immersive interaction and content generation; Data feedback closed loop module, used to establish a "cloud-edge-end" collaborative operation mechanism, return user interaction data and offline workshop sales and reservation data to the construction system for knowledge graph update and artificial intelligence model iteration, forming a self-evolution closed loop mechanism.
[0012] The application adopts the above structure and method, constructs a rural intangible cultural heritage digital twin base through a digital twin construction module, collects rural intangible cultural heritage project static asset parameters and dynamic adjustment processes through a multi-modal sensing device, and fuses village geographic space data to construct a high-fidelity digital twin body; constructs a rural intangible cultural heritage knowledge graph through a knowledge graph module, extracts rural intangible cultural heritage entities and their semantic relationships from historical documents, oral history, and folk customs records to form a structured knowledge base; trains a rural intangible cultural heritage generative artificial intelligence large model through an artificial intelligence module, uses the knowledge graph as a constraint condition, fine-tunes the model using a multi-modal data set, so that the rural intangible cultural heritage generative artificial intelligence large model can generate culture-compliant content; uses an application interaction module to interact with the intangible cultural heritage digital twin village through natural language or gestures, and the generative artificial intelligence generates personalized narratives, virtual inheritor dialogues, and derivative digital collectibles in real time; and establishes a 'cloud-edge-end' collaborative operation mechanism through a data feedback closed loop module, and returns user interaction data and offline workshop sales and reservation data to the construction system, which has the advantages of intelligence, efficiency, accuracy and practicality. BRIEF DESCRIPTION OF DRAWINGS
[0013] Figure 1 The overall architecture and data flow diagram of the application.
[0014] Figure 2 The rural intangible cultural heritage multi-modal data acquisition and digital twin body construction flowchart of the application.
[0015] Figure 3 The rural intangible cultural heritage knowledge graph construction and TransE embedding diagram of the application.
[0016] Figure 4 The AI content generation and user interaction flowchart of the application.
[0017] Figure 5 The self-evolution closed loop mechanism diagram of the application. DETAILED DESCRIPTION
[0018] To clearly illustrate the technical features of the present application, the application will be described in detail below with reference to the specific embodiments and the accompanying drawings.
[0019] As shown in Figures 1-5 The intangible cultural heritage digital village construction method based on digital twin and generative artificial intelligence includes the following steps: S1, construct a rural intangible cultural heritage digital twin base, collect rural intangible cultural heritage project static asset parameters and dynamic adjustment processes through a multi-modal sensing device, and fuse village geographic space data to construct a high-fidelity digital twin body; S2, construct a knowledge graph of rural intangible cultural heritage, extract rural intangible cultural heritage entities and their semantic relationships from historical documents, oral histories, and folklore to form a structured knowledge base; S3 trains a large-scale AI model for the generative development of rural intangible cultural heritage. Using knowledge graphs as constraints, it fine-tunes the model using LoRA with a multimodal dataset, enabling the model to generate culturally compliant content. S4 interacts with the digital twin village of intangible cultural heritage through natural language or gestures. Generative artificial intelligence generates personalized narratives, virtual inheritor dialogues and derivative digital collections in real time, realizing immersive interaction and content generation. S5 establishes a "cloud-edge-device" collaborative operation mechanism, which feeds back user interaction data and offline workshop sales and reservation data to the construction system for knowledge graph updates and artificial intelligence model iterations, forming a self-evolving closed-loop mechanism. This enables high-fidelity mapping, intelligent generation, and immersive interactive experiences of rural intangible cultural heritage resources, constructs self-evolving digital twin villages of intangible cultural heritage, and enhances the effective protection of rural intangible cultural heritage content.
[0020] The multimodal sensing device includes: A static acquisition component, which uses a 3D laser scanner or photogrammetry to acquire millimeter-level 3D models of rural intangible cultural heritage artifacts; The motion capture component uses an inertial motion capture kit to record the joint angles and tool trajectories of the craftsman during the crafting process, and simultaneously collects ambient sound and narration voice. A spatial mapping component is used to construct a 3D model of the village landscape through UAV oblique photography and integrate it with BIM data or GIS data.
[0021] Constructing a knowledge graph of rural intangible cultural heritage involves extracting rural intangible cultural heritage entities and their semantic relationships from historical documents, oral histories, and folklore to form a structured knowledge base. This includes the following steps: S2.1, using the BERT-BiLSTM-CRF model to identify intangible cultural heritage entities from unstructured text; S2.2 Extracting triplet relations based on predefined ontology; S2.3 forms a directed graph where nodes are intangible cultural heritage entities and edges are semantic relationships.
[0022] A total loss function is constructed corresponding to the large-scale generative artificial intelligence model of rural intangible cultural heritage. The total loss function is as follows: Among them, L CE For cross-entropy loss, L align For multimodal alignment loss, L preference For user preference loss, L culture=||h+rt|| 2 λ1, λ2, and λ3 are the knowledge graph embedding constraint loss, where h, r, and t are the TransE embedding vectors of the head entity, relation, and tail entity, respectively; λ1, λ2, and λ3 are dynamic weights used to balance cultural consistency with other objectives, and their value range is [0.1, 1.0].
[0023] The immersive interaction includes: The virtual inheritor engages in dialogue, combining knowledge graphs and motion data to generate voice responses and drive the virtual craftsman to demonstrate actions; Generate AI-generated digital collectibles and output 3D models and NFT metadata that conform to cultural symbol standards.
[0024] The self-evolutionary closed-loop mechanism includes: For every physical product sold in the offline workshop, the system automatically triggers an online digital collectible airdrop. When a user's rating of the virtual experience falls below a preset threshold, a knowledge graph review and incremental training of the AI model are triggered.
[0025] The construction system includes: The digital twin construction module is used to build a digital twin base for rural intangible cultural heritage. It collects static asset parameters and dynamic adjustment processes of rural intangible cultural heritage projects through multimodal sensing devices and integrates village geospatial data to construct a high-fidelity digital twin. The knowledge graph module is used to construct a knowledge graph of rural intangible cultural heritage, extracting rural intangible cultural heritage entities and their semantic relationships from historical documents, oral histories, and folklore records to form a structured knowledge base; The artificial intelligence module is used to train a large-scale generative artificial intelligence model for rural intangible cultural heritage. It uses a knowledge graph as a constraint and LoRA fine-tuning with a multimodal dataset to enable the large-scale generative artificial intelligence model for rural intangible cultural heritage to generate culturally compliant content. The application interaction module is used to interact with the digital twin village of intangible cultural heritage through natural language or gestures. Generative artificial intelligence generates personalized narratives, virtual inheritor dialogues and derivative digital collections in real time, realizing immersive interaction and content generation. The data feedback closed-loop module is used to establish a "cloud-edge-device" collaborative operation mechanism, which feeds back user interaction data and offline workshop sales and reservation data to the construction system for knowledge graph updates and artificial intelligence model iterations, forming a self-evolving closed-loop mechanism.
[0026] The working principle of the intangible cultural heritage digital village construction method based on digital twins and generative artificial intelligence in this invention embodiment is as follows: Through the combined effect of digital twin technology and generative artificial intelligence technology, the dynamic process flow, artisan interaction mode and rural cultural context of rural intangible cultural heritage skills can be accurately restored, and derivative content with cultural accuracy can be automatically generated. At the same time, real-time linkage is established between the digital content of rural intangible cultural heritage and physical villages, workshops and inheritors. Thus, the artificial intelligence content generation of intangible cultural heritage knowledge and the offline rural entities are linked in real time and form a self-evolving closed loop mechanism, realizing high-fidelity mapping, intelligent generation and immersive interactive experience of rural intangible cultural heritage resources, constructing a self-evolving digital twin village of intangible cultural heritage, and improving the effective protection of rural intangible cultural heritage content.
[0027] The overall solution includes the following steps: constructing a digital twin foundation for rural intangible cultural heritage, collecting static asset parameters and dynamic adjustment processes of rural intangible cultural heritage projects through multimodal sensing devices, and integrating village geospatial data to construct a high-fidelity digital twin; constructing a knowledge graph of rural intangible cultural heritage, extracting rural intangible cultural heritage entities and their semantic relationships from historical documents, oral histories, and folklore to form a structured knowledge base; and training a large-scale generative artificial intelligence model for rural intangible cultural heritage, using the knowledge graph as a constraint and fine-tuning it using a multimodal dataset to enable the large-scale generative artificial intelligence model for rural intangible cultural heritage to generate texts. The system complies with regulations regarding content creation; through natural language or gestures, users interact with digital twin villages of intangible cultural heritage, and generative artificial intelligence generates personalized narratives, virtual inheritor dialogues, and derivative digital collectibles in real time, achieving immersive interaction and content generation; a "cloud-edge-device" collaborative operation mechanism is established, feeding user interaction data and offline workshop sales and reservation data back to the system for knowledge graph updates and artificial intelligence model iterations, forming a self-evolving closed-loop mechanism to achieve high-fidelity mapping, intelligent generation, and immersive interactive experiences of rural intangible cultural heritage resources, constructing self-evolving digital twin villages of intangible cultural heritage, and enhancing the effective protection of rural intangible cultural heritage content.
[0028] Correspondingly, we will take a digital twin village with intangible cultural heritage in Shandong as an example. The specific intangible cultural heritage elements include three major projects: Weifang kites, Taishan shadow puppetry, and Lu embroidery.
[0029] Step S1: Deploy a Faro 3D laser scanner at the Weifang Kite Workshop to acquire more than 1,000 point cloud data of kite frames; a virtual craftsman wearing an Xsens MVN motion capture suit records the entire process of "kite making, pasting, painting, and flying" and uses a DJI M300 RTK drone to perform 0.5cm resolution oblique photography of the village to generate a real-world 3D model.
[0030] Step S2: Crawl 10 local documents such as "Weifang Kite Chronicle", use BERT-BiLSTM-CRF model to extract triple data such as "kite - making - bamboo strip splitting - inheritor Zhang Yanlu", and construct a knowledge graph containing more than 5,000 entities and more than 8,000 relationships.
[0031] Step S3: Based on the LLaVA-1.6-13B open-source model, LoRA fine-tuning was performed on 100,000 image-text pairs and 5,000 action-text pairs, and TransE embedding was introduced as a regularization term to ensure the cultural accuracy of the generated content.
[0032] Step S4: The user scans the physical kite with the mobile AR application, triggering a virtual inheritor to demonstrate; the user inputs "design a simplified version of Lu embroidery pattern suitable for children", and AI generates a pattern that conforms to the principles and characteristics of "flat, even, fine and dense".
[0033] Step S5: For every AI-designed Lu embroidery piece sold in the offline workshop, the system automatically airdrops the corresponding digital collectible to the buyer's wallet; when the user's actual rating is lower than the preset threshold, the system automatically generates a knowledge graph conflict report.
[0034] After experimental comparison, the accuracy of the content generated by this application is improved by 26.7% compared with the baseline model without knowledge graph constraints, and the user immersive experience satisfaction is improved by 41.2%, which shows significant improvement and innovation.
[0035] For some functional components of this application, the modal sensing device includes: a static acquisition component, which uses a 3D laser scanner or photogrammetry to acquire millimeter-level 3D models of rural intangible cultural heritage artifacts; a dynamic capture component, which uses an inertial motion capture kit to record joint angles and tool trajectories during the craftsman's production process, and simultaneously acquires ambient sound and narration voice; and a spatial mapping component, which constructs a 3D model of the village landscape through UAV oblique photography and integrates it with BIM data or GIS data.
[0036] Because digital content of rural intangible cultural heritage lacks real-time linkage with physical villages, workshops, and inheritors, it is necessary to establish a "cloud-edge-terminal" collaborative operation mechanism. User interaction data and offline workshop sales and reservation data will be fed back to the system for knowledge graph updates and artificial intelligence model iterations, forming a self-evolving closed-loop mechanism to achieve high-fidelity mapping, intelligent generation, and immersive interactive experience of rural intangible cultural heritage resources.
[0037] In this application, immersive interaction includes: engaging in dialogue with virtual inheritors, generating voice responses by combining knowledge graphs and motion data, and driving virtual artisans to demonstrate actions; simultaneously generating AI digital collectibles and outputting 3D models and NFT metadata that conform to cultural symbol norms.
[0038] Because existing models are too expensive, have too long a production cycle, and are difficult to cover a large number of rural intangible cultural heritage projects, this application trains a large-scale generative artificial intelligence model for rural intangible cultural heritage. With knowledge graphs as constraints, it uses multimodal datasets for LoRA fine-tuning, which shortens the cycle, reduces costs, and improves the user's experience with intangible cultural heritage.
[0039] Furthermore, the total loss function corresponding to the generative AI model for rural intangible cultural heritage is: Among them, L CE For cross-entropy loss, L align For multimodal alignment loss, L preference For user preference loss, L culture =||h+rt|| 2 The knowledge graph embedding constraint loss is denoted by h, r, and t, which are the TransE embedding vectors of the head entity, relation, and tail entity, respectively. λ1, λ2, and λ3 are dynamic weights used to balance cultural consistency with other objectives, and their value range is [0.1, 1.0], thereby achieving accurate quantification.
[0040] In order to achieve deep semantic understanding of the historical texts, folk symbols, and oral histories behind intangible cultural heritage and automatically generate derivative content with cultural accuracy, it is necessary to construct a knowledge graph of rural intangible cultural heritage, extracting rural intangible cultural heritage entities and their semantic relationships from historical documents, oral histories, and folklore records to form a structured knowledge base.
[0041] Specifically, the process includes the following steps: using the BERT-BiLSTM-CRF model to identify intangible cultural heritage entities from unstructured text; extracting triple relations based on a predefined ontology; and forming a directed graph where nodes are intangible cultural heritage entities and edges are semantic relations.
[0042] Corresponding to the construction method, the construction system of this application includes: a digital twin construction module, used to construct a digital twin foundation for rural intangible cultural heritage, which collects static asset parameters and dynamic adjustment processes of rural intangible cultural heritage projects through multimodal sensing devices, and integrates village geospatial data to construct a high-fidelity digital twin; a knowledge graph module, used to construct a knowledge graph of rural intangible cultural heritage, extracting rural intangible cultural heritage entities and their semantic relationships from historical documents, oral histories, and folklore to form a structured knowledge base; and an artificial intelligence module, used to train a generative artificial intelligence model of rural intangible cultural heritage, using the knowledge graph as a constraint condition, and leveraging... LoRA fine-tuning using a multimodal dataset enables a large-scale generative AI model for rural intangible cultural heritage to generate culturally compliant content. An interactive module allows interaction with digital twin villages of intangible cultural heritage via natural language or gestures. Generative AI generates personalized narratives, virtual inheritor dialogues, and derivative digital collectibles in real time, achieving immersive interaction and content generation. A data feedback closed-loop module establishes a "cloud-edge-device" collaborative operation mechanism, feeding user interaction data and offline workshop sales and reservation data back to the system for knowledge graph updates and AI model iterations, forming a self-evolving closed-loop mechanism.
[0043] It should be noted that this application involves the interdisciplinary field of digital twins and generative artificial intelligence, which can realize high-fidelity mapping, intelligent generation and immersive interactive experience of rural intangible cultural heritage resources, build a self-evolving "digital twin village of intangible cultural heritage", and improve the effective protection of rural intangible cultural heritage content.
[0044] In summary, the method for constructing digital villages based on digital twins and generative artificial intelligence in this invention, through the combined use of digital twin and generative artificial intelligence technologies, can accurately reproduce the dynamic process flow, artisan interaction methods, and rural cultural context of rural intangible cultural heritage skills. It automatically generates derivative content with cultural accuracy and establishes real-time linkage between the digital content of rural intangible cultural heritage and physical villages, workshops, and inheritors. This forms a self-evolving closed-loop mechanism that links the AI-generated content of intangible cultural heritage knowledge with offline rural entities, achieving high-fidelity mapping, intelligent generation, and immersive interactive experience of rural intangible cultural heritage resources. It constructs self-evolving digital twin villages of intangible cultural heritage and enhances the effective protection of rural intangible cultural heritage content.
[0045] The above specific embodiments should not be construed as limiting the scope of protection of the present invention. For those skilled in the art, any alternative improvements or modifications made to the embodiments of the present invention shall fall within the scope of protection of the present invention.
[0046] Any aspects of this invention not described in detail are well-known to those skilled in the art.
Claims
1. A method for constructing digital villages for intangible cultural heritage based on digital twins and generative artificial intelligence, characterized in that: The construction method includes the following steps: S1, constructing a digital twin foundation for rural intangible cultural heritage, collecting static asset parameters and dynamic adjustment processes of rural intangible cultural heritage projects through multimodal sensing devices, and integrating village geospatial data to construct a high-fidelity digital twin; S2, construct a knowledge graph of rural intangible cultural heritage, extract rural intangible cultural heritage entities and their semantic relationships from historical documents, oral histories, and folklore to form a structured knowledge base; S3 trains a large-scale AI model for the generative development of rural intangible cultural heritage. Using knowledge graphs as constraints, it fine-tunes the model using LoRA with a multimodal dataset, enabling the model to generate culturally compliant content. S4 interacts with the digital twin village of intangible cultural heritage through natural language or gestures. Generative artificial intelligence generates personalized narratives, virtual inheritor dialogues and derivative digital collections in real time, realizing immersive interaction and content generation. S5 establishes a "cloud-edge-device" collaborative operation mechanism, which feeds back user interaction data and offline workshop sales and reservation data to the construction system for knowledge graph updates and artificial intelligence model iterations, forming a self-evolving closed-loop mechanism. This enables high-fidelity mapping, intelligent generation, and immersive interactive experiences of rural intangible cultural heritage resources, constructs self-evolving digital twin villages of intangible cultural heritage, and enhances the effective protection of rural intangible cultural heritage content.
2. The method for constructing intangible cultural heritage digital villages based on digital twins and generative artificial intelligence according to claim 1, characterized in that, The multimodal sensing device includes: A static acquisition component, which uses a 3D laser scanner or photogrammetry to acquire millimeter-level 3D models of rural intangible cultural heritage artifacts; The motion capture component uses an inertial motion capture kit to record the joint angles and tool trajectories of the craftsman during the crafting process, and simultaneously collects ambient sound and narration voice. A spatial mapping component is used to construct a 3D model of the village landscape through UAV oblique photography and integrate it with BIM data or GIS data.
3. The method for constructing intangible cultural heritage digital villages based on digital twins and generative artificial intelligence according to claim 1, characterized in that, Constructing a knowledge graph of rural intangible cultural heritage involves extracting rural intangible cultural heritage entities and their semantic relationships from historical documents, oral histories, and folklore to form a structured knowledge base. This includes the following steps: S2.1, using the BERT-BiLSTM-CRF model to identify intangible cultural heritage entities from unstructured text; S2.2 Extracting triplet relations based on predefined ontology; S2.3 forms a directed graph where nodes are intangible cultural heritage entities and edges are semantic relationships.
4. The method for constructing intangible cultural heritage digital villages based on digital twins and generative artificial intelligence according to claim 1, characterized in that, A total loss function is constructed corresponding to the large-scale generative artificial intelligence model of rural intangible cultural heritage. The total loss function is as follows: Among them, L CE For cross-entropy loss, L align For multimodal alignment loss, L preference For user preference loss, L culture =||h+rt|| 2 λ1, λ2, and λ3 are the knowledge graph embedding constraint loss, where h, r, and t are the TransE embedding vectors of the head entity, relation, and tail entity, respectively; λ1, λ2, and λ3 are dynamic weights used to balance cultural consistency with other objectives, and their value range is [0.1, 1.0].
5. The method for constructing intangible cultural heritage digital villages based on digital twins and generative artificial intelligence according to claim 1, characterized in that, The immersive interaction includes: The virtual inheritor engages in dialogue, combining knowledge graphs and motion data to generate voice responses and drive the virtual craftsman to demonstrate actions; Generate AI-generated digital collectibles and output 3D models and NFT metadata that conform to cultural symbol standards.
6. The method for constructing intangible cultural heritage digital villages based on digital twins and generative artificial intelligence according to claim 1, characterized in that, The self-evolutionary closed-loop mechanism includes: For every physical product sold in the offline workshop, the system automatically triggers an online digital collectible airdrop. When a user's rating of the virtual experience falls below a preset threshold, a knowledge graph review and incremental training of the AI model are triggered.
7. The method for constructing digital villages for intangible cultural heritage based on digital twins and generative artificial intelligence according to claim 1, characterized in that, The construction system includes: The digital twin construction module is used to build a digital twin base for rural intangible cultural heritage. It collects static asset parameters and dynamic adjustment processes of rural intangible cultural heritage projects through multimodal sensing devices and integrates village geospatial data to construct a high-fidelity digital twin. The knowledge graph module is used to construct a knowledge graph of rural intangible cultural heritage, extracting rural intangible cultural heritage entities and their semantic relationships from historical documents, oral histories, and folklore records to form a structured knowledge base; The artificial intelligence module is used to train a large-scale generative artificial intelligence model for rural intangible cultural heritage. It uses a knowledge graph as a constraint and LoRA fine-tuning with a multimodal dataset to enable the large-scale generative artificial intelligence model for rural intangible cultural heritage to generate culturally compliant content. The application interaction module is used to interact with the digital twin village of intangible cultural heritage through natural language or gestures. Generative artificial intelligence generates personalized narratives, virtual inheritor dialogues and derivative digital collections in real time, realizing immersive interaction and content generation. The data feedback closed-loop module is used to establish a "cloud-edge-device" collaborative operation mechanism, which feeds back user interaction data and offline workshop sales and reservation data to the construction system for knowledge graph updates and artificial intelligence model iterations, forming a self-evolving closed-loop mechanism.