An artificial intelligence-based virtual cultural relic management method and system

Abstract

The application relates to an artificial intelligence-based virtual cultural relic management method and system, belonging to the technical field of virtual cultural relic management; the system comprises a central control and coordination module, a cultural relic digitization module, a virtual display module, a cultural relic protection and restoration suggestion module, a system security and privacy protection module, an artificial intelligence analysis module and a data management and storage module. Through integration of the high-precision cultural relic digitization module and the advanced artificial intelligence analysis module, the application can quickly and accurately convert physical cultural relics into digital formats and perform deep analysis, thereby effectively predicting and identifying potential damage of the cultural relics and providing timely protection and restoration suggestions, which not only improves the efficiency of cultural relic protection, but also reduces the risk of cultural relic damage caused by human factors.

Description

Technical Field

[0001] This application belongs to the field of virtual cultural relic management technology, specifically involving a virtual cultural relic management method and system based on artificial intelligence. Background Technology

[0002] With the rapid development of information technology, virtual cultural relic management systems are gaining increasing attention as an emerging means of cultural heritage protection and display. Existing cultural relic management systems mostly employ traditional digital methods, including photography and scanning, to electronically store images and information about cultural relics. These methods have, to some extent, achieved the preservation and dissemination of cultural relic information. However, in current technology, cultural relic protection mainly relies on manual inspections and simple monitoring equipment. These methods suffer from problems such as blind spots in monitoring, slow response times, and limited data analysis capabilities.

[0003] Furthermore, traditional video surveillance systems can only provide limited environmental information and cannot comprehensively monitor various risk factors faced by cultural relics, such as changes in temperature and humidity, light intensity, and vibration. Once the monitoring equipment malfunctions or data transmission is interrupted, it will lead to delays or even failure of monitoring work. Most existing cultural relic protection systems lack intelligent and automated early warning mechanisms. Potential threats and abnormal situations often require manual analysis and judgment, which is not only inefficient but also prone to oversights due to human factors.

[0004] In addition, existing cultural relic management systems often lack effective multilingual support, which limits users from different countries and regions from understanding and learning about cultural relics. At the same time, system security issues are becoming increasingly prominent, with the risks of data leakage and unauthorized access posing a threat to the security of cultural relic information. Summary of the Invention

[0005] In view of the above-mentioned shortcomings of the existing technology, this application proposes a virtual cultural relics management system based on artificial intelligence.

[0006] The first aspect of this application provides a virtual cultural relic management method based on artificial intelligence, the method comprising: The digitization process employs high-precision 3D laser scanning and multispectral imaging technology to non-contactly collect the geometric structure and texture information of the surface of physical cultural relics, which is then used to convert the physical cultural relics into digital format through high-precision scanning technology. The virtual display process is built on a real-time graphics rendering engine and uses virtual reality or augmented reality technology to display digital cultural relics. The analysis steps, based on artificial intelligence analysis technology, employ deep learning algorithms to automatically identify and classify the characteristics of cultural relics; The proposed steps for the protection and restoration of cultural relics, based on the data provided in the analysis steps, employ image segmentation networks and time series prediction models to quantitatively assess the damage to cultural relics and predict their degradation, and provide protection and restoration recommendations. Security and privacy protection measures employ homomorphic encryption and differential privacy technologies to protect the system from the risks of unauthorized access and data breaches; The data management and storage steps are based on a hierarchical storage architecture built on a distributed file system and a time-series database. Columnar storage is used to store and manage digitized cultural relic data, ensuring data security and accessibility. The central control and coordination mechanism coordinates and manages the activities of all other modules, receives user input requests and task requests from other modules, distributes tasks to the appropriate modules according to needs, and collects and integrates data and results from each module to provide a unified system status and feedback.

[0007] The second aspect of this application provides an artificial intelligence-based virtual cultural relic management system, which includes: a central control and coordination module, a cultural relic digitization module, a virtual display module, a cultural relic protection and restoration suggestion module, a system security and privacy protection module, an artificial intelligence analysis module, and a data management and storage module; The central control and coordination module is responsible for coordinating and managing the activities of all other modules, receiving user input requests and task requests from other modules, distributing tasks to the corresponding modules according to needs, and collecting and integrating data and results from each module to provide a unified system status and feedback. The cultural relic digitization module uses high-precision three-dimensional laser scanning and multispectral imaging technology to non-contactly collect the geometric structure and texture information of the surface of physical cultural relics, which is used to convert physical cultural relics into digital format through high-precision scanning technology. The virtual display module is built on a real-time graphics rendering engine and displays digital cultural relics through virtual reality or augmented reality technology. The artificial intelligence analysis module uses deep learning algorithms to automatically identify and classify the characteristics of cultural relics; The cultural relic protection and restoration suggestion module is based on the data output by the artificial intelligence analysis module. It uses image segmentation network and time series prediction model to quantitatively assess and predict the degradation of cultural relics, and provides protection and restoration suggestions. The system security and privacy protection module employs homomorphic encryption and differential privacy technology to protect the system from the risks of unauthorized access and data leakage. The data management and storage module is based on a distributed file system and a time-series database to build a hierarchical storage architecture. It uses columnar storage to store and manage digital cultural relic data, ensuring data security and accessibility.

[0008] Furthermore, the system also includes: A multilingual support module is provided to offer multilingual interfaces and information displays to suit users in different countries and regions. The user behavior analysis module collects and analyzes user interaction data with the system, analyzes user behavior, and feeds this data back to the central control and coordination module and the virtual display module in order to continuously optimize the user experience and the displayed content. The security and privacy protection module is used to protect the system from the risks of unauthorized access and data leakage; The copyright and ethics management module is used to ensure that copyright laws and ethical standards are followed during the digitization and display of cultural relics.

[0009] Furthermore, the specific data conversion process of the cultural relic digitization module is as follows: Data acquisition submodule: Utilizes a high-precision 3D laser scanner to perform multi-view, all-around scanning of the physical artifact. Based on the time or phase difference of the laser return, it calculates the 3D coordinates (X, Y, Z) of each sampling point on the artifact's surface, thereby obtaining a point cloud set. Where p1, p2, ..., pn represent the spatial location information of each point in the set; Multi-view point cloud registration submodule: Employs an iterative nearest-point algorithm to accurately transform all local point clouds to the same global coordinate system. Its mathematical objective function is:

[0010] in, Indicates the registration error. It is a rotation matrix. It is a translation vector. It is the source cloud The point in the middle, It is a target point cloud Zhongyu For the corresponding point, where n is a natural number, calculate the optimal rigid body transformation. To minimize the error function The final output is a global point cloud model; The triangular mesh reconstruction submodule uses the Poisson surface reconstruction algorithm to convert discrete point clouds into continuous triangular mesh models with well-defined geometric and topological relationships. ,in V For vertex set, E For edge set, F A set of face pieces; Texture Mapping Submodule: Through texture mapping technology, the UV mapping correspondence between each triangular facet on the mesh and the two-dimensional texture image is calculated, and the color texture image is accurately wrapped on the surface of the three-dimensional mesh model to restore all visual details of the color, pattern and inscription on the surface of the cultural relic. Final output submodule: After the above steps, a complete, high-fidelity digital cultural relic model is output.

[0011] Furthermore, the triangular mesh reconstruction submodule further includes: Input submodule: Estimates the global point cloud from point cloud data. and the normal vector of each of its points; Processing submodule: The Poisson surface reconstruction algorithm treats the point cloud and its normal vector as the gradient field of an indicator function, and reconstructs the implicit surface by solving the Poisson equation, which is: ,in, Denotes the divergence operator, Represents the gradient operator. It is the implicit surface function to be determined. It is a vector field composed of point cloud normal vectors; Output submodule: Outputs the implicit surfaces obtained from the solution. To extract isosurfaces, the moving cube algorithm can be used to generate a set of vertices. Edge set and triangular dough set The constructed manifold triangular mesh model The mesh model accurately describes the geometric shape of the artifact.

[0012] Furthermore, the virtual display module further includes: receiving standardized 3D cultural relic model data in glTF and FBX formats from the data management and storage module, and initiating physically based rendering (PBR) technology for realistic rendering. The PBR technology accurately calculates reflectivity, roughness, metallicity, and normal perturbation optical properties by simulating the physical interaction between light and the object's surface material. The rendering equation is: ,in Indicates from surface point Along the direction of launch The brightness of the emitted light, It is the bidirectional reflectance distribution function (BRDF), representing the reflection distribution at a point on the surface. How is incident light reflected? Indicates direction Incident point on the surface The brightness; n represents the surface point. Surface normal vector at the location; It is the Lambert cosine law term, representing the cosine value of the angle between the incident light direction and the normal; Indicates all possible incident directions In hemispherical space Integrating the results.

[0013] For the brightness of the emitted light, It is a bidirectional reflection distribution function (BRDF). i Let n be the incident light intensity, and n be the surface normal vector. i and o These are the incident and exit directions, respectively.

[0014] Furthermore, the artificial intelligence analysis module employs two parallel deep neural network architectures to process different types of cultural relic data: for two-dimensional image data, a convolutional neural network (CNN) based on a residual network (ResNet) is used for feature extraction and image classification; for three-dimensional point cloud or mesh data, a hierarchical point cloud processing network (PointNet++) is used for three-dimensional geometric feature learning and classification.

[0015] Furthermore, for two-dimensional image data, the use of a convolutional neural network (CNN) based on a residual network (ResNet) for feature extraction and image classification further includes: Input is a digital image of a cultural relic. First, low-level features are extracted through convolutional layers, and then deep feature transformation is performed through multiple residual blocks. The calculation of each residual block can be represented as: ,in, It is the first Feature map of the layer It is determined by weight The parameterized residual mapping function, which includes multiple convolutions, batch normalization, and activation functions, is finally passed through global average pooling and fully connected layers to output the feature vector of the image. and classification probability distribution ,in This represents the total number of cultural relics categories. It represents the set of all real numbers.

[0016] A third aspect of this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of a virtual cultural relic management method based on artificial intelligence.

[0017] A fourth aspect of this application provides a computer device, comprising: a memory for storing a computer program; and a processor for executing the computer program to implement the steps of an artificial intelligence-based virtual cultural relic management method.

[0018] The innovation of this application lies in the fact that by integrating a high-precision cultural relic digitization module and an advanced artificial intelligence analysis module, it is possible to quickly and accurately convert physical cultural relics into digital format and conduct in-depth analysis, thereby effectively predicting and identifying potential damage to cultural relics and providing timely protection and restoration suggestions. This not only improves the efficiency of cultural relic protection, but also reduces the risk of damage to cultural relics caused by human factors.

[0019] This application can achieve the following technical effects: (1) By integrating a high-precision cultural relic digitization module and an advanced artificial intelligence analysis module, physical cultural relics can be quickly and accurately converted into digital format and analyzed in depth, thereby effectively predicting and identifying potential damage to cultural relics and providing timely protection and repair suggestions. This not only improves the efficiency of cultural relic protection, but also reduces the risk of damage to cultural relics caused by human factors. (2) By combining the virtual display module with VR / AR technology, users are provided with an immersive cultural relic viewing experience, so that users can feel the charm of cultural relics up close no matter where they are. At the same time, the user behavior analysis module collects user interaction data, continuously optimizes the display content and methods, and further enhances user participation and satisfaction. (3) The system security and privacy protection module adopts advanced encryption technology and access control mechanism to ensure the security and privacy of cultural relics data, while the copyright and ethics management module ensures that the use and display of all digital cultural relics comply with relevant laws, regulations and ethical standards, thus protecting the legitimate rights and interests of cultural relics owners and creators; (4) The multilingual support module enables the system to overcome language barriers and provide services to users in different countries and regions, thereby promoting communication and understanding in different cultural backgrounds. The education and research support module provides scholars and students with rich research resources and tools, supports academic research and education activities related to cultural relics, and helps to inherit and promote human cultural heritage. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a flowchart of the operation of the AI-based virtual cultural relics management system proposed in this application. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] In one embodiment, please refer to Figure 1 The artificial intelligence-based virtual cultural relic management system of this application includes a central control and coordination module, a cultural relic digitization module, a virtual display module, a multilingual support module, a cultural relic protection and restoration suggestion module, a system security and privacy protection module, an artificial intelligence analysis module, a data management and storage module, a user behavior analysis module, and a copyright and ethics management module. Central control and coordination module: The central control and coordination module is responsible for coordinating and managing the activities of all other modules. This module is responsible for receiving user input requests and task requests from other modules, and then distributing these tasks to the appropriate modules according to the needs. It also collects and integrates data and results from each module to provide a unified system status and feedback.

[0024] As the core scheduling hub of the system, the central control and coordination module, based on an event-driven microservice architecture and message middleware technology, is responsible for receiving, parsing, and distributing various task requests within the system. Through task queue management and resource scheduling algorithms, it coordinates data flow and state synchronization between functional modules, monitors the system's operational status in real time, collects and integrates feedback data from various modules, forms unified control commands, and outputs standardized task execution status signals and coordination instructions, providing a stable and efficient operating environment and control interface for other modules within the system.

[0025] Cultural Relics Digitization Module: This artifact digitization module is responsible for converting physical artifacts into digital format using high-precision scanning technology. First, the module utilizes a high-precision 3D laser scanner and multispectral imaging equipment to perform non-contact, multi-view data acquisition of the artifacts. High-precision 3D laser scanning and multispectral imaging technology are used to non-contactly acquire the surface geometry and texture information of the artifacts, obtaining massive amounts of point cloud data containing spatial coordinates and color information. Next, an iterative nearest-point algorithm is used to precisely register all local point clouds to a unified global coordinate system, forming a complete and seamless global point cloud model. Through point cloud registration and triangular mesh reconstruction algorithms, the acquired raw data is converted into a high-fidelity 3D digital model, with multi-view point cloud registration achieved through the iterative nearest-point algorithm (ICP). Then, a Poisson surface reconstruction algorithm is used to convert the discrete point clouds into a continuous triangular mesh model with a clear geometric topological relationship, thereby constructing the 3D geometric shape of the artifact. After registration, the point clouds are reconstructed using Poisson surface reconstruction to generate a triangular mesh model with a topological structure. M(V,E,F) ,in V For vertex set, E For edge set, F The module outputs a digital artifact model containing geometric, material, and texture mapping information, serving as a set of facets and providing structured data for the virtual display and AI analysis modules. Finally, the module uses texture mapping technology to precisely wrap the previously collected multi-view texture photos onto the surface of the 3D mesh model, restoring all visual details such as color and pattern on the artifact's surface. The final output is a high-fidelity digital artifact model containing geometric, material, and texture information, providing a structured data foundation for subsequent virtual display and AI analysis.

[0026] The specific data conversion process for the cultural relics digitization module is as follows: Data Acquisition: A high-precision 3D laser scanner is used to scan the physical artifact from multiple perspectives and in all directions. The scanner emits a laser beam, and based on the time or phase difference of the returning laser beam, the 3D coordinates (X, Y, Z) of each sampling point on the artifact's surface are calculated, thus obtaining a massive, discrete set of points, called a point cloud, which can be represented as... ,in, This represents the collected point cloud set. p1, p2, ..., pn represent each point in the set. This represents its spatial location information. An integrated high-resolution multispectral or RGB camera simultaneously captures color and texture information of the artifact's surface, and for each spatial point... Assign the corresponding color values ​​(R, G, B).

[0027] Multi-view point cloud registration: Since a single scan can only acquire point clouds from one angle of the artifact, multiple scans are necessary to obtain a complete model. Because point clouds obtained from different viewpoints reside in their respective independent coordinate systems, this invention employs an iterative nearest-point algorithm to accurately transform all local point clouds to the same global coordinate system. The core of the iterative nearest-point algorithm is to minimize the sum of distances between corresponding points in the source and target point clouds. Its mathematical objective function is:

[0028] in, Indicates the registration error. It is a rotation matrix. It is a translation vector. It is the source cloud The point in the middle, It is a target point cloud Zhongyu The corresponding point, where n is a natural number. The algorithm iteratively finds the nearest pair of corresponding points and calculates the optimal rigid body transformation. To minimize the error function The process continues until convergence, and the final output is a complete and seamless global point cloud model.

[0029] Triangular mesh reconstruction: The point cloud obtained after registration only contains discrete points and lacks the topological structure (i.e., the connection relationships between points) necessary to represent the surface of an object. The Poisson surface reconstruction algorithm is used to convert the discrete point cloud into a continuous triangular mesh model with well-defined geometric and topological relationships. .

[0030] The specific implementation process is as follows: Input: Global point cloud And the normal vector of each of its points (estimated from point cloud data).

[0031] Processing: This algorithm treats the point cloud and its normal vectors as the gradient field of an indicator function. The implicit surface is reconstructed by solving a Poisson equation, which is: ,in, Denotes the divergence operator, Represents the gradient operator. It is the implicit surface function to be determined. It is a vector field composed of point cloud normal vectors.

[0032] Output: The implicit surface obtained from the solution. To extract isosurfaces, the moving cube algorithm can be used to generate a set of vertices. Edge set and triangular dough set The constructed manifold triangular mesh model The grid model accurately describes the geometric shape of the artifact.

[0033] Texture mapping: The generated geometric mesh is a pure gray-white geometry. Using previously acquired multi-view texture photographs as input, texture mapping technology is used to calculate the UV mapping correspondence between each triangular facet on the mesh and the two-dimensional texture image. The color texture image is then precisely wrapped around the surface of the three-dimensional mesh model. This process restores all visual details of the artifact's surface, such as color, patterns, and inscriptions.

[0034] Final Output: After the above steps, a complete, high-fidelity digital model of the cultural relic is output. The model file is available in OBJ, FBX, and gITF formats, and contains the following information: Geometric Information: Triangular Mesh Model This defines the three-dimensional shape of the artifact.

[0035] Material and texture information: including diffuse map, specular map, and normal map, which define the visual appearance of the artifact's surface.

[0036] Metadata: coordinate system, unit scale.

[0037] The structured 3D model generated by this module is directly output to the data management and storage module for persistent storage, and provides input for the "virtual display module" and the "artificial intelligence analysis module". It provides the geometric and texture data required for rendering in the virtual display module, which is a prerequisite for realizing virtual display, and provides the raw 3D data for analysis, identification and classification in the artificial intelligence analysis module.

[0038] Virtual display module: This virtual display module showcases digital cultural relics using virtual reality or augmented reality technologies. Built on a real-time graphics rendering engine, it receives 3D model data of cultural relics from the data management module and uses physically based rendering (PBR) technology to simulate the optical properties of real materials. It combines virtual reality (VR) and augmented reality (AR) devices to achieve multimodal interactive display. At the same time, it optimizes rendering performance through frustum culling and level of detail (LOD) technology, allowing users to immerse themselves in viewing and operating the relics through gestures, voice, and other means. The processed output is a highly realistic real-time rendered image and interactive response events, making it easy for users to obtain a visual experience close to that of viewing the real relics.

[0039] The virtual display module serves as the interface for immersive interaction between the system and the user. Its functionality relies on third-party hardware devices, and it ensures broad compatibility and adaptability through standardized data interfaces and protocols. It receives standardized 3D cultural relic model data in glTF and FBX formats from the data management and storage module, and initiates the physically based rendering pipeline for realistic rendering. PBR technology simulates the physical interaction between light and the surface materials of objects, accurately calculating optical properties such as reflectivity, roughness, metallicity, and normal perturbation. The rendering equation is: ,in, Indicates from surface point Along the direction of launch The brightness of the emitted light, It is the bidirectional reflectance distribution function (BRDF), representing the reflection distribution at a point on the surface. How is incident light reflected? Indicates direction Incident point on the surface The brightness; n represents the surface point. Surface normal vector at the location; It is the Lambert cosine law term, representing the cosine value of the angle between the incident light direction and the normal; Indicates all possible incident directions In hemispherical space Integrating is performed on the surface. Parallel computation is performed in the GPU to generate artifact appearances with high physical fidelity.

[0040] To ensure real-time rendering performance of large-scale scenes and high-precision models on consumer-grade hardware, this module is optimized using view frustum culling technology. This automatically excludes model components outside the viewport, reducing rendering calls. A Level of Detail (LOD) system is employed to dynamically switch model precision based on the distance between objects and the viewpoint. The detail selection strategy in this module is as follows: ,in The distance from the object to the camera. d 0 is the baseline distance parameter. The module also employs visibility technology to determine the truly visible portion of data from the current viewpoint, avoiding wasting computational resources on occluded or out-of-view areas and further reducing computational load.

[0041] This module generates high-frame-rate, low-latency rendered video streams and encapsulates them into common video stream formats, including H.264, H.265, and Volumetric Video data formats. These streams are then output via a TCP / IP interface to third-party display devices such as VR, AR, and interactive large screens. Simultaneously, user interaction functions are implemented through the integration of third-party peripheral operating devices, including gesture recognition sensors, voice input devices, haptic feedback controllers, and motion capture systems. For example, this module receives raw interaction data, such as hand joint coordinates, voice waveforms, and controller button states, and converts them into standardized interaction events, such as "select," "rotate," "zoom," and "voice commands." These events trigger corresponding response logic within the system, enabling actions such as rotating artifact models, playing narration, and displaying trending information, forming a closed-loop interactive experience.

[0042] The entire virtual display process operates by using system-output rendered data and interactive events, with third-party hardware responsible for presentation and input acquisition. Through cross-platform standards such as OpenXR and WebXR, it ensures that its output data and interaction protocols are seamlessly adapted to mainstream third-party hardware devices.

[0043] Artificial intelligence analysis module: The artificial intelligence analysis module employs deep learning algorithms to automatically identify and classify the features of stored cultural relic data, facilitating the retrieval and research of these artifacts. Its aim is to transform digitized cultural relic data into structured knowledge with clear semantic information, thereby providing quantitative and objective decision-making basis for the intelligent management, protection, restoration, and value discovery of cultural relics. Through automated and intelligent analysis methods, it replaces traditional subjective appraisal methods relying on expert experience, achieving efficient and accurate identification and classification of cultural relic materials, craftsmanship, age, category, condition, and even cultural attributes. This provides reliable data support for subsequent restoration suggestions, risk warnings, content display, and user behavior analysis.

[0044] The artificial intelligence analysis module constructs a multi-task deep learning framework, using convolutional neural networks (CNNs) and graph neural networks (GNNs) to process artifact image data and 3D geometric data, respectively. For image classification tasks, a ResNet architecture is used for feature extraction. .in The set of learnable weight parameters in the l-th residual block of a deep convolutional neural network; Indicates weight Parameterized residual mapping function For the feature map of layer l, This is the residual mapping function. For 3D data processing, the PointNet++ network is used to directly process point cloud data, achieving rotation-invariant feature extraction.

[0045] This module employs a deep learning-based multimodal fusion analysis method. It first receives standardized cultural relic data from the data management module, including 2D images and 3D models. Then, it processes the image information using deep convolutional neural networks to extract visual features such as texture, color, and form. Simultaneously, a specialized 3D point cloud processing network is used to analyze geometric structure and spatial morphological features. These features are further fused and abstracted before being input into a classifier for discrimination. Finally, it outputs structured results including cultural relic category labels, feature vectors, and similarity information, completing the transformation from raw data to semantic knowledge. The entire process relies on pre-trained model parameters and achieves end-to-end intelligent recognition through forward computation, enabling automated identification and summarization of multi-dimensional attributes of cultural relics without human intervention.

[0046] The AI ​​analysis module classifies cultural relics based on a highly complex and multi-layered feature discrimination system automatically learned and internally constructed through training with cultural relic data. The module's classification is based on two aspects: first, the multimodal information contained in the input data itself, namely the visual and spatial data inherent in the two-dimensional digital images and three-dimensional geometric models of the cultural relics; second, the deep features abstracted and extracted from this raw data by the deep learning model. The module's classification is based on a multi-layered feature system progressing from shallow to deep. At the most basic level, the module perceives and extracts low-level visual and geometric attributes of the cultural relics, such as edges, corners, color distribution, basic geometric shapes, and surface roughness. Building upon this, the module further combines these basic elements to form more complex mid-level feature patterns, such as specific decorative patterns, inscription fonts, glaze crack patterns, the shape of specific parts of the object, and the microscopic morphology of damage. Finally, the module integrates and abstracts these mid-level and low-level features at a higher dimension, forming high-level semantic features capable of globally defining and distinguishing different categories of cultural relics. These features collectively constitute a multi-dimensional "feature space," within which each type of cultural relic forms a unique cluster. The module determines the category of the relic by judging the specific location of its features within this space and its distance from the cluster centers of typical features for each category. The entire process relies on the knowledge learned by the model from labeled cultural relic data during the training phase. By continuously adjusting its internal parameters, the model effectively separates different types of cultural relics within its internal feature space, thus possessing the ability to automatically identify and classify cultural relics.

[0047] This module employs two parallel deep neural network architectures to process different types of cultural relic data: for two-dimensional image data, a convolutional neural network (CNN) based on residual network (ResNet) is used for feature extraction and image classification; for three-dimensional point cloud or mesh data, a hierarchical point cloud processing network (PointNet++) is used for three-dimensional geometric feature learning and classification.

[0048] In the image data processing branch, the input is a digital image of the cultural relic. First, low-level features are extracted through convolutional layers, and then deep feature transformation is performed through multiple residual blocks. The computation of each residual block can be represented as: ,in, It is the first Feature map of the layer It is determined by weight The parameterized residual mapping function, which incorporates multiple convolutions, batch normalization, and activation functions, effectively mitigates the vanishing gradient problem during deep network training, ensuring the network can learn highly discriminative visual features such as artifact patterns, inscriptions, color distribution, and surface damage. Finally, after global average pooling and fully connected layers, the output image's feature vector is determined. and classification probability distribution ,in This represents the total number of cultural relics categories. The overall meaning is: the input image I is a three-dimensional tensor belonging to the set of real numbers, with height H, width W, and number of channels C. It represents the set of all real numbers.

[0049] In the 3D geometric data processing branch, the input is a set of point clouds. Alternatively, a triangular mesh model can be used. This module employs the PointNet++ network architecture, which primarily expands the receptive field through hierarchical sampling and grouping operations, and utilizes a multilayer perceptron to learn the local and global features of the point cloud. At each level, the farthest point of the input point set is sampled to select the center point, and then a local region is constructed through a ball query. The point features within each local region are aggregated using a shared MLP. ,in, For point In the Features of the layer To query the neighborhood, It is a non-linear activation function. The term "feature concatenation" is used, and "MAX" represents the symmetric max pooling operation. Through multi-layer stacking, the network can progressively extract 3D features with rotational and geometric invariance, effectively characterizing the morphological structure, geometric details, and surface topology of cultural relics. The final output is a feature vector of the 3D shape. And classification prediction.

[0050] Furthermore, this module supports cross-modal feature fusion. When the same cultural relic has both a two-dimensional image and a three-dimensional model, the extracted feature vectors can be fused together. and By splicing or weighted fusion, a more comprehensive and robust multimodal representation of cultural relics can be formed, further improving the accuracy of classification and identification.

[0051] The entire processing is end-to-end trainable, and the loss function typically uses cross-entropy loss for classification tasks. ,in One-hot encoding for the true category, To predict probabilities, metric learning losses such as triplet loss are also introduced to optimize the feature space, causing similar artifacts to cluster together and dissimilar artifacts to move away from each other, thereby outputting more discriminative feature vectors and similarity matrices.

[0052] The AI ​​analysis module outputs the following: artifact feature vectors (for retrieval and comparison), category labels (such as era, kiln, and artifact type), classification confidence scores, and similarity matrices between artifacts.

[0053] V. Suggestions for Cultural Relics Protection and Restoration: This cultural relic protection and restoration suggestion module analyzes the current state of cultural relics and provides protection and restoration recommendations. Based on the output data of the artificial intelligence analysis module, an image segmentation network and a time series prediction model are used to quantitatively assess the damage to cultural relics and predict their degradation. Specifically, the U-Net architecture is used to achieve pixel-level segmentation of defect areas on the surface of cultural relics, while the ARIMA model is used to predict the time series aging process of the material of cultural relics. The analysis yields a health status assessment report of the cultural relics, restoration priority suggestions, and preventive protection plans, providing data support for cultural relic protection decisions.

[0054] The specific implementation process is as follows: The cultural relic protection and restoration suggestion module receives structured output data from the artificial intelligence analysis module, including high-resolution images of the cultural relic, three-dimensional geometric features, material composition data, and historical environmental monitoring time series. It comprehensively utilizes computer vision and time series analysis techniques to achieve quantitative assessment of cultural relic damage and prediction of its future state. First, a U-Net-based semantic segmentation network is used to perform pixel-level recognition of surface defects on the cultural relic. U-Net employs an encoder-decoder structure; the encoder progressively extracts multi-scale features through convolutional layers, and the decoder recovers spatial details through upsampling and skip connections. Let the input image be... The encoder outputs a feature map. The decoder generates a segmentation mask. ,in and These are the encoder and decoder parameters, respectively. The loss function combines cross-entropy loss and Dice coefficient loss: ,in, i For pixels The true label (0 or 1). To predict probabilities, and These are the real and predicted segmented regions, respectively. To balance the hyperparameters, N is a natural number. This loss function can simultaneously optimize pixel classification accuracy and region overlap, improving the ability to detect small defects.

[0055] The artifact status identified by segmentation networks mainly includes the following categories: Surface wear status: The surface wear area is identified through texture feature analysis and edge detection algorithms. The wear degree is quantified by the following formula: in This is the area of ​​wear. The magnitude of the image gradient. This represents the average gradient value for the normal region.

[0056] Structural crack status: Crack networks were identified using directional filter banks and skeleton extraction algorithms, and crack indices were calculated. ,in and The first The length and average width of the crack, Let n be the total area of ​​the cultural relics, and n be a natural number.

[0057] Material degradation state: The degree of degradation is defined by analyzing changes in material composition through multispectral imaging. Where v is the material feature vector, This indicates the current degree of material degradation in the target area of ​​the cultural relic. It is the final quantitative result calculated by the formula. The higher the value, the more severe the degradation and the greater the damage to the material of the cultural relic. The calculation result ranges from [0,1], where 0 represents no degradation and 1 represents complete degradation and failure. It represents the current residual healthy material volume (effective strength value) in the target area, which is a quantitative indicator of the remaining intact and usable material in that area of ​​the cultural relic after deterioration and damage. It represents the original, intact material volume (initial strength value) of the target area, which is the initial quantitative indicator of the area when the cultural relic has not suffered any deterioration or damage.

[0058] Bio-erosion status: Microbial contamination areas are identified using anomaly detection algorithms, with the contaminated area ratio being... .

[0059] Color fading status: Color deviation is calculated through colorimetric analysis. in This is the current Lab color value. For reference only.

[0060] To predict the long-term degradation of cultural relic materials, an ARIMA (Autoregressive Integrated Moving Average) model was used to model the time series data. Let... For the observation sequence, The model is represented as: ,in For lag operators, Let be the difference order. and For model parameters, It is white noise, where p and q are natural numbers.

[0061] The output of the cultural relic protection and restoration recommendation module is the "Cultural Relic Health Status Assessment and Protection Recommendation Report." This report integrates the output from the artificial intelligence analysis module, specifically including high-precision image segmentation results, 3D geometric data, material spectral feature sequences, and time-series prediction data. For each identified cultural relic status category, a specific status severity quantification algorithm is used for calculation. The core of the algorithm is to define a normalized severity index for each status, represented as... , among which, SI k Indicates the first A severity index for various states of damage, such as surface wear and structural cracks, with a value range of [0,1], where 0 represents no damage and 1 represents extremely severe damage. k This represents the core indicator of the state calculated by the aforementioned analytical formula, such as... A. k This represents the proportion of area affected by the state. (T) k The factor representing the time evolution trend of this state comes from the prediction results of the ARIMA model. It is a set of weighted parameters pre-defined through an expert knowledge base for different types of cultural relics (such as ceramics, bronzes, and murals) to ensure the adaptability of the evaluation criteria. k It is a specific function used to integrate multi-dimensional indicators and output normalized results, usually a weighted linear combination or a membership function based on fuzzy logic.

[0062] The "Report on Health Status Assessment and Protection Recommendations for Cultural Relics" utilizes a rule-based expert system. This system incorporates professional knowledge and best practices in the field of cultural relic protection, and its core decision-making logic is shown in the following formula:

[0063] Among them, Action k,m Indicates the first The recommended course of action for this condition Such measures (e.g., monitoring, protection, restoration). k,m,i Decision thresholds are predefined for different severity levels. k,m,i These are the corresponding specific and actionable entries in the expert advice text library.

[0064] Surface wear condition, including: Decision criteria, based on The values ​​and wear distribution characteristics.

[0065] Example of protection recommendations, when Below the threshold When this occurs, the output recommendation is "Strengthen environmental monitoring, control relative humidity below 55%, and reduce physical contact." Exceeding the threshold The output suggestion is to "apply nanoscale reinforcing materials (such as ethyl silicate) for infiltration reinforcement; it is recommended to install physical barriers during exhibition."

[0066] Structural crack conditions, including: Decision criteria, based on The values, crack depth (calculated from three-dimensional geometric data), and activity (determined by time series analysis).

[0067] Example of repair recommendations for microcracks ( Lower), output "Capillary penetration reinforcement using low-viscosity epoxy resin". For macroscopic cracks ( For cracks that are relatively high or in the expansion stage, the output should be "It is recommended to use micro-anchors for structural reinforcement and color-code the reinforced area".

[0068] Material degradation conditions include: Decision criteria, based on The value and type of degradation products (identified by spectral features).

[0069] Example of protection recommendations: For metal artifacts, if chlorides are identified and The output for elevated levels is "Dechlorination treatment is required, along with the application of corrosion inhibitors (such as benzotriazole), and finally, sealing with microcrystalline wax." For inorganic silicate materials, the output is "It is recommended to use silicon-based hydrophobic materials for surface protection to slow down the hydrolysis rate."

[0070] Biological erosion states include: Decision criteria, based on The value (preliminarily determined through image morphological analysis).

[0071] Example of protection recommendations: The output recommendations will clearly define the treatment process, such as "Phase 1: Use mechanical methods to remove visible microbial communities; Phase 2: Use low-toxicity broad-spectrum bactericides (such as thiabendazole) for fumigation or spraying treatment; Phase 3: Improve environmental ventilation and lighting conditions to eliminate breeding grounds."

[0072] Color fading states include: Decision criteria based on color deviation The distribution of .

[0073] Example of protection recommendations: The output recommendations are mainly preventive measures, such as "controlling the ultraviolet intensity of the lighting source below 75μW / lm and the visible light illuminance below 50 lux; it is recommended to use digital imaging technology for color restoration and virtual display instead of physical reproduction."

[0074] The "Report on the Assessment of the Health Status of Cultural Relics and Recommendations for Conservation" integrates all recommendations for different statuses using a multi-objective optimization ranking algorithm to avoid conflicting measures and determine priorities. This algorithm considers the urgency, intrusiveness, and cost of the measures. The final structured cultural relic conservation report will include the following: Comprehensive health score of cultural relics ,in k The weights of each state reflect their impact on overall stability.

[0075] Repair the priority matrix, a 5x5 matrix, to clarify the order of processing each state and their dependencies.

[0076] The protection and restoration measures list is a detailed and actionable list of measures, with each item linked to its corresponding status, severity, and scientific basis.

[0077] Preventative protection plans are based on future maintenance plans and environmental control parameter settings based on time-series forecasts.

[0078] Predictive maintenance schedules provide suggested times for the next inspection or maintenance based on predictions from the ARIMA model.

[0079] This module receives raw data from the upstream AI analysis module and transforms it into an executable, structured report using embedded expert knowledge and decision-making algorithms. This report can be directly provided to cultural heritage preservation experts as decision support, or transmitted to the system security and privacy protection module to ensure the secure storage and controlled access of this critical decision data. It achieves a closed loop from data perception to intelligent decision-making.

[0080] System security and privacy protection module: The system security and privacy protection module employs homomorphic encryption and differential privacy technology to ensure the security of cultural relic data throughout the entire process, protecting the system from the risks of unauthorized access and data leakage. For sensitive data... and y, encryption function and Satisfies the homomorphic property: It supports basic operations in encrypted form. Access control employs an attribute-based encryption scheme (ABE) to ensure that only users meeting specific attributes can decrypt and access data.

[0081] This module receives raw cultural relic data, user identity information, and operation requests from the data management and storage module. By integrating modern cryptography and privacy computing technologies, it achieves end-to-end protection of sensitive data. Its output includes encrypted data ciphertext, access control policy execution results, and system security audit logs, providing a secure and reliable data service foundation for subsequent modules.

[0082] The specific implementation process is as follows: An additive homomorphic encryption algorithm (specifically the Paillier encryption scheme) is used to achieve privacy-preserving computation of the cultural relic data. First, the key is generated by selecting two large prime numbers. and ,calculate and Choose a random integer satisfy Then the public key is The private key is in ,function Regarding sensitive cultural relic data For measurements such as texture details and material composition, the encryption process involves selecting a random number. Calculate ciphertext This encryption scheme satisfies the additive homomorphic property: given any two ciphertexts... Existence operation Make This feature allows the AI ​​analysis module to perform aggregation, statistics, and other calculations on the cultural relic feature data directly in encrypted form, without decryption and exposing the original information.

[0083] For user access behavior data (such as query records and browsing history), the module employs differential privacy technology to provide privacy protection. For statistical query functions... This is achieved by adding Laplace noise. Differential privacy protection: The published results are represented as Among them, noise Follows Laplace distribution , For function global sensitivity, This is a privacy budget parameter. This technology ensures that attackers cannot infer sensitive information about specific individuals from published statistical results.

[0084] Access control employs Cited Policy Attribute-Based Encryption (CP-ABE) technology. A central authority generates the master key MK and public parameters PK. The data owner defines an access policy tree, such as job title + department corresponding to a high security level, and uses this policy to encrypt the artifact data, generating ciphertext CT. The user's private key SK is associated with their set of attributes; the ciphertext can only be decrypted if the user's attributes satisfy the policy. The encryption and decryption process involves bilinear pairing operations: Let... For a bilinear mapping, choose the group generator The encryption process selects random numbers. Calculate the ciphertext components Set a ciphertext component for each attribute. Decryption is performed using bilinear pairing calculations. Restore the session key. This mechanism ensures that only users whose attributes meet the policy requirements can access the cultural relic data at the corresponding security level.

[0085] The system security and privacy protection module receives raw cultural relic data and user query requests from the data management module, outputs encrypted data streams to the network transmission channel or storage system, and simultaneously outputs access decision results and security audit events to the central control module. The encrypted data is directly used for encrypted computation or controlled decryption by subsequent modules such as the artificial intelligence analysis module and the virtual display module. Audit logs are used by the system operation and maintenance module for security posture analysis. This ensures both data utility and privacy security, supporting the system's compliance with the legal and ethical requirements for cultural heritage data management.

[0086] Data management and storage module: This data management and storage module stores and manages digitized cultural relic data, ensuring data security and accessibility. It employs a hierarchical storage architecture built upon a distributed file system and time-series database, using columnar storage to optimize query efficiency for multi-dimensional cultural relic data. Data indexing utilizes B+ trees and inverted indexes for fast retrieval, while also supporting data version management and snapshot functionality. This module outputs a standardized data access interface, providing unified and efficient data services to all modules within the system.

[0087] The data management and storage module is responsible for unified, secure, and efficient full lifecycle management of all types of cultural relic digital assets and system operation data. It aims to build a reliable, scalable, and high-performance data infrastructure to meet the storage, organization, and access needs of massive, multi-source, and heterogeneous cultural relic data, and to provide stable and consistent data services for various upper-level intelligent analysis, virtual display, and management applications. It achieves centralized control, standardized processing, and efficient utilization of data resources, ensuring the smooth, secure, and consistent data flow of the entire system.

[0088] This module primarily stores and manages four main categories of data: First, the data on the artifact itself, namely high-precision 3D models, multispectral textures, point cloud data, and their metadata generated through the artifact digitization module; second, the data on the artifact's environment and condition, including time-series monitoring data such as temperature, humidity, light, and vibration collected by sensors; third, business process data, including user interaction logs, system operation records, and permission allocation information; and fourth, system-derived data, including feature vectors, model parameters, protection and restoration suggestion reports output by the artificial intelligence analysis module, as well as digital watermarks and authorization information required for copyright management.

[0089] In terms of data management methods, a hierarchical storage architecture and a unified metadata management strategy are adopted. For different data types and access characteristics, the system divides data into multiple levels, such as hot data, warm data, and cold data, and optimizes each level using different storage media and strategies. For large-volume unstructured data such as 3D models of cultural relics, a distributed file system is used for storage. This works by dividing the data into blocks and redundantly storing it on multiple physical nodes to achieve high reliability and high throughput access. For time-series data generated by environmental sensors, a time-series database is used for management. This database leverages the high write and low update characteristics of time-series data, employing special data compression and storage structures to support efficient time-range queries and aggregation analysis. For structured data such as cultural relic metadata, user information, and index information, the transaction characteristics and complex query capabilities of relational databases are utilized for management. To achieve efficient data retrieval, this module comprehensively utilizes multiple indexing technologies, including B+ tree indexes suitable for range queries and quick sorting, and inverted indexes suitable for full-text search and multi-keyword matching, achieving millisecond-level response times for massive amounts of data.

[0090] The specific implementation process is as follows: All data flowing into this module must undergo a series of standardized preprocessing steps, including data cleaning, format validation, metadata extraction, and tagging. After automatic classification, the data is allocated to the corresponding storage pools according to preset strategies. The system catalogs and registers all storage assets through a unified metadata catalog service, assigning a globally unique identifier to each data object and recording its attributes, lineage, and access permissions. Version management is implemented by creating immutable snapshots for each change to a data object, ensuring that any data modification history is traceable and providing a foundation for collaborative work and data auditing. Regarding data security, the module integrates an access control mechanism; all data access requests must pass identity authentication and permission verification to ensure that data can only be operated on by authorized modules or users in a prescribed manner.

[0091] This module provides standardized and encapsulated data access interfaces (APIs). These APIs offer transparent and consistent data read and write services to upper-layer application modules such as AI analysis, virtual display, and user behavior analysis. It receives data requests from the upper layers, locates the physical location of the data through the metadata directory, retrieves the data using an optimized query engine, and returns the results to the caller in a unified format, thus shielding the complexity of the underlying storage.

[0092] This module receives raw digital assets from the cultural relics digitization module and streaming data from various sensors and user interactions, and properly stores, organizes and manages them; at the same time, it provides structured data for training and inference for the artificial intelligence analysis module, delivers model and texture data for rendering in real time to the virtual display module, provides historical interaction logs for the user behavior analysis module, and provides evidence storage for the copyright and ethics management module.

[0093] User behavior analysis module: The user behavior analysis module collects and analyzes user interaction data with the system, analyzes user behavior, and feeds this data back to the central control and coordination module and the virtual display module to continuously optimize the user experience and displayed content. This module receives raw user behavior logs generated in real-time by the user interaction interface in the virtual display module, outputs structured user profiles and behavior pattern analysis reports, and directly feeds the analysis results back to the virtual display module to enable dynamic and personalized adjustments to the displayed content, thereby significantly improving the system's user experience and engagement.

[0094] The specific implementation process is as follows: First, through front-end event tracking and log collection technologies, user interaction behavior sequence data in the virtual environment is continuously collected. This raw data includes, but is not limited to, user click locations, browsing duration, navigation paths, zoom operations, content collection and sharing events, forming raw behavior sequences. Each event It includes the event type, timestamp, operation object identifier, and related parameters.

[0095] Subsequently, the module employs a Latent Dirichlet Allocation (LDA) topic model to mine the distribution of topics of interest from a large number of user behavior records. Assuming there are... There are several potential interest topics, each of which is a multi-distribution of artifact content. Each user's behavioral sequence can be viewed as a mixture of topics. For each behavioral event in the user's behavioral sequence... Corresponding to a hidden variable The LDA (Local Descriptor Analysis) is generated as follows, indicating its associated topic: For each topic Generate topic-behavior item distribution For each user Generate user-topic distribution For users Each behavioral event Generate topic tags Generate behavioral events ;in, and These are the prior hyperparameters. After estimating the parameters through variational inference sampling, the topic distribution for each user is obtained. This indicates the current user's preference intensity for different categories of cultural relics, historical periods, cultural attributes, etc.

[0096] Furthermore, to further capture dynamic sequence patterns of user access, this module employs a Long Short-Term Memory (LSTM) network to model temporal behavioral data. LSTM effectively learns long-term dependencies through a gating mechanism; its computation process is as follows: Input Gate: ; Forgotten Gate: Cell state update: c~t=tanh(Wc [ht 1,xt]+bc) ct=ft⊙ct 1+it⊙c~t Output gate:

[0097]

[0098] in, yes Moment-time behavior event embedding representation, It is in a hidden state. It is a cellular state. It is the sigmoid function. This represents element-wise multiplication. and These are the model parameters. The final model output predicts the content or behavior the user might be interested in next.

[0099] The output of this module is a structured user profile model and behavioral pattern analysis report, mainly including: user interest topic distribution vectors. This indicates the degree of user preference for various cultural relic themes; the implicit representation of user behavior sequences. This encodes the user's current interest state; it also predicts future user behavior, such as the artifacts or preference types they might visit next. These analytical results are fed back to the virtual display module, primarily for the following optimizations: Personalized content recommendations: Based on the distribution of user interests and themes, dynamically adjust the presentation priority of exhibits in the virtual exhibition hall, recommend related cultural relics, or plan personalized visit routes.

[0100] Adaptive interface adjustments: Based on user behavior patterns, such as preference for zooming or reading details, adjust the interactive complexity, information display density, or navigation method of the virtual environment.

[0101] Experience smoothness optimization: By predicting the user's next action, relevant cultural relic models or resources are preloaded to reduce interaction latency and improve the smoothness of the experience.

[0102] In this invention, the user behavior analysis module receives raw data such as user behavior logs provided by the data management module and real-time interactive events transmitted by the virtual display module, and transforms them into actionable knowledge through machine learning and deep learning technologies. Its output directly guides the decision-making and rendering process of the downstream virtual display module, realizing a closed loop from "user behavior perception" to "adaptive display content" and improving user experience satisfaction.

[0103] Copyright and Ethics Management Module: This copyright and ethics management module ensures compliance with copyright laws and ethical standards during the digitization and display of cultural relics. It aims to fulfill the key function of ensuring that the entire process of digital reconstruction, dissemination, and utilization of cultural relics conforms to intellectual property laws and regulations and cultural ethics requirements. This module integrates key technologies such as blockchain notarization, contracts, and digital watermarking to construct a complete and automated digital copyright management and ethical risk control system.

[0104] This module first receives the digitized artifact models and related metadata, such as artifact name, era, acquisition time, and copyright holder information, from the front-end data management module. This information is then sent to the blockchain-based copyright registration sub-module. This sub-module uses distributed ledger technology to generate the artifact's digital fingerprint, i.e., its hash value, using the SHA-256 algorithm, representing it as... ,in The core data of the digital cultural relic, along with ownership information, registration timestamps, and other data, are constructed into an immutable copyright certificate transaction and written into the blockchain. This step establishes proof of ownership of the digital cultural relic at the moment of its production, providing a legally valid technical basis for subsequent copyright determination.

[0105] Regarding copyright transfer and usage control, this model employs smart contract technology. A smart contract is a piece of code logic deployed on the blockchain that can be executed automatically. When the system or other modules need to display, copy, or perform secondary creation on a digital artifact, the smart contract automatically verifies the identity and permissions of the user or module initiating the request. Its execution process can be formally represented as follows: given an operation request... Smart contracts query predefined rule bases and return the authorization decision. This process ensures that any operation on digital cultural relics must be authorized in advance; otherwise, the request will be automatically rejected, thus technically preventing unauthorized use and strictly adhering to copyright regulations.

[0106] To address the challenges of copyright tracking once digital content is leaked, this module integrates robust digital watermarking technology. It employs an algorithm combining Discrete Wavelet Transform (DWT) and Singular Value Decomposition (SVD) to embed watermarks into images of cultural relics or model textures. The principle is to leverage the stability of image features using SVD to embed watermark information within singular values, thus resisting common image processing attacks. The implementation process is as follows: First, the original image matrix... Perform multi-level DWT decomposition to obtain the low-frequency subband. And high-frequency subband; for low-frequency subband Perform SVD decomposition: ; encrypt the watermark information With strength coefficient Embedded into singular value matrix In the middle, we obtained Then, perform inverse SVD operation to obtain the low-frequency subband with embedded watermark. Finally, the final image with the embedded watermark is reconstructed through inverse DWT transformation. This watermark information typically contains copyright identifiers and authorized user information, and is invisible to the human visual system. When a suspected infringing copy is discovered, the watermark information can be detected through appropriate extraction algorithms, thereby enabling copyright tracing and verification.

[0107] Regarding ethical standards and compliance, this module incorporates an ethical rules knowledge base. These rules are derived from the *Law of the People's Republic of China on the Protection of Cultural Relics*, relevant policy documents from the State Administration of Cultural Heritage, and other cultural relics-related policy documents issued by various national ministries and commissions, serving as ethical guidelines for cultural relics protection and digitalization. Before any digital cultural relics to be publicly displayed are released, their metadata and display content, especially descriptions that may involve sensitive cultural customs or sacred objects, are automatically compared and screened against this knowledge base. If potential ethical risks are identified, the system will issue a risk alert to the administrator and suspend the automated process, switching to manual review to ensure the accuracy and respect for cultural expression.

[0108] The final outputs of this module include: 1) a digital copyright certificate, serving as legal proof of ownership; 2) a real-time authorization decision flow, controlling all operations and access to digital cultural relics within the system; 3) a final published copy of the digital cultural relics with embedded watermarks, for external sharing and dissemination; and 4) an ethics and compliance screening report, recording the review results and potential risk warnings.

[0109] Within the entire system, this module is a crucial safeguard ensuring the legality of commercial applications and cultural respect. It receives standardized digital cultural relic data from the data management and storage module, and its output authorization instructions directly influence the behavior of the virtual display module and the multilingual support module; only authorized content can be displayed and translated. Simultaneously, its output watermarked digital copy serves as a secure version for the system's interaction with the external environment, effectively protecting the copyright and ethical compliance of digital assets throughout their entire lifecycle.

[0110] Multilingual support module: This multilingual support module provides multilingual interfaces and information displays to accommodate users from different countries and regions. Its aim is to achieve multimodal and multi-language generation and presentation of cultural relic-related content, significantly improving the system's international service capabilities and cultural dissemination efficiency. Based on content type and application scenario differences, the module employs a differentiated translation strategy: for highly specialized texts requiring precise and accurate descriptions of cultural relics, historical background, and archaeological information, pre-stored, manually proofread standard multilingual translations are used for matching and retrieval, ensuring academic rigor and authority; for more real-time content such as system interactive interface elements, operation prompts, and user Q&A, neural machine translation (NMT) technology is used for dynamic translation to support user-selectable languages; simultaneously, a speech synthesis system is integrated to provide users with voice guidance and interactive support.

[0111] In terms of technical implementation, the following key technologies and algorithms are used: For descriptions of cultural relics that require manual proofreading Standardize the text and establish a multilingual standard terminology database and translation database. Assume the original text exists. and its corresponding target language translation set Through semantic feature vectors With a vectorized retrieval mechanism, the optimal translation can be quickly matched from multiple corpora. ,in To calculate the semantic similarity between the original text and the candidate translations, a pre-trained language model is used to generate sentence vectors and calculate the cosine similarity.

[0112] For machine translation of interactive text, a neural machine translation model with an encoder-decoder architecture is adopted. Let the source language sequence be... The translation model generates target language sequences through an encoder-decoder architecture. The model aims to maximize the conditional probability. The encoder encodes the input sequence into a contextual semantic vector, and the decoder predicts the next word autoregressively based on this vector and the generated results.

[0113] A speech synthesis system receives text input and generates natural speech using an acoustic model and a vocoder. Given a text sequence... The process of generating a speech signal can be described as follows: in For acoustic feature sequences, This is the final speech waveform. The system supports multiple speakers, multiple languages, and emotional speech output.

[0114] This module receives cultural relic-related text data from the data management module and language selection signals from the user interaction module. Its output includes three categories: first, professional cultural relic explanation text matched to the user's language, with the translated text sourced from a manually proofread translation database; second, real-time translated interface text and dialogue content; and third, corresponding audio signals. The output results are sent to the user interaction module and the virtual display module, ultimately forming a multilingual user experience combining visual and auditory elements.

[0115] This module outputs text and voice content adapted to the user's language preferences. It works in conjunction with the user interaction module to break down language barriers and achieve barrier-free dissemination of cultural content. By integrating three technologies—human text proofreading, machine translation, and speech synthesis—this module ensures the accuracy of key content while also possessing good scalability and real-time interactive capabilities.

[0116] This application integrates a high-precision cultural relic digitization module and an advanced artificial intelligence analysis module, which can quickly and accurately convert physical cultural relics into digital format and perform in-depth analysis. This effectively predicts and identifies potential damage to cultural relics and provides timely protection and restoration suggestions. This not only improves the efficiency of cultural relic protection but also reduces the risk of damage to cultural relics caused by human factors.

[0117] The virtual cultural relic management system uses traditional digital methods, including photography and scanning technologies, to electronically store images and information about cultural relics. This enables the preservation and dissemination of cultural relic information.

[0118] This application provides a new business model for the protection and management of cultural relics through digital and virtual technologies. The system can reduce the risks and maintenance costs of displaying physical cultural relics, while creating new cultural consumption markets such as virtual exhibitions, educational games, and interactive experiences, thereby driving the digital transformation of the cultural industry and economic growth.

[0119] This application presents an AI-based virtual cultural relic management system that integrates high-precision digitization technology and advanced AI analysis modules, enabling efficient digital management and virtual restoration of cultural relics. The application of this technology enhances the accessibility of cultural relic information and the depth of research, while simultaneously reducing the risk of damage caused by human factors, demonstrating its technological innovation and application potential in the field of cultural relic conservation.

[0120] Furthermore, this application utilizes enhanced sensor networks and more advanced sensor technologies, such as high-sensitivity environmental monitoring sensors, to provide more comprehensive environmental data and enhance the monitoring capabilities of the environment in which cultural relics are located.

[0121] Furthermore, this application stores and analyzes the collected data through a cloud data processing platform, leveraging the powerful data processing capabilities of cloud computing to improve the efficiency and accuracy of data analysis.

[0122] Furthermore, this application integrates artificial intelligence algorithms, and uses an AI-assisted decision-making system to perform deep learning and pattern recognition on monitoring data to achieve more intelligent early warning and decision support.

[0123] In summary, this application discloses an artificial intelligence-based virtual cultural relic management system, including a central control and coordination module, a cultural relic digitization module, a virtual display module, a multilingual support module, a cultural relic protection and restoration suggestion module, a system security and privacy protection module, an artificial intelligence analysis module, a data management and storage module, a user behavior analysis module, and a copyright and ethics management module. By integrating a high-precision cultural relic digitization module and an advanced artificial intelligence analysis module, it can quickly and accurately convert physical cultural relics into digital formats and perform in-depth analysis, thereby effectively predicting and identifying potential damage to cultural relics and providing timely protection and restoration suggestions. This not only improves the efficiency of cultural relic protection but also reduces the risk of damage to cultural relics caused by human factors. This artificial intelligence-based virtual cultural relic management system, through digitization and virtualization technologies, provides a new business model for cultural relic protection and management. The system can reduce the display risks and maintenance costs of physical cultural relics while creating new cultural consumption markets, such as virtual exhibitions, educational games, and interactive experiences, driving the digital transformation of the cultural industry and economic growth.

[0124] This application utilizes technologies such as digital twins and AI analysis to enhance the interactivity and educational value of cultural relic displays. By strengthening the digital management of cultural relic information, it aims to improve the scientific and standardized nature of cultural relic protection.

[0125] Furthermore, this application describes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of a virtual cultural relic management method based on artificial intelligence.

[0126] Furthermore, this application describes a computer device comprising: a memory for storing a computer program; and a processor for executing the computer program to implement the steps of an artificial intelligence-based virtual cultural relic management method.

[0127] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0128] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0129] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0130] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0131] The above is an example of the working process and results of one specific embodiment of this application, for reference only. This application is not limited to this embodiment, and other variations or equivalent substitutions are possible. All such variations or substitutions, without departing from the technical solution and inventive concept of this application, fall within the protection scope of this application.

[0132] It should be noted that the above detailed descriptions are exemplary and intended to provide further explanation of this application. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0133] In the detailed description above, reference has been made to the accompanying drawings, which form part of this document. In the drawings, similar symbols typically identify similar parts unless the context otherwise indicates otherwise. The illustrated embodiments described in the detailed specification, drawings, and claims are not intended to be limiting. Other embodiments may be used and other changes may be made without departing from the spirit or scope of the subject matter presented herein.

Claims

1. A virtual cultural relic management method based on artificial intelligence, characterized in that, The method includes: The digitization process employs high-precision 3D laser scanning and multispectral imaging technology to non-contactly collect the geometric structure and texture information of the surface of physical cultural relics, which is then used to convert the physical cultural relics into digital format through high-precision scanning technology. The virtual display process is built on a real-time graphics rendering engine and uses virtual reality or augmented reality technology to display digital cultural relics. The analysis steps, based on artificial intelligence analysis technology, employ deep learning algorithms to automatically identify and classify the characteristics of cultural relics; The proposed steps for the protection and restoration of cultural relics, based on the data provided in the analysis steps, employ image segmentation networks and time series prediction models to quantitatively assess the damage to cultural relics and predict their degradation, and provide protection and restoration recommendations. Security and privacy protection measures employ homomorphic encryption and differential privacy technologies to protect the system from the risks of unauthorized access and data breaches; The data management and storage steps are based on a hierarchical storage architecture built on a distributed file system and a time-series database. Columnar storage is used to store and manage digitized cultural relic data, ensuring data security and accessibility. The central control and coordination mechanism coordinates and manages the activities of all other modules, receives user input requests and task requests from other modules, distributes tasks to the appropriate modules according to needs, and collects and integrates data and results from each module to provide a unified system status and feedback.

2. A virtual cultural relic management system based on artificial intelligence, characterized in that, The system includes: a central control and coordination module, a cultural relic digitization module, a virtual display module, a cultural relic protection and restoration suggestion module, a system security and privacy protection module, an artificial intelligence analysis module, and a data management and storage module; The central control and coordination module is responsible for coordinating and managing the activities of all other modules, receiving user input requests and task requests from other modules, distributing tasks to the corresponding modules according to needs, and collecting and integrating data and results from each module to provide a unified system status and feedback. The cultural relic digitization module uses high-precision three-dimensional laser scanning and multispectral imaging technology to non-contactly collect the geometric structure and texture information of the surface of physical cultural relics, which is used to convert physical cultural relics into digital format through high-precision scanning technology. The virtual display module is built on a real-time graphics rendering engine and displays digital cultural relics through virtual reality or augmented reality technology. The artificial intelligence analysis module uses deep learning algorithms to automatically identify and classify the characteristics of cultural relics; The cultural relic protection and restoration suggestion module is based on the data output by the artificial intelligence analysis module. It uses image segmentation network and time series prediction model to quantitatively assess and predict the degradation of cultural relics, and provides protection and restoration suggestions. The system security and privacy protection module employs homomorphic encryption and differential privacy technology to protect the system from the risks of unauthorized access and data leakage. The data management and storage module is based on a distributed file system and a time-series database to build a hierarchical storage architecture. It uses columnar storage to store and manage digital cultural relic data, ensuring data security and accessibility.

3. The artificial intelligence-based virtual cultural relic management system as described in claim 2, characterized in that, The system also includes: A multilingual support module is provided to offer multilingual interfaces and information displays to suit users in different countries and regions. The user behavior analysis module collects and analyzes user interaction data with the system, analyzes user behavior, and feeds this data back to the central control and coordination module and the virtual display module in order to continuously optimize the user experience and the displayed content. The security and privacy protection module is used to protect the system from the risks of unauthorized access and data leakage; The copyright and ethics management module is used to ensure that copyright laws and ethical standards are followed during the digitization and display of cultural relics.

4. The virtual cultural relic management system based on artificial intelligence as described in claim 3, characterized in that, The specific data conversion process of the cultural relic digitization module is as follows: Data acquisition submodule: Utilizes a high-precision 3D laser scanner to perform multi-view, all-around scanning of the physical artifact. Based on the time or phase difference of the laser return, it calculates the 3D coordinates (X, Y, Z) of each sampling point on the artifact's surface, thereby obtaining a point cloud set. ,in, p1, p2, ..., pn represent the spatial location information of each point in the set; Multi-view point cloud registration submodule: Employs an iterative nearest-point algorithm to accurately transform all local point clouds to the same global coordinate system. Its mathematical objective function is: in, Indicates the registration error. It is a rotation matrix. It is a translation vector. It is the source cloud The point in the middle, It is a target point cloud Zhongyu For the corresponding point, where n is a natural number, calculate the optimal rigid body transformation. To minimize the error function The final output is a global point cloud model; The triangular mesh reconstruction submodule uses the Poisson surface reconstruction algorithm to convert discrete point clouds into continuous triangular mesh models with well-defined geometric and topological relationships. ,in V For vertex set, E For edge set, F A set of face pieces; Texture Mapping Submodule: Through texture mapping technology, the UV mapping correspondence between each triangular facet on the mesh and the two-dimensional texture image is calculated, and the color texture image is accurately wrapped on the surface of the three-dimensional mesh model to restore all visual details of the color, pattern and inscription on the surface of the cultural relic. Final output submodule: After the above modules, a complete, high-fidelity digital cultural relic model is output.

5. The virtual cultural relic management system based on artificial intelligence as described in claim 4, characterized in that, The triangular mesh reconstruction submodule further includes: Input submodule: Estimates the global point cloud from point cloud data. and the normal vector of each of its points; Processing submodule: The Poisson surface reconstruction algorithm treats the point cloud and its normal vector as the gradient field of an indicator function, and reconstructs the implicit surface by solving the Poisson equation, which is: ,in, Denotes the divergence operator, Represents the gradient operator. It is the implicit surface function to be determined. It is a vector field composed of point cloud normal vectors; Output submodule: Outputs the implicit surfaces obtained from the solution. To extract isosurfaces, the moving cube algorithm can be used to generate a set of vertices. Edge set and triangular dough set The constructed manifold triangular mesh model The mesh model accurately describes the geometric shape of the artifact.

6. The virtual cultural relic management system based on artificial intelligence as described in claim 2, characterized in that, The virtual display module further includes: receiving standardized 3D cultural relic model data in glTF and FBX formats from the data management and storage module, and initiating physically based rendering (PBR) technology for realistic rendering. The PBR technology simulates the physical interaction between light and the surface material of an object, accurately calculating optical properties such as reflectivity, roughness, metallicity, and normal perturbation. The rendering equation is: ,in denoted by , represents the luminance emitted from the surface point along the emission direction; it is the bidirectional reflection distribution function (BRDF), indicating how the incident light is reflected at the surface point; represents the luminance incident on the surface point from the direction; n represents the surface normal vector at the surface point; is the Lambert cosine law term, representing the cosine value of the angle between the incident light direction and the normal; represents the integral over all possible incident directions in hemispherical space.

7. The virtual cultural relic management system based on artificial intelligence as described in claim 2, characterized in that, The artificial intelligence analysis module employs two parallel deep neural network architectures to process different types of cultural relic data: for two-dimensional image data, a convolutional neural network (CNN) based on a residual network (ResNet) is used for feature extraction and image classification; for three-dimensional point cloud or mesh data, a hierarchical point cloud processing network (PointNet++) is used for three-dimensional geometric feature learning and classification.

8. The virtual cultural relic management system based on artificial intelligence as described in claim 7, characterized in that, For two-dimensional image data, the method of using a convolutional neural network (CNN) based on a residual network (ResNet) for feature extraction and image classification further includes: Input is a digital image of a cultural relic. First, low-level features are extracted through convolutional layers, and then deep feature transformation is performed through multiple residual blocks. The calculation of each residual block can be represented as: ,in, It is the first Feature map of the layer It is determined by weight The parameterized residual mapping function, which includes multiple convolutions, batch normalization, and activation functions, is finally passed through global average pooling and fully connected layers to output the feature vector of the image. and classification probability distribution ,in This represents the total number of cultural relics categories. It represents the set of all real numbers.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the artificial intelligence-based virtual cultural relic management method as described in claim 1.

10. A computer device, characterized in that, include: Memory, used to store computer programs; A processor for executing the computer program to implement the steps of the artificial intelligence-based virtual cultural relic management method as described in claim 1.

Images