A non-material cultural heritage digitization method using VR panorama technology

Abstract

The application discloses a kind of non-material cultural heritage digitization method of application VR panorama technology, it is related to non-material cultural heritage protection and digitization technical field, the method includes following component parts: S1, data acquisition and pre-processing, S2, real-time rendering engine building, S3, user operation and time perception module and S4, dynamic scene generation algorithm;The application is realized by real-time rendering engine, user operation and time perception module and dynamic scene generation algorithm advanced technology, the real-time generation and smooth display of non-heritage scene, user can freely explore in VR environment, according to different time and operation trigger different scene changes, obtain highly personalized experience, simultaneously, by introducing the feedback optimization mechanism and scene preloading technology based on reinforcement learning, system can quickly adjust scene according to the behavior and time change of user, reduce loading delay, further improve the smoothness and immersion of user experience.

Description

Technical Field

[0001] This invention relates to the field of intangible cultural heritage protection and digitization technology, specifically to a method for digitizing intangible cultural heritage using VR panoramic technology. Background Technology

[0002] With the rapid development of digital technology, the digital protection and inheritance of intangible cultural heritage has become a hot topic. As a treasure of national culture, intangible cultural heritage contains rich historical value and cultural connotation. However, traditional protection methods are often limited to physical display and oral transmission, which cannot meet the needs of modern people for diversified and interactive cultural experiences.

[0003] Traditional technologies have shortcomings. When processing large amounts of data and high-resolution images, traditional real-time rendering engines often suffer from slow rendering speed, low image quality, and high resource consumption. These problems limit the realism and smoothness of intangible cultural heritage digital scenes and affect the user experience.

[0004] In conclusion, in order to overcome the shortcomings of traditional technologies and enhance users' immersion and experience, it is particularly important to develop a digital method for intangible cultural heritage that applies VR panoramic technology. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and provide a method for digitizing intangible cultural heritage using VR panoramic technology. This method can achieve real-time generation and smooth display of intangible cultural heritage scenes by using advanced real-time rendering engines, user operation and time perception modules, and dynamic scene generation algorithms.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a method for digitizing intangible cultural heritage using VR panoramic technology, the specific steps of which are as follows: S1. Data Collection and Preprocessing: Comprehensively collect images, audio and video materials related to intangible cultural heritage, including materials from different periods of traditional festivals and celebrations, and organize and preprocess these data. S2. Real-time rendering engine construction: Using real-time rendering technology, a rendering engine capable of quickly processing large amounts of data is built to ensure the real-time generation and smooth display of scenes; S3, User Operation and Time Awareness Module: Used to monitor the user's operation behavior in the VR environment in real time, and to obtain the system time, which is used as the trigger condition for dynamic scene generation; S4. Dynamic Scene Generation Algorithm: Based on user operations and time changes, select materials from preprocessed data to generate intangible cultural heritage scenes in real time. When displaying traditional festival celebrations, display sacrificial preparation scenes in the morning, celebration scenes at noon, and bonfire party scenes in the evening. Also, display different details based on the user's proximity to different areas.

[0007] Furthermore, in the data acquisition stage, multi-source heterogeneous data fusion technology is used to collect images, videos, and 3D model data of traditional festival celebrations at different times through high-definition cameras and 3D scanners. Environmental data of intangible cultural heritage sites are also collected with the help of IoT sensors. For the collected text data, keyword extraction and semantic annotation are performed using natural language processing technology. In the preprocessing stage, duplicate, erroneous, and incomplete data are removed through data cleaning algorithms. The cleaned data is then standardized to unify the format and units, providing a high-quality data foundation for subsequent scene construction.

[0008] Furthermore, when building the real-time rendering engine, a technique combining physically based lighting models and deep learning super-resolution reconstruction is adopted. In terms of physically based lighting simulation, an improved ray tracing algorithm is used, considering multiple reflections and refractions, as shown in the formula: ; in It is a point Along the direction The emitted light emissivity, Representing a point in space, it is the specific location where light interacts with objects in the scene. Indicates the direction of the emitted light, from point The direction vector of the light ray's propagation from the origin. It is a two-way reflection distribution function. It is the direction of the incident light. It is a hemisphere. It is a surface normal. A deep learning super-resolution reconstruction model is introduced to enhance the low-resolution rendering results in real time. The model is trained with a large number of high and low resolution image pairs, and its parameters are continuously optimized through the backpropagation algorithm to achieve fast super-resolution reconstruction of the rendered image. While ensuring the rendering quality, it meets the real-time requirements of VR scenes.

[0009] Furthermore, the user operation and time awareness module utilizes multimodal data fusion and adaptive filtering algorithms to acquire user motion posture data through accelerometers, gyroscopes, and magnetometers, while simultaneously combining this with eye-tracking devices to collect user gaze focus data. The multimodal data is then fused using the following formula: ,in These are the filter coefficients from the previous time step. This allows for accurate perception of user actions and time, based on the current data variance.

[0010] Furthermore, the dynamic scene generation algorithm introduces a scene element weight allocation model based on cultural semantic understanding. It constructs an intangible cultural heritage semantic network using a knowledge graph, linking various elements of the intangible cultural heritage project as nodes and edges. Text mining is then used to determine the cultural semantic importance score of each element, using the following formula: ; in The feature weights are determined using the analytic hierarchy process (AHP) combined with expert opinions. It is the first Based on the scores of each feature, during scene generation, scene elements are loaded and rendered according to the semantic score weights of the elements based on user operations and time changes, highlighting the core cultural elements of intangible cultural heritage and enhancing the cultural expressiveness of the scene.

[0011] Furthermore, in the data preprocessing stage, an image inpainting and enhancement algorithm based on a combination of generative adversarial networks and attention mechanisms is adopted for the image data. The generator network uses the attention mechanism to focus on damaged and blurred areas in the image, as shown in the formula: ,in To query the matrix and key matrix, In the first dimension, an attention map generated through an attention mechanism guides the generator to repair and enhance key regions, while the discriminator judges the difference between the generated image and the real image. Both are continuously optimized through adversarial training. For audio data, a denoising and classification algorithm combining wavelet transform and a deep learning classification model is used. Wavelet transform decomposes the audio signal into different frequency components, and the deep learning classification model identifies and removes noise components. The loss function of the classification model is: ,in It's a real label. It predicts probabilities, thereby improving the quality of audio data.

[0012] Furthermore, the real-time rendering engine employs a resource allocation algorithm based on hardware performance awareness and task priority scheduling. By real-time monitoring of the CPU, GPU, and memory hardware performance indicators of the VR device, the performance parameters of the device are obtained. These parameters are determined through device performance testing and user experience feedback. Based on the performance indicators, rendering tasks are prioritized. During resource allocation, a dynamic adjustment strategy is adopted, dynamically adjusting the resource allocation ratio according to the task execution progress and the real-time performance of the device to ensure the high efficiency and stability of the rendering process.

[0013] Furthermore, a reinforcement learning-based feedback optimization mechanism is established between the user operation and time awareness module and the dynamic scene generation algorithm. User operation, time information, and the current scene state are taken as inputs, and scene adjustment strategies are generated through reinforcement learning algorithms. Through continuous learning and optimization, the system can quickly and accurately generate intangible cultural heritage dynamic scenes that meet user expectations based on user operation and time changes.

[0014] Furthermore, the dynamic scene generation algorithm employs a scene preloading technique based on spatiotemporal sequence prediction. By analyzing user historical operation data and time series data, it uses a long short-term memory network to predict the user's possible next operation and the current time point. The formula is as follows: ,in It is the currently hidden state. This is the current input. It is a weight matrix. It is a bias term. It is an activation function that preloads possible scene elements and related data based on the prediction results. When the user actually interacts with the scene, it can quickly display the complete scene, reduce scene loading delay, and improve the smoothness and immersion of the user experience.

[0015] Compared with existing technologies, this method for digitizing intangible cultural heritage using VR panoramic technology has the following advantages: First, this method utilizes advanced technologies such as a real-time rendering engine, user operation and time awareness modules, and dynamic scene generation algorithms to achieve real-time generation and smooth display of intangible cultural heritage scenes. Users can freely explore the VR environment, triggering different scene changes based on different times and operations, thus obtaining a highly personalized experience. At the same time, by introducing a reinforcement learning-based feedback optimization mechanism and scene preloading technology, the system can quickly adjust the scene according to user behavior and time changes, reducing loading delays and further enhancing the smoothness and immersion of the user experience. This immersive experience helps to stimulate users' interest and love for intangible cultural heritage and promotes cultural dissemination and exchange.

[0016] Second, this method uses VR panoramic technology to comprehensively collect and digitize relevant images, audio, and video materials of intangible cultural heritage, enabling these precious cultural heritages to be presented to the public in a more intuitive and vivid way. Users can experience traditional festival celebrations and handicraft production processes in an immersive way, thereby deepening their understanding and awareness of intangible cultural heritage. This not only helps to enhance public interest in and awareness of the protection of cultural heritage, but also provides new ways and means for the transmission of intangible cultural heritage, ensuring that these precious cultural heritages can continue.

[0017] Other advantages, objectives and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from the practice of the invention. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0019] Figure 1 A flowchart illustrating a method for digitizing intangible cultural heritage using VR panoramic technology; Figure 2 Detailed diagram of key steps in a method for digitizing intangible cultural heritage using VR panoramic technology. Detailed Implementation

[0020] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below. Example

[0021] This embodiment describes how to create an immersive VR experience using this digital method during the Spring Festival.

[0022] This study utilizes multi-source heterogeneous data fusion technology, employs high-definition cameras to capture dragon and lion dances and temple fair activities during the Spring Festival, and 3D scanners to acquire 3D model data of traditional buildings such as ancestral halls. IoT sensors collect environmental data on temperature and humidity at the temple fair sites. For the collected Spring Festival couplets and poems, natural language processing techniques are used for keyword extraction and semantic annotation. In the preprocessing stage, data cleaning algorithms remove duplicate, erroneous, and incomplete data. The cleaned data undergoes standardization to unify format and units. For image data, an image inpainting and enhancement algorithm based on a combination of generative adversarial networks and attention mechanisms is employed. The generator network utilizes the attention mechanism to focus on damaged and blurred areas in the image. The formula is as follows: ,in To query the matrix and key matrix, In the first dimension, an attention map generated through an attention mechanism guides the generator to repair and enhance key regions, while the discriminator judges the difference between the generated image and the real image. Both are continuously optimized through adversarial training. For audio data, a denoising and classification algorithm combining wavelet transform and a deep learning classification model is used. Wavelet transform decomposes the audio signal into different frequency components, and the deep learning classification model identifies and removes noise components. The loss function of the classification model is: ,in It's a real label. It predicts probabilities, thereby improving the quality of audio data.

[0023] A real-time rendering engine is built using a combination of physically based lighting models and deep learning-based super-resolution reconstruction. In terms of physically based lighting simulation, an improved ray tracing algorithm is used, taking into account multiple reflections and refractions. The formula is: ; in It is a point Along the direction The emitted light emissivity, Representing a point in space, it is the specific location where light interacts with objects in the scene. Indicates the direction of the emitted light, from point The direction vector of the light ray's propagation from the origin. It is a two-way reflection distribution function. It is the direction of the incident light. It is a hemisphere. It uses surface normals and introduces a deep learning super-resolution reconstruction model to enhance low-resolution rendering results in real time. The model is trained on a large number of high- and low-resolution image pairs, and its parameters are continuously optimized through the backpropagation algorithm to achieve fast super-resolution reconstruction of the rendered image. While ensuring rendering quality, it meets the real-time requirements of VR scenes. At the same time, it adopts a resource allocation algorithm based on hardware performance awareness and task priority scheduling. By monitoring the hardware performance indicators of the VR device's CPU, GPU, and memory in real time, it obtains the device's performance parameters. Based on the performance indicators, the rendering tasks are prioritized. During the resource allocation process, a dynamic adjustment strategy is adopted to dynamically adjust the resource allocation ratio according to the task execution progress and the device's real-time performance.

[0024] By utilizing multimodal data fusion and adaptive filtering algorithms, user motion and posture data are acquired through accelerometers, gyroscopes, and magnetometers, while eye-tracking devices collect user gaze focus data. The multimodal data is then fused, and an adaptive filtering algorithm is used to remove noise interference, dynamically adjusting the filtering parameters based on real-time data changes.

[0025] A scene element weighting model based on cultural semantic understanding is introduced. A semantic network of Spring Festival intangible cultural heritage is constructed using a knowledge graph, linking elements such as dragon and lion dances, Spring Festival couplets, and temple fairs through nodes and edges. Text mining is used to determine the cultural semantic importance score of each element, using the following formula: ; in The feature weights are determined using the analytic hierarchy process (AHP) combined with expert opinions. It is the first The scores of each feature are used to generate scenes. In the morning, the scene shows people posting Spring Festival couplets and preparing sacrificial offerings; at noon, it shows the lively celebrations of a temple fair; and in the evening, it shows dragon and lion dances and fireworks displays. Different details are displayed based on the user's proximity to different areas. Simultaneously, a scene preloading technique based on spatiotemporal sequence prediction is used. By analyzing the user's historical operation data and time series data, a Long Short-Term Memory (LSTM) network is used to predict the user's next possible action and their current time point. The formula is: ,in It is the currently hidden state. This is the current input. It is a weight matrix. It is a bias term. It is an activation function that preloads possible scene elements and related data based on prediction results. When the user actually operates, it can quickly display the complete scene. In addition, a reinforcement learning-based feedback optimization mechanism is established between the user operation, the time awareness module, and the dynamic scene generation algorithm. The user operation, time information, and the current scene state are taken as inputs, and the scene adjustment strategy is generated through the reinforcement learning algorithm. This enables the system to quickly and accurately generate a dynamic Spring Festival intangible cultural heritage scene that meets the user's expectations based on the user operation and time changes.

[0026] Example 2 This embodiment describes a deep and immersive intangible cultural heritage experience for users, using VR panoramic technology against the backdrop of the Dragon Boat Festival. In this experience, from data collection and processing to the construction of various modules and the application of algorithms, everything is closely centered around the unique customs of the Dragon Boat Festival, creating a wonderful Dragon Boat Festival journey for users.

[0027] Using high-definition cameras and 3D scanners, images, videos, and 3D model data of dragon boat races and rice dumpling making activities were collected from all angles. Simultaneously, IoT sensors were used to collect environmental data on humidity and wind speed at the riverside competition venue to enhance the realism of the scene. For textual materials describing Dragon Boat Festival customs, natural language processing techniques were used for keyword extraction and semantic annotation, such as extracting keywords like "dragon boat racing," "rice dumpling," and "Qu Yuan," and labeling their semantic relationships. In the data preprocessing stage, data cleaning algorithms were first used to remove duplicate, erroneous, and incomplete data. Then, the cleaned data was standardized to unify the format and units. For dragon boat race image data, an image inpainting and enhancement algorithm based on generative adversarial networks and attention mechanisms was used to focus on restoring blurred dragon boat details and rider movements caused by lighting and shooting angle issues. For audio data, a denoising and classification algorithm based on wavelet transform and deep learning classification models was used to accurately remove noisy and irrelevant sounds from the riverside. Simultaneously, sounds related to the Dragon Boat Festival, such as dragon boat drumbeats and people's shouts, were classified, identified, and enhanced.

[0028] A rendering engine is built using a combination of physically based lighting models and deep learning super-resolution reconstruction. An improved ray tracing algorithm simulates the reflection and refraction of sunlight on the water surface during the Dragon Boat Festival, making the shimmering water effect more realistic. Multiple reflections and refractions between dragon boats and riverside structures are considered to enhance the scene's realism and three-dimensionality. A deep learning super-resolution reconstruction model is introduced to enhance low-resolution rendering results in real time. The model is trained using a large number of high- and low-resolution images, and parameters are optimized using backpropagation to make the dragon boat colors more vibrant and the textures of the rice dumplings clearer in the VR scene. Furthermore, a resource allocation algorithm based on hardware performance awareness and task priority scheduling is employed. The CPU, GPU, and memory hardware performance indicators of the VR device are monitored in real time. Based on device performance testing and user experience feedback, rendering tasks are prioritized. During resource allocation, the resource allocation ratio is dynamically adjusted according to task execution progress and real-time device performance to ensure efficient and stable rendering.

[0029] By using accelerometers, gyroscopes, magnetometers, and eye-tracking devices, the system comprehensively acquires users' motion posture and gaze focus data. Utilizing multimodal data fusion technology, it organically integrates data from different sensors to provide comprehensive information for subsequent analysis. An adaptive filtering algorithm dynamically adjusts filtering parameters based on real-time data changes, effectively removing noise interference and ensuring the accuracy of the acquired data. In conjunction with the system's timeline, the system displays scenes of people preparing ingredients for making zongzi (sticky rice dumplings) and crafting sachets in the morning. When users turn their heads to observe their surroundings in the VR scene, the system captures this action in real time and adjusts the viewing angle accordingly. At noon, it displays a scene of a fierce dragon boat race; user actions, such as body rotation and walking, cause the viewing angle to change accordingly, creating an immersive experience. In the evening, it displays a heartwarming scene of a family sitting around eating zongzi and chatting.

[0030] A scene element weighting model based on cultural semantic understanding is introduced to construct a Dragon Boat Festival cultural semantic network. Dragon boat racing, making zongzi (sticky rice dumplings), commemorating Qu Yuan, and wearing sachets are linked as nodes and edges. Text mining and expert opinions are used to determine the cultural semantic importance score of each element. For example, dragon boat racing holds an important position in Dragon Boat Festival culture and has a high semantic score weight. During scene generation, scene elements are loaded and rendered according to their semantic score weights based on user actions and time changes. When a user approaches the dragon boat racing area, the system prioritizes loading and finely rendering key elements such as the details of the dragon boats, the paddling actions of the competitors, and the splashing water. Scene preloading technology based on spatiotemporal sequence prediction is used to deeply analyze user historical operation data and time series data, utilizing Long Short-Term Memory (LSTM) networks. The system predicts the user's next actions and their current time point. For example, if it predicts that the user might walk towards the rice dumpling-making area after watching the dragon boat race, the system will preload relevant elements and data of the rice dumpling-making scene, including tools, ingredients, and character models. When the user actually arrives, the system can quickly display the complete scene, reducing waiting time and improving user experience. At the same time, a reinforcement learning-based feedback optimization mechanism is established between the user's actions, the time perception module, and the dynamic scene generation algorithm. The system takes the user's actions, time information, and the current scene state as input and generates scene adjustment strategies through reinforcement learning algorithms. As the user continues to operate and experience, the system continuously learns and optimizes, enabling it to generate dynamic intangible cultural heritage scenes that meet the user's expectations more quickly and accurately, allowing the user to experience the rich cultural connotations of the Dragon Boat Festival in VR.

[0031] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for digitizing intangible cultural heritage using VR panoramic technology, characterized in that, The specific steps of this method are as follows: S1. Data Collection and Preprocessing: Comprehensively collect images, audio and video materials related to intangible cultural heritage, including materials from different periods of traditional festivals and celebrations, and organize and preprocess these data. S2. Real-time rendering engine construction: Using real-time rendering technology, a rendering engine capable of quickly processing large amounts of data is built to ensure the real-time generation and smooth display of scenes; S3, User Operation and Time Awareness Module: Used to monitor the user's operation behavior in the VR environment in real time, and to obtain the system time, which is used as the trigger condition for dynamic scene generation; S4. Dynamic Scene Generation Algorithm: Based on user operations and time changes, select materials from preprocessed data to generate intangible cultural heritage scenes in real time. When displaying traditional festival celebrations, display sacrificial preparation scenes in the morning, celebration scenes at noon, and bonfire party scenes in the evening. Also, display different details based on the user's proximity to different areas.

2. The method for digitizing intangible cultural heritage using VR panoramic technology according to claim 1, characterized in that, In the data acquisition stage, multi-source heterogeneous data fusion technology is used to collect images, videos, and 3D model data of traditional festival celebrations at different times through high-definition cameras and 3D scanners. Environmental data of intangible cultural heritage sites are also collected with the help of IoT sensors. For the collected text data, keyword extraction and semantic annotation are performed using natural language processing technology. In the preprocessing stage, duplicate, erroneous, and incomplete data are removed through data cleaning algorithms. The cleaned data is then standardized to unify the format and units.

3. The method for digitizing intangible cultural heritage using VR panoramic technology according to claim 1, characterized in that, When building the real-time rendering engine, a technique combining physically based lighting models and deep learning super-resolution reconstruction is adopted. In terms of physically based lighting simulation, an improved ray tracing algorithm is used, considering multiple reflections and refractions. The formula is as follows: ； in It is a point Along the direction The emitted light emissivity, Representing a point in space, it is the specific location where light interacts with objects in the scene. Indicates the direction of the emitted light, from point The direction vector of the light ray's propagation from the origin. It is a two-way reflection distribution function. It is the direction of the incident light. It is a hemisphere. It is a surface normal. A deep learning super-resolution reconstruction model is introduced to enhance the low-resolution rendering results in real time. The model is trained with a large number of high and low resolution image pairs, and its parameters are continuously optimized through the backpropagation algorithm to achieve fast super-resolution reconstruction of the rendered image. While ensuring the rendering quality, it meets the real-time requirements of VR scenes.

4. The method for digitizing intangible cultural heritage using VR panoramic technology according to claim 1, characterized in that, The user operation and time awareness module utilizes multimodal data fusion and adaptive filtering algorithms to acquire user action and posture data through accelerometers, gyroscopes, and magnetometers, while simultaneously combining eye-tracking devices to collect user gaze focus data. The multimodal data is fused, and an adaptive filtering algorithm is used to remove noise interference, dynamically adjusting the filtering parameters according to real-time changes in the data.

5. A method for digitizing intangible cultural heritage using VR panoramic technology according to claim 1, characterized in that, The dynamic scene generation algorithm introduces a scene element weight allocation model based on cultural semantic understanding. It constructs an intangible cultural heritage semantic network using a knowledge graph, linking various elements of the intangible cultural heritage project as nodes and edges. Text mining is then used to determine the cultural semantic importance score of each element, using the following formula: ; in The feature weights are determined using the analytic hierarchy process (AHP) combined with expert opinions. It is the first Based on the scores of each feature, scene elements are loaded and rendered according to their semantic score weights during scene generation, taking into account user actions and time changes.

6. The method for digitizing intangible cultural heritage using VR panoramic technology according to claim 1, characterized in that, In the data preprocessing stage, an image inpainting and enhancement algorithm based on a combination of generative adversarial networks and attention mechanisms is adopted for the image data. The generator network uses the attention mechanism to focus on damaged and blurred areas in the image, as shown in the formula: ,in To query the matrix and key matrix, In the first dimension, an attention map generated through an attention mechanism guides the generator to repair and enhance key regions, while the discriminator judges the difference between the generated image and the real image. Both are continuously optimized through adversarial training. For audio data, a denoising and classification algorithm combining wavelet transform and a deep learning classification model is used. Wavelet transform decomposes the audio signal into different frequency components, and the deep learning classification model identifies and removes noise components. The loss function of the classification model is: ,in It's a real label. It is a predicted probability.

7. The method for digitizing intangible cultural heritage using VR panoramic technology according to claim 1, characterized in that, The real-time rendering engine employs a resource allocation algorithm based on hardware performance awareness and task priority scheduling. By monitoring the hardware performance indicators of the VR device's CPU, GPU, and memory in real time, the system obtains the device's performance parameters. These parameters are determined through device performance testing and user experience feedback. Based on these performance indicators, rendering tasks are prioritized. During resource allocation, a dynamic adjustment strategy is adopted, dynamically adjusting the resource allocation ratio according to the task execution progress and the device's real-time performance.

8. The method for digitizing intangible cultural heritage using VR panoramic technology according to claim 1, characterized in that, The system establishes a reinforcement learning-based feedback optimization mechanism between the user operation and time awareness module and the dynamic scene generation algorithm. It takes user operation, time information, and the current scene state as input, and generates scene adjustment strategies through reinforcement learning algorithms. Through continuous learning and optimization, the system can quickly and accurately generate intangible cultural heritage dynamic scenes that meet user expectations based on user operation and time changes.

9. The method for digitizing intangible cultural heritage using VR panoramic technology according to claim 1, wherein the dynamic scene generation algorithm employs scene preloading technology based on spatiotemporal sequence prediction, and through the analysis of user historical operation data and time series data, uses a long short-term memory network to predict the user's possible next operation and the current time point, as shown in the formula: ,in It is the currently hidden state. This is the current input. It is a weight matrix. It is a bias term. It is an activation function that preloads possible scene elements and related data based on the prediction results, so that the complete scene can be quickly displayed when the user actually interacts with it.

Images