Enterprise culture immersive experience method and system based on virtual reality

Abstract

The invention relates to the technical field of virtual reality, in particular to an enterprise culture immersive experience method and system based on virtual reality, and the method comprises the following steps: calculating to obtain rendering batch merging potential scores for independent enterprise culture exhibit model pairs in a virtual reality scene, iteratively selecting the model pair with the highest rendering batch merging potential score, and selecting the model pair with the highest rendering batch merging potential score; and the vertexes and the index data are connected in series, a unified texture image set is generated, UV coordinates of all the vertexes are re-calculated to be matched with new positions in the texture image set, and an enterprise culture scene integration model is established. According to the method, multi-dimensional calculation is carried out on the geometrical relationship and the rendering relevance of the enterprise culture exhibit models in the virtual reality scene, so that the merging priority among the models can be more efficiently determined, redundant drawing is reduced, and the utilization rate of rendering resources is improved; geometric features of vertexes and patches are extracted in a three-dimensional space, and a culture element matching relation is combined.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of virtual reality, in particular to a virtual reality-based enterprise culture immersive experience method and system. BACKGROUND

[0002] The virtual reality technology field is a cross-research direction of computer graphics, human-computer interaction and multimedia integration, mainly through constructing a highly realistic three-dimensional virtual environment and combining visual, auditory, tactile and other multi-sensory feedback to realize immersive human-computer interaction experience. The prior art has multiple limitations in processing the virtual scene of enterprise culture. The prior art has a lag in responding to the user's gaze point and the handle pointing vector, resulting in insufficient real-time and accuracy of dynamic attention semantic labels. At the same time, the prior art adopts a uniform standard in mesh subdivision and simplification strategy, which easily causes the problems of insufficient details in high attention areas and rendering waste in low attention areas. Therefore, improvement is needed. SUMMARY

[0003] The purpose of the present application is to solve the shortcomings in the prior art and to provide a virtual reality-based enterprise culture immersive experience method and system.

[0004] In order to achieve the above-mentioned purpose, the present application adopts the following technical scheme: a virtual reality-based enterprise culture immersive experience method, comprising the following steps: For a pair of independent enterprise culture exhibit models in the virtual reality scene, a rendering batch merging potential score is calculated, the model pair with the highest rendering batch merging potential score is iteratively selected, vertex and index data are concatenated, a unified texture atlas is generated, all vertex UV coordinates are recalculated to match the new position in the texture atlas, and an enterprise culture scene integration model is established; All vertices and faces of the enterprise culture scene integration model are traversed, geometric parts associated with specific cultural elements of the honor wall and the innovation idea display area are identified, preset cultural semantic labels are assigned to the vertices and faces, a vertex semantic label set is obtained, each cultural semantic label in the vertex semantic label set is associated with a static numerical weight, and an enterprise culture semantic weight graph is constructed; According to the enterprise culture semantic weight graph, the gaze point coordinates and the handle pointing vector of the user are tracked in real time in the virtual reality interaction, the enterprise culture exhibit geometric area pointed by the input is judged, the cultural semantic label associated with the hit vertex is queried from the enterprise culture semantic weight graph, the real-time attention semantic label is obtained, the static numerical weight corresponding to the real-time attention semantic label is calculated according to the continuous duration of the user's gaze or pointing, and a user interaction dynamic attention weight is generated; Based on the user interaction dynamic attention weight and the corporate culture scenario integration model, the user interaction dynamic attention weight value of each region on the corporate culture scenario integration model grid is retrieved and compared with the preset subdivision threshold and simplification threshold. New vertices and edges are inserted in the grid regions where the weight value exceeds the subdivision threshold to generate a dynamic subdivision vertex coordinate set. At the same time, edge folding and vertex removal are performed in the regions where the weight value is lower than the simplification threshold to output an adaptive corporate culture model topology structure.

[0005] Preferably, the steps for obtaining the corporate culture scenario integration model are as follows: The independent corporate culture exhibit model pairs are analyzed, the shader identifier of each model is extracted and converted into a numerical vector, the texture hash is mapped to a fixed interval, the material parameter vector is normalized and calculated, and the original Euclidean space distance is calculated based on the 3D coordinate difference and the material similarity function value and Euclidean space distance of each pair of models are recorded. Calculate the rendering batch merging potential score based on the material similarity function value and the Euclidean space distance; Based on the rendering batch merging potential score, a priority queue is used to retain only the top three model pairs in terms of rendering batch merging potential score. The vertex data and index data of the model pairs are concatenated and stitched together to form a unified texture atlas. The vertex UV coordinates of the selected model pairs are recalculated and mapped to the newly generated texture atlas position to form an integrated model of corporate culture scene.

[0006] Preferably, the step of obtaining the vertex semantic tag set is as follows: Traverse all vertices and faces of the corporate culture scene integration model, compare the geometric position and shape features of each vertex and face one by one, identify the geometric parts associated with the honor wall and the geometric parts associated with the innovation concept display area, and establish matching relationships between vertices and faces and corresponding specific cultural elements to obtain a set of vertices and faces. Assign a preset cultural semantic label to each vertex and face in the vertex and face set, bind the vertex and face identifiers to the cultural semantic labels, and organize them in sequence to form a vertex semantic label set, and ensure that each cultural semantic label in the vertex semantic label set has a unique correspondence with the corresponding vertex and face.

[0007] Preferably, the steps for obtaining the corporate culture semantic weight graph are as follows: Each cultural semantic tag in the vertex semantic tag set is associated with a static numerical weight. Vertices, patches, cultural semantic tags, and static numerical weights are organized into a unified structured data table. All content is stored in a structured storage system to form a corporate culture semantic weight map.

[0008] Preferably, the step of obtaining the real-time attention semantic tags is as follows: In virtual reality interaction, the coordinates of the user's gaze point and the pointer vector of the controller are captured in real time. The coordinates of the gaze point and the pointer vector are used as input. The spatial acceleration structure is used to traverse all vertices and faces of the corporate culture scene integration model frame by frame. The intersection position of the ray and the geometric region is determined to obtain the sequence of hit vertices and the sequence of hit faces corresponding to the user input. Based on the hit vertex sequence and hit face sequence, the corresponding cultural semantic tags are queried in the corporate culture semantic weight graph. The number of hits of each cultural semantic tag in different frames is counted and the cumulative duration is combined with the timestamp. Cultural semantic tags with a cumulative duration of zero are removed to obtain the real-time attention semantic tags and their cumulative duration.

[0009] Preferably, the steps for obtaining the dynamic attention weight of user interaction are as follows: Based on the real-time attention semantic tags, static numerical weights are read from the corporate culture semantic weight map to calculate the dynamic attention weights of user interactions.

[0010] Preferably, the step of obtaining the dynamically subdivided vertex coordinate set is as follows: Based on the user interaction dynamic attention weight and the corporate culture scenario integration model, the user interaction dynamic attention weight of each region in the corporate culture scenario integration model grid is retrieved region by region. The user interaction dynamic attention weight is compared with the preset subdivision threshold and simplification threshold one by one to obtain the set of grid regions that exceed the subdivision threshold and the set of grid regions that are lower than the simplification threshold. Based on the set of grid regions that exceed the subdivision threshold, new vertices are inserted into the regions and new edges are generated. The three-dimensional coordinates of the inserted vertices are recorded one by one and organized into a dynamic subdivision vertex coordinate set, while maintaining the topological relationship with the original vertices and edges, thus obtaining the dynamic subdivision vertex coordinate set.

[0011] Preferably, the steps for obtaining the adaptive corporate culture model topology are as follows: Based on the set of grid regions below the simplification threshold, edge folding is performed within the regions and redundant vertices are removed. Combining the dynamically subdivided vertex coordinate set with the simplified region information, the structure that conforms to the topological constraints is sorted out and the adaptive corporate culture model topology is generated.

[0012] This invention also provides an immersive corporate culture experience system, including: The scene integration and rendering optimization module is used to calculate the rendering batch merging potential score for independent corporate culture exhibit model pairs in virtual reality scenes, iteratively select the model pair with the highest rendering batch merging potential score, concatenate vertex and index data, generate a unified texture atlas, recalculate the UV coordinates of all vertices to match the new positions in the texture atlas, and establish a corporate culture scene integration model. The cultural semantic annotation and weight graph module is used to traverse all vertices and faces of the corporate culture scene integration model, identify the geometric parts associated with specific cultural elements of the honor wall and innovation concept display area, assign preset cultural semantic labels to vertices and faces, obtain the vertex semantic label set, and associate each cultural semantic label in the vertex semantic label set with a static numerical weight to construct a corporate culture semantic weight graph. The interaction tracking and dynamic attention calculation module is used to track the coordinates of the user's gaze point and the pointer vector of the controller in real time in virtual reality interaction based on the corporate culture semantic weight map, determine the geometric area of ​​the corporate culture exhibit that the input is pointing to, query the cultural semantic tags associated with the hit vertex from the corporate culture semantic weight map, obtain the real-time attention semantic tags, and calculate the static numerical weights corresponding to the real-time attention semantic tags according to the duration of the user's gaze or pointing, so as to generate the user interaction dynamic attention weight. The adaptive topology subdivision and simplification module is used to retrieve the dynamic attention weight values ​​of user interactions in each region of the corporate culture scene integration model grid based on the dynamic attention weights of user interactions and the corporate culture scene integration model. The values ​​are compared with preset subdivision thresholds and simplification thresholds. New vertices and edges are inserted in the grid regions where the weight values ​​exceed the subdivision thresholds to generate a dynamic subdivision vertex coordinate set. At the same time, edge folding and vertex removal are performed in the regions where the weight values ​​are lower than the simplification thresholds to output the adaptive corporate culture model topology structure.

[0013] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, by performing multi-dimensional calculations on the geometric relationships and rendering correlations of corporate culture exhibit models in a virtual reality scene, the merging priority between models can be determined more efficiently, reducing redundant rendering and improving the utilization rate of rendering resources. By extracting the geometric features of vertices and faces in three-dimensional space and combining them with cultural element matching relationships, regions corresponding to specific corporate culture elements can be identified, avoiding incorrect assignment of semantic labels and ensuring a more accurate correspondence between semantics and geometry. By introducing joint analysis of user gaze point coordinates and controller pointing vectors during real-time interaction, user interaction behavior is highly correlated with the virtual scene content, enabling real-time capture of user attention on different cultural elements. This approach enhances the responsiveness and interactivity of dynamic data by dynamically updating the static numerical weights corresponding to real-time attention semantic tags and combining them with the duration of continuous user interaction. This makes the attention calculation results more consistent with the user's actual interest distribution, achieving a more personalized and immersive experience. In the dynamic adjustment of the virtual scene mesh, by performing fine-grained retrieval and partition comparison of the dynamic attention weight values ​​of user interaction in different areas, it can control the generation of new vertices and edges in areas with high attention, improving the accuracy of local detail rendering. At the same time, it automatically performs edge folding and vertex removal in areas with low attention, reducing the rendering complexity of irrelevant areas, thereby achieving dynamic adaptive model topology optimization. Attached Figure Description

[0014] Figure 1 This is a schematic diagram of the steps of the present invention. Detailed Implementation

[0015] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0016] Please see Figure 1 This invention provides a technical solution for an immersive corporate culture experience based on virtual reality, comprising the following steps: For independent corporate culture exhibit model pairs in virtual reality scenes, the rendering batch merging potential score is calculated, the model pair with the highest rendering batch merging potential score is selected iteratively, the vertex and index data are concatenated to generate a unified texture atlas, and all vertex UV coordinates are recalculated to match the new positions in the texture atlas to establish an integrated corporate culture scene model. Traverse all vertices and faces of the corporate culture scene integration model, identify the geometric parts associated with specific cultural elements such as the honor wall and the innovation concept display area, assign preset cultural semantic labels to vertices and faces, obtain the vertex semantic label set, and associate each cultural semantic label in the vertex semantic label set with a static numerical weight to construct a corporate culture semantic weight map. Based on the corporate culture semantic weight map, the coordinates of the user's gaze point and the pointer vector of the controller are tracked in real time during virtual reality interaction. The geometric area of ​​the corporate culture exhibit that the input is pointing to is determined. The cultural semantic tags associated with the hit vertex are queried from the corporate culture semantic weight map to obtain the real-time attention semantic tags. Based on the duration of the user's gaze or pointing, the static numerical weights corresponding to the real-time attention semantic tags are calculated to generate the dynamic attention weights of user interaction. Based on the user interaction dynamic attention weight and corporate culture scenario integration model, the user interaction dynamic attention weight value of each region on the corporate culture scenario integration model grid is retrieved and compared with the preset subdivision threshold and simplification threshold. New vertices and edges are inserted in the grid regions where the weight value exceeds the subdivision threshold to generate a dynamic subdivision vertex coordinate set. At the same time, edge folding and vertex removal are performed in the regions where the weight value is lower than the simplification threshold to output an adaptive corporate culture model topology structure.

[0017] The steps to obtain the corporate culture scenario integration model are as follows: The independent corporate culture exhibit model pairs are analyzed, the shader identifier of each model is extracted and converted into a numerical vector, the texture hash is mapped to a fixed interval, the material parameter vector is normalized and calculated, and the original Euclidean space distance is calculated based on the 3D coordinate difference and the material similarity function value and Euclidean space distance of each pair of models are recorded. The rendering batch merging potential score is calculated based on the material similarity function value and the Euclidean distance. The calculation formula is as follows: ; in, To render batch merged potential scores, The material similarity function value is normalized to... interval, This represents the original Euclidean space distance between the two models. is the base of the natural logarithm. The effective merging radius varies with the material similarity function value, and is calculated as follows: , Based on the merging radius, Material influencing factors; Based on the rendering batch merging potential score, a priority queue is used to retain only the top three model pairs in terms of rendering batch merging potential score. The vertex data and index data of the model pairs are concatenated and stitched together to form a unified texture atlas. The vertex UV coordinates of the selected model pairs are recalculated and mapped to the newly generated texture atlas position to form an integrated model of corporate culture scene.

[0018] Specifically, the process involves parsing independent corporate culture exhibit model pairs. First, all corporate culture exhibits in the scene are traversed, and each pair is combined to form a model pair. For each model file, such as FBX or OBJ files, data structure parsing is performed to extract geometric, material, and texture information. For each model's shader identifier, such as the string name "PBRMetallicRoughness" or "BlinnPhong", a global shader type dictionary is created. Each unique shader name is mapped to an integer index, which is then converted into a fixed-length one-hot encoded numerical vector. For example, if there are 5 different shaders in the scene, "PBRMetallicRoughness" can be encoded as [1,0, 0, 0, 0], while "BlinnPhong" is [0, 1, 0, 0, 0], used for subsequent similarity calculations. For the textures used by the models, their image data is extracted and a hash value is calculated using the SHA-256 algorithm, resulting in a 64-bit hexadecimal string. To map this hash value to a fixed numerical range, such as [0, ... [1] First, convert the hexadecimal string into an unsigned long integer value, then divide it by 2 to the power of 64 minus 1, which is the maximum possible value in hexadecimal, thus obtaining a normalized floating-point number representing the unique identifier of the texture. Next, extract the detailed material parameter vector for each model. This vector contains parameters such as albedo color, metallic, roughness, and emissive color. Normalize these parameters using the min-max normalization method. The calculation formula is as follows: ,in These are the original parameter values. and This refers to the theoretical or practical range of the parameter type. For example, the range of RGB color components is [0, 255], and the range of metallicity and roughness is [0, 1]. This method uniformly scales all material parameters to the [0, 1] interval, facilitating subsequent unified measurement. Simultaneously, to calculate the spatial relationships between models, the 3D bounding box of each model is determined, and the 3D coordinates of its geometric center point are calculated. Based on the coordinate difference between the center points of two models, the 3D coordinates of the two models are calculated. and The distance between them, i.e. The original Euclidean space distance is obtained. Finally, the material similarity function value and Euclidean space distance of each pair of models are calculated and recorded by combining the numerical vectors processed above.

[0019] formula: ,in, ; The above formula constructs a comprehensive evaluation index for merging potential by combining material similarity and spatial proximity. The design idea of ​​this formula is that the optimization of rendering batch merging depends not only on the similarity of model materials, but also on their physical location in the scene, and uses a Gaussian function form of decay term. To simulate the impact of spatial distance on the merge value, when the distance between two models is... As the value increases, this factor decays exponentially, leading to a decrease in the final score. This reduction aligns with the reality that excessive spatial span in the rendering pipeline leads to increased merging overhead, while also introducing a dynamically changing effective merging radius. It is not a fixed value, but rather depends on the material similarity function value. Positive correlation; the more similar the materials of the two models ( The higher the value, the more spatial distance between them can be tolerated, because the performance gains from merging are greater enough to offset the negative impact of distance. This dynamic adjustment mechanism allows the scoring model to adapt to the complexity and object distribution of different scenarios, resulting in a final rendering batch merging potential score. Material similarity The product with the spatial decay term ensures that only model pairs that perform well in both material and spatial dimensions can achieve high scores.

[0020] This is the material similarity function value. This parameter integrates the similarity of shader type, texture content, and other material attributes. Its value is normalized to the range [0,1], where 1 indicates that the materials are completely identical and 0 indicates that they are completely different. This value is obtained based on the previous processing step, which involves one-hot encoding of the shader identifier for each model, hash mapping of the texture map, and normalization of material parameters (such as color, roughness, and metallicity) to form multiple feature vectors. Subsequently, the final similarity is obtained by calculating the weighted similarity between the corresponding feature vectors of the two models. The value is calculated as follows: , among which The term represents the cosine similarity or distance similarity of the corresponding features, with weights... The setting is based on performance analysis of rendering overhead caused by changes in different material properties. In a standard test scene containing 1000 objects, the shader, texture, and other material parameters were changed independently, and the resulting increase in average frame time was monitored. The results showed that switching shaders increased the average frame time by 0.2 milliseconds, switching textures by 0.5 milliseconds, and switching other parameters by 0.1 milliseconds. Based on these overheads, the weights were set to be proportional to the overhead and normalized to obtain... , , In this example, the material similarity function value of a certain model pair is calculated through the aforementioned steps. It is 0.85.

[0021] The original Euclidean distance between two models represents the physical distance between the two corporate culture exhibit models in virtual 3D space, measured in meters. This distance is calculated based on the coordinates of the center point of the bounding box of each model's geometry. The coordinates of each vertex are obtained by parsing the model data, and the average coordinate value of all vertices is calculated, which is the geometric center of the model. This parameter directly reflects the spatial proximity of the models and is a fundamental physical constraint for determining the feasibility of merging. Models that are closer together are more likely to have more local vertex data in memory, resulting in higher cache utilization after merging. For example, in a virtual corporate culture exhibition hall, the center coordinates of one trophy model A are (10.5, 1.2, 3.4), and the center coordinates of another trophy model B are (11.0, 1.8, 4.5). The original Euclidean distance between them is... pass The calculation yielded the following result: In this example, the original Euclidean space distance of a model pair is defined as meters. It is 1.2 meters.

[0022] is the base of the natural logarithm.

[0023] As an effective merging radius that varies with the material similarity function value, it defines a dynamic spatial range within which the distance between models is considered "acceptable." This radius is not a fixed value, but rather a measure of the material similarity. The function, when material similarity At higher levels, An increase in the value of indicates that the system is more willing to merge objects that are spatially farther apart; conversely, a decrease in indicates a lower material similarity. The reduction requirement necessitates that objects be physically very close before merging is considered. This design makes merging decisions more flexible and intelligent, dynamically adjusting spatial constraints based on the performance gains from material matching. This effectively balances the benefits of merging with potential overhead. The calculation of this parameter depends on the base merging radius. Material Influence Factor .

[0024] The base merging radius is the effective merging radius. The baseline value when material similarity is zero represents the maximum spatial distance scale that the system is willing to merge models without considering any material similarity gain. This parameter needs to be set with reference to the average density and size of objects in the virtual scene. A reasonable setting method is based on the nearest neighbor distance statistics of all objects to be optimized in the scene. The specific steps are as follows: First, calculate the Euclidean distance between each exhibit model and its nearest neighbor model in the scene, forming a distance list. Then, calculate the average and standard deviation of all distances in this list. The basic merging radius can be set to this average distance plus a multiple of the standard deviation, such as 1.5 times, to cover most compact object clusters. For example, in a scene containing 200 exhibits, if the calculated average nearest neighbor distance of all models is 2.0 meters and the standard deviation is 0.8 meters, then... Can be set to Meters, in this example, the basic merging radius is set. It is 3.0 meters.

[0025] The material influence factor is a dimensionless adjustment coefficient used to control material similarity. For effective merging radius The magnitude of the improvement, and the value of this factor, determines the sensitivity of the merging decision to material similarity. The higher the value, the more significant the impact of material similarity, making it possible for highly similar models to be merged even if they are far apart. Conversely, a lower value indicates a lower similarity. The smaller the value, the greater the weight of spatial distance. The optimal value of this parameter is obtained through experimental calibration. Specifically, a series of test scenarios with different combinations of material similarity and spatial distance are constructed. For each scenario, under different conditions... Run the merging algorithm under the specified values ​​and measure rendering performance (e.g., average frame rate) to draw. The relationship curve between the value and performance improvement; select the value that brings the greatest average performance improvement. The value is used as the final set value. For example, by testing in increments of 0.1 between 0.1 and 2.0, it was found that when... When the value was set to 0.6, the average frame rate of the test scene set increased by a peak of 15%, and further increases thereafter... This will cause a slight decrease in performance, therefore it is determined The optimal value is 0.6. In this example, the material influence factor is set as follows: It is 0.5.

[0026] Calculations based on parameters: First, substitute the material similarity function value. Basic merging radius and material influencing factors Calculate the effective merging radius : ; ; Next, the calculated effective merging radius will be... and material similarity function value Distance from the original Euclidean space Substitute the potential score into the rendering batch. In the calculation formula: ; ; ; The results show that the rendering batch merging potential score of this model pair is 0.81716, which is a relatively high score (ranging from 0 to 1), meaning that this model pair has strong merging value. This score will be used as the basis for ranking in subsequent steps, compared with the potential scores of all other model pairs. If this score can rank among the top (e.g., in the top three), then this model pair will be selected, and its geometric and texture data will be integrated together to form part of the corporate culture scene integration model.

[0027] Based on the render batch merging potential score, calculate all model pairs and their corresponding render batch merging potential scores. As an element, a min-heap data structure of fixed size 3 is constructed as a priority queue. The scores of all model pairs are traversed. For each score, it is compared with the top element of the heap (i.e., the lowest score in the current queue). If the current score is higher than the top score, the top element is removed, and the current model pair and its score are inserted into the heap, maintaining the heap property. After traversing all model pairs, the priority queue retains the top three model pairs with the highest potential scores for rendering batch merging. Subsequently, these three model pairs are processed sequentially for merging. For the first selected model pair (e.g., model A and model B), a new vertex buffer and index buffer are created. All vertex data of model A are directly copied to the new vertex buffer, and its index data is also copied to the new index buffer. The number of vertices of model A is recorded, denoted as . Next, all vertex data of model B is appended to the end of the new vertex buffer. Then, each index value of model B is traversed and its offset is added. Then, it is appended to the end of the new index buffer to complete the concatenation of geometric data. After that, all texture maps used by the two models are collected, and the MaxRects algorithm is used to stitch these small textures into a unified texture atlas. This algorithm maintains a list of free rectangles to efficiently find the best placement for each texture in the atlas space to minimize the total area of ​​the final atlas. During the stitching process, the offset coordinates of each original texture in the new atlas are recorded. Finally, the UV coordinates of all vertices must be recalculated to match their new positions in the texture atlas. This involves iterating through all vertices of the merged model, querying the position information of each vertex in the new atlas based on its original model (A or B) and the original texture it referenced, and then using the formula... and Perform coordinate transformation, where These are the original UV coordinates. and It is the width and height of the original texture. and The width and height of the new atlas will be used to calculate the new UV coordinates. The attributes of the corresponding vertices are updated, and the merged geometric data is combined with the newly generated unified texture atlas to form an integrated model of corporate culture scene.

[0028] The steps for obtaining the vertex semantic tag set are as follows: Traverse all vertices and faces of the corporate culture scene integration model, compare the geometric position and shape features of each vertex and face one by one, identify the geometric parts associated with the honor wall and the geometric parts associated with the innovation concept display area, and establish matching relationships between vertices and faces and corresponding specific cultural elements to obtain a set of vertices and faces. Assign a preset cultural semantic label to each vertex and face in the vertex and face set, bind the vertex and face identifiers to the cultural semantic labels, organize them in sequence to form a vertex semantic label set, and ensure that each cultural semantic label in the vertex semantic label set has a unique correspondence with the corresponding vertex and face.

[0029] Specifically, the process iterates through all vertices and faces of the integrated corporate culture scenario model. First, it predefines geometric constraints associated with specific cultural areas. For example, it defines a 3D axis-aligned bounding box for the honor wall area, with its minimum corner coordinates being (10.0, 0.5, 1.0) and its maximum corner coordinates being (15.0, 3.0, 1.2). Simultaneously, it defines its primary surface normal direction as the positive Z-axis direction (0, 0, 1). For the innovation concept display area, it defines a spherical influence range with a radius of 5 meters, centered at (-5.0, 1.5, 4.0). Then, it initiates an iterative process. For each vertex in the integrated model, its 3D world coordinates are read, and it is determined whether these coordinates are within the bounding box of the honor wall or the spherical range of the innovation concept display area. For vertices that pass the positional screening, the geometric characteristics of their respective faces are further analyzed. Specifically, the average normal vector of the face is calculated, and this vector is compared with the preset surface normal direction of the honor wall (0, 0, 1). 1) Perform a dot product operation. If the dot product result is greater than a preset normal alignment threshold of 0.95, then all vertices and the face itself that make up the face are initially classified as being associated with the Wall of Honor. For the Innovation Concept Display Area, check whether the vertices that make up the face reference a specific material identifier, such as a material named "Concept_Hologram". If so, classify it. Finally, establish a one-to-one matching relationship between all vertices and faces that have passed the dual verification of the above geometric position and shape features and their object names in the original model, such as "Annual Outstanding Employee Trophy_2023" or "Core Technology Patent A Demonstration Model". Summarize all successfully matched geometries to obtain a set of vertices and faces.

[0030] For each vertex and face in the set of vertices and faces, a "Cultural Element-Semantic Label Mapping Table" is first constructed. This table is compiled by the corporate culture building department based on the company's values ​​and historical events. It details the abstract cultural semantics that each specific cultural exhibit (specific cultural element) should be assigned. For example, "Annual Outstanding Employee Trophy_2023" is mapped to "Employee Honors," "Core Technology Patent A Demonstration Model" is mapped to "Technological Innovation," and "Founder's Early Office Restoration Model" is mapped to "Striving Spirit." Next, each element in the vertex and face set obtained in the previous step is traversed, and each vertex or face is identified by the specific cultural element associated with it, such as "Annual Outstanding Employee Trophy." The lookup table for "Cup_2023" is used to retrieve its corresponding preset cultural semantic tag "Employee Honor". Then, a parallel data array of the same size as the vertex buffer of the integrated model is created to store the semantic tags. The retrieved cultural semantic tag "Employee Honor" is written into the array at the position corresponding to the current vertex index. This operation is repeated for all vertices and faces in the set to accurately assign its cultural semantic tag to each geometric element. This process strongly binds the unique identifier of the vertex or face (such as its index in the buffer) with the retrieved cultural semantic tag. Finally, this parallel data array containing all vertex identifiers and their corresponding cultural semantic tags is organized into a vertex semantic tag set.

[0031] The steps to obtain the semantic weight graph of corporate culture are as follows: Each cultural semantic tag in the vertex semantic tag set is associated with a static numerical weight. Vertices, patches, cultural semantic tags, and static numerical weights are organized into a unified structured data table. All content is stored in a structured storage system to form a corporate culture semantic weight map.

[0032] Specifically, each cultural semantic tag in the vertex semantic tag set is weighted using a multi-dimensional scoring system. This system comprises three dimensions: relevance to core values ​​(weight coefficient 0.5), influence on employee motivation (weight coefficient 0.3), and importance of historical events (weight coefficient 0.2). These weights are derived from statistical analysis of the annual corporate culture awareness survey conducted on all employees. The relevance to core values ​​is considered the most important and is therefore assigned the highest weight. For each cultural semantic tag, such as "employee honor," a review panel consisting of five senior executives and five senior employee representatives anonymously scores each of the three dimensions on a scale of 1 to 10. After removing the highest and lowest scores for each dimension, the average score is calculated. For example, the average scores for the "employee honor" tag are: relevance to core values ​​8.5, influence on employee motivation 9.2, and importance of historical events 7.5. The final static weight is calculated as follows: The calculation result is used as the static numerical weight of the label. This process is repeated for all cultural semantic labels. The index of each vertex in the model, the index of its corresponding face, the cultural semantic label assigned to the vertex, and the calculated corresponding static numerical weight are integrated into a unified structured data table containing four columns (vertex ID, face ID, cultural semantic label, and static numerical weight). This table is stored in a key-value pair database or a similar efficient query structure in memory to form a corporate cultural semantic weight map.

[0033] The steps for obtaining semantic tags in real time are as follows: In virtual reality interaction, the coordinates of the user's gaze point and the pointer vector of the controller are captured in real time. The coordinates of the gaze point and the pointer vector are used as input. The spatial acceleration structure is used to traverse all vertices and faces of the corporate culture scene integration model frame by frame. The intersection position of the ray and the geometric region is determined to obtain the sequence of hit vertices and the sequence of hit faces corresponding to the user input. Based on the sequence of hit vertices and the sequence of hit faces, the corresponding cultural semantic tags are queried in the corporate culture semantic weight graph. The number of hits of each cultural semantic tag in different frames is counted and the cumulative duration is combined with the timestamp. Cultural semantic tags with a cumulative duration of zero are removed to obtain the real-time attention semantic tags and their cumulative duration.

[0034] Specifically, in real-time capture of the user's gaze coordinates and controller pointing vector during virtual reality interaction, firstly, to improve ray detection efficiency, a bounding volume hierarchy (BVH) is pre-constructed as a spatial acceleration structure for the integrated model of the corporate culture scene. This structure recursively wraps complex scene geometry within a series of simple bounding boxes, forming a tree-like data structure. In each frame's rendering loop, the 3D spatial position of the VR headset is obtained as the starting point of the gaze ray, and its forward vector is obtained as the direction of the ray. Simultaneously, the 3D spatial position and pointing of the VR controller are obtained as the starting point and direction of the controller pointing ray. Then, these two rays are intersected with the bounding box of the root node of the BVH tree. If the ray does not intersect with the bounding box of the current node, then the ray is considered not to have intersected. All child nodes under a node and their contained geometry are pruned and do not require subsequent detection. If they intersect, the same intersection test is recursively performed on the child nodes of that node until a leaf node is reached. When a ray intersects the bounding box of a leaf node, the Möller-Trumbore ray-triangle intersection algorithm is used to accurately determine the intersection of all triangles contained in that leaf node. All successfully intersecting triangles and their intersection point information are recorded. Finally, the distances of all intersection points to the ray origin are compared, and the triangle containing the closest intersection point and its three vertices are selected. Their unique identifiers (e.g., indices in the vertex buffer) are recorded to obtain the sequence of hit vertices and the sequence of hit faces corresponding to the user input.

[0035] Based on the sequence of hit vertices and hit faces, for each hit vertex or face, its unique identifier is used as the query key to quickly search within the previously constructed corporate culture semantic weight graph. This graph is a hash table with vertex or face IDs as keys and structures containing cultural semantic tags and static weights as values. Through this query operation, the cultural semantic tags associated with the user interaction points in the current frame are obtained, such as "technological innovation" or "employee honor." Next, a global attention statistics table is maintained. This table also uses cultural semantic tags as keys, recording the total number of hits and cumulative duration for each tag. For each cultural semantic tag hit in the current frame, its corresponding number of hits is... The numerical segment is incremented by one, and the cumulative duration is updated using the difference between the timestamp of the current frame and the timestamp of the previous frame (i.e., the frame interval time, for example, about 11.1 milliseconds at a 90Hz refresh rate). This frame interval time is added to the duration field of the corresponding tag. This process is repeated in every frame to dynamically track changes in user attention to different cultural elements. Before proceeding to the next calculation, the attention statistics table is filtered once. All cultural semantic tag entries in the table are traversed, and their cumulative duration fields are checked. All entries with a duration of zero, i.e., tags that have never been paid attention to by users in the current statistical period, are removed from the subsequent processing list to obtain the real-time attention semantic tags and cumulative duration.

[0036] The steps to obtain the dynamic attention weight of user interaction are as follows: Based on real-time attention semantic tags, static numerical weights are read from the corporate culture semantic weight map to calculate the dynamic attention weights of user interactions. The calculation formula is as follows: ; in, Dynamically weighted user interactions For static numerical weights, To monitor the cumulative duration of semantic tags in real time, Based on the attention time constant, It is a very small positive number used to prevent the denominator from being zero.

[0037] Specifically, the formula: The above formula is used to construct a dynamic user attention weight model, which can reflect the user's brief, real-time interactive behavior (represented as attention duration). The inherent, static importance of the exhibits themselves (reflected in static weights) By combining these factors, a dynamic weight value can be generated that reflects the user's current level of interest in real time. It utilizes the hyperbolic tangent function The function exhibits an S-shaped curve characteristic, starting from 0 and rapidly increasing before eventually saturating and approaching 1. This simulates the cognitive process of attention shifting from initial attraction to gradual saturation, while the denominator... This makes the time required for attention to saturate related to the static importance of the exhibits. Proportional, that is, the more important the exhibit ( The higher the value, the longer it takes for the user's attention to reach saturation, ultimately leading to... Results and static weights Multiplication ensures that the final upper limit of the dynamic weight will not exceed its inherent importance level, thus achieving a system that can quickly respond to user behavior.

[0038] This is a static numerical weight. This parameter represents the inherent importance of a corporate culture exhibit associated with a certain cultural semantic tag within the corporate culture system. Its value is pre-calculated and stored in the corporate culture semantic weight map during system initialization using the aforementioned multi-dimensional scoring system. This value does not change with real-time user interaction; it is a preset, stable benchmark value that reflects the company's value judgment and emphasis on different cultural elements. For example, an exhibit symbolizing a company milestone event might be assigned a high static weight, while a conventional decorative element would have a relatively low weight. In this example, using the aforementioned multi-dimensional scoring system, the cultural semantic tag "employee honor" is comprehensively scored, and the calculated static numerical weight is 8.51. Therefore, it is substituted here... .

[0039] To monitor the cumulative duration of semantic tags in real time, measured in seconds (s), this parameter dynamically records the total duration from when a user begins to focus on a specific cultural semantic tag until the current moment. It is obtained frame-by-frame by combining timestamps from the previous step and directly reflects the user's attention span on a particular exhibit. It is the most direct input variable for measuring the user's level of interest. When the user's gaze or the controller points to a new exhibit, the corresponding new tag... The value accumulates starting from 0. If the user's attention remains focused, The value increases linearly over time; if the user loses attention, the tag's value will decrease. The value stops increasing. This parameter gives the entire weight calculation model timeliness and dynamic responsiveness, enabling it to capture the real-time flow of user attention. In this example, after the aforementioned steps, the user has been continuously paying attention to the exhibits associated with the "Employee Honors" tag for 3.2 seconds. Therefore, we substitute this value here. .

[0040] The basic attention time constant, measured in seconds (s), is a calibration parameter used to adjust the base time scale required for user attention to reach saturation. It is defined in static weights. When the value is 1, the attention duration The impact of the dynamic weight growth rate was determined based on research data from user experience testing and human factors engineering. Specifically, 50 participants were recruited to freely explore a virtual reality environment, and eye-tracking technology was used to record their gaze at different exhibits (each with known static weights). Analyze the data to determine the average time users spend on a moderately important (e.g., ...) item. For exhibits where stable perception is achieved (e.g., a 95% accuracy rate in post-exhibition questionnaire responses), statistics show that the average time for this perception to develop is 6 seconds. Based on this, [the following is a continuation of the previous sentence]. The value is set to 1.2 to ensure that... hour, The input value of the function is close to 1, causing the dynamic weights to approach saturation. In this example, this empirical value is used to set... .

[0041] For a very small positive number, set .

[0042] Calculations based on parameters: First, substitute the static numerical weights. Basic attention time constant and extremely small positive numbers The denominator in the calculation formula: ; Next, we substitute the cumulative duration of real-time attention semantic tags. Calculate the hyperbolic tangent function Input parameters: ; Then, calculate the value of the hyperbolic tangent function: ; Finally, the above results are compared with static numerical weights. Multiply them to obtain the final dynamic user interaction attention weight. : ; The results show that for the "Employee Honors" exhibit with a static importance of 8.51, after a user has been paying attention for 3.2 seconds, its current dynamic user interaction attention weight is 2.5833. This value dynamically reflects the user's level of interest. It is much greater than zero, indicating that the user has generated clear attention. However, since its value (2.5833) is far from reaching the upper limit of the static weight (8.51), it indicates that the user's attention is still in the initial stage and has not yet reached saturation. In the subsequent adaptive topology adjustment steps, this weight value will be compared with the preset subdivision threshold and simplification threshold to determine whether the grid density of the area where the exhibit model is located needs to be adjusted.

[0043] The steps for obtaining the dynamically subdivided vertex coordinate set are as follows: Based on the user interaction dynamic attention weight and corporate culture scenario integration model, the user interaction dynamic attention weight of each region in the corporate culture scenario integration model grid is retrieved region by region. The user interaction dynamic attention weight is compared with the preset subdivision threshold and simplification threshold one by one to obtain the set of grid regions that exceed the subdivision threshold and the set of grid regions that are below the simplification threshold. Based on the set of mesh regions that exceed the subdivision threshold, new vertices are inserted into the regions and new edges are generated. The 3D coordinates of the inserted vertices are recorded one by one and organized into a dynamic subdivision vertex coordinate set, while maintaining the topological relationship with the original vertices and edges, thus obtaining the dynamic subdivision vertex coordinate set.

[0044] Specifically, based on the user interaction dynamic attention weight and corporate culture scenario integration model, the mesh of the integrated model is first divided into multiple local regions. This division can be based on semantic labels, where all vertices and faces with the same cultural semantic label constitute an independent region. Then, for each region, its average user interaction dynamic attention weight is calculated. Specifically, all vertices within the region are traversed, and the user interaction dynamic attention weight value corresponding to each vertex is read from the corporate culture semantic weight map. The average of these values ​​is then calculated as the overall attention weight for that region. Next, preset subdivision and simplification thresholds are set. The subdivision threshold is set based on the principle that when user attention is sufficiently high, more visual details are needed. User experience testing revealed that when the dynamic weight value exceeds 60% of its corresponding static weight value, users expect to see a more precise model. Therefore, the subdivision threshold is set to 0.6 times the static weight value. For example, for the "Employee Honors" region with a static weight of 8.51, its subdivision threshold is... The simplified threshold setting is intended to save performance in areas where users are not interested, and is set to 10% of the static weight value, i.e. If the average dynamic weight of a region is lower than this value, it is considered that the user has basically ignored the region and it can be simplified. Then, the average user interaction dynamic attention weight of each region is compared with these two thresholds one by one. If the region weight, for example, the previously calculated 2.5833, is less than 5.106 but greater than 0.851, the region retains its current topology. If it is greater than 5.106, it is added to the set of grid regions to be subdivided. If it is less than 0.851, it is added to the set of grid regions to be simplified. After comparing all regions, the set of grid regions that exceed the subdivision threshold and the set of grid regions that are lower than the simplification threshold are obtained.

[0045] Based on the set of mesh regions exceeding the subdivision threshold, a cyclic subdivision algorithm is executed for each region in the set. This algorithm is an iterative surface subdivision method. For each triangular facet within the region, three new vertices are first inserted at the midpoints of its three edges. The 3D coordinates of these new vertices are calculated by linear interpolation of the coordinates of the two endpoints of their respective edges. Simultaneously, other attributes of the new vertices, such as normals and UV coordinates, are also calculated using the same interpolation method. Then, the original triangular facet is connected to the original three vertices using these three new vertices, splitting it into four smaller sub-triangles. This process is executed iteratively, with the number of iterations depending on the current region's dynamic user interaction focus. The degree to which the subdivision threshold is exceeded is increased by one iteration for every 20% exceeding the threshold, thus achieving progressive detail enhancement. After each insertion of a new vertex and generation of a new edge, the 3D coordinates of the new vertex are immediately recorded and appended to a global dynamic vertex coordinate list. This list is separate from the original vertex buffer and is specifically used to store dynamically generated geometric data. At the same time, the index buffer is updated to add new triangle connections. However, during this process, the original vertex and edge indices and topological relationships remain unchanged, and new elements are only inserted between them. After processing all the regions to be subdivided, the dynamic vertex coordinate list is organized to obtain the dynamic subdivided vertex coordinate set.

[0046] The steps to obtain the topology of the adaptive corporate culture model are as follows: Based on the set of grid regions below the simplification threshold, edge lines are folded and redundant vertices are removed within the regions. Combining the dynamically subdivided vertex coordinate set with the simplified region information, the structure that conforms to the topological constraints is sorted out and the adaptive corporate culture model topology is generated.

[0047] Specifically, based on the set of mesh regions below the simplification threshold, a quadratic error folding algorithm is applied to each region in the set for mesh simplification. This algorithm determines the simplification order and method by calculating a folding cost for each edge. Edges with lower costs have a smaller impact on the model's appearance after folding, and are therefore processed first. Specifically, all edges within a region are traversed, the geometric error introduced by shrinking the edge into a new vertex is calculated, and all edges and their folding costs are stored in a priority queue. Then, the edge with the lowest cost is retrieved from the queue, and a folding operation is performed, i.e., the edge and its two endpoints are removed, and a new vertex is generated at a calculated optimal position (the position that minimizes the quadratic error). The indices of all faces associated with the removed vertex are updated to point to this new vertex. This process continues. The process continues until the number of faces in the region is reduced to a preset target percentage. This percentage is also related to the dynamic attention weight of user interaction. The lower the weight, the higher the target percentage for simplification. For example, the number of faces in a region with a weight close to zero can be reduced by 90%. After completing the edge folding and vertex removal operations for all regions to be simplified, the simplified region geometry information is integrated with the dynamically subdivided vertex coordinate set generated by the subdivision operation in the previous step, as well as the region information that remains unchanged. A complete vertex buffer and index buffer are then reconstructed. This new data structure contains both the rich details of the high-attention regions and the simplified representation of the low-attention regions. Finally, the structure that conforms to all topological constraints (e.g., no non-manifold edges, no overlapping faces) is output, generating the adaptive corporate culture model topology structure.

[0048] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications 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 protection scope of the present invention.

Claims

1. A method for immersive corporate culture experience based on virtual reality, characterized in that: Includes the following steps: For independent corporate culture exhibit model pairs in virtual reality scenes, the rendering batch merging potential score is calculated, the model pair with the highest rendering batch merging potential score is selected iteratively, the vertex and index data are concatenated to generate a unified texture atlas, and all vertex UV coordinates are recalculated to match the new positions in the texture atlas to establish an integrated corporate culture scene model. Traverse all vertices and faces of the corporate culture scene integration model, identify the geometric parts associated with specific cultural elements such as the honor wall and the innovation concept display area, assign preset cultural semantic labels to vertices and faces, obtain the vertex semantic label set, and associate each cultural semantic label in the vertex semantic label set with a static numerical weight to construct a corporate culture semantic weight map. Based on the corporate culture semantic weight map, the coordinates of the user's gaze point and the pointer vector of the controller are tracked in real time during virtual reality interaction. The geometric area of ​​the corporate culture exhibit that the input is pointing to is determined. The cultural semantic tags associated with the hit vertex are queried from the corporate culture semantic weight map to obtain the real-time attention semantic tags. Based on the duration of the user's gaze or pointing, the static numerical weights corresponding to the real-time attention semantic tags are calculated to generate the dynamic attention weights of user interaction. Based on the user interaction dynamic attention weight and the corporate culture scenario integration model, the user interaction dynamic attention weight value of each region on the corporate culture scenario integration model grid is retrieved and compared with the preset subdivision threshold and simplification threshold. New vertices and edges are inserted in the grid regions where the weight value exceeds the subdivision threshold to generate a dynamic subdivision vertex coordinate set. At the same time, edge folding and vertex removal are performed in the regions where the weight value is lower than the simplification threshold to output an adaptive corporate culture model topology structure.

2. The method for immersive corporate culture experience based on virtual reality according to claim 1, characterized in that, The steps for obtaining the corporate culture scenario integration model are as follows: The independent corporate culture exhibit model pairs are analyzed, the shader identifier of each model is extracted and converted into a numerical vector, the texture hash is mapped to a fixed interval, the material parameter vector is normalized and calculated, and the original Euclidean space distance is calculated based on the 3D coordinate difference and the material similarity function value and Euclidean space distance of each pair of models are recorded. Calculate the rendering batch merging potential score based on the material similarity function value and the Euclidean space distance; Based on the rendering batch merging potential score, a priority queue is used to retain only the top three model pairs in terms of rendering batch merging potential score. The vertex data and index data of the model pairs are concatenated and stitched together to form a unified texture atlas. The vertex UV coordinates of the selected model pairs are recalculated and mapped to the newly generated texture atlas position to form an integrated model of corporate culture scene.

3. The method for immersive corporate culture experience based on virtual reality according to claim 1, characterized in that, The steps for obtaining the vertex semantic tag set are as follows: Traverse all vertices and faces of the corporate culture scene integration model, compare the geometric position and shape features of each vertex and face one by one, identify the geometric parts associated with the honor wall and the geometric parts associated with the innovation concept display area, and establish matching relationships between vertices and faces and corresponding specific cultural elements to obtain a set of vertices and faces. Assign a preset cultural semantic label to each vertex and face in the vertex and face set, bind the vertex and face identifiers to the cultural semantic labels, and organize them in sequence to form a vertex semantic label set, and ensure that each cultural semantic label in the vertex semantic label set has a unique correspondence with the corresponding vertex and face.

4. The method for immersive corporate culture experience based on virtual reality according to claim 1, characterized in that, The steps for obtaining the corporate culture semantic weight graph are as follows: Each cultural semantic tag in the vertex semantic tag set is associated with a static numerical weight. Vertices, patches, cultural semantic tags, and static numerical weights are organized into a unified structured data table. All content is stored in a structured storage system to form a corporate culture semantic weight map.

5. The method for immersive corporate culture experience based on virtual reality according to claim 1, characterized in that, The steps for obtaining the real-time attention semantic tags are as follows: In virtual reality interaction, the coordinates of the user's gaze point and the pointer vector of the controller are captured in real time. The coordinates of the gaze point and the pointer vector are used as input. The spatial acceleration structure is used to traverse all vertices and faces of the corporate culture scene integration model frame by frame. The intersection position of the ray and the geometric region is determined to obtain the sequence of hit vertices and the sequence of hit faces corresponding to the user input. Based on the hit vertex sequence and hit face sequence, the corresponding cultural semantic tags are queried in the corporate culture semantic weight graph. The number of hits of each cultural semantic tag in different frames is counted and the cumulative duration is combined with the timestamp. Cultural semantic tags with a cumulative duration of zero are removed to obtain the real-time attention semantic tags and their cumulative duration.

6. The method for immersive corporate culture experience based on virtual reality according to claim 1, characterized in that, The steps for obtaining the dynamic attention weight of user interaction are as follows: Based on the real-time attention semantic tags, static numerical weights are read from the corporate culture semantic weight map to calculate the dynamic attention weights of user interactions.

7. The method for immersive corporate culture experience based on virtual reality according to claim 1, characterized in that, The steps for obtaining the dynamically subdivided vertex coordinate set are as follows: Based on the user interaction dynamic attention weight and the corporate culture scenario integration model, the user interaction dynamic attention weight of each region in the corporate culture scenario integration model grid is retrieved region by region. The user interaction dynamic attention weight is compared with the preset subdivision threshold and simplification threshold one by one to obtain the set of grid regions that exceed the subdivision threshold and the set of grid regions that are lower than the simplification threshold. Based on the set of grid regions that exceed the subdivision threshold, new vertices are inserted into the regions and new edges are generated. The three-dimensional coordinates of the inserted vertices are recorded one by one and organized into a dynamic subdivision vertex coordinate set, while maintaining the topological relationship with the original vertices and edges, thus obtaining the dynamic subdivision vertex coordinate set.

8. The method for immersive corporate culture experience based on virtual reality according to claim 1, characterized in that, The steps for obtaining the topology of the adaptive corporate culture model are as follows: Based on the set of grid regions below the simplification threshold, edge folding is performed within the regions and redundant vertices are removed. Combining the dynamically subdivided vertex coordinate set with the simplified region information, the structure that conforms to the topological constraints is sorted out and the adaptive corporate culture model topology is generated.

9. The corporate culture immersive experience system based on the virtual reality-based corporate culture immersive experience method according to any one of claims 1-8, characterized in that, include: The scene integration and rendering optimization module is used to calculate the rendering batch merging potential score for independent corporate culture exhibit model pairs in virtual reality scenes, iteratively select the model pair with the highest rendering batch merging potential score, concatenate vertex and index data, generate a unified texture atlas, recalculate the UV coordinates of all vertices to match the new positions in the texture atlas, and establish a corporate culture scene integration model. The cultural semantic annotation and weight graph module is used to traverse all vertices and faces of the corporate culture scene integration model, identify the geometric parts associated with specific cultural elements of the honor wall and innovation concept display area, assign preset cultural semantic labels to vertices and faces, obtain the vertex semantic label set, and associate each cultural semantic label in the vertex semantic label set with a static numerical weight to construct a corporate culture semantic weight graph. The interaction tracking and dynamic attention calculation module is used to track the coordinates of the user's gaze point and the pointer vector of the controller in real time in virtual reality interaction based on the corporate culture semantic weight map, determine the geometric area of ​​the corporate culture exhibit that the input is pointing to, query the cultural semantic tags associated with the hit vertex from the corporate culture semantic weight map, obtain the real-time attention semantic tags, and calculate the static numerical weights corresponding to the real-time attention semantic tags according to the duration of the user's gaze or pointing, so as to generate the user interaction dynamic attention weight. The adaptive topology subdivision and simplification module is used to retrieve the dynamic attention weight values ​​of user interactions in each region of the corporate culture scene integration model grid based on the dynamic attention weights of user interactions and the corporate culture scene integration model. The values ​​are compared with preset subdivision thresholds and simplification thresholds. New vertices and edges are inserted in the grid regions where the weight values ​​exceed the subdivision thresholds to generate a dynamic subdivision vertex coordinate set. At the same time, edge folding and vertex removal are performed in the regions where the weight values ​​are lower than the simplification thresholds to output the adaptive corporate culture model topology structure.

Images