Metacosm scene automatic generation and optimization platform based on AIGC technology

Abstract

The invention discloses a meta universe scene automatic generation and optimization platform based on an AIGC technology, and relates to the technical field of meta universe. The platform comprises a resource management module, a resource matching module, a scene generation module and a scene optimization module which are in communication connection. The resource matching module is used for transmitting a resource acquisition request to the resource management module according to the meta universe user portrait; the resource management module returns framework resources, component resources and texture resources to the resource matching module according to the resource acquisition request; the scene generation module comprises a framework construction module, a component layout module and a texture synthesis module; and the scene optimization module is used for optimizing the global illumination effect of the scene in real time. According to the method and the device, customized scene resources meeting individual requirements of users can be automatically acquired and efficiently combined, the complexity and the time cost of manual intervention are reduced, efficient, customized and excellent-visual-effect meta-universe scene generation is realized, and the development efficiency and the user experience feeling are improved.

Description

Technical Field

[0001] The present invention relates to the field of metaverse technology, and in particular to an automatic generation and optimization platform for metaverse scenarios based on AIGC technology. Background Art

[0002] With the development of science and technology, technologies such as virtual reality and augmented reality have gradually matured, and the metaverse has emerged as a new digital world concept around the world. However, traditional metaverse scene generation often requires a lot of manpower and time costs. Designers need to manually create building frameworks, design component layouts, render textures, etc. This process is complicated and time-consuming. In addition, manual processing of personalized needs is difficult to quickly meet the diverse needs of users, making the user experience unsatisfactory.

[0003] In response to the above problems, the industry has adopted some technical means in recent years to optimize the metaverse scene generation process, such as improving design efficiency through parametric design and prefabricated component libraries to shorten development time and improve generation effects. At the same time, the development of cloud computing and edge computing technologies also provides computing resources to support real-time rendering and interaction of large-scale scenes.

[0004] Nevertheless, these methods still have shortcomings in efficiency and effectiveness in processing customized metaverse scene automatic generation. Existing technologies are difficult to quickly obtain and combine resources that meet the personalized needs of users, resulting in the generated scenes being difficult to fully meet the personalized expectations of users. Summary of the invention

[0005] The present invention provides an automatic generation and optimization platform for metaverse scenes based on AIGC technology to solve at least one of the problems mentioned in the above background technology.

[0006] The specific technical solutions provided by this application are as follows: An AIGC-based metaverse scene automatic generation and optimization platform, including a communication connection resource management module, a resource matching module, a scene generation module, and a scene optimization module; The resource matching module is used to transmit a resource acquisition request to the resource management module according to the metaverse user profile; The resource management module returns the framework resources, component resources and texture resources to the resource matching module according to the resource acquisition request; The scene generation module includes a framework construction module, a component layout module and a texture synthesis module; the framework construction module is used to construct a basic building framework according to the framework resources; the component layout module is used to determine the component position according to the component resources, and add corresponding components to the basic building framework according to the component position; the texture synthesis module is used to perform texture rendering on the basic building framework and components according to the texture resources; The scene optimization module is used to optimize the global illumination effect of the scene in real time.

[0007] As a preferred solution, it also includes a user management module; the resource matching module is further used to initiate a preference information acquisition request to the user management module; the user management module is used to: Parsing the preference information acquisition request to obtain a user identifier; Query the user preference information corresponding to each scenario type in the database through the user ID; If the user preference information corresponding to the scene type is empty, the reference user preference information with the highest similarity to the user basic information in the database is obtained to complete it; Transmit user preference information corresponding to each scene type to the resource matching module.

[0008] As a preferred solution, the user management module is also used to set user preference information through user behavior history information; The setting of user preference information through user behavior history information includes: If the amount of the target user's user behavior history information reaches a set threshold, the user behavior history information and its corresponding scenario resource combination are obtained based on the user's basic information; Acquire scene features of scene resource combinations through feature extraction algorithms; Obtain several reference users in the database whose basic information is most similar to that of the target user, and construct a user analysis matrix based on the user behavior history information and scenario characteristics of the target user and the reference users; The user analysis matrix is ​​input into the user preference model to obtain the user preference information of the target user and stored in the database.

[0009] As a preferred scheme, the user preference model includes a matrix analysis module, a collaborative filtering module, a scene similarity recommendation module and a joint decision-making module; the matrix analysis module is used to obtain the scene features of the target user in which elements are missing in the user analysis matrix, which are recorded as the missing element column; the collaborative filtering module is used to calculate the first reference preference value of the missing element column based on the user analysis matrix and the collaborative filtering algorithm; the scene similarity recommendation module is used to obtain the scene feature with the highest similarity to the missing element column, and record the preference value of the scene feature with the highest similarity as the second reference preference value of the missing element column; the joint decision-making module is used to fill the preference value of the missing element column according to the first reference preference value and the second reference preference value.

[0010] As a preferred solution, the framework construction module is used to parse the framework resources to obtain three-dimensional design models and basic structural elements, set several sub-areas and their attribute information in the basic building framework according to the three-dimensional design models and basic structural elements, construct a path structure through a path generation algorithm, and form a basic building framework based on several sub-areas and path structures.

[0011] As a preferred solution, the component layout module is used to allocate components to corresponding sub-areas according to attribute information, and determine the component position in each sub-area by using a particle swarm algorithm.

[0012] As a preferred solution, the method of determining the component position in each sub-area by using a particle swarm algorithm includes: S11, obtaining the preset optimal spacing and the preset minimum spacing of each component, creating a group of particles for each sub-area and assigning an initial velocity; each particle represents a component layout mode; S12, using the fitness function to calculate the fitness value of each particle; S13, update the personal best position and group best position of each particle; S14, updating the speed and position of each particle according to the individual best position and the group best position; S15, repeating steps S12 to S14 until the maximum number of iterations is reached or the fitness value of a particle meets the preset conditions, and outputting the optimal position of the group; The fitness function is expressed as: ; ; in, represents the fitness function of the ith particle; n is the number of components in the sub-region; j and k both represent component numbers; represents the preset optimal spacing of the jth component; Indicates the distance between the jth component and the kth component; Indicates the preset minimum spacing of the j-th component.

[0013] As a preferred solution, the texture synthesis module is used to obtain texture samples corresponding to the basic building framework and components, synthesize texture images according to the size and texture samples of the basic building framework and components, and render the basic building framework and components in real time according to the texture images.

[0014] As a preferred solution, the method of synthesizing a texture image according to the size of the basic building frame and components and the texture sample image comprises the steps of: Creating a blank target texture image and setting a synthesis start position of the target texture image; the size of the target texture image is the size of the selected basic structural element or component; Obtaining the domain error value of each sample image pixel point in the texture sample image, and generating a sample image error set; the sample image error set stores the coordinates of each sample image pixel point and its domain error value; Randomly take values ​​from the texture sample image and fill them into the synthesis starting position of the target texture image; Traversing the pixels to be filled in the target texture image based on the synthesis starting position and the preset scan line sequence; obtaining the neighborhood error value of the pixels to be filled; The pixel filling strategy is executed cyclically until the last pixel; the pixel filling strategy is: based on the domain error value of the pixel to be filled, the sample image pixel corresponding to the domain error value with the smallest relative error in the sample image error set is matched, and the pixel value of the sample image pixel is used as the pixel value of the pixel to be filled.

[0015] As a preferred solution, the relative error is expressed as: ; in, The area of ​​pixel t to be filled, is the domain of the sample pixel s, Relative error between the filling pixel t and the sample pixel s; To represent the domain, it can be expressed in matrix form; express The pixel value of the pixel point (i, j) in the middle area; express The pixel value of the pixel point (i, j) in the domain; N represents the number of pixels in the domain.

[0016] Compared with the prior art, the present invention has the following beneficial effects: In the embodiment of the present invention, the resource matching module transmits a resource acquisition request to the resource management module according to the metaverse user portrait, and the resource management module returns the framework resources, component resources and texture resources to the resource matching module according to the resource acquisition request, so that customized scene resources that meet the individual needs of users can be automatically acquired and efficiently combined, reducing the complexity and time cost of manual intervention; the scene generation module automatically constructs the scene according to the framework resources, component resources and texture resources, and the scene optimization module realizes global illumination optimization, which enhances the overall visual depth and immersion. This application realizes efficient, customized, and visually superior metaverse scene generation, improving development efficiency and user experience.

[0017] The embodiments of the present invention obtain comprehensive user behavior history information by tracking and recording user activities in different scenarios, clarify the user's specific needs in different virtual scenarios, optimize scene resource matching, and provide a personalized and satisfactory user experience.

[0018] The component layout module of the embodiment of the present invention allocates components to corresponding sub-areas through attribute information, and optimizes the position of components in each sub-area using a particle swarm algorithm, thereby ensuring that the components meet user needs and scene characteristics, and maintaining visual and functional consistency between components in the scene, thereby enhancing space utilization and making the automatic generation of scenes more reasonable. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0020] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0021] Figure 1 A schematic diagram of the structure of a Metaverse scene automatic generation and optimization platform based on AIGC technology provided in one embodiment of the present invention; Figure 2 A schematic diagram of the structure of a Metaverse scene automatic generation and optimization platform based on AIGC technology provided in another embodiment of the present invention; Figure 3 A schematic diagram of the structure of a user preference model provided by an embodiment of the present invention; Figure 4 A schematic diagram of a process for synthesizing a texture image according to the size and texture sample of a basic building frame and components provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0022] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0023] It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present invention are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0024] In addition, the descriptions of "first", "second", etc. in the present invention are only used for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least the feature. In addition, the technical solutions between the various embodiments can be combined with each other, but they must be based on the ability of ordinary technicians in the field to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0025] With the development of science and technology, technologies such as virtual reality and augmented reality have gradually matured, and the metaverse has emerged as a new digital world concept around the world. However, traditional metaverse scene generation often requires a lot of manpower and time costs. Designers need to manually create building frameworks, design component layouts, render textures, etc. This process is complicated and time-consuming. In addition, manual processing of personalized needs is difficult to quickly meet the diverse needs of users, making the user experience unsatisfactory.

[0026] In response to the above problems, the industry has adopted some technical means in recent years to optimize the metaverse scene generation process, such as improving design efficiency through parametric design and prefabricated component libraries to shorten development time and improve generation effects. At the same time, the development of cloud computing and edge computing technologies also provides computing resources to support real-time rendering and interaction of large-scale scenes.

[0027] Nevertheless, these methods still have shortcomings in efficiency and effectiveness in processing customized metaverse scene automatic generation. Existing technologies are difficult to quickly obtain and combine resources that meet the personalized needs of users, resulting in the generated scenes being difficult to fully meet the personalized expectations of users.

[0028] To this end, the present invention provides an automatic generation and optimization platform for metaverse scenes based on AIGC (generative artificial intelligence, Artificial Intelligence Generated Content) technology. The specific implementation process of the present invention will be described in detail below.

[0029] See also Figure 1 , the present invention provides a metaverse scene automatic generation and optimization platform based on AIGC technology, including a communication-connected resource management module, a resource matching module, a scene generation module, and a scene optimization module; The resource matching module is used to transmit a resource acquisition request to the resource management module according to the metaverse user profile; The resource management module returns the framework resources, component resources and texture resources to the resource matching module according to the resource acquisition request; The scene generation module includes a framework construction module, a component layout module and a texture synthesis module; the framework construction module is used to construct a basic building framework according to the framework resources; the component layout module is used to determine the component position according to the component resources, and add corresponding components to the basic building framework according to the component position; the texture synthesis module is used to perform texture rendering on the basic building framework and components according to the texture resources; The scene optimization module is used to optimize the global illumination effect of the scene in real time.

[0030] This application transmits a resource acquisition request to the resource management module according to the Metaverse user portrait through the resource matching module, and the resource management module returns the framework resources, component resources and texture resources to the resource matching module according to the resource acquisition request, so that customized scene resources that meet the user's individual needs can be automatically acquired and efficiently combined, reducing the complexity and time cost of manual intervention; the scene generation module automatically builds the scene according to the framework resources, component resources and texture resources, and the scene optimization module realizes global lighting optimization, which enhances the overall visual depth and immersion. This application realizes efficient, customized, and visually superior Metaverse scene generation, improving development efficiency and user experience.

[0031] Furthermore, the Metaverse user portrait includes basic user information and user preference information corresponding to each scenario type. For example, basic user information includes user ID, age, gender, occupation, several style preferences, etc. Scenario types may include virtual conference and event venues, virtual education environments, virtual museums, and virtual shopping environments. User preference information corresponding to each scenario type can be constructed through user behavior history information.

[0032] As a preferred embodiment, please refer to Figure 2 The automatic generation and optimization platform of the present application further includes a user management module; the resource matching module is further used to initiate a preference information acquisition request to the user management module; the user management module is used to: Parsing the preference information acquisition request to obtain a user identifier; Query the user preference information corresponding to each scenario type in the database through the user ID; If the user preference information corresponding to the scene type is empty, the reference user preference information with the highest similarity to the user basic information in the database is obtained to complete it; Transmit user preference information corresponding to each scene type to the resource matching module.

[0033] Based on the above content, in order to obtain the user portrait of the metaverse, the resource matching module will initiate a preference information acquisition request to the user management module. After receiving the preference information acquisition request, the user management module will identify the user identifier (such as user ID) by parsing the content. Using the user identifier, the user management module searches the database to find the user's preference information under different scene types. If the user preference information of certain scene types is missing, the user management module will find similar users through the user's basic information and obtain their preference information as a reference; this step aims to find users with high similarity by analyzing the user's basic information (such as age, gender, occupation, etc.), so as to draw on their preference information for completion. Finally, the user management module returns the complete or completed user preference information to the resource matching module, ensuring that even in the case of incomplete user preference information, the intelligent completion mechanism can be used to improve the personalization of scene generation and user satisfaction.

[0034] Based on the above embodiments, the Metaverse scene automatic generation and optimization platform can be used to build virtual conference and event venues, virtual education environments, virtual museums, virtual shopping environments, etc. Taking the virtual museum as an example, the basic building framework constructed by the framework construction module includes the walls, floors and ceilings of each exhibition hall, as well as basic layouts such as roads, stairs and corridors. At this stage, the museum shows the overall spatial distribution and dynamic lines inside, and users can know where the exhibition halls and rest areas are, but the elements in the basic building framework are relatively simple, lacking exhibits and detailed decorations. After completing the construction of the basic building framework, the component layout module is used to determine the locations of components such as exhibits, signboards, information screens, guardrails, indicator lights and rest area furniture according to component resources and the basic building framework, and automatically configure the interactive elements between users and each component. The texture synthesis module provides basic textures for major components such as walls, floors, and ceilings, such as basic patterns of wallpaper, wooden floors or stone floors, and adds surface textures to components, such as the fabric texture of sofas, the wood grain and gloss of coffee tables, etc.

[0035] Furthermore, the user management module is also used to set user preference information through user behavior history information. Specifically, setting user preference information through user behavior history information includes: If the amount of information of the target user's user behavior history information reaches the set threshold, the user behavior history information and its corresponding scene resource combination are obtained according to the user's basic information. The purpose of this step is to observe and record the user's activities in a specific virtual scene to collect and analyze the process of user behavior; the scene resources include framework resources, component resources and texture resources; the user behavior history information includes the user's specific activities, operations and corresponding residence time in each scene type. For example, in virtual conferences and activity venues, the number of times users attend meetings, duration, and interaction frequency are counted; in virtual education environments, the number of times users attend courses, learning progress and completion rate are counted; in virtual museums, the number of user visits, residence time, and types of exhibits browsed are counted; in virtual shopping environments, the user's shopping behavior, browsed commodity types, and purchase history are counted. In this step, the user's identity is identified by the user identifier in the user's basic information, and each obtained user's basic information corresponds to multiple user behavior history information and its corresponding scene resource combination, that is, the user behavior history information collected by the user under different scene resource combinations. It can be understood that when executing this step, if the amount of user behavior history information does not reach the set threshold, the subsequent steps will not be executed, but the user preference information will be directly supplemented by obtaining the reference user preference information with the highest similarity to the user basic information in the database as mentioned above.

[0036] Acquire scene features of the scene resource combination through feature extraction algorithms; the scene features include visual elements (color, shape, layout), interactive elements (buttons, links), functional elements (educational tools, exhibits), etc. of the scene; Obtain several reference users in the database whose basic information is most similar to that of the target user, and construct a user analysis matrix based on the user behavior history information and scene features of the target user and several reference users; the user analysis matrix is ​​a two-dimensional matrix, with rows representing users and columns representing scene features. Each element of the matrix represents the preference value of the target user or several reference users for the scene feature, and the preference value is obtained by referring to the preference value of the user or by quantifying the user behavior history information. Among them, the quantification and weighting processing are specifically as follows: extracting several behavioral features based on the user behavior history information, quantifying each behavioral feature to obtain a behavioral feature preference score, and weighting the behavioral feature preference score according to the importance of the behavioral feature to obtain the preference value of the scene feature. For example, the number of clicks, browsing time, interaction frequency, etc. can be standardized to make them suitable for comparison, and the weights can be learned based on experience, business needs, or through machine learning models.

[0037] The user analysis matrix is ​​input into the user preference model to obtain the user preference information of the target user and stored in the database. The user preference model of this embodiment can identify the user's preference for different scenarios, thereby predicting the user's reaction or rating when encountering similar feature scenarios in the future.

[0038] For further information, see Figure 3 The user preference model includes a matrix analysis module, a collaborative filtering module, a scene similarity recommendation module and a joint decision module; the matrix analysis module is used to obtain scene features of target users whose elements are missing in the user analysis matrix, recorded as the missing element column; the collaborative filtering module is used to calculate the first reference preference value of the missing element column according to the user analysis matrix and the collaborative filtering algorithm; the scene similarity recommendation module is used to obtain the scene feature with the highest similarity to the missing element column, and record the preference value of the scene feature with the highest similarity as the second reference preference value of the missing element column; the joint decision module is used to fill the preference value of the missing element column according to the first reference preference value and the second reference preference value.

[0039] In this embodiment, the matrix analysis module is used to identify which scene features in the user analysis matrix have missing preference values. For example, a user has never generated user behavior history information for a certain scene feature, so it is impossible to directly calculate the preference value corresponding to the scene feature based on the user behavior history information. The matrix analysis module will mark the scene feature of the user as an element missing column.

[0040] The collaborative filtering module is used to calculate the first reference preference value of the missing element column based on the user analysis matrix and the collaborative filtering algorithm, that is, to predict the target user's preference based on similar user behaviors. The target user's preference value for unexposed scene features is predicted by analyzing the similarity between the target user and other users on non-empty scene features. For example, if two users exhibit similar behaviors in most cases, the system will assume that they may also have similar preferences in other features.

[0041] The scene similarity recommendation module is used to obtain the scene feature with the highest similarity to the missing element column, and record the preference value of the scene feature with the highest similarity as the second reference preference value of the missing element column, that is, to estimate the user's preference value for the missing element column based on the similarity of the scene features. When the missing element column has a significant correlation with other scene features, the scene similarity recommendation module uses the other scene features as the second reference preference value of the missing element column.

[0042] The joint decision module combines the prediction results from the collaborative filtering module and the scene similarity recommendation module to finally determine the preference value of the missing element column. This module will evaluate and integrate the preference values ​​generated by the collaborative filtering module and the scene similarity recommendation module to balance the recommendation results of the user's historical behavior and scene characteristics, so as to obtain a more accurate and personalized user preference information.

[0043] This embodiment tracks and records the user's activities in different scenes, obtains comprehensive user behavior history information, clarifies the user's specific needs in different virtual scenes, optimizes scene resource matching, and provides a personalized and satisfactory user experience.

[0044] Furthermore, the framework construction module is used to parse the framework resources to obtain the three-dimensional design model and basic structural elements, set several sub-areas and their attribute information in the basic building framework according to the three-dimensional design model and the basic structural elements, construct the path structure through the path generation algorithm, and form the basic building framework according to the several sub-areas and the path structure. Among them, the three-dimensional design model is used to identify the boundaries and positions of each sub-area; the basic structural elements include elements such as the exhibition hall location, walls, floors and ceilings, and the size and position information can be extracted according to the basic structural elements; the path structure can include roads, corridors, stairs, etc. to ensure the connectivity between the sub-areas in the basic building framework. Among them, the path generation algorithm is the Dijkstra algorithm; by generating a path through the Dijkstra algorithm, the shortest path in the internal space of the building can be found to ensure efficient connectivity between the sub-areas.

[0045] Furthermore, the component layout module is used to allocate components to corresponding sub-areas according to the attribute information, and determine the component positions in each sub-area by using a particle swarm algorithm. The component positions in each sub-area determined by using a particle swarm algorithm include: S11, obtaining the preset optimal spacing and the preset minimum spacing of each component, creating a group of particles for each sub-area and assigning an initial velocity; each particle represents a component layout mode; S12. Calculate the fitness value of each particle using a fitness function; the fitness function is used to evaluate the layout spacing of each component in the particle and to establish a preset optimal spacing.

[0046] The fitness function is expressed as: ; ; in, represents the fitness function of the ith particle; n is the number of components in the sub-region; j and k both represent component numbers; represents the preset optimal spacing of the jth component; Indicates the distance between the jth component and the kth component; Indicates the preset minimum spacing of the j-th component.

[0047] S13, update the personal best position and group best position of each particle; S14, updating the speed and position of each particle according to the individual best position and the group best position; The velocity of the particle is expressed as: , The position of the particle is expressed as: , in, is the inertia weight, which indicates the influence of the velocity of the previous generation of particles on the velocity of the current generation of particles; t indicates the iteration round; i indicates the sequence number of the particle; Specifically, it represents the velocity of the i-th particle at the t-th iteration; Specifically represents the position of the i-th particle at the t-th iteration; and They are the first learning factor and the second learning factor. The first learning factor is used to control the amplitude of the particle moving toward the individual optimal position, and the second learning factor is used to control the amplitude of the particle moving toward the group optimal position. is the particle's personal best position, is the group optimal position of the group to which the particle belongs. and are all random numbers in the interval [0,1], used to provide randomness for each iteration.

[0048] S15. Repeat steps S12 to S14 until the maximum number of iterations is reached or the fitness value of a particle meets a preset condition, and output the optimal position of the group.

[0049] The component layout module of this embodiment allocates components to corresponding sub-areas through attribute information, and optimizes the position of components in each sub-area using the particle swarm algorithm, thereby ensuring that the components meet user needs and scene characteristics, and maintaining the visual and functional consistency between components in the scene, thereby enhancing space utilization and making the automatic generation of scenes more reasonable.

[0050] Furthermore, the texture synthesis module is used to obtain texture sample images corresponding to the basic building framework and components, synthesize texture images according to the size of the basic building framework and components and the texture sample images, and render the basic building framework and components in real time according to the texture images.

[0051] Among them, the texture sample is a two-dimensional image file used to define the appearance of a surface in three-dimensional modeling and rendering. It provides visual information about surface details, such as color, pattern, material, etc. The texture image is a large-scale image generated by splicing, expanding or synthesizing texture samples. In one embodiment, the synthesized texture image is generated by repeatedly splicing and tiling texture samples to generate a large-scale image based on the size of the basic building frame and components to cover the surface of the building frame and components. In another embodiment, the synthesized texture image analyzes and blends adjacent areas through a domain search algorithm to gradually synthesize a large-scale texture image to avoid obvious boundaries and seams.

[0052] For details, please refer to Figure 4 The step of synthesizing the texture image according to the size of the basic building frame and components and the texture sample image comprises the following steps: S21, creating a blank target texture image and setting a synthesis start position of the target texture image; the size of the target texture image is the size of the selected basic structural element or component; S22, obtaining the domain error value of each sample image pixel point in the texture sample image, and generating a sample image error set; the sample image error set stores the coordinates of each sample image pixel point and its domain error value; S23, randomly taking values ​​from the texture sample image and filling them into the synthesis starting position of the target texture image; S24, traversing the pixels to be filled in the target texture image based on the synthesis starting position and the preset scan line sequence; obtaining the neighborhood error value of the pixels to be filled; S25. Loop through the pixel filling strategy until the last pixel. The pixel filling strategy is: match the sample image pixel corresponding to the domain error value with the smallest relative error in the sample image error set based on the domain error value of the pixel to be filled, and use the pixel value of the sample image pixel as the pixel value of the pixel to be filled.

[0053] In one embodiment, the domain error value is an L-shaped domain error value. In this embodiment, the neighborhood of the pixel to be filled is not a regular complete rectangular window, but an L-shaped area. This selection method is more suitable for filling textures in scan line order, because in the L-shaped area, many filled pixels can be used to calculate the error value. The L-shaped neighborhood consists of two parts: a horizontal line segment and a vertical line segment, which together form an L shape. For example, the L-shaped neighborhood of the pixel to be filled is set to a vertical line segment consisting of two pixels and a horizontal line segment consisting of five pixels surrounding the pixel to be filled.

[0054] The relative error is expressed as: ; in, The area of ​​pixel t to be filled, is the domain of the sample pixel s, Relative error between the filling pixel t and the sample pixel s; To represent the domain, it can be expressed in matrix form; express The pixel value of the pixel point (i, j) in the middle area; express The pixel value of the pixel point (i, j) in the domain; N represents the number of pixels in the domain.

[0055] This embodiment reduces seams and boundaries through point-by-point matching and blending, so that the final texture image is naturally continuous; randomness and local feature processing are appropriately introduced through the domain search algorithm to avoid generating monotonous and repetitive texture images, so that the texture image has higher heterogeneity.

[0056] In one embodiment, the scene optimization module, said real-time optimization of scene global illumination, comprises: Starting from the eye position of the Metaverse user, an initial ray is generated and projected into the scene; Tracing the path of a ray of light through multiple interactions of reflection, refraction, and scattering to ultimately reach a light source; where a is the entity or point in the scene from which the ray of light originated.

[0057] When light interacts with the surface of an object, the surface illumination properties of the intersection are calculated; the surface illumination properties include normal, reflectivity, refractive index, etc.

[0058] Use BSDF (bidirectional scattering distribution function) model to determine the direction of light propagation; The final color of pixels in the scene is determined by accumulating the rays traced to the light sources; The Russian roulette method is used to determine the stopping condition for ray tracing. In this embodiment, when the light is traced in the scene, it will continuously reflect, refract or scatter with the surface of the object, and each interaction will increase the amount of calculation. In order to avoid endless tracing, each ray needs a termination condition. The Russian roulette method terminates the light after randomly selecting a certain interaction. Specifically, the termination probability can be calculated based on the reflectivity and refractive index of the light at the current interaction point. If the termination probability is high, the light is more likely to continue tracing, otherwise it is more likely to terminate. In specific operations, a random number between 0 and 1 is generated. If the random number is less than the termination probability, the light continues to be traced; otherwise, the light is terminated. Furthermore, in order to maintain the consistency and physical authenticity of the results, the contribution of the terminated light needs to be weighted according to the termination probability of determining the continued tracing of the light. The contribution of the light that continues to be traced will be amplified, that is, divided by the termination probability, to make up for the loss caused by random termination.

[0059] This embodiment reduces the average tracing depth of light rays through the Russian roulette method. This reduces the resources and time required for each frame calculation, thereby increasing the rendering speed of the scene, and can reasonably allocate the calculation focus of tracing rays under limited computing resources, ensuring that the light and shadow changes in the high dynamic range are effectively expressed. The technical effect of this embodiment is particularly significant in large-scale and complex scenes.

[0060] The foregoing is merely a specific embodiment of the present invention, which enables those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A Metaverse scene automatic generation and optimization platform based on AIGC technology, characterized by: It includes a communication connection resource management module, a resource matching module, a scenario generation module and a scenario optimization module; The resource matching module is used to transmit a resource acquisition request to the resource management module according to the metaverse user profile; The resource management module returns the framework resources, component resources and texture resources to the resource matching module according to the resource acquisition request; The scene generation module includes a framework construction module, a component layout module and a texture synthesis module; the framework construction module is used to construct a basic building framework according to the framework resources; the component layout module is used to determine the component position according to the component resources, and add corresponding components to the basic building framework according to the component position; the texture synthesis module is used to perform texture rendering on the basic building framework and components according to the texture resources; The scene optimization module is used to optimize the global illumination effect of the scene in real time.

2. The AIGC-based Metaverse scene automatic generation and optimization platform according to claim 1, characterized in that: It also includes a user management module; the resource matching module is further used to initiate a preference information acquisition request to the user management module; the user management module is used to: Parsing the preference information acquisition request to obtain a user identifier; Query the user preference information corresponding to each scenario type in the database through the user ID; If the user preference information corresponding to the scene type is empty, the reference user preference information with the highest similarity to the user basic information in the database is obtained to complete it; Transmit user preference information corresponding to each scene type to the resource matching module.

3. The AIGC-based Metaverse scene automatic generation and optimization platform according to claim 2, characterized in that: The user management module is also used to set user preference information through user behavior history information; The setting of user preference information through user behavior history information includes: If the amount of the target user's user behavior history information reaches a set threshold, the user behavior history information and its corresponding scenario resource combination are obtained based on the user's basic information; Acquire scene features of scene resource combinations through feature extraction algorithms; Obtain several reference users in the database whose basic information is most similar to that of the target user, and construct a user analysis matrix based on the user behavior history information and scenario characteristics of the target user and the reference users; The user analysis matrix is ​​input into the user preference model to obtain the user preference information of the target user and stored in the database.

4. The AIGC-based Metaverse scene automatic generation and optimization platform according to claim 3 is characterized by: The user preference model includes a matrix analysis module, a collaborative filtering module, a scene similarity recommendation module and a joint decision module; the matrix analysis module is used to obtain scene features of target users whose elements are missing in the user analysis matrix, recorded as element missing columns; the collaborative filtering module is used to calculate the first reference preference value of the element missing column according to the user analysis matrix and the collaborative filtering algorithm; the scene similarity recommendation module is used to obtain the scene feature with the highest similarity to the element missing column, and record the preference value of the scene feature with the highest similarity as the second reference preference value of the element missing column; The joint decision module is used to fill in the preference value of the missing element column according to the first reference preference value and the second reference preference value.

5. The AIGC-based Metaverse scene automatic generation and optimization platform according to claim 1, characterized in that: The framework construction module is used to parse the framework resources to obtain the three-dimensional design model and basic structural elements, set several sub-areas and their attribute information in the basic building framework according to the three-dimensional design model and the basic structural elements, construct the path structure through the path generation algorithm, and form the basic building framework according to the several sub-areas and the path structure.

6. The AIGC-based Metaverse scene automatic generation and optimization platform according to claim 1, characterized in that: The component layout module is used to allocate components to corresponding sub-areas according to attribute information, and to determine the component positions in each sub-area by using a particle swarm algorithm.

7. The AIGC-based Metaverse scene automatic generation and optimization platform according to claim 6, characterized in that: The determining the component position in each sub-area by using a particle swarm algorithm includes: S11, obtaining the preset optimal spacing and the preset minimum spacing of each component, creating a group of particles for each sub-area and assigning an initial velocity; each particle represents a component layout mode; S12, using the fitness function to calculate the fitness value of each particle; S13, update the personal best position and group best position of each particle; S14, updating the speed and position of each particle according to the individual best position and the group best position; S15, repeating steps S12 to S14 until the maximum number of iterations is reached or the fitness value of a particle meets the preset conditions, and outputting the optimal position of the group; The fitness function is expressed as: ； ； in, represents the fitness function of the ith particle; n is the number of components in the sub-region; j and k both represent component numbers; represents the preset optimal spacing of the jth component; Indicates the distance between the jth component and the kth component; Indicates the preset minimum spacing of the j-th component.

8. The AIGC-based Metaverse scene automatic generation and optimization platform according to claim 1, characterized in that: The texture synthesis module is used to obtain texture sample images corresponding to the basic building framework and components, synthesize texture images according to the size of the basic building framework and components and the texture sample images, and render the basic building framework and components in real time according to the texture images.

9. The AIGC-based Metaverse scene automatic generation and optimization platform according to claim 8, characterized in that: The method of synthesizing a texture image according to the size of the basic building frame and components and the texture sample image comprises the steps of: Creating a blank target texture image and setting a synthesis start position of the target texture image; the size of the target texture image is the size of the selected basic structural element or component; Obtaining the domain error value of each sample image pixel point in the texture sample image, and generating a sample image error set; the sample image error set stores the coordinates of each sample image pixel point and its domain error value; Randomly take values ​​from the texture sample image and fill them into the synthesis starting position of the target texture image; Traversing the pixels to be filled in the target texture image based on the synthesis starting position and the preset scan line sequence; obtaining the neighborhood error value of the pixels to be filled; The pixel filling strategy is executed cyclically until the last pixel; the pixel filling strategy is: based on the domain error value of the pixel to be filled, the sample image pixel corresponding to the domain error value with the smallest relative error in the sample image error set is matched, and the pixel value of the sample image pixel is used as the pixel value of the pixel to be filled.

10. The AIGC-based Metaverse scene automatic generation and optimization platform according to claim 9, characterized in that: The relative error is expressed as: ； in, The area of ​​pixel t to be filled, is the domain of the sample pixel s, Relative error between the filling pixel t and the sample pixel s; To represent the domain, it can be expressed in matrix form; express The pixel value of the pixel point (i, j) in the middle area; express The pixel value of the pixel point (i, j) in the domain; N represents the number of pixels in the domain.

Images