Multi-source data driven three-dimensional model texture rendering system and method thereof

Abstract

The invention relates to a multi-source data driven three-dimensional model texture rendering system and method, and the system comprises a multi-source input parameter processing module which is used for receiving and processing multi-source input data from a user instruction and a light and shadow environment; the AI texture calculation decision module is connected to the multi-source input parameter processing module and is used for analyzing a user intention through an artificial intelligence technology based on the multi-source input data and generating a texture decision; the texture state and application rendering module is connected to the AI texture calculation decision module and is used for dynamically managing and rendering the texture state of the three-dimensional model in real time according to the texture decision, so that multi-source data driving of natural interaction, environment self-adaption and intelligent decision can be realized, intelligent association between texture expression and environment factors is established, and the application rendering module is used for rendering the texture state of the three-dimensional model according to the texture decision. And intelligent fusion and coordination of multi-source data are realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of computer vision and image processing, in particular to a multi-source data driven three-dimensional model texture rendering system and method thereof. BACKGROUND

[0002] In the field of three-dimensional graphics rendering, texture management is a key factor in determining rendering quality and efficiency. Traditional texture management systems mainly use the following technical solutions: (1) Static texture mapping technology: the mapping relationship between texture maps and models is defined in advance, and the loading and rendering are performed according to fixed rules at runtime. This method is simple but lacks flexibility. (2) Parameterized texture control: a large number of parameters are provided for user adjustment through a GUI interface, such as roughness, metallicity, etc. This method is accurate in control but complex in operation. (3) Preset scheme switching: several sets of predefined texture schemes are provided for user selection. This method is simple to use but has limited choices. (4) Programmed texture generation: noise functions and other algorithms are used to generate textures in real time. This method is flexible but computationally expensive. The main defects of the prior art include: single and unnatural interaction mode, lack of environmental perception ability, difficulty in multi-source data fusion, lack of intelligent decision-making ability, poor system scalability. In particular, it cannot understand natural language descriptions (such as "make the material look more aged"), and cannot automatically adjust the texture performance according to time, weather, and other environmental factors.

[0003] Based on this, the present application is proposed.

[0004] In addition, on the one hand, there are differences in understanding between those skilled in the art; on the other hand, the applicant has studied a large number of literatures and patents when making the present application, but due to space limitations, all details and contents are not listed in detail, which does not mean that the present application does not have these characteristics of the prior art. On the contrary, the present application already has all the characteristics of the prior art, and the applicant reserves the right to add relevant prior art in the background art. SUMMARY

[0005] In view of the deficiencies of the prior art, the present application provides a multi-source data driven three-dimensional model texture rendering system and method thereof, which can realize natural interaction, environmental adaptation, intelligent decision-making of multi-source data driving, establish intelligent association between texture performance and environmental factors, and realize intelligent fusion and coordination of multi-source data.

[0006] The present application discloses a multi-source data driven three-dimensional model texture rendering system, comprising: A multi-source input parameter processing module is used to receive and process multi-source input data from user instructions and light and shadow environments; An AI texture calculation decision module is connected to the multi-source input parameter processing module and is used to analyze user intent and generate texture decisions based on multi-source input data through artificial intelligence technology. a texture state and application rendering module connected to the AI texture computation decision module, configured to dynamically manage and render the texture state of the three-dimensional model according to the texture decision.

[0007] In some embodiments, the multi-source input parameter processing module comprises: a user instruction processing unit configured to parse different types of user instructions, including at least explicit instructions with explicit texture parameters, ambiguous instructions based on natural language descriptions, and style instructions with artistic style descriptions; a light and shadow environment setting unit configured to process parameters related to time, weather, lighting, and scene environment to adapt the texture performance to environmental changes.

[0008] In some embodiments, the user instruction processing unit comprises: an ambiguous instruction parser for converting natural language or imprecise descriptions into specific texture adjustment parameters; a style instruction parser for mapping abstract artistic style descriptions to corresponding texture rendering parameter sets.

[0009] In some embodiments, the light and shadow environment setting unit comprises: a time parameter processor for calculating sun angle, lighting intensity, and color temperature parameters according to the time of day; a weather parameter analyzer for calculating environmental light shading coefficients and generating texture adjustment parameters according to weather types.

[0010] In some embodiments, the AI texture computation decision module comprises: an intent parsing module that uses deep learning-based techniques to understand the semantics of user instructions and classifies and maps abstract intentions to specific rendering parameters; a texture decision module that uses a reinforcement learning-based intelligent decision system to select the optimal texture adjustment action from a pre-set action space based on the current texture state, environmental parameters, and user history.

[0011] In some embodiments, the intent parsing module uses a pre-trained model for natural language processing and unifies text instructions and environmental parameters through a multi-modal fusion network; and / or, the texture decision module is based on a reinforcement learning framework, with a reward function configured to optimize rendering quality, user satisfaction, and system performance overhead simultaneously.

[0012] In some embodiments, the texture state and application rendering module comprises: a texture state manager for managing smooth transitions, caching, and version control of texture states; A texture rendering parameter controller is configured to dynamically update physics-based rendering parameters and perform parameter binding with GPU shaders. A particle system linkage module is configured to trigger and coordinate corresponding particle effects when the texture state changes.

[0013] The application further discloses a multi-source data-driven three-dimensional model texture rendering method, comprising the following steps: Receiving and fusing multi-source input data from user instructions and light and shadow environments; Performing intent analysis and texture decision based on artificial intelligence technology on the fused multi-source input data to generate texture control instructions; According to the texture control instructions, dynamically managing and real-time rendering the texture state of the three-dimensional model.

[0014] In some embodiments, the step of receiving and fusing multi-source input data comprises: Analyzing user instructions, the user instructions including at least one of explicit instructions with explicit texture parameters, fuzzy instructions based on natural language descriptions, and style instructions with artistic style descriptions; Obtaining and processing current light and shadow environment parameters, the environment parameters including at least one of time, weather, illumination, and scene environment information.

[0015] In some embodiments, the step of analyzing user instructions comprises: Using a pre-trained natural language processing model to extract semantic features of the user instructions; Converting the extracted semantic features into specific texture adjustment parameters through a pre-established "natural language-texture attribute" mapping knowledge base.

[0016] In some embodiments, the step of performing intent analysis and texture decision based on artificial intelligence technology comprises: Using a deep learning model to analyze user intent, identifying intent categories and mapping them to rendering parameters; Using a reinforcement learning model to make texture decisions, the state space of reinforcement learning including current texture state, target intent, and environment parameters, and the action space including texture selection, parameter adjustment, and mixed mode operation.

[0017] In some embodiments, the step of dynamically managing and real-time rendering the texture state comprises: Managing smooth transitions of the texture state and supporting state rollback; When the texture state changes, linkage control of the particle system is performed to generate corresponding visual effects; According to the visual range, dynamically adjusting the detail level of the texture. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 This is a schematic diagram of the overall system architecture of the present invention; Figure 2 This is a schematic diagram illustrating the overall architecture of the method of the present invention; Figure 3 This is a flowchart of the user instruction processing of the present invention; Figure 4 This is a schematic diagram of parameter conflict detection according to the present invention; Figure 5 This is a schematic diagram of the time parameters of the present invention; Figure 6 This is a schematic diagram of weather parameters for the present invention; Figure 7 This is a schematic diagram of the illumination parameters of the present invention; Figure 8 This is a schematic diagram of the model environment processing of the present invention; Figure 9 A schematic diagram illustrating the multi-parameter fusion calculation of the light and shadow settings of this invention; Figure 10 This is a schematic diagram of the operation of the intent parsing module of the present invention; Figure 11 This is a schematic diagram of the texture decision module of the present invention. Figure 12 This is a schematic diagram illustrating the final selection of texture decision in this invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] In the field of 3D graphics rendering, texture management is a key factor in determining rendering quality and efficiency. Traditional texture management systems mainly adopt the following technical solutions: (1) Static texture mapping technology: predefine the mapping relationship between texture maps and models, and load and render according to fixed rules at runtime. This method is simple but lacks flexibility. (2) Parametric texture control: provide a large number of parameters for users to adjust through the GUI interface, such as roughness and metallicity. This method is precise but complex to operate. (3) Preset scheme switching: provide several predefined texture schemes for users to choose from. This method is simple to use but has limited choices. (4) Procedural texture generation: use algorithms such as noise functions to generate textures in real time. This method is flexible but has high computational overhead. The main defects of existing technologies include: single and unnatural interaction methods, lack of environmental awareness, difficulty in multi-source data fusion, lack of intelligent decision-making ability, and poor system scalability. In particular, it cannot understand natural language descriptions (such as "make the material look more vintage") and cannot automatically adjust texture performance according to environmental factors such as time and weather.

[0021] As attached Figure 1 As shown, a multi-source data-driven 3D model texture rendering system is provided, including: The multi-source input parameter processing module is used to receive and process multi-source input data from user commands and the lighting environment; The AI ​​texture calculation and decision module is connected to the multi-source input parameter processing module and is used to analyze user intent and generate texture decisions based on the multi-source input data and artificial intelligence technology. The texture state and application rendering module is connected to the AI ​​texture calculation and decision module, and is used to dynamically manage and render the texture state of the 3D model in real time according to the texture decision.

[0022] This project constructs a data-driven, closed-loop intelligent rendering framework through a three-stage architecture: "Multi-source Input -> AI Intelligent Decision-Making -> Precise Rendering." By clearly defining the responsibilities of the three modules and connecting them, the system can transform multi-source, heterogeneous, and potentially ambiguous input data from users and the environment into accurate, reasonable, and real-time rendering outputs. The AI ​​texture calculation and decision-making module can use conventional large AI models, such as the DEEPSEEK large model and the GTP large model. This architecture tightly integrates human creative intent with the machine's computational optimization capabilities, enabling it to understand and execute various user instructions, from precise technical parameters to abstract artistic descriptions. This significantly reduces the operational threshold for non-professional users and improves the naturalness and efficiency of interaction. It establishes an intelligent correlation between texture representation and environmental factors, achieving intelligent fusion and coordination of multi-source data.

[0023] In some embodiments, the multi-source input parameter processing module includes: The user instruction processing unit is configured to parse different types of user instructions, which include at least: explicit instructions with explicit texture parameters, fuzzy instructions with fuzzy descriptions based on natural language, and style instructions with artistic style descriptions. The lighting and environment setting unit is configured to handle parameters related to time, weather, lighting, and scene environment so that the texture representation adapts to environmental changes.

[0024] It achieves systematic support for diverse user interactions and dynamic environments. Through different command input methods, such as explicit commands with clear texture parameters, fuzzy commands with fuzzy descriptions based on natural language, and style commands with artistic style descriptions, it provides greater convenience and understandability for human-computer interaction, offering a comprehensive and structured information foundation for subsequent AI decision-making, ensuring that the decision results meet user needs. By setting parameters through the lighting and shadow environment setting unit, it ensures that the obtained textures conform to the physical logic of the scene, enhancing the rationality and immersiveness of the rendering results.

[0025] In some embodiments, the user instruction processing unit includes: A fuzzy instruction parser is used to convert natural language or imprecise descriptions into specific texture adjustment parameters; A style instruction parser is used to map abstract art style descriptions to corresponding texture rendering parameter sets.

[0026] Fuzzy instruction parsers and style instruction parsers essentially act as bridges mapping the natural language space to the texture parameter space, greatly improving the system's ease of use and expressiveness.

[0027] In some embodiments, the lighting environment setting unit includes: The time parameter processor is used to calculate the sun angle, light intensity, and color temperature parameters based on the time of day. The weather parameter analyzer is used to calculate the ambient light shading coefficient based on the weather type and generate texture adjustment parameters.

[0028] The time parameter processor and weather parameter analyzer establish a physical causal relationship model between environmental conditions and rendering parameters, making the system's adaptive rendering based on evidence rather than simply looking up tables, resulting in more realistic textures.

[0029] Explanation of user commands: Explicit instructions with explicit texture parameters, for example: a. Texture resolution (512x512, 1024x1024, 2048x2048, etc.); b. Texture type (diffuse map, normal map, roughness map, etc.); c. Specific numerical parameters (e.g., roughness 0.8, metallicity 0.2, etc.).

[0030] Fuzzy instructions based on fuzzy descriptions in natural language, for example: a. "Make the surface rougher"; b. "Add some wear and tear"; c. "Make it look like it's been used for many years".

[0031] Style instructions with artistic style descriptions include, for example: a. Realistic style, cartoon style, watercolor style; b. Sci-fi feel, retro feel, futuristic feel; c. Specific art styles (such as cyberpunk, steampunk).

[0032] Instructions regarding lighting and shadow settings: 1) Time parameter processor: a. Analyze the time of day (0-24 hours); b. Calculate the sun's angle and light intensity; c. Generate time-dependent color temperature parameters.

[0033] 2) Weather parameter analyzer: a. Identify weather types (sunny, cloudy, rainy, snowy, etc.); b. Calculate ambient light occlusion coefficient; c. Generate weather-related texture adjustment parameters.

[0034] 3) Lighting parameter calculator: a. Direction and intensity of the main light source; b. Ambient light intensity; c. Contribution of indirect lighting.

[0035] 4) Model Environment Adapter: a. Scene type recognition (indoor / outdoor); b. Reflection effects of surrounding objects; c. Atmospheric scattering effect parameters.

[0036] Time parameter processor: Assuming the system time is a virtual "3 pm", the calculated solar altitude angle is 35 degrees. However, since the weather is cloudy, the actual dominant light is ambient diffuse light.

[0037] Weather parameter analyzer: Based on the user's instruction of "cloudy and humid", the ambient light shading coefficient is calculated to be 0.7 (dark) and the ambient light color temperature is 6500K (cool tone).

[0038] Output: Environment parameter vector In some implementations, the AI ​​texture calculation decision module includes: The intent parsing module uses deep learning-based technology to understand the semantics of user commands and classifies and maps abstract intents into specific rendering parameters; The texture decision module employs an intelligent decision-making system based on reinforcement learning, which selects the optimal texture adjustment action from a preset action space based on the current texture state, environmental parameters, and user history.

[0039] The intent parsing module employs deep learning technology, which excels at extracting deep semantic features from complex, unstructured inputs (such as natural language), achieving an "understanding" of user intent rather than merely a "match." The texture decision module utilizes reinforcement learning technology, endowing the system with the ability to make sequential decisions and perform long-term optimizations under complex constraints, thus formulating a comprehensive and optimal adjustment strategy.

[0040] The intent parsing module uses deep learning technology to accurately understand user intent: 1) Natural Language Processing Engine: Pre-trained model based on Transformer architecture.

[0041] 2) Multimodal fusion network: Unifies the encoding of text instructions and environmental parameters.

[0042] 3) Intent classifier: Identifies user intent categories (adjustment, replacement, mixing, etc.).

[0043] 4) Parameter mapper: Converts abstract intentions into specific rendering parameters.

[0044] Texture decision module, an intelligent decision-making system based on reinforcement learning: 1) State space definition: current texture state, environment parameters, user history.

[0045] 2) Motion space design: Texture adjustment motion set.

[0046] 3) Reward function: based on rendering quality and user satisfaction.

[0047] 4) Decision network: Deep Q-network (DQN).

[0048] In some implementations, the intent parsing module uses a pre-trained model for natural language processing and encodes textual instructions and environmental parameters uniformly through a multimodal fusion network to form a combined instruction. The user_intent dictionary from the user instruction processing unit and the env_params vector from the environment awareness unit are fused using multimodal methods. Specifically, they are projected onto a feature space of the same dimension through fully connected layers, and then the features are concatenated.

[0049] By employing pre-trained models for natural language processing (NLP), prior knowledge learned from large-scale corpora is utilized, ensuring the depth and breadth of semantic understanding.

[0050] And / or, the texture decision module is based on a reinforcement learning framework, and its reward function is configured to simultaneously optimize rendering quality, user satisfaction, and system performance overhead.

[0051] By explicitly incorporating user satisfaction and system performance overhead into the reward function of reinforcement learning, the AI ​​agent not only pursues the ultimate visual effects during the learning process, but also actively considers real-time engineering constraints and user experience, preventing unrealistic decisions and demonstrating the system's practicality and intelligence level.

[0052] In some implementations, the texture state and application rendering module includes: Texture State Manager, used to manage smooth transitions, caching, and version control of texture states; Texture rendering parameter controller, used to dynamically update physically based rendering parameters and bind parameters to GPU shaders; The particle system linkage module is used to trigger and coordinate corresponding particle effects when the texture state changes.

[0053] The texture state manager manages smooth transitions and version control, avoiding visual flickering caused by parameter mutations and supporting "undo" operations, thus improving the user experience. The transition algorithm uses non-linear interpolation (ease-in-out curves) to interpolate scalar parameters such as roughness and specular intensity. For texture maps, alpha blending is used to create gradients between two sets of maps. The texture rendering parameter controller ensures seamless and efficient integration with underlying graphics APIs (such as GPU shaders). The introduction of the particle system linkage module enables intelligent linkage between texture changes and dynamic visual effects (such as rust particles falling when metal rusts), upgrading from "static materials" to "dynamic scenarios," greatly enhancing visual expressiveness and realism.

[0054] As attached Figure 2 As shown, a multi-source data-driven 3D model texture rendering method is also provided, including the following steps: It receives and integrates multi-source input data from user commands and the lighting environment; Based on artificial intelligence technology, intent parsing and texture decision-making are performed on the fused multi-source input data to generate texture control instructions; Based on the texture control instructions, the texture state of the 3D model is dynamically managed and rendered in real time.

[0055] This method's intelligent rendering process consists of three key steps, clearly defining the data flow and processing logic. This provides clear guidance for software implementation and ensures that even with different programming languages ​​or hardware platforms, as long as the method's flow is followed, it falls within the scope of this invention. This solution can understand and execute various user commands, ranging from precise technical parameters to abstract artistic descriptions, greatly reducing the operational threshold for non-professional users and improving the naturalness and efficiency of interaction. It establishes an intelligent correlation between texture representation and environmental factors, achieving intelligent fusion and coordination of multi-source data.

[0056] This solution employs multi-source data intelligent fusion technology, achieving for the first time a deep fusion of user commands (explicit / fuzzy / style) and environmental parameters (time / weather / lighting / scene), thus overcoming the limitations of single input in traditional systems.

[0057] Multidimensional data fusion algorithms and weight calculation methods: The weights include visual quality score, predicted user satisfaction value, performance overhead, etc., and the weight coefficients are preset according to requirements. For example: Parameter fusion result = (Visual Quality Score)+ (Predicted User Satisfaction Value) (Performance overhead), where the values ​​of w1, w2 and w3 are all between 0.1 and 0.9.

[0058] In some implementations, the step of receiving and fusing multi-source input data includes: Parse user instructions, which include at least one of the following: explicit instructions with explicit texture parameters, fuzzy instructions with fuzzy descriptions based on natural language, and style instructions with artistic style descriptions; Acquire and process current lighting and shadow environment parameters, including at least one of time, weather, lighting, and scene environment information.

[0059] This approach ensures the practicality and flexibility of the method, enabling it to handle complex input scenarios in real-world applications.

[0060] In some implementations, the step of parsing the user instruction includes: Use a pre-trained natural language processing model to extract semantic features of user commands; The extracted semantic features are converted into specific texture adjustment parameters through a pre-established "natural language-texture attribute" mapping knowledge base.

[0061] By explicitly using a pre-trained NLP model and a "natural language-texture attribute" mapping knowledge base, a feasible solution for concretizing fuzzy instructions is presented.

[0062] It adopts natural language-driven texture control, uses NLP technology to understand fuzzy instructions and artistic style descriptions, greatly reduces the user's operating threshold, and supports natural expressions such as "make the surface rougher".

[0063] Methods for mapping natural language to texture parameters, for example: Build the database: Prepare a large dataset of paired texture images and text descriptions.

[0064] Feature extraction: Use a pre-trained text encoder (such as BERT) to convert text descriptions into vectors; use a pre-trained image encoder (such as ResNet) to convert texture images into vectors.

[0065] Alignment space: The model is trained to make the distance between paired (text vectors, image vectors) in the vector space as close as possible.

[0066] Retrieval: When a user enters a new description, it is encoded as a vector, and then the nearest vector is found in the texture vector library, returning the corresponding texture image.

[0067] In some implementations, the steps of performing intent parsing and texture decision-making based on artificial intelligence technology include: A deep learning model is used to parse user intent, identify intent categories, and map them to rendering parameters. Reinforcement learning models are used for texture decision-making. The state space of reinforcement learning includes the current texture state, target intent, and environmental parameters, while the action space includes texture selection, parameter adjustment, and blending mode operations.

[0068] By defining the state space and action space of reinforcement learning, the information on which the AI ​​agent bases its decisions and the actions it can take are clearly outlined, demonstrating the method's advanced nature and feasibility.

[0069] This case utilizes an AI dual-model decision-making architecture, combining deep learning (intent understanding) and reinforcement learning (decision optimization), to both understand user intent and make optimal decisions.

[0070] In some implementations, the step of dynamically managing and rendering the texture state in real time includes: Manage smooth transitions in texture states and support state rollback; When the texture state changes, the particle system is controlled in conjunction to generate corresponding visual effects; The level of detail in the texture is dynamically adjusted based on the viewing distance.

[0071] Managing smooth transitions and supporting state rollback enhances the fluidity and fault tolerance of the interaction. The interconnected particle system enriches the visual presentation. Dynamically adjusting the texture detail level based on view distance is a key performance optimization measure. It ensures that the system can intelligently allocate computing resources while maintaining visual quality, focusing limited computing power on the near details that users care about most, thus guaranteeing overall smooth operation. This is a crucial guarantee that this method can be applied to real-time rendering scenarios.

[0072] The specific workflow principle is as follows: Figures 3 to 12 As shown below, a detailed explanation will follow.

[0073] As attached Figure 3 The diagram shown is a user command processing flowchart, which details the processing paths for explicit commands, ambiguous commands, and style commands.

[0074] For explicit input instructions, they can be directly parsed, and textures can be generated directly through parameter normalization texture control by verifying parameters, validating numerical ranges, and checking formats.

[0075] For fuzzy input commands, after parsing, they undergo NLP processing, word segmentation, and semantic understanding, followed by intent recognition, classifier classification, and confidence calculation to obtain the user's true intent. This is then processed through parameter standardization and fusion. For example, a pre-trained BERT model (trained on a large text corpus and fine-tuned on a dataset containing tens of thousands of "material description-rendering parameter" pairs) is used to encode the commands.

[0076] The model outputs key semantic labels and their confidence scores, for example: {“Degree of weathering”: 0.95 (extremely high), “Time span”: “300 years”, “Weather”: “Cloudy”, “Humidity”: “Damp”}.

[0077] For the input style instruction, after parsing, it searches in the preset style library, performs style database matching until a corresponding style is found, extracts style features such as hue, texture, and feel, and performs parameter standardization and fusion processing. It also queries the built-in "art style-parameter" mapping knowledge base, which is stored in JSON format, and outputs a structured intent description object, such as a Python dictionary.

[0078] Sometimes, user input commands may conflict with the actual texture. In such cases, it is necessary to determine the conflict. This can be done based on weighted average, fuzzy logic, or neural networks. For example, by setting system priorities, such as prioritizing user commands or physical realism, or by describing how to handle the conflict through a "conflict detection and resolution module".

[0079] As attached Figure 4As shown, after parsing user input commands, parameter conflict detection is performed. When conflicts exist, they are resolved through priority judgment and parameter reconciliation. Parameters without conflicts are generated into a unified command format, such as JSON structured data, and then output to the AI ​​texture calculation and decision module. At the same time, logs are saved to record user preferences, which are used as training data for user preference learning.

[0080] Figure 5 This is a schematic diagram of the time parameters of the present invention; Time parameter processing: Start time parameter input is 0-24 hours; Calculate the solar angle, including azimuth angle and altitude angle; Determine the time period, such as early morning / morning / noon / afternoon / dusk / night; Calculate the natural light color temperature, such as 2800K-6500K; Calculate the light intensity: 0-100%.

[0081] Figure 6 This is a schematic diagram of weather parameters for this invention; weather parameter processing: cloud cover calculation 0-100%; atmospheric scattering coefficient is Rayleigh scattering / Mie scattering; precipitation impact analysis, such as reflection / refraction changes; visibility calculation, such as fog effect attenuation coefficient.

[0082] Figure 7 This is a schematic diagram of the lighting parameters of the present invention; lighting parameter inputs include: direct lighting analysis, such as main light source settings; indirect lighting calculation of ambient light / reflected light, light source color correction, such as RGB / color temperature conversion; and shadow parameter settings, such as hardness / offset.

[0083] Figure 8 This is a schematic diagram of the model environment processing of the present invention. Model environment inputs include: indoor environment, involving indoor lighting models, such as window light and lamplight; outdoor environment, involving outdoor lighting models, such as skylight and ground reflected light; and special environment models, such as underwater, cave, and space.

[0084] Figure 9 This diagram illustrates the multi-parameter fusion calculation for lighting and shadow settings. The multi-parameter fusion calculation includes: weight coefficient calculation, PBR parameter generation (metallicity / roughness adjustment), ambient occlusion calculation, and obtaining global illumination parameters.

[0085] Then, the global illumination parameters are checked for reasonableness. If any abnormalities are found, abnormal parameter correction and physical constraint checks are performed. If no abnormalities are found, a lighting parameter set is generated.

[0086] The output lighting parameters set includes final lighting parameters (such as intensity / direction / color), shadow parameters (such as type / intensity / blur), environmental parameters (fog / scattering / reflection), and texture adjustment suggestions (brightness / contrast / saturation, etc.), and is then output to the AI ​​texture decision module.

[0087] Figure 10This is a schematic diagram of the intent parsing module. User commands are segmented and encoded into text, then interpreted using the BERT semantic understanding pre-trained model. A multi-label classification is performed using an intent classifier, and a confidence score (0-1) is calculated to obtain the intent with confidence. The parsed intent label and the extracted lighting and environmental parameters are then fused and imported into the texture decision module for decision-making.

[0088] Multidimensional features are extracted and combined into feature vectors, including: user intent features (128-dimensional vector, encoding the semantics of user commands); environment features (64-dimensional vector, representing environment parameters); context features (32-dimensional vector, incorporating historical context); and style features (64-dimensional vector, capturing artistic style preferences). These features are then fed into the texture decision module.

[0089] Figure 11 This is a schematic diagram of the texture decision module.

[0090] State space: Current texture state: Records the current rendering state parameters.

[0091] Decision network: Deep Q-network: Uses a 5-layer fully connected network and Experience Replay to handle discrete actions.

[0092] Policy Network: Employs the DDPG algorithm to handle the continuous action space.

[0093] Action Space: Generates three types of actions: Texture Selection: Selects a suitable texture from the knowledge base. Parameter Adjustment: Adjusts rendering parameters (such as metallicity and roughness). Blending: Combines multiple texture blending schemes.

[0094] Reward function: Evaluates the quality of the action and generates a reward signal.

[0095] Target state reasoning: Infer the target texture state based on actions and rewards.

[0096] The target texture state is evaluated and optimized to obtain better and more texture images. Optimization strategies include rendering quality scoring, performance cost scoring, and user preference matching. After optimization, multiple texture results can be generated for the user to choose from.

[0097] Figure 12 For the final texture decision, multiple texture results are selected, and the final selection is output as parameters for the application rendering module to perform rendering operations. The output parameters include a list of texture IDs, a set of rendering parameters, a transition animation scheme, and a decision confidence level.

[0098] We will demonstrate how the system works by focusing on a specific user command input.

[0099] For example, a user inputs the command: "Make this metal surface look older, with a steampunk feel."

[0100] analyze: I. User Input: 1. User natural language commands: "Make this metal surface look older, with a bit of a steampunk feel." 2. Environmental parameter input: Time: 18:30 (dusk); Weather: Cloudy; Scene type: Outdoor.

[0101] II. System Processing Flow 1. User instruction processing unit Fuzzy instruction parser recognition: "Older" → Increases wear and rust, reduces metallicity, and increases roughness.

[0102] Style instruction interpreter recognition: "Steampunk" → patina, rivets, mechanical textures, dark tones, vintage metal.

[0103] Define the command interface: Texture types: diffuse map, normal map, metallic map, roughness map.

[0104] Resolution: 1024x1024.

[0105] 2. Lighting and Shadow Environment Setting Unit Time parameter processor: 18:30 → The sun is at a low angle, and the color temperature is warm (orange-red tone).

[0106] Weather parameter analyzer: Cloudy conditions increase the ambient light shading factor and reduce overall brightness.

[0107] Lighting parameter calculator: The main light source is positioned low, resulting in soft ambient light.

[0108] Model environment adapter: Outdoors → There may be reflections and atmospheric scattering → which increases the ambiguity of surface reflections.

[0109] 3. AI Texture Calculation Decision Module Intent parsing module: The Transformer model is used to fuse the "older" and "steampunk" elements with environmental parameters.

[0110] The intent classifier identified it as "style replacement + surface aging".

[0111] Parameter mapper output: Metallicity: 0.2; Roughness: 0.85; The color is a dark reddish-copper. Add rust and scratch layers.

[0112] Texture Decision Network (DQN): Current status: Default metal material.

[0113] Action selection: Add rust layer, adjust PBR parameters, replace texture style.

[0114] Reward function: Predicts user satisfaction (based on historical feedback or a pre-trained model).

[0115] Output action sequence: Generate new texture parameters and call render update.

[0116] 4. Texture State and Application Rendering Module Texture State Manager: Save the current state → Create a new state → Smooth transition.

[0117] Model texture map library: Search for "steampunk" style textures → synthesize rust + copper textures.

[0118] Texture rendering parameter controller: Update PBR parameters (metallicity, roughness, AO, etc.).

[0119] Bind GPU shader parameters.

[0120] Application rendering engine interface: Apply the new texture to the model → Real-time rendering.

[0121] Particle system linkage module (optional): If the user enables → Add steam jet particle effect.

[0122] Material parameter rendering updater: Real-time interpolation transition → Smooth transition from the original material to the new material.

[0123] III. Output Results The model surface exhibits: Dark bronze tone, mottled rust, and obvious scratches; High surface roughness and low metallicity; Due to the influence of twilight light, the surface reflects a warm color.

[0124] The overall style fits the "steampunk + old metal + twilight environment" aesthetic.

[0125] System feedback: It displays "Steampunk style texture applied"; Offers "Cancel" and "Further Adjustment" options; It can be saved as texture version 1.1.

[0126] IV. Summary of Judgment Logic Fuzzy language → Intent parsing → Parameter mapping; Environmental parameters → Lighting / Color temperature / Occlusion → Rendering adjustment factors; AI decision-making → Selecting the optimal texture adjustment action; Rendering system → Real-time updates of texture and material parameters.

[0127] V. Visualization Effects (Add optional) Original material: Silver-white metal, smooth.

[0128] New material: copper rust color, rough, scratched, and rusted edges.

[0129] Lighting effects: warm-toned reflections, soft shadows.

[0130] Through the above specific embodiments, three-dimensional textures can be accurately generated based on natural language commands input by the user.

[0131] It should be noted that the specific embodiments described above are exemplary. Those skilled in the art can devise various solutions inspired by the disclosure of this invention, and these solutions all fall within the scope of this invention and its protection. Those skilled in the art should understand that this specification and its accompanying drawings are illustrative and do not constitute a limitation on the claims. The scope of protection of this invention is defined by the claims and their equivalents. This specification contains multiple inventive concepts; phrases such as "preferred" or "according to a preferred embodiment" indicate that the corresponding paragraph discloses an independent concept. The applicant reserves the right to file divisional applications based on each inventive concept. Throughout the text, the feature introduced by "preferred" is only an optional mode and should not be construed as mandatory. Therefore, the applicant reserves the right to abandon or delete relevant preferred features at any time.

Claims

1. A multi-source data-driven 3D model texture rendering system, characterized in that, include: The multi-source input parameter processing module is used to receive and process multi-source input data from user commands and the lighting environment; The AI ​​texture calculation and decision module is connected to the multi-source input parameter processing module and is used to analyze user intent and generate texture decisions based on the multi-source input data and artificial intelligence technology. The texture state and application rendering module is connected to the AI ​​texture calculation and decision module, and is used to dynamically manage and render the texture state of the 3D model in real time according to the texture decision.

2. The three-dimensional model texture rendering system according to claim 1, characterized in that, The multi-source input parameter processing module includes: The user instruction processing unit is configured to parse different types of user instructions, which include at least: explicit instructions with explicit texture parameters, fuzzy instructions with fuzzy descriptions based on natural language, and style instructions with artistic style descriptions. The lighting and environment setting unit is configured to handle parameters related to time, weather, lighting, and scene environment so that the texture representation adapts to environmental changes.

3. The three-dimensional model texture rendering system according to claim 1 or 2, characterized in that, The user instruction processing unit includes: A fuzzy instruction parser is used to convert natural language or imprecise descriptions into specific texture adjustment parameters; A style instruction parser is used to map abstract art style descriptions to corresponding texture rendering parameter sets; And / or, the lighting environment setting unit includes: The time parameter processor is used to calculate the sun angle, light intensity, and color temperature parameters based on the time of day. The weather parameter analyzer is used to calculate the ambient light shading coefficient based on the weather type and generate texture adjustment parameters.

4. The three-dimensional model texture rendering system according to any one of claims 1 to 3, characterized in that, The AI ​​texture calculation decision module includes: The intent parsing module uses deep learning-based technology to understand the semantics of user commands and classifies and maps abstract intents into specific rendering parameters; The texture decision module employs an intelligent decision-making system based on reinforcement learning, which selects the optimal texture adjustment action from a preset action space based on the current texture state, environmental parameters, and user history.

5. The three-dimensional model texture rendering system according to any one of claims 1 to 4, characterized in that, The intent parsing module uses a pre-trained model for natural language processing and encodes text instructions and environmental parameters in a unified manner through a multimodal fusion network. And / or, the texture decision module is based on a reinforcement learning framework, and its reward function is configured to simultaneously optimize rendering quality, user satisfaction, and system performance overhead.

6. The three-dimensional model texture rendering system according to any one of claims 1 to 5, characterized in that, The texture state and application rendering module includes: Texture State Manager, used to manage smooth transitions, caching, and version control of texture states; Texture rendering parameter controller, used to dynamically update physically based rendering parameters and bind parameters to GPU shaders; The particle system linkage module is used to trigger and coordinate corresponding particle effects when the texture state changes.

7. A multi-source data-driven method for rendering textures in 3D models, characterized in that, Includes the following steps: Receive and fuse multi-source input data from user commands and the lighting environment; Based on artificial intelligence technology, intent parsing and texture decision-making are performed on the fused multi-source input data to generate texture control instructions; Based on the texture control instructions, the texture state of the 3D model is dynamically managed and rendered in real time.

8. The three-dimensional model texture rendering method according to claim 7, characterized in that, The steps of receiving and fusing multi-source input data include: Parse user instructions, which include at least one of the following: explicit instructions with explicit texture parameters, fuzzy instructions with fuzzy descriptions based on natural language, and style instructions with artistic style descriptions; Acquire and process current lighting and shadow environment parameters, including at least one of time, weather, lighting, and scene environment information.

9. The three-dimensional model texture rendering method according to claim 7 or 8, characterized in that, The steps for parsing user instructions include: Use a pre-trained natural language processing model to extract semantic features of user commands; The extracted semantic features are converted into specific texture adjustment parameters through a pre-established "natural language-texture attribute" mapping knowledge base; And / or, The steps for intent parsing and texture decision-making based on artificial intelligence technology include: A deep learning model is used to parse user intent, identify intent categories, and map them to rendering parameters. Reinforcement learning models are used for texture decision-making. The state space of reinforcement learning includes the current texture state, target intent, and environmental parameters, while the action space includes texture selection, parameter adjustment, and blending mode operations.

10. The three-dimensional model texture rendering method according to any one of claims 7 to 9, characterized in that, The steps for dynamically managing and rendering texture states in real time include: Manage smooth transitions in texture states and support state rollback; When the texture state changes, the particle system is controlled in conjunction to generate corresponding visual effects; The level of detail in the texture is dynamically adjusted based on the viewing distance.

Images