An instant light color rendering system for a game scene

Abstract

The application belongs to the technical field of game rendering, and discloses a kind of instant light color rendering system for game scene, comprising: scene illumination information acquisition based on game engine, generate initial illumination data;Heterogeneous light source data fusion processing, generate integrated illumination data;Multi-dimensional illumination attribute clustering analysis and feature extraction, generate illumination feature data;Scene region division processing, construct regional illumination feature data;Regional material-illumination interaction feature analysis, obtain interaction feature data;Illumination propagation feature analysis, generate propagation feature data;Design area rendering priority prediction engine;Real-time transmission feature data for rendering requirement intelligent analysis;Global instant light color rendering adaptive decision analysis, form rendering decision;According to the decision adjustment initial illumination data, execute the instant light color rendering of game scene;Greatly improve the visual performance and rendering performance of game.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of game rendering, more particularly, the present application relates to a real-time light color rendering system for game scenes. BACKGROUND

[0002] The existing real-time light color rendering system for game scenes has significant deficiencies in handling multiple dynamic light source changes and accurate rendering of visual focus areas in open world environments, especially in handling sunrise and sunset transitions, extreme weather event transitions, or scenes with complex interactions between light sources and media. Traditional rendering methods cannot effectively capture the subtle gradual changes in light intensity and color. Existing technologies usually use global uniform rendering parameters, which cannot be differentiated according to the importance and particularity of the scene area, resulting in unreasonable resource allocation, missing details in key visual areas, or excessive consumption of computing resources in non-key areas. Existing technologies lack intelligent correlation analysis capabilities for light propagation characteristics and material properties, and cannot make adaptive rendering decisions based on the unique response characteristics of different materials to light, making it difficult to achieve a good balance between rendering quality and performance when handling complex scenes with highly heterogeneous material distribution. Especially in high-end games that require accurate simulation of light propagation effects in multiple media (such as semi-transparent fabrics, multi-level water bodies, and volumetric smoke), traditional rendering systems cannot meet the dual requirements of physical accuracy and visual expressiveness.

[0003] In view of the above, the present application proposes a real-time light color rendering system for game scenes to solve the above problems. SUMMARY

[0004] In order to overcome the above-mentioned defects of the prior art, in order to achieve the above-mentioned purpose, the present application provides the following technical scheme: a real-time light color rendering system for game scenes, comprising:

[0005] The acquisition processing module: based on the game engine, the scene lighting information is collected to generate initial lighting data; according to the initial lighting data, the heterogeneous light source data fusion processing is carried out to generate integrated lighting data;

[0006] The data extraction module: according to the integrated lighting data, multi-dimensional lighting attribute clustering analysis is carried out to generate multi-dimensional lighting attribute clustering data; according to the multi-dimensional lighting attribute clustering data, multi-dimensional lighting feature extraction is carried out to generate multi-dimensional lighting feature data;

[0007] The analysis module: according to the multi-dimensional lighting feature data, the scene area division processing is carried out to generate regional multi-dimensional lighting feature data; based on the regional multi-dimensional lighting feature data, the regional material-light interaction feature analysis is carried out to generate regional material-light interaction feature data;

[0008] Real-time rendering module: This module performs lighting propagation feature analysis based on multi-dimensional regional lighting characteristic data to generate lighting propagation feature data; it designs an intelligent judgment engine based on regional material-lighting interaction feature data to predict regional rendering priorities, resulting in a regional rendering priority prediction engine; it transmits the lighting propagation feature data to the regional rendering priority prediction engine in real time for intelligent analysis of regional dynamic rendering needs, generating regional dynamic rendering requirement data; it performs adaptive decision analysis for global real-time lighting and color rendering based on the regional dynamic rendering requirement data, generating global real-time lighting and color rendering decisions; it adjusts the initial lighting data for real-time lighting and color rendering based on the global real-time lighting and color rendering decisions, generating real-time lighting and color rendering data, and then executes real-time lighting and color rendering of the game scene using this data.

[0009] Furthermore, the process of collecting scene lighting information based on the game engine to generate initial lighting data, and then performing heterogeneous light source data fusion processing based on the initial lighting data to generate integrated lighting data, includes:

[0010] Scene lighting information is collected based on the game engine to generate initial lighting data;

[0011] Based on the initial illumination data, perform light source type identification processing to generate light source type feature data;

[0012] Design a heterogeneous light source representation model based on light source type characteristic data, and generate a heterogeneous light source representation model;

[0013] Heterogeneous light source data is fused based on the heterogeneous light source representation model to generate integrated illumination data.

[0014] Further, the step of performing multi-dimensional illumination attribute clustering analysis based on integrated illumination data to generate multi-dimensional illumination attribute clustering data; and performing multi-dimensional illumination feature extraction based on the multi-dimensional illumination attribute clustering data to generate multi-dimensional illumination feature data, includes:

[0015] Illumination intensity features are extracted from the integrated illumination data to generate illumination intensity feature data.

[0016] Spectral distribution features are extracted from integrated illumination data to generate spectral distribution feature data.

[0017] The initial multidimensional light attribute clustering analysis is performed on the light intensity feature data and spectral distribution feature data using a preset multidimensional clustering algorithm to generate initial multidimensional light attribute clustering data.

[0018] Visually irrelevant lighting data is filtered out based on the initial multidimensional lighting attribute clustering data to generate multidimensional lighting attribute clustering data;

[0019] Multidimensional illumination feature data is generated by extracting multidimensional illumination features from clustered data of multidimensional illumination attributes.

[0020] Furthermore, the step of filtering out visually irrelevant lighting data based on the initial multi-dimensional lighting attribute clustering data to generate multi-dimensional lighting attribute clustering data includes:

[0021] Visual perception threshold analysis is performed based on the initial multi-dimensional illumination attribute clustering data to generate visual perception threshold data, which includes illumination intensity threshold data and spectral sensitivity data.

[0022] Low-impact light source aggregation processing is performed based on illuminance intensity threshold data to generate aggregated low-impact light source data.

[0023] Visual redundancy illumination data is generated by performing visual redundancy illumination recognition processing based on aggregated low-impact light source data and spectral sensitivity data.

[0024] Based on the visually redundant lighting data, the initial multi-dimensional lighting attribute clustering data is filtered out to generate multi-dimensional lighting attribute clustering data. The visually redundant lighting data is then transmitted to the rendering engine to perform a simplified rendering job.

[0025] Further, the process of dividing the scene into regions based on multi-dimensional illumination feature data to generate multi-dimensional illumination feature data for each region; and performing material-lighting interaction feature analysis on the region based on the multi-dimensional illumination feature data to generate material-lighting interaction feature data for each region, including:

[0026] The scene region is divided based on the multi-dimensional illumination feature data to generate multi-dimensional illumination feature data for the region.

[0027] Based on the multi-dimensional illumination feature data of the region, a regional material property mapping analysis is performed to generate regional material property data.

[0028] Based on the regional material property data, perform regional material-lighting interaction feature analysis to generate regional material-lighting interaction feature data.

[0029] Furthermore, the step of performing regional material characteristic mapping analysis based on regional multi-dimensional illumination feature data to generate regional material characteristic data includes:

[0030] Surface material reflection characteristics are analyzed based on multi-dimensional regional illumination feature data to generate surface material reflection characteristic data.

[0031] Surface material scattering characteristics are analyzed based on multi-dimensional regional illumination feature data to generate surface material scattering characteristic data.

[0032] The material-light interaction response matrix is ​​calculated based on the surface material reflection characteristics data and the surface material scattering characteristics data to generate the material-light interaction response matrix.

[0033] Based on the material-lighting interaction response matrix, regional material property mapping analysis is performed to generate regional material property data.

[0034] Furthermore, the step of performing regional material-lighting interaction feature analysis based on regional material characteristic data to generate regional material-lighting interaction feature data includes:

[0035] Perform physically based rendering model feature analysis on the regional material property data to generate physically based rendering model feature data;

[0036] Based on the physical rendering model feature data, region material-lighting interaction feature analysis is performed to generate region material-lighting interaction feature data.

[0037] Further, the process includes: analyzing light propagation characteristics based on multi-dimensional regional lighting feature data to generate lighting propagation feature data; designing an intelligent judgment engine for regional rendering priority prediction based on regional material-lighting interaction feature data to obtain a regional rendering priority prediction engine; transmitting the lighting propagation feature data to the regional rendering priority prediction engine in real time for intelligent analysis of regional dynamic rendering needs to generate regional dynamic rendering need data; performing adaptive decision analysis for global real-time lighting and color rendering based on the regional dynamic rendering need data to generate global real-time lighting and color rendering decisions; adjusting the initial lighting data for real-time lighting and color rendering based on the global real-time lighting and color rendering decisions to generate real-time lighting and color rendering data; and executing real-time lighting and color rendering of the game scene using the real-time lighting and color rendering data.

[0038] Light propagation characteristic analysis is performed based on multi-dimensional regional light characteristic data to generate light propagation characteristic data.

[0039] Visual importance scoring nodes are analyzed based on regional material-lighting interaction feature data to generate visual importance scoring node data, and decision tree priority nodes are designed based on the visual importance scoring node data.

[0040] Based on a pre-defined deep neural network model, a neural network rendering priority tree is designed to allocate lighting effects and rendering resources based on decision tree priority nodes and regional material-lighting interaction feature data, so as to obtain a rendering priority neural network model.

[0041] A smart judgment engine for predicting region rendering priority is designed based on a rendering priority neural network model to obtain a region rendering priority prediction engine.

[0042] The light propagation characteristic data is transmitted in real time to the region rendering priority prediction engine to perform intelligent analysis of region dynamic rendering requirements and generate region dynamic rendering requirement data.

[0043] Based on the regional dynamic rendering requirement data, perform adaptive decision analysis for global real-time lighting and color rendering, and generate global real-time lighting and color rendering decisions;

[0044] Based on global real-time lighting and color rendering decisions, the initial lighting data is adjusted and processed in real-time lighting and color rendering to generate real-time lighting and color rendering data, and then the real-time lighting and color rendering of the game scene is executed using the real-time lighting and color rendering data.

[0045] Furthermore, the neural network rendering priority tree design based on the preset deep neural network model for allocating lighting effects and rendering resources to decision tree priority nodes and region material-lighting interaction feature data, to obtain a rendering priority neural network model, includes:

[0046] Based on a pre-defined deep neural network model, a neural network rendering priority tree architecture is designed to allocate lighting effects and rendering resources to the priority nodes of the decision tree, generating a rendering priority neural network model architecture.

[0047] Based on the regional material-lighting interaction feature data, user visual attention prediction analysis is performed to generate user visual attention prediction data.

[0048] Based on user visual attention prediction data, the rendering priority neural network model architecture is processed by learning the priority feature weights of lighting effects and rendering resources to generate a rendering priority neural network model.

[0049] The technical effects and advantages of a real-time lighting and color rendering system for game scenes according to the present invention are as follows:

[0050] This invention establishes a heterogeneous light source representation model through lighting data acquisition and light source type identification, enabling unified management and coordinated rendering of multiple light source types. The system employs multi-dimensional lighting attribute clustering analysis and feature extraction to accurately capture changes in lighting intensity and spectral distribution within the scene. A visually irrelevant lighting data filtering mechanism effectively eliminates redundant lighting information with minimal impact on visual effects, significantly improving rendering efficiency. Based on scene region division and material-lighting interaction feature analysis, the system achieves precise lighting rendering for different region material characteristics and presents more realistic lighting effects through physically based rendering model analysis. A region rendering priority prediction engine analyzes lighting propagation characteristics in real time and intelligently allocates rendering resources using a deep neural network model, ensuring efficient processing capabilities during dynamic scene changes. Based on user visual attention prediction data and regional dynamic rendering needs analysis, the system generates adaptive rendering decisions, prioritizing visually important regions and effectively reducing computational resource consumption while maintaining high-quality visual performance. This adaptive real-time lighting and color rendering technology, designed for heterogeneous light source data and complex rendering requirements, achieves more efficient and realistic lighting effects in game scenes. Attached Figure Description

[0051] Figure 1 This is a schematic diagram of a real-time light and color rendering system for game scenes according to the present invention;

[0052] Figure 2 This is a schematic diagram of a real-time light and color rendering method for game scenes according to the present invention. Detailed Implementation

[0053] 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.

[0054] Example 1;

[0055] Please see Figure 1 As shown in this embodiment, a real-time lighting and color rendering system for game scenes includes: a data acquisition and processing module: acquiring scene lighting information based on the game engine to generate initial lighting data; and performing heterogeneous light source data fusion processing based on the initial lighting data to generate integrated lighting data.

[0056] Data extraction module: Performs multi-dimensional illumination attribute clustering analysis on the integrated illumination data to generate multi-dimensional illumination attribute clustering data; extracts multi-dimensional illumination features from the multi-dimensional illumination attribute clustering data to generate multi-dimensional illumination feature data;

[0057] Feature Analysis Module: Performs scene region segmentation based on multi-dimensional illumination feature data to generate multi-dimensional illumination feature data for each region; performs region material-lighting interaction feature analysis based on the region multi-dimensional illumination feature data to generate region material-lighting interaction feature data.

[0058] Real-time rendering module: This module performs lighting propagation feature analysis based on multi-dimensional regional lighting characteristic data to generate lighting propagation feature data; it designs an intelligent judgment engine based on regional material-lighting interaction feature data to predict regional rendering priorities, resulting in a regional rendering priority prediction engine; it transmits the lighting propagation feature data to the regional rendering priority prediction engine in real time for intelligent analysis of regional dynamic rendering needs, generating regional dynamic rendering requirement data; it performs adaptive decision analysis for global real-time lighting and color rendering based on the regional dynamic rendering requirement data, generating global real-time lighting and color rendering decisions; it adjusts the initial lighting data for real-time lighting and color rendering based on the global real-time lighting and color rendering decisions, generating real-time lighting and color rendering data, and then executes real-time lighting and color rendering of the game scene using this data.

[0059] Preferably, scene lighting information is collected based on the game engine to generate initial lighting data; heterogeneous light source data fusion processing is performed based on the initial lighting data to generate integrated lighting data, including:

[0060] (1) Collect scene lighting information based on the game engine and generate initial lighting data;

[0061] (2) Perform light source type identification processing based on the initial illumination data to generate light source type feature data;

[0062] (3) Design a heterogeneous light source representation model based on the light source type characteristic data, and generate a heterogeneous light source representation model;

[0063] (4) Perform heterogeneous light source data fusion processing based on the heterogeneous light source representation model to generate integrated illumination data.

[0064] In this embodiment of the invention, a lighting acquisition framework is deployed in the game engine, supporting six basic light source types: parallel light, point light, spotlight, area light, ambient light, and volumetric light. The acquisition framework captures lighting parameters each frame, including intensity (lumen value), RGB color values, attenuation coefficient, illumination range (and shadow resolution). The system acquires data before rendering each frame using the engine's UpdateLightData() function, with a sampling frequency of 60Hz to ensure real-time data transmission. The acquired data is stored in a 512KB circular buffer indexed by light source ID and timestamp, allowing for retrospective analysis of lighting changes over the last 240 frames to generate initial lighting data containing a complete parameter set. Light source type identification uses a feature matching algorithm, with a defined identification rule base: parallel light (without...) The system identifies various light sources, including those with attenuation and unidirectional emission, point light sources (omnidirectional attenuation, point emission), and spotlights (conical attenuation, angle limitation). The system executes a three-step identification process: parameter pattern matching, outlier detection, and type confirmation. For hybrid light sources, the dominant characteristic determination method is used to identify their primary type. The system sets parameter rationality thresholds: luminous intensity range 0 to 8000 lumens, color temperature range 1500K to 15000K, and parameters exceeding these ranges are pruned. The identification accuracy reaches 98.5%, outputting standardized light source type characteristic data. The heterogeneous light source representation model adopts a layered design: a basic illumination layer, a spatial layer, and a hierarchical layer. The model consists of a base layer and a time-varying characteristic layer. The base layer uses uniform photometric parameters (luminous flux, luminous intensity, illuminance) to describe all light sources; the spatial layer uses spherical harmonic functions (3rd order, 16 coefficients) to represent complex illumination distributions; and the time-varying layer uses spline interpolation curves to record the changes in illumination parameters over time. The model supports four decay functions: linear, inverse square, exponential, and a custom curve, all controllable through parameterization. Model validation tests were conducted in four scenarios (outdoor daylight, indoor mixed lighting, dynamic weather, and nighttime city), achieving a spatial accuracy deviation of less than 3% and a temporal consistency score of 4.8, generating a standardized heterogeneous light source representation model. The light source data fusion employs a partitioned processing strategy, dividing the scene into an 8×8×4 voxel grid. Within each voxel, the individual contributions of six light source types are calculated, and the physical lighting superposition principle (linear superposition + saturation control) is applied for synthesis. The fusion process includes direct lighting synthesis, indirect reflection calculation, and volumetric light scattering processing, with a computational precision of 16-bit floating-point. The system sets a lighting importance scoring metric, ignoring secondary light sources with contributions below 2% to improve performance. The fused lighting data is stored in the form of a lighting probe grid with a resolution of 32³, covering the entire scene space and generating globally consistent integrated lighting data.

[0065] Preferably, multi-dimensional illumination attribute clustering analysis is performed based on the integrated illumination data to generate multi-dimensional illumination attribute clustering data; multi-dimensional illumination feature extraction is performed based on the multi-dimensional illumination attribute clustering data to generate multi-dimensional illumination feature data, including:

[0066] (1) Extract light intensity features from integrated illumination data to generate light intensity feature data;

[0067] (2) Extract spectral distribution features from the integrated illumination data to generate spectral distribution feature data;

[0068] (3) Use the preset multidimensional clustering algorithm to perform initial multidimensional light attribute clustering analysis on the light intensity feature data and spectral distribution feature data to generate initial multidimensional light attribute clustering data;

[0069] (4) Visually irrelevant lighting data is filtered out based on the initial multidimensional lighting attribute clustering data to generate multidimensional lighting attribute clustering data;

[0070] (5) Extract multi-dimensional illumination features based on multi-dimensional illumination attribute clustering data to generate multi-dimensional illumination feature data.

[0071] In this embodiment of the invention, illumination intensity features are first extracted. The system employs a three-stage processing flow: parameter identification, numerical normalization, and distribution calculation. Four key intensity parameters are extracted from the integrated illumination data: peak illuminance (unit: lux), average luminance (unit: cd / m²), attenuation coefficients (α, β, γ parameters), and spatial distribution function. The system assigns standardized identifiers (DL_01, PL_03, and SL_02) to each light source and sets the effective range of the parameters (illuminance: 0-120000 lux, luminance: 0-10000 cd / m²). Spatial illumination distribution is recorded using a sampling grid (32×32×16 resolution), generating illumination intensity feature data containing a complete set of intensity parameters. Spectral distribution feature extraction employs color space conversion technology, converting RGB illumination data to the CIE-Lab color space for processing. The system extracts three key color features: color temperature parameters (1500K to 15000K, accuracy ±50K), hue angle (0° to 360°, accuracy ±0.5°), and color saturation (0 to 100%, accuracy ±1%). For high dynamic range light sources, color gamut coverage and color transition curves are additionally recorded. The system uses a four-point sampling method to analyze the color space distribution. Each sampling point records a complete color vector, and a continuous color distribution map is generated through cubic spline interpolation, outputting standardized spectral distribution feature data. The initial clustering analysis uses an improved K-means++ algorithm, combining light intensity and spectral distribution features to construct an 8-dimensional feature vector. The system presets the number of clusters K=7, and the initial cluster centers are selected using a maximum distance optimization strategy. The maximum number of iterations for clustering is set to 100, and the convergence threshold is 0.001. For mixed light sources, a fuzzy clustering coefficient (0-1.0) is introduced to represent the degree of belonging. The system evaluates the clustering results using the Davies-Bouldin index to ensure clustering quality. After clustering, each lighting group is assigned a unique ID (LC_01 to LC_07), generating initial multi-dimensional lighting attribute clustering data containing complete clustering information. Visually irrelevant lighting data is filtered using a triple-screening mechanism: brightness contribution assessment (threshold 1.5%), field-of-view coverage analysis (threshold 2%), and visual saliency calculation (threshold 0.15). The system performs ray tracing detection on occluded light sources and calculates the occlusion rate. The actual contribution of each light source to the final image is evaluated using a visual perception model, and light sources with a contribution below the threshold are marked as "low priority." The filtering process retains the light source data but reduces its sampling rate in rendering, generating simplified multi-dimensional lighting attribute clustering data. In the multi-dimensional lighting feature extraction stage, the system extracts five key features from the optimized clustering: lighting gradient map (16-bit precision), color transition matrix (4×4×4 resolution), shadow boundary feature map, specular distribution map, and ambient occlusion map. Feature extraction uses spherical harmonic functions (3rd order, 16 coefficients) to represent complex lighting distributions, generating complete multi-dimensional lighting feature data, which provides a foundation for subsequent scene region segmentation and rendering optimization.

[0072] Preferably, visually irrelevant illumination data is filtered out based on the initial multi-dimensional illumination attribute clustering data to generate multi-dimensional illumination attribute clustering data, including:

[0073] (1) Visual perception threshold analysis is performed based on the initial multidimensional illumination attribute clustering data to generate visual perception threshold data, wherein the visual perception threshold data includes illumination intensity threshold data and spectral sensitivity data.

[0074] (2) Perform low-impact light source aggregation processing based on the light intensity threshold data to generate aggregated low-impact light source data;

[0075] (3) Visual redundancy illumination recognition processing is performed based on aggregated low-impact light source data and spectral sensitivity data to generate visual redundancy illumination data;

[0076] (4) Based on the visual redundant lighting data, the initial multi-dimensional lighting attribute clustering data is filtered out to generate multi-dimensional lighting attribute clustering data, and the visual redundant lighting data is transmitted to the rendering engine to perform simplified rendering work.

[0077] In this embodiment of the invention, visual perception threshold analysis is performed first. Based on initial multi-dimensional illumination attribute clustering data, the system calculates the illumination perception threshold using an improved Weber-Fechner model. The processing flow includes two parts: brightness contribution assessment and spectral sensitivity analysis. Brightness contribution assessment uses a six-point sampling method, setting a 32×32 sampling grid within the main viewpoint of the scene, and calculating the illuminance contribution of each light source to the grid points. The system sets the brightness contribution threshold to 1.2%, and light sources with contributions below this value are marked as "low priority." Spectral sensitivity analysis uses the CIE1931 standard color observer function to map the human eye's sensitivity to different wavelengths, calculating the perception weight based on the color temperature and color gamut distribution of the light source. The system generates visual perception threshold data containing brightness threshold mapping and spectral sensitivity curves. Low-impact light source aggregation processing employs a spatial partitioning and characteristic grouping strategy. The system divides the scene into an 8×8×4 voxel grid and performs cluster analysis on low-impact light sources within each voxel. The aggregation conditions are set as follows: spatial distance less than 15% of the scene's unit length and illumination characteristic similarity greater than 85%. The system employs a hierarchical clustering algorithm to iteratively merge light sources that meet certain conditions. Each aggregated light source group is assigned a unique identifier (ALCG_001~ALCG_n) and its original parameters are recorded. Aggregation processing replaces multiple low-impact light sources with simplified light source models of equivalent parameters, generating aggregated low-impact light source data. Visual redundancy illumination identification uses an image quality assessment method. The system sets three key viewpoints, renders the scene from different angles, and compares the scene using full illumination and illumination after removing aggregated light source groups. The Structural Similarity Index (SSIM) is used to evaluate image differences, with a redundancy threshold of 0.96 (out of 1.0). Aggregated light source groups with an SSIM score higher than the threshold are marked as "visually redundant," and the system records their ID, location, impact range, and visual contribution, generating visually redundant illumination data. In the filtering stage, the system removes light source groups marked as visually redundant from the initial multi-dimensional illumination attribute clustering data, retaining their metadata for possible subsequent recovery. The system employs adaptive filtering strategies for different rendering scenes and viewpoints: weak light sources within the main field of view are downsampled (reduced to 25%); non-main field of view light sources are completely replaced with simplified lighting models; and distant light sources are replaced with pre-computed lightmaps. The processed data and visually redundant lighting data are sent to the rendering engine separately; the former is used for the standard rendering pipeline, and the latter for simplified rendering. The system achieves its goal of significantly improving rendering efficiency while maintaining visual quality.

[0078] Preferably, scene regions are divided based on multi-dimensional illumination feature data to generate multi-dimensional illumination feature data for each region; based on the multi-dimensional illumination feature data for each region, material-lighting interaction feature analysis is performed to generate material-lighting interaction feature data for each region, including:

[0079] (1) Based on the multi-dimensional illumination feature data, the scene region is divided and processed to generate multi-dimensional illumination feature data of the region;

[0080] (2) Perform regional material property mapping analysis based on regional multi-dimensional illumination feature data to generate regional material property data;

[0081] (3) Perform regional material-lighting interaction feature analysis based on regional material characteristic data to generate regional material-lighting interaction feature data.

[0082] In this embodiment of the invention, scene region segmentation is performed first. Based on multi-dimensional lighting feature data, the system uses an adaptive quadtree segmentation algorithm to hierarchically segment the scene. The segmentation process sets three main evaluation indicators: lighting gradient change rate (threshold 20% / meter), material complexity (threshold 0.65, full score 1.0), and visual importance (based on a center-periphery attenuation model). The system initially divides the scene into four basic regions: core interaction area, environment transition area, background area, and special effects area. Fine-grained segmentation (minimum grid size 2×2 meters) is used for areas with drastic lighting changes (gradient > 30% / meter) to ensure accurate capture of lighting boundaries; coarse-grained segmentation (maximum grid size 10×10 meters) is used for areas with uniform lighting to improve efficiency. Each region is assigned a unique identifier (e.g., RA_001 to RA_128) and its boundary data, lighting feature vector, and spatial location are recorded, generating multi-dimensional lighting feature data for the region containing complete spatial partitioning information. The region material characteristic mapping analysis uses physically based rendering parameter extraction technology. The system parameterizes the material characteristics within each defined region, extracting four key parameters: base color / albedo (RGB value precision 16 bits), metallicity (0 to 1.0, precision 0.01), roughness (0 to 1.0, precision 0.01), and subsurface scattering coefficient (0 to 1.0, precision 0.05). The system sets the sampling density to 4 points per square meter for each region, increasing to 16 points per square meter for highly complex regions (such as crystal chandelier areas). Material characteristic clustering employs the K-Means++ algorithm (K=8), grouping similar materials (similarity greater than 92%) within the same region into the same category. The system establishes separate high-precision parameter models for special materials (such as translucent glass and metallic reflective surfaces), generating regional material characteristic data containing material microstructure and classification information. Material-light interaction feature analysis utilizes simplified ray tracing and BRDF evaluation methods. The system sets the interaction sampling rate for each material category to 1024 samples / square meter, recording key interaction parameters: direct reflection intensity (0 to 1.0), specular concentration (0 to 1.0), and diffuse reflection coefficient (0 to 1.0). The analysis considers three typical lighting conditions: high-intensity directional light (greater than 7500 lumens), medium-intensity ambient light (2000 to 5000 lumens), and a combination of low-intensity point light sources (less than 1500 lumens). The system calculates the lighting response curves for each material at different incident angles (0°, 45°, 75°) and evaluates the viewpoint dependence. Interaction features are recorded with 32-bit floating-point precision, including a complete angle-response mapping table, forming regional material-lighting interaction feature data, providing a scientific basis for subsequent rendering optimization.

[0083] Preferably, regional material property mapping analysis is performed based on multi-dimensional illumination characteristic data of the region to generate regional material property data, including:

[0084] (1)Analyze the surface material reflection characteristics based on the regional multi-dimensional light characteristic data to generate surface material reflection characteristic data;

[0085] (2)Analyze the surface material scattering characteristics based on the regional multi-dimensional light characteristic data to generate surface material scattering characteristic data;

[0086] (3)Calculate the material-light interaction response matrix based on the surface material reflection characteristic data and the surface material scattering characteristic data to generate the material-light interaction response matrix;

[0087] (4)Perform regional material characteristic mapping analysis based on the material-light interaction response matrix to generate regional material characteristic data.

[0088] In the embodiment of the present invention, first, extract the light parameters from the regional multi-dimensional light characteristic data for surface material reflection characteristic analysis. The system uses a four-parameter reflection model, including the incident angle (0° to 90°, accuracy 0.5°), light intensity (0 to 10000 lumens, accuracy 5 lumens), color temperature (1500K to 10000K, accuracy 50K), and light source size (point light source to surface light source, radius 0 to 5 meters). The improved Cook-Torrance micro-surface model is used in the analysis process, and four types of reflection type identifiers are set: high specular reflection (SR>0.75), medium specular reflection (0.4<SR≤0.75), mixed reflection (0.2<SR≤0.4), and high diffuse reflection (SR≤0.2), where SR is the specular reflection ratio. The system constructs an incident-outgoing angle mapping table (resolution 18×18), records the BRDF response curves of various materials, sets the material analysis grid density to 9 sampling points per square meter, and enhances the key surfaces to 25 points / square meter to generate standardized surface material reflection characteristic data. The surface material scattering characteristic analysis uses the volume light transport equation, and the system sets the scattering analysis parameters: scattering coefficient σs (0 to 20, accuracy 0.1), absorption coefficient σa (0 to 10, accuracy 0.1), and phase function g (-0.9 to 0.9, accuracy 0.05). The analysis process includes light penetration depth measurement and scattering direction distribution calculation. The system applies a 16-direction scattering detector to translucent materials (such as wax, skin, cloud), measures the light propagation characteristics within the material, and sets the scattering sampling rate to 2048 samples / material type. Material classification uses the scattering threshold criterion: high scattering material (σs>8), medium scattering material (3<σs≤8), and low scattering material (σs≤3). The system records the scattering curves and depth distribution maps of each material to generate surface material scattering characteristic data. The material-light interaction response matrix calculation integrates the reflection and scattering characteristic data. The system constructs an M×L-dimensional response matrix, where M represents the number of material types (set to 12 types), and L represents the number of light condition types (set to 8 types). The matrix calculation formula is where and These are the weighting coefficients. and These are the normalized reflection and scattering response values. The system applies additional weighting factors for special visual effects (such as metallic highlights and glass refraction). The response threshold is set to 0.65 (out of 1.0), and material-lighting combinations exceeding this threshold are marked as "high-priority rendering objects." Matrix calculations are GPU-accelerated, with processing time controlled within 5ms, generating a complete material-lighting interaction response matrix. Regional material characteristic mapping analysis is based on the interaction response matrix for material classification and optimization. The system employs a four-dimensional optical characteristic scoring system: visual saliency (0 to 1.0), computational complexity (1 to 10), surface detail fidelity (0 to 1.0), and lighting sensitivity (0 to 1.0). Material classification uses a hierarchical clustering algorithm, setting the maximum number of categories to 8 and the minimum intra-class similarity to 0.8. The system assigns rendering processing strategies to each material category: high-priority materials use high-precision shaders; medium-priority materials use the standard PBR pipeline; and low-priority materials use a simplified lighting model. Rendering parameter adjustments are based on feature mapping results. For example, the reflection sampling rate of high-mirror metal surfaces is increased to 16x, while the standard diffuse surface maintains a 4x sampling rate, generating optimized regional material characteristic data and providing a scientific basis for resource allocation in the rendering system.

[0089] Preferably, based on the regional material characteristic data, regional material-lighting interaction feature analysis is performed to generate regional material-lighting interaction feature data, including:

[0090] (1) Perform physical-based rendering model feature analysis on the regional material property data to generate physical-based rendering model feature data;

[0091] (2) Perform regional material-lighting interaction feature analysis based on the physical rendering model feature data to generate regional material-lighting interaction feature data.

[0092] In this embodiment of the invention, regional material property data is applied to a physically based rendering model (PBR) to analyze parameters such as the material's BRDF, refractive index, metallicity, and roughness. The system uses the Cook-Torrance micro-surface model or DisneyBRDF to physically model the material properties, identify differences from the ideal model, and generate physically based rendering model feature data. Subsequently, ray tracing algorithms and radiometric techniques are used to simulate the interaction process between the material and lighting, ultimately generating regional material-lighting interaction feature data that accurately expresses the visual performance of the material under different lighting conditions.

[0093] Preferably, the real-time rendering module includes the following steps:

[0094] (1) Perform light propagation characteristic analysis based on multi-dimensional regional light characteristic data to generate light propagation characteristic data;

[0095] (2) Based on the regional material-lighting interaction feature data, visual importance scoring nodes are analyzed to generate visual importance scoring node data, and decision tree priority nodes are designed based on the visual importance scoring node data.

[0096] (3) Based on the preset deep neural network model, a neural network rendering priority tree is designed for the priority nodes of the decision tree and the regional material-lighting interaction feature data to allocate lighting effects and rendering resources, so as to obtain the rendering priority neural network model.

[0097] (4) Design an intelligent judgment engine for predicting region rendering priority based on the rendering priority neural network model, so as to obtain the region rendering priority prediction engine.

[0098] (5) Transmit the light propagation characteristic data to the region rendering priority prediction engine in real time to perform intelligent analysis of region dynamic rendering requirements and generate region dynamic rendering requirement data.

[0099] (6) Perform adaptive decision analysis for global real-time lighting and color rendering based on regional dynamic rendering requirement data, and generate global real-time lighting and color rendering decisions;

[0100] (7) Adjust the initial lighting data according to the global real-time lighting and color rendering decision, generate real-time lighting and color rendering data, and execute the real-time lighting and color rendering of the game scene through the real-time lighting and color rendering data.

[0101] In this embodiment of the invention, the system first performs light propagation feature analysis on the multi-dimensional lighting feature data of the region. The system extracts key parameters such as light angle, intensity, and color temperature from the raw data and uses an optimized bidirectional path tracing algorithm to analyze the propagation behavior of light in the scene. The light sampling density is set to 32 sampling points per square meter, the maximum path tracing depth is 5 bounces, and the lighting propagation accuracy threshold is 4%. The system records the reflection, scattering, and absorption coefficients, generating 16-bit precision light density distribution maps, shadow projection maps, and indirect lighting contribution maps, forming a complete lighting propagation feature data matrix. Visual importance scoring node analysis is based on the region's material-lighting interaction feature data. The system uses an improved visual saliency algorithm, combining four key indicators: brightness contrast (weight 0.35), specular quality (weight 0.25), texture complexity (weight 0.2), and gaze attention (weight 0.2) for comprehensive scoring. The scoring threshold is set to 0.7 (out of 1.0), and regions exceeding this threshold are marked as high-priority rendering regions. The system constructs a four-layer decision tree structure with 32 leaf nodes and a minimum split sample size of 10. Information gain ratio is used as the splitting criterion to generate priority nodes in the decision tree, containing the priority states of each region in the scene. The neural network rendering priority tree design uses a pre-defined deep convolutional neural network model. The network structure includes four convolutional layers (number of filters: 48, 96, 192, and 256) and three fully connected layers (number of nodes: 512, 256, and 64), using the LeakyReLU activation function (α=0.15) and batch normalization layers. The training process uses the Adam optimizer (β1=0.9, β2=0.999), an initial learning rate of 0.008, and a batch size of 96. The model is trained for 150 epochs on a 200,000-sample set, using an early stopping strategy (patience=12) to avoid overfitting. Validation uses SSIM (threshold ≥0.85) and performance evaluation (target frame rate ≥60fps) to generate the rendering priority neural network model. The intelligent decision-making engine adopts a three-stage pipeline architecture, including a feature analyzer, a priority predictor, and a resource scheduler. The engine has a built-in high-speed cache (64MB), keeping real-time processing latency within 8ms. Optimized with SIMD instructions, the engine can process up to 120 scene regions per frame. Its stability was tested and verified in four typical scenarios (daytime outdoor, nighttime indoor, underwater, and areas with dense special effects), generating a region rendering priority prediction engine. Light propagation feature data is sent to the prediction engine in real-time via an asynchronous transmission channel, employing an incremental update strategy that transmits only changed data, achieving a compression ratio of 4:1. The prediction engine applies a pre-trained model to the input data, allocating rendering resource percentages and processing priorities to each scene region, generating dynamic rendering requirement data for each region. Global real-time lighting and color rendering decision analysis uses a hierarchical optimization algorithm, integrating the requirements of each region and considering performance budgets (target frame rate, video memory limitations).The system defines three levels of rendering quality, dynamically balancing visual quality and performance. The decision-making process updates 15 times per second, generating global real-time lighting and color rendering decisions. During the rendering adjustment phase, the system dynamically adjusts light source parameters, material response characteristics, and sampling rate allocation based on the decision data. Visually important areas receive high-quality processing (8x sampling rate), secondary areas use standard processing (4x), and background areas use simplified processing (2x). Finally, rendering is performed through an efficient shader pipeline, achieving a balance between excellent visual effects and high performance.

[0102] Preferably, a neural network rendering priority tree is designed based on a preset deep neural network model to allocate lighting effects and rendering resources to decision tree priority nodes and region material-lighting interaction feature data, thereby obtaining a rendering priority neural network model, including:

[0103] (1) Based on the preset deep neural network model, design a neural network rendering priority tree architecture for lighting effects and rendering resource allocation of decision tree priority nodes, and generate a rendering priority neural network model architecture.

[0104] (2) Perform user visual attention prediction analysis based on regional material-lighting interaction feature data to generate user visual attention prediction data;

[0105] (3) Based on the user visual attention prediction data, the rendering priority neural network model architecture is processed to learn the priority feature weights of lighting effects and rendering resources, and a rendering priority neural network model is generated.

[0106] In this embodiment of the invention, a deep neural network model is first preset, using a VGG variant structure design, comprising 4 convolutional layers and 3 fully connected layers. The number of filters in the convolutional layers is 32-64-128-256, and the number of nodes in the fully connected layers is 512-256-128. LeakyReLU is selected as the activation function (α=0.2) to avoid neuron death; the output layer uses the Softmax function to implement a multi-region priority probability distribution. The initial learning rate is set to 0.003, and the Adam optimizer (β1=0.9, β2=0.999) is used to improve training stability, with a batch size of 48. Specialized designs are implemented for the priority nodes of the decision tree, configuring differentiated network parameters for regions with different lighting conditions: nodes in strong lighting regions use more convolutional kernels (increased to 96) to capture highlight details; nodes in shadow regions are configured with a deeper receptive field (9×9 pixels) to improve edge detection capabilities; and an attention mechanism module (attention heads=4) is added to dynamic light source regions to handle time-varying lighting. A residual connection structure is used between nodes to mitigate the gradient vanishing problem. The neural networks of each priority node are integrated into a unified tree architecture, with a three-level rendering priority hierarchy: high-priority layer (core visual region), medium-priority layer (secondary visual region), and basic layer (background region). Each level controls the resource allocation ratio through a gating unit (gating threshold = 0.65) to ensure that key regions receive sufficient rendering resources. In the tree architecture, the decision results of upper-level nodes affect the weight parameters of lower-level nodes, achieving global collaborative optimization and generating a complete rendering priority neural network model architecture. Key visual attention elements are extracted from the region material-lighting interaction feature data, including brightness contrast (weight 0.3), color saturation (weight 0.25), texture complexity (weight 0.25), and motion saliency (weight 0.2). An improved Itti-Koch model is used to analyze visual saliency, and a deep learning enhancement module is introduced to improve prediction accuracy. A region sampling grid (16×16) is set, and a saliency score is calculated for each region. Regions with a score greater than 0.7 are marked as high-attention regions. The prediction model was calibrated using eye-tracking validation (50 participants), achieving an accuracy of 85%, and a user visual attention prediction data matrix was generated. Based on the generated visual attention prediction data, weights were learned for the rendering priority neural network model. Cross-validation (K=5) was used to ensure model generalization, with a training set to validation set ratio of 7:3. The loss function was designed to integrate mean squared error (MSE) and structural similarity (SSIM) losses, balancing numerical accuracy and visual consistency. A dynamic learning rate adjustment strategy was implemented: if the validation loss decreased for three consecutive rounds, the learning rate increased by 10%; if the validation loss increased, the learning rate decreased by 15%.Resource allocation coefficients are set for different priority regions: high-attention regions (coefficient = 1.8) receive the most rendering sampling points; medium-attention regions (coefficient = 1.0) receive a standard sampling rate; and low-attention regions (coefficient = 0.6) reduce the sampling rate to save resources. An early stopping mechanism (patience = 10) is implemented during the weight learning process to stop training when the validation loss stabilizes, preventing overfitting. The model's effectiveness is verified through A / B testing (with a 20% improvement in user satisfaction), ultimately generating an optimized rendering priority neural network model that achieves intelligent allocation of lighting effects and rendering resources.

[0107] This invention establishes a heterogeneous light source representation model through lighting data acquisition and light source type identification, enabling unified management and coordinated rendering of multiple light source types. The system employs multi-dimensional lighting attribute clustering analysis and feature extraction to accurately capture changes in lighting intensity and spectral distribution within the scene. A visually irrelevant lighting data filtering mechanism effectively eliminates redundant lighting information with minimal impact on visual effects, significantly improving rendering efficiency. Based on scene region division and material-lighting interaction feature analysis, the system achieves precise lighting rendering for different region material characteristics and presents more realistic lighting effects through physically based rendering model analysis. A region rendering priority prediction engine analyzes lighting propagation characteristics in real time and intelligently allocates rendering resources using a deep neural network model, ensuring efficient processing capabilities during dynamic scene changes. Based on user visual attention prediction data and regional dynamic rendering needs analysis, the system generates adaptive rendering decisions, prioritizing visually important regions and effectively reducing computational resource consumption while maintaining high-quality visual performance. This adaptive real-time lighting and color rendering technology, designed for heterogeneous light source data and complex rendering requirements, achieves more efficient and realistic lighting effects in game scenes.

[0108] Example 2;

[0109] Please see Figure 2 As shown, for parts not described in detail in this embodiment, please refer to the description in Embodiment 1. A real-time lighting and color rendering method for game scenes is provided, including:

[0110] S1. Collect scene lighting information based on the game engine to generate initial lighting data; perform heterogeneous light source data fusion processing based on the initial lighting data to generate integrated lighting data;

[0111] S2. Perform multi-dimensional illumination attribute clustering analysis based on the integrated illumination data to generate multi-dimensional illumination attribute clustering data; extract multi-dimensional illumination features based on the multi-dimensional illumination attribute clustering data to generate multi-dimensional illumination feature data.

[0112] S3. Perform scene region segmentation processing based on multi-dimensional illumination feature data to generate region multi-dimensional illumination feature data; perform region material-lighting interaction feature analysis based on region multi-dimensional illumination feature data to generate region material-lighting interaction feature data.

[0113] S4. Perform light propagation feature analysis based on multi-dimensional regional lighting feature data to generate light propagation feature data; design an intelligent judgment engine for regional rendering priority prediction based on regional material-lighting interaction feature data to obtain a regional rendering priority prediction engine; transmit the light propagation feature data to the regional rendering priority prediction engine in real time for intelligent analysis of regional dynamic rendering requirements to generate regional dynamic rendering requirement data; perform adaptive decision analysis for global real-time lighting and color rendering based on the regional dynamic rendering requirement data to generate global real-time lighting and color rendering decisions; perform real-time lighting and color rendering adjustment processing on the initial lighting data based on the global real-time lighting and color rendering decisions to generate real-time lighting and color rendering data, and execute real-time lighting and color rendering of the game scene using the real-time lighting and color rendering data.

[0114] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0115] The above description is merely a preferred embodiment of the present invention, and the scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for users of ordinary technical skills, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A real-time lighting and color rendering system for game scenes, characterized in that, include: Acquisition and Processing Module: Acquires scene lighting information based on the game engine and generates initial lighting data; Heterogeneous light source data fusion processing is performed based on the initial illumination data to generate integrated illumination data; Data extraction module: Performs multi-dimensional illumination attribute clustering analysis based on integrated illumination data to generate multi-dimensional illumination attribute clustering data; Multi-dimensional illumination feature data is generated by extracting multi-dimensional illumination features from multi-dimensional illumination attribute clustering data. Analysis module: Performs scene region segmentation based on multi-dimensional illumination feature data to generate multi-dimensional illumination feature data for each region; performs region material property mapping analysis based on the region multi-dimensional illumination feature data to generate region material property data; performs region material-lighting interaction feature analysis based on the region material property data to generate region material-lighting interaction feature data. Real-time rendering module: Performs lighting propagation feature analysis based on multi-dimensional regional lighting feature data to generate lighting propagation feature data; performs visual importance scoring node analysis based on regional material-lighting interaction feature data to generate visual importance scoring node data, and designs decision tree priority nodes based on the visual importance scoring node data; designs a neural network rendering priority tree for lighting effects and rendering resource allocation based on the preset deep neural network model for the decision tree priority nodes and regional material-lighting interaction feature data, thus obtaining a rendering priority neural network model; A smart judgment engine for predicting region rendering priority is designed based on a rendering priority neural network model to obtain a region rendering priority prediction engine. The light propagation characteristic data is transmitted in real time to the region rendering priority prediction engine to perform intelligent analysis of region dynamic rendering requirements and generate region dynamic rendering requirement data. Based on the regional dynamic rendering requirement data, perform adaptive decision analysis for global real-time lighting and color rendering, and generate global real-time lighting and color rendering decisions; Based on global real-time lighting and color rendering decisions, the initial lighting data is adjusted and processed in real-time lighting and color rendering to generate real-time lighting and color rendering data, and then the real-time lighting and color rendering of the game scene is executed using the real-time lighting and color rendering data.

2. The real-time lighting and color rendering system for game scenes according to claim 1, characterized in that, The scene lighting information is collected based on the game engine to generate initial lighting data; Heterogeneous light source data fusion processing is performed based on the initial illumination data to generate integrated illumination data, including: Scene lighting information is collected based on the game engine to generate initial lighting data; Based on the initial illumination data, perform light source type identification processing to generate light source type feature data; Design a heterogeneous light source representation model based on light source type characteristic data, and generate a heterogeneous light source representation model; Heterogeneous light source data is fused based on the heterogeneous light source representation model to generate integrated illumination data.

3. The real-time lighting and color rendering system for game scenes according to claim 1, characterized in that, The process involves performing multi-dimensional illumination attribute clustering analysis based on integrated illumination data to generate multi-dimensional illumination attribute clustering data. Multi-dimensional illumination feature extraction is performed based on multi-dimensional illumination attribute clustering data to generate multi-dimensional illumination feature data, including: Illumination intensity features are extracted from the integrated illumination data to generate illumination intensity feature data. Spectral distribution features are extracted from integrated illumination data to generate spectral distribution feature data. The initial multidimensional light attribute clustering analysis is performed on the light intensity feature data and spectral distribution feature data using a preset multidimensional clustering algorithm to generate initial multidimensional light attribute clustering data. Visually irrelevant lighting data is filtered out based on the initial multidimensional lighting attribute clustering data to generate multidimensional lighting attribute clustering data; Multidimensional illumination feature data is generated by extracting multidimensional illumination features from clustered data of multidimensional illumination attributes.

4. The real-time lighting and color rendering system for game scenes according to claim 3, characterized in that, The step of filtering out visually irrelevant lighting data based on the initial multi-dimensional lighting attribute clustering data to generate multi-dimensional lighting attribute clustering data includes: Visual perception threshold analysis is performed based on the initial multi-dimensional illumination attribute clustering data to generate visual perception threshold data, which includes illumination intensity threshold data and spectral sensitivity data. Low-impact light source aggregation processing is performed based on illuminance intensity threshold data to generate aggregated low-impact light source data. Visual redundancy illumination data is generated by performing visual redundancy illumination recognition processing based on aggregated low-impact light source data and spectral sensitivity data. Based on the visually redundant lighting data, the initial multi-dimensional lighting attribute clustering data is filtered out to generate multi-dimensional lighting attribute clustering data. The visually redundant lighting data is then transmitted to the rendering engine to perform a simplified rendering job.

5. The real-time lighting and color rendering system for game scenes according to claim 1, characterized in that, The step of performing regional material characteristic mapping analysis based on multi-dimensional regional illumination feature data to generate regional material characteristic data includes: Surface material reflection characteristics are analyzed based on multi-dimensional regional illumination feature data to generate surface material reflection characteristic data. Surface material scattering characteristics are analyzed based on multi-dimensional regional illumination feature data to generate surface material scattering characteristic data. The material-light interaction response matrix is ​​calculated based on the surface material reflection characteristics data and the surface material scattering characteristics data to generate the material-light interaction response matrix. Based on the material-lighting interaction response matrix, regional material property mapping analysis is performed to generate regional material property data.

6. The real-time lighting and color rendering system for game scenes according to claim 1, characterized in that, The step of performing regional material-lighting interaction feature analysis based on regional material characteristic data to generate regional material-lighting interaction feature data includes: Perform physically based rendering model feature analysis on the regional material property data to generate physically based rendering model feature data; Based on the physical rendering model feature data, region material-lighting interaction feature analysis is performed to generate region material-lighting interaction feature data.

7. The real-time lighting and color rendering system for game scenes according to claim 1, characterized in that, The aforementioned neural network rendering priority tree design, based on a preset deep neural network model, allocates lighting effects and rendering resources to decision tree priority nodes and region material-lighting interaction feature data to obtain a rendering priority neural network model, including: Based on a pre-defined deep neural network model, a neural network rendering priority tree architecture is designed to allocate lighting effects and rendering resources to the priority nodes of the decision tree, generating a rendering priority neural network model architecture. Based on the regional material-lighting interaction feature data, user visual attention prediction analysis is performed to generate user visual attention prediction data. Based on user visual attention prediction data, the rendering priority neural network model architecture is processed by learning the priority feature weights of lighting effects and rendering resources to generate a rendering priority neural network model.

Images