Optimization method for rapid rendering of three-dimensional model in animation design

Abstract

The invention discloses an optimization method for rapid rendering of a three-dimensional model in animation design, and particularly relates to the technical field of three-dimensional model rendering, comprising the steps of S1, model lightweight processing, S2, open source material library access, S3, illumination rendering parameter optimization, S4, agent model and scene layering, S5, hardware acceleration and rendering, S6, cache precomputation optimization, and S7, real-time rendering interaction optimization. By means of a series of optimization measures, efficient rendering of the three-dimensional model is achieved, wiring is conducted through the optimization model, redundant patches are reduced, pre-optimized material packages are downloaded, complex material nodes are automatically created, the rendering calculation amount is reduced, texture compression, resolution adaptation and UV layout are optimized, texture waste is avoided, and the rendering efficiency is improved. According to the method, the sampling rate and the global illumination rebound frequency are dynamically adjusted, and the noise is quickly reduced, so that the processing is simpler, the lagging, the crash and the data loss are effectively avoided, and the rendering efficiency is remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of 3D model rendering technology, and more specifically, to an optimized method for rapid rendering of 3D models in animation design. Background Technology

[0002] In animation design, optimized methods for rapid rendering of 3D models significantly improve project efficiency, reduce resource consumption, optimize detail rendering, and enhance team collaboration by shortening rendering time, reducing hardware costs, improving visual quality, supporting complex scenes and real-time interaction, ultimately achieving an efficient, high-quality, and low-cost animation production process.

[0003] Among the existing publicly available documents, patent publication number CN108921920A discloses a method for producing 3D animation of hydropower projects. This invention meets the accuracy requirements of the design model, improves production efficiency and effect, and fully utilizes the powerful modeling function of Catia, the powerful animation production function of 3ds Max, and the ability of Vue to quickly build outdoor scenes. On the other hand, it uses a plugin in 3ds Max to reference Vue's algorithm for rendering, which greatly improves rendering efficiency. However, this patent has the following defects.

[0004] In the process of rapid rendering of 3D models in animation design, the processing of a large amount of 3D animation data results in a large amount of processing in each rendering step, which is more complex and prone to stuttering, crashes, or even data loss, making it difficult to achieve efficient rendering. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art, the present invention provides the following technical solution: an optimized method for rapid rendering of 3D models in animation design, comprising the following specific steps:

[0006] S1. Lightweight model processing: The model wiring is optimized 5-10 times using the automatic topology tool TopoGun, reducing redundant facets by 50-80. The engine performs 5-10 geometry processing operations. For recurring objects, instantiation technology is used to store only one copy of the model data and back it up 5-10 times.

[0007] S2. Access the open-source material library, download 20-30 pre-optimized material packages, automatically create 1000-2000 complex material nodes, directly call metal material presets, replace custom reflection and refraction parameters 5-20 times with the calculator, reduce the amount of rendering calculations by 5-10 times, and optimize texture compression, resolution adaptation, and UV layout.

[0008] S3. Optimize lighting rendering parameters. Use the generated sky system 5-10 times from low-resolution textures to replace complex light source arrays 2-5 times. Adjust the sampling rate and global illumination bounce times according to the scene complexity within 5-10 minutes, and perform noise reduction processing within 5-30 seconds.

[0009] S4. Proxy Model and Scene Layering: Generate 5-10 proxy models for complex character models and high-precision props. Load complete geometric data only during rendering and unfold within 5-10 minutes. Scene layer rendering: Divide the scene into foreground, midground and background layers and repeat 5-10 times, setting 6-15 different rendering parameters for each layer.

[0010] S5, hardware acceleration and rendering, GPU renderer selection, using a GPU-based renderer, utilizing internet accelerators and layered accelerators for acceleration, and multi-machine parallel rendering.

[0011] S6. Optimize cache pre-computation: Enable light caching in the renderer and save indirect lighting calculation results 5-10 times to avoid repeated calculations. Use point cloud caching for dynamic scene character animations 20-30 times.

[0012] S7. Real-time rendering interaction optimization: The real-time rendering engine is integrated and used for 50-60 previews, and 5-10 sets of high-precision models are imported through the plugin.

[0013] In a preferred embodiment, the geometry processing of the engine in S1 is divided into static geometry processing, dynamic geometry processing, instantiated geometry processing, level of detail processing, and procedural geometry processing. The processing order is the synchronous processing of static geometry processing, dynamic geometry processing, instantiated geometry processing, level of detail processing, and procedural geometry processing, with 3-5 synchronous processing cycles.

[0014] In a preferred embodiment, the texture compression and resolution adaptation in S2 involves compressing the texture 5-10 times using an online tool, and dynamically adjusting the texture resolution 10-20 times based on the size of the model in the scene. The UV layout optimization in S2 involves UV unwrapping without overlap 5-10 times, and using an automatic UV tool to reduce texture waste 2-5 times.

[0015] In a preferred embodiment, the noise reduction process in S3 uses the denoiser built into the renderer to achieve high-quality repeated noise reduction 5-10 times at a low sampling rate, with each time lasting 2-5 seconds.

[0016] In a preferred embodiment, in step S4, the rendering layer is divided into 5-10 layers, and after 5-10 renderings per channel, 10-20 groups are synthesized. Among them, Gaussian noise pixel values ​​follow a Gaussian distribution and camera sensor thermal noise is used. Salt and pepper noise is then used to make pixel values ​​suddenly change to extreme values, compressing distortion and sensor failure. At the same time, Poisson noise is related to signal intensity and photon noise in low-light imaging. The color image is converted to grayscale to simplify the processing. The channels are separated and denoised 5-10 times individually.

[0017] In a preferred embodiment, in step S5, the task is distributed to multiple computers for parallel processing 5-10 times through rendering management software, and then cloud rendering service is used. 10-20 cloud rendering platforms are used to expand computing power as needed 5-10 times. Emergency projects or large-scale renderings are performed 30-60 times. Among them, large-scale rendering is for film and animation production, through emergency demand scenarios, in the post-production of movies, concentrated rendering tasks caused by approaching release dates or special effects modifications, and the need to process high resolution, complex lighting effects, hair, and fluid simulations, with a single frame rendering time exceeding 5-10 minutes.

[0018] In a preferred embodiment, in step S6, the lighting and shadow information of key frames is recorded 6-10 times. The ambient light occlusion, reflection, and refraction rendering elements are saved as cache files using the rendering element cache. The refraction rendering elements are divided into refraction colors, and the color information carried by the refracted light is recorded. The filtering of light by colored glass and the dispersion effect of the medium itself are recorded.

[0019] In a preferred embodiment, in step S6, the refractive roughness records the roughness of the refractive surface and blurs the refractive effect in 20-30 groups, usually represented by a grayscale image where black represents smoothness and white represents roughness. The refractive depth records the penetration depth of the refracted light in the scene and is often used to control the recursive limit of refraction to 5-10 times, with subsequent modification time controlled within 5-10 seconds.

[0020] In a preferred embodiment, in step S7, the level of detail dynamically switches the model level of detail by 5-10 levels according to the camera distance, reducing the number of polygons rendered in real time by 20-50 times. For large scenes, 5-10 sets of model data are loaded using network streaming, the model is simplified and constructed, and the polygons of the model are reduced using tools. Multiple versions of high, medium and low models are generated according to the accuracy requirements, and the texture is compressed and divided into 100-200 blocks.

[0021] In a preferred embodiment, in step S7, the texture is converted to a compressed format to reduce video memory usage. Large texture maps are divided into small pixel blocks. The model data is converted from the original format to a lightweight format and compressed 10-20 times according to the format optimization. The mesh data is encoded in 5-10 groups using a compression algorithm, which can reduce the model file size by 50-70%, reduce the peak memory usage during loading by 10-30%, and reduce memory pressure.

[0022] The technical effects and advantages of this invention are as follows:

[0023] 1. This invention achieves efficient rendering of 3D models through a series of optimization measures. It reduces redundant facets by optimizing model topology and rationally plans the engine's geometric processing order and synchronization times, lowering memory and processing burden. It accesses open-source material libraries, downloads pre-optimized material packages, and automatically creates complex material nodes, replacing custom parameters and reducing rendering computation. Simultaneously, it optimizes texture compression, resolution adaptation, and UV layout to avoid texture waste. In terms of lighting rendering, it uses a low-resolution texture mapping system to generate the sky instead of a complex light source array, dynamically adjusts the sampling rate and global illumination bounce count, and performs rapid noise reduction. These optimizations simplify processing, effectively avoid stuttering, crashes, and data loss, and significantly improve rendering efficiency.

[0024] 2. This invention adopts a proxy model and scene layering strategy to generate multiple proxy models for complex character models and high-precision props, renders them in layers and sets different parameters, and combines multi-channel rendering and compositing technology to effectively simplify the processing flow. It uses GPU renderers, the Internet and layered accelerators, and multi-machine parallel rendering and cloud rendering services to efficiently handle urgent projects and large-scale rendering tasks, as well as complex lighting effects in film and animation production. These optimization measures make the processing simpler and greatly reduce the risk of lag, crashes and data loss.

[0025] 3. This invention significantly improves rendering efficiency through cache pre-computation and real-time rendering interaction optimization. It enables lighting cache, point cloud cache, and rendering element cache to save key calculation results and lighting information, avoids redundant calculations, simplifies the processing flow, integrates a real-time rendering engine, supports high-precision model import, dynamically switches detail levels, reduces the number of polygons, and uses network streaming to load large scene data. Combined with model simplification, texture compression and chunking techniques, as well as format optimization and compression algorithms, it greatly reduces model file size and video memory usage. These optimization measures make processing simpler, effectively reduce the risk of lag, crashes, and data loss, and achieve a highly efficient and smooth real-time rendering interaction experience. Attached Figure Description

[0026] Figure 1 This is a schematic diagram illustrating the operation of the optimized method for rapid rendering of 3D models in animation design according to the present invention. Detailed Implementation

[0027] 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.

[0028] Example 1:

[0029] Reference Appendix Figure 1 An optimized method for fast rendering of 3D models in animation design, including the following specific steps:

[0030] S1. Lightweight model processing: The model topology is optimized 5 times using the automatic topology tool TopoGun, reducing redundant facets by 50. The engine performs 5 geometry processing operations, which are divided into static geometry processing, dynamic geometry processing, instantiated geometry processing, level of detail processing, and procedural geometry processing. The processing order is simultaneous processing of static geometry processing, dynamic geometry processing, instantiated geometry processing, level of detail processing, and procedural geometry processing, with 3 simultaneous processing operations. Instantiation technology is used for recurring objects, storing only one copy of the model data and backing it up 5 times.

[0031] S2. Access the open-source material library, download 20 pre-optimized material packages, automatically create 1000 complex material nodes, directly call the metal material preset, replace the custom reflection and refraction parameters 5 times with the calculator, reduce the amount of rendering calculations 5 times, and optimize texture compression, resolution adaptation, and UV layout. Texture compression and resolution adaptation use online tools to compress textures 5 times, and dynamically adjust texture resolution 10 times according to the size of the model in the scene. Among them, the UV layout optimization in S2 is to unwrap UVs without overlap 5 times, and use the automatic UV tool to reduce texture waste 2 times.

[0032] S3. Lighting rendering parameters are optimized by generating a sky system 5 times from low-resolution textures, replacing complex light source arrays 2 times. Within 5 minutes, the sampling rate and global illumination bounce times are adjusted according to the scene complexity. At the same time, noise reduction is performed within 5 seconds. The noise reduction uses the denoiser built into the renderer to achieve high-quality noise reduction 5 times at a low sampling rate, with each time taking 2 seconds.

[0033] S4. Proxy Model and Scene Layering: Generate 5 proxy models for complex character models and high-precision props. Unfold within 5 minutes of loading complete geometric data during rendering. Scene layer rendering divides the scene into foreground, midground, and background layers, repeating 5 times. Set 6 different rendering parameters for each layer. Use rendering layers to divide into 5 layers. Render 5 times per channel and then synthesize 10 groups. Among them, Gaussian noise pixel values ​​follow a Gaussian distribution. Camera sensor thermal noise is used, and salt and pepper noise is used to make pixel values ​​jump to extreme values ​​to compress distortion and sensor failure. At the same time, Poisson noise is related to signal strength and photon noise in low-light imaging. The color image is converted to grayscale to simplify the processing. Separate channels are denoised 5 times.

[0034] S5. Hardware acceleration and rendering: GPU renderer selection, using a GPU-based renderer, accelerating with internet accelerators and layered accelerators; multi-machine parallel rendering, distributing tasks to multiple computers for parallel processing 5 times through rendering management software; cloud rendering service, using 10 cloud rendering platforms, scaling computing power as needed 5 times; 30 urgent projects or large-scale renderings, including large-scale renderings for film and animation production, through urgent demand scenarios, film post-production, concentrated rendering tasks caused by approaching release dates or special effects modifications, and the need to handle high resolution, complex lighting effects, hair, and fluid simulation, with single frame rendering time exceeding 5 minutes.

[0035] S6. Optimize pre-calculation of caching. Lighting caching is enabled in the renderer, saving indirect lighting calculation results 5 times to avoid duplicate calculations. Point cloud caching is used for dynamic scene character animations 20 times, recording the lighting and shadow information of key frames 6 times. Rendering element caching is used to save ambient light occlusion, reflection, and refraction rendering elements as cache files. Among them, refraction rendering elements are divided into refraction color, which records the color information carried by the refracted light, the filtering of light by colored glass, and the dispersion effect of the medium itself. Refraction roughness records the roughness of the refraction surface. There are 20 sets of blurred refraction effects, usually represented by grayscale. Black is smooth and white is rough. Refraction depth records the penetration depth of the refracted light in the scene. It is often used to control the recursion limit of refraction to 5 times, and the subsequent modification time is controlled within 5 seconds.

[0036] S7. Real-time rendering interaction optimization: The real-time rendering engine is integrated and previewed 50 times. Five sets of high-precision models are imported through plugins. The level of detail is dynamically switched 5 times according to the camera distance, reducing the number of polygons in real-time rendering 20 times. Network streaming is used to load five sets of model data for large scenes. The model is simplified and constructed. The polygon reduction tool is used to reduce the number of polygons in the model. Multiple versions of high, medium and low-poly models are generated according to the accuracy requirements. 100 textures are compressed and segmented. The textures are converted to a compressed format to reduce the use of video memory. Large texture maps are divided into small-sized pixel blocks. The model data is converted from the original format to a lightweight format and compressed 10 times according to the format optimization. Five sets of mesh data are encoded using compression algorithms, which can reduce the model file size by 50% and reduce the peak memory usage during loading by 10%, thus reducing memory pressure.

[0037] Example 2:

[0038] Reference Appendix Figure 1 An optimized method for fast rendering of 3D models in animation design, including the following specific steps:

[0039] S1. Lightweight model processing: The model topology is optimized 8 times using the automatic topology tool TopoGun, reducing redundant faces by 60. The engine performs 8 geometry processing operations, which are divided into static geometry processing, dynamic geometry processing, instantiated geometry processing, level of detail processing, and procedural geometry processing. The processing order is the simultaneous processing of static geometry processing, dynamic geometry processing, instantiated geometry processing, level of detail processing, and procedural geometry processing, with 4 simultaneous processing operations. Instantiation technology is used for recurring objects, storing only one copy of the model data and backing it up 8 times.

[0040] S2. Access the open-source material library, download 25 pre-optimized material packages, automatically create 1500 complex material nodes, directly call the metal material preset, replace the custom reflection and refraction parameters 10 times with the calculator, reduce the renderer's calculation load 8 times, optimize texture compression and resolution adaptation and UV layout. Texture compression and resolution adaptation use online tools to compress textures 8 times, dynamically adjust texture resolution according to the size of the model in the scene 15 times, among which UV layout optimization is performed to achieve UV unpacking without overlap 8 times, and use the automatic UV tool to reduce texture waste 3 times.

[0041] S3. Lighting rendering parameters are optimized by generating a sky system 8 times from low-resolution textures, replacing complex light source arrays 4 times. Within 8 minutes, the sampling rate and global illumination bounce times are adjusted according to the scene complexity. At the same time, noise reduction is performed within 15 seconds. The noise reduction uses the denoiser built into the renderer to achieve high-quality repeated noise reduction 8 times at a low sampling rate, with each time taking 3 seconds.

[0042] S4. Proxy Model and Scene Layering: Generate 8 proxy models for complex character models and high-precision props. Unfold within 8 minutes of loading complete geometric data during rendering. Scene layer rendering divides the scene into foreground, midground, and background layers, repeating 8 times. Set 12 different rendering parameters for each layer. Use rendering layers to divide into 8 layers. Render 8 times in each channel and then synthesize 15 groups. Among them, Gaussian noise pixel values ​​follow a Gaussian distribution. Camera sensor thermal noise is used, and salt and pepper noise is used to make pixel values ​​jump to extreme values ​​to compress distortion and sensor failure. At the same time, Poisson noise is related to signal strength and photon noise in low-light imaging. The color image is converted to grayscale to simplify the processing. Separate channels are denoised 8 times.

[0043] S5. Hardware acceleration and rendering: GPU renderer selection, using a GPU-based renderer, accelerating with internet accelerators and layered accelerators; multi-machine parallel rendering, distributing tasks to multiple computers for parallel processing 8 times through rendering management software; cloud rendering service, using 15 cloud rendering platforms, scaling computing power as needed 8 times; 50 urgent projects or large-scale renderings, including large-scale renderings for film and animation production, through urgent demand scenarios, film post-production, concentrated rendering tasks due to approaching release dates or special effects modifications, and the need to handle high resolution, complex lighting effects, hair, and fluid simulation, with single frame rendering time exceeding 8 minutes;

[0044] S6. Optimize pre-calculation of caching. Lighting caching is enabled in the renderer, saving indirect lighting calculation results 5 times to avoid duplicate calculations. Point cloud caching is used for dynamic scene character animations 20 times, recording the lighting and shadow information of key frames 6 times. Rendering element caching is used to save ambient light occlusion, reflection, and refraction rendering elements as cache files. Among them, refraction rendering elements are divided into refraction color, which records the color information carried by the refracted light, the filtering of light by colored glass, and the dispersion effect of the medium itself. Refraction roughness records the roughness of the refraction surface. There are 20 sets of blurred refraction effects, usually represented by grayscale. Black is smooth and white is rough. Refraction depth records the penetration depth of the refracted light in the scene. It is often used to control the recursion limit of refraction to 5 times, and the subsequent modification time is controlled within 5 seconds.

[0045] S7. Real-time rendering interaction optimization: The real-time rendering engine is integrated and previewed 50 times. Five sets of high-precision models are imported through plugins. The level of detail is dynamically switched 5 times according to the camera distance, reducing the number of polygons in real-time rendering 20 times. Network streaming is used to load five sets of model data for large scenes. The model is simplified and constructed. The polygon reduction tool is used to reduce the number of polygons in the model. Multiple versions of high, medium and low-poly models are generated according to the accuracy requirements. 100 textures are compressed and segmented. The textures are converted to a compressed format to reduce the use of video memory. Large texture maps are divided into small-sized pixel blocks. The model data is converted from the original format to a lightweight format and compressed 10 times according to the format optimization. Five sets of mesh data are encoded using compression algorithms, which can reduce the model file size by 50% and reduce the peak memory usage during loading by 10%, thus reducing memory pressure.

[0046] Example 3:

[0047] Reference Appendix Figure 1 An optimized method for fast rendering of 3D models in animation design, including the following specific steps:

[0048] S1. Lightweight model processing: The model topology is optimized 10 times using the automatic topology tool TopoGun, reducing redundant faces by 80. The engine performs 10 geometry processing operations, which are divided into static geometry processing, dynamic geometry processing, instantiated geometry processing, level of detail processing, and procedural geometry processing. The processing order is simultaneous processing of static geometry processing, dynamic geometry processing, instantiated geometry processing, level of detail processing, and procedural geometry processing, with 5 simultaneous processing operations. For recurring objects, instantiation technology is used to store only one copy of the model data, with 10 backups.

[0049] S2. Access the open-source material library, download 30 sets of pre-optimized material packages, automatically create 2000 complex material nodes, directly call the metal material preset, replace the custom reflection and refraction parameters 20 times with the calculator, reduce the amount of rendering calculations by 10 times, optimize texture compression and resolution adaptation and UV layout. Texture compression and resolution adaptation use online tools to compress textures 10 times, dynamically adjust texture resolution according to the size of the model in the scene 20 times, optimize UV ​​layout to UV unroll without overlap 10 times, and use automatic UV tools to reduce texture waste by 5 times.

[0050] S3. Lighting rendering parameters are optimized by generating a sky system 10 times from low-resolution textures, replacing complex light source arrays 5 times. Within 10 minutes, the sampling rate and global illumination bounce times are adjusted according to the scene complexity. At the same time, noise reduction is performed within 30 seconds. The noise reduction uses the denoiser built into the renderer. High-quality noise reduction is repeated 10 times at a low sampling rate, with each time lasting 5 seconds.

[0051] S4. Proxy Model and Scene Layering: Generate 10 proxy models for complex character models and high-precision props. Load complete geometric data only during rendering and unfold within 10 minutes. Scene layer rendering divides the scene into foreground, midground, and background layers, repeating 10 times. Set 15 different rendering parameters for each layer. Use rendering layers to divide into 10 layers. Render 10 times per channel and then synthesize 20 groups. Gaussian noise pixel values ​​follow a Gaussian distribution. Camera sensor thermal noise is used, and salt and pepper noise is used to make pixel values ​​jump to extreme values ​​to compress distortion and sensor failure. At the same time, Poisson noise is related to signal strength and photon noise in low-light imaging. Convert color images to grayscale to simplify processing. Separate channels for noise reduction 10 times.

[0052] S5. Hardware acceleration and rendering: GPU renderer selection, using a GPU-based renderer, accelerating with internet accelerators and layered accelerators; multi-machine parallel rendering, distributing tasks to multiple computers for parallel processing 10 times through rendering management software; cloud rendering service, using 20 cloud rendering platforms, scaling computing power as needed 10 times; 60 urgent projects or large-scale renderings, including large-scale renderings for film and animation production, through urgent demand scenarios, film post-production, concentrated rendering tasks caused by approaching release dates or special effects modifications, and the need to handle high resolution, complex lighting effects, hair, and fluid simulations, with single-frame rendering time exceeding 10 minutes.

[0053] S6. Optimize pre-calculation of caching. Lighting caching is enabled in the renderer, saving indirect lighting calculation results 10 times to avoid duplicate calculations. Point cloud caching is used for dynamic scene character animations 30 times, recording the lighting and shadow information of key frames 10 times. Rendering element caching is used to save ambient light occlusion, reflection, and refraction rendering elements as cache files. Among them, refraction rendering elements are divided into refraction color, which records the color information carried by the refracted light, the filtering of light by colored glass, and the dispersion effect of the medium itself. Refraction roughness records the roughness of the refraction surface. There are 30 sets of blur refraction effects, usually represented by grayscale. Black is smooth and white is rough. Refraction depth records the penetration depth of the refracted light in the scene. It is often used to control the recursion limit of refraction 10 times, and the subsequent modification time is controlled within 10 seconds.

[0054] S7. Real-time rendering interaction optimization: The real-time rendering engine is integrated and previewed 60 times. 10 sets of high-precision models are imported through plugins. The level of detail is dynamically switched by 10 levels based on the camera distance, reducing the number of polygons in real-time rendering by 50 times. Network streaming is used to load 10 sets of model data for large scenes. Model construction is simplified. The polygon reduction tool is used to reduce the number of polygons in the model. Multiple versions of high, medium, and low-poly models are generated according to the accuracy requirements. 200 textures are compressed and segmented. Textures are converted to compressed formats to reduce video memory usage. Large texture maps are divided into small pixel blocks. Based on format optimization, the model data is converted from the original format to a lightweight format and compressed 20 times. 10 sets of mesh data are encoded using compression algorithms, which can reduce the model file size by 70% and reduce the peak memory usage during loading by 30%, thus reducing memory pressure.

[0055] After conducting experiments on the above three sets of embodiments based on complexity processing efficiency, crash rate, and data processing improvement, the following table was obtained:

[0056] Complex processing efficiency (%) Failure rate (%) Data processing improvement rate (%) Example 1 89.5 6.5 26 Example 2 86.5 10.5 24 Example 3 83.5 8.5 23.5

[0057] In summary, the complex processing efficiency in Example 1 is significantly higher than that in Examples 2-3. Specifically, the crash rate in Example 1 is significantly lower than that in Examples 2-3, and the data processing improvement rate in Example 1 is significantly higher than that in Examples 2-3. Therefore, through a series of optimization measures, efficient rendering of 3D models is achieved. By optimizing model wiring, reducing redundant faces, and rationally planning the engine geometry processing order and synchronization times, memory and processing burden are reduced. Complex material nodes are automatically created to reduce the amount of rendering computation. In terms of lighting rendering, a low-resolution texture generation sky system is used to replace the complex light source array. These optimizations make the processing simpler, effectively avoid stuttering, crashes, and data loss, and significantly improve rendering efficiency.

[0058] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An optimized method for rapid rendering of 3D models in animation design, characterized in that: The specific steps are as follows: S1. Lightweight model processing: The model wiring is optimized 5-10 times using the automatic topology tool TopoGun, reducing redundant facets by 50-80. The engine performs 5-10 geometry processing operations. For recurring objects, instantiation technology is used to store only one copy of the model data and back it up 5-10 times. S2. Access the open-source material library, download 20-30 pre-optimized material packages, automatically create 1000-2000 complex material nodes, directly call metal material presets, replace custom reflection and refraction parameters 5-20 times with the calculator, reduce the amount of rendering calculations by 5-10 times, and optimize texture compression, resolution adaptation, and UV layout. S3. Optimize lighting rendering parameters. Use the generated sky system 5-10 times from low-resolution textures to replace complex light source arrays 2-5 times. Adjust the sampling rate and global illumination bounce times according to the scene complexity within 5-10 minutes, and perform noise reduction processing within 5-30 seconds. S4. Proxy Model and Scene Layering: Generate 5-10 proxy models for complex character models and high-precision props. Load complete geometric data only during rendering and unfold within 5-10 minutes. Scene layer rendering: Divide the scene into foreground, midground and background layers and repeat 5-10 times, setting 6-15 different rendering parameters for each layer. S5, hardware acceleration and rendering, GPU renderer selection, using a GPU-based renderer, utilizing internet accelerators and layered accelerators for acceleration, and multi-machine parallel rendering. S6. Optimize cache pre-computation: Enable light caching in the renderer and save indirect lighting calculation results 5-10 times to avoid repeated calculations. Use point cloud caching for dynamic scene character animations 20-30 times. S7. Real-time rendering interaction optimization: The real-time rendering engine is integrated and used for 50-60 previews, and 5-10 sets of high-precision models are imported through the plugin.

2. The optimization method for rapid rendering of 3D models in animation design according to claim 1, characterized in that: In S1, the engine's geometry processing is divided into static geometry processing, dynamic geometry processing, instantiated geometry processing, level of detail processing, and procedural geometry processing. The processing order is the synchronous processing of static geometry processing, dynamic geometry processing, instantiated geometry processing, level of detail processing, and procedural geometry processing, with 3-5 synchronous processing cycles.

3. The optimization method for rapid rendering of 3D models in animation design according to claim 1, characterized in that: In S2, texture compression and resolution adaptation are performed by compressing textures 5-10 times using online tools and dynamically adjusting texture resolution 10-20 times based on the size of the model in the scene. In S2, UV layout optimization involves UV unwrapping without overlap 5-10 times and using automatic UV tools to reduce texture waste 2-5 times.

4. The optimized method for rapid rendering of 3D models in animation design according to claim 1, characterized in that: The noise reduction process in S3 uses the denoiser built into the renderer to achieve high-quality repeated noise reduction 5-10 times at a low sampling rate, with each time taking 2-5 seconds.

5. The optimization method for rapid rendering of 3D models in animation design according to claim 1, characterized in that: In S4, the rendering layer is divided into 5-10 layers. After 5-10 renderings per channel, 10-20 groups are synthesized. Among them, Gaussian noise pixel values ​​follow a Gaussian distribution and camera sensor thermal noise is used. Salt and pepper noise is then used to make pixel values ​​suddenly become extreme values, compressing distortion and sensor failure. At the same time, Poisson noise is related to signal intensity and photon noise in low-light imaging. The color image is converted to grayscale to simplify the processing. The channels are separated and denoised 5-10 times individually.

6. The optimization method for rapid rendering of 3D models in animation design according to claim 1, characterized in that: In S5, the rendering management software distributes the task to multiple computers for parallel processing 5-10 times, and then utilizes cloud rendering services, using 10-20 cloud rendering platforms to expand computing power as needed 5-10 times. For urgent projects or large-scale rendering, there are 30-60 renderings. Among them, large-scale rendering is for film and animation production, through urgent demand scenarios, in the post-production of movies, concentrated rendering tasks caused by the approaching release date or special effects modification, and the need to process high resolution, complex lighting effects, hair, and fluid simulation, with a single frame rendering time exceeding 5-10 minutes.

7. The optimization method for rapid rendering of 3D models in animation design according to claim 1, characterized in that: In S6, the lighting and shadow information of key frames are recorded 6-10 times. The ambient light occlusion, reflection and refraction rendering elements are saved as cache files using the rendering element cache. The refraction rendering elements are divided into refraction colors, which record the color information carried by the refracted light. The filtering of light by colored glass and the dispersion effect of the medium itself.

8. The optimization method for rapid rendering of 3D models in animation design according to claim 1, characterized in that: In S6, the refractive roughness records the roughness of the refractive surface and the blurred refractive effect in 20-30 groups, usually represented by a grayscale image. Black represents smoothness and white represents roughness. The refractive depth records the penetration depth of the refracted light in the scene. It is often used to control the recursive limit of refraction to 5-10 times, and the subsequent modification time is controlled within 5-10 seconds.

9. An optimization method for rapid rendering of 3D models in animation design according to claim 1, characterized in that: In S7, the level of detail dynamically switches the model detail level by 5-10 levels based on the camera distance, reducing the number of polygons rendered in real time by 20-50 times. For large scenes, network streaming is used to load 5-10 sets of model data, simplifying model construction. Tools are used to reduce the polygon count of the model, generating multiple versions of high, medium, and low-poly models according to accuracy requirements, and compressing and dividing textures into 100-200 blocks.

10. An optimization method for rapid rendering of 3D models in animation design according to claim 1, characterized in that: In S7, textures are converted to a compressed format to reduce video memory usage. Large texture maps are divided into small pixel blocks. The model data is converted from the original format to a lightweight format and compressed 10-20 times according to the format optimization. The mesh data is encoded in 5-10 groups using a compression algorithm, which can reduce the model file size by 50-70% and reduce the peak memory usage during loading by 10-30%, thus reducing memory pressure.

Images