Animation scene rendering system based on three-dimensional computer aided design

Abstract

The invention discloses a cartoon scene rendering system based on three-dimensional computer aided design, and relates to the technical field of cartoon scene rendering, the cartoon scene rendering system comprises a cartoon scene three-dimensional modeling unit, and the cartoon scene rendering system based on the three-dimensional computer aided design is applied to cartoon natural scene rendering. The animation scene three-dimensional modeling unit obtains animation scene design requirements, carries out early-stage preparation work of animation scene making, carries out three-dimensional modeling on animation characters and animation scenes by using 3D modeling software, and gradually adds details, so that animation character scenes are more exquisite. The invention provides a cartoon scene rendering system based on three-dimensional computer aided design, which is applied to cartoon natural scene rendering, and carries out three-dimensional modeling on cartoon characters and cartoon scenes through an arranged cartoon scene three-dimensional modeling unit.

Description

Technical Field

[0001] The present invention relates to the technical field of anime scene rendering, and specifically provides an anime scene rendering system based on three-dimensional computer-aided design. Background Art

[0002] Anime rendering is a key link in realizing creativity and preliminary design concepts, directly determining the visual effects of anime works. In the Chinese patent with the application number 202310993358.8, there is disclosed "An anime scene real-time rendering system, including: a modeling module for constructing an anime scene model and an anime character model; a rendering module connected to the modeling module, including a splitting unit for splitting the rendering frame image and a rendering unit connected to the splitting unit for rendering the image to be rendered; a signal transmission module connected to the rendering module for real-time transmitting the real-time anime image output by the rendering module back to the local user terminal; a central control module for splitting the image centered on the maximum granularity position in a single rendering frame image to determine the number of split blocks of a single rendering frame image, and for adjusting the granularity of non-edge simple frame blocks according to the delay duration of the rendering image. The present invention realizes the improvement of rendering efficiency."; This existing technology only solves the problem of the decline in rendering efficiency caused by the imbalance between the complexity of the rendering image and the number of image splits and rendering nodes in the current anime scene rendering production method. It does not consider that when rendering an anime natural scene, it is necessary to ensure the rendering effect of the scene from multiple aspects and further optimize the rendering effect of the anime natural scene. Moreover, for the problem that the foreground appears blurred and semi-transparent when focusing on the background, a rendering framework based on background filling is designed for rendering optimization, and a renderability parameter is defined to determine the rendering quality of the anime scene. Summary of the Invention

[0003] The purpose of the present invention is to provide an anime scene rendering system based on three-dimensional computer-aided design to solve the problems raised in the above background art.

[0004] To achieve the above purpose, the present invention provides the following technical solution: An anime scene rendering system based on three-dimensional computer-aided design, including an anime scene three-dimensional modeling unit. An anime scene rendering system based on three-dimensional computer-aided design is applied to the rendering of anime natural scenes. The anime scene three-dimensional modeling unit obtains the design requirements of the anime scene and conducts preliminary preparation work for the production of the anime scene. It uses 3D modeling software to perform three-dimensional modeling on anime characters and anime scenes and gradually adds details to make the anime character scene more exquisite; Anime scene rendering unit. The anime scene rendering unit differentiates the texture rendering features of the anime natural scene by constructing a texture rendering network, thereby optimizing the rendering effect of the terrain in the anime natural scene. When rendering the same type of vegetation in the anime natural scene, a rendering strategy is designed. In order to improve the lighting rendering effect of the anime natural scene during the day-night alternation process, an illuminance correction strategy is designed; Anime scene optimization unit. The anime scene optimization unit aims to optimize the rendering effect of the anime natural scene. By combining a traditional renderer and a neural renderer, a multi-scale rendering strategy is provided to efficiently process scene images. For the problem that when the foreground should be focused, the foreground appears blurred and semi-transparent, resulting in some originally occluded background areas being revealed, a multi-plane image representation method is adopted, and a rendering framework based on background filling is designed for rendering optimization; Rendering quality management unit. The rendering quality management unit defines a renderability parameter based on the source of the rendering error of the anime scene image, and then determines the rendering quality of the anime scene through this parameter of renderability.

[0005] Preferably, the anime scene 3D modeling unit includes an anime scene requirement acquisition module and a character scene model construction module. The anime scene requirement acquisition module obtains the design requirements of the anime scene and conducts the preliminary preparation work for the production of the anime scene, including drawing a sketch of the scene, determining the overall style layout of the scene, and collecting relevant reference images. The character scene model construction module is electrically connected to the anime scene requirement acquisition module. The character scene model construction module uses 3D modeling software to perform 3D modeling on the anime characters and the anime scene, and gradually adds details, including buildings, vegetation, and props in the scene, to make the anime character scene more exquisite.

[0006] Preferably, the anime scene rendering unit includes a texture rendering network construction module. The texture rendering network construction module constructs a texture rendering network to differentiate the texture rendering features of the anime natural scene for the rendering of the terrain in the anime natural scene. The specific operation is as follows: First, a composite terrain layer is constructed by adding terrain attributes. The natural environment includes various terrains such as soil, rock, and vegetation. The texture rendering network uses multiple different types of layers to cover the surface of the terrain model in the anime scene. Secondly, to ensure the texture quality, PBR-standardized texture materials are selected from the 3D scan resource library Quixel Megascans. During the material rendering process, considering that the shader needs to spend a lot of time sampling each texture map, in order to optimize the size of the texture space, the number of textures sampled in the material is kept to a minimum, and only two texture maps, namely the base color and the normal, are retained. Finally, the texture features are differentiated from two aspects: texture coordinate scaling and noise mixing. The operation of differentiating the texture features from the aspect of texture coordinate scaling is that the texture forms a two-dimensional coordinate system with the UV channel, and the material will assign a default coordinate system to the input texture so as to map it to the bound mesh. When the texture density cannot correctly match the size of the mesh, the spare texture coordinate system is applied to the texture sample through the TexCoord node to avoid occupying additional texture memory. For the input texture, the texture is scaled 2 times according to the near-distance UV coefficient and the far-distance UV coefficient respectively to obtain the near-distance display texture and the far-distance display texture. The operation of differentiating the texture features from the aspect of noise mixing is that since the noise texture is also a grayscale texture, as the pixel value ranges from 0 to 1, the pixel color transitions from black 0sRGB to white 255sRGB, and the three parts of black, white, and gray are randomly distributed in the texture area. Therefore, applying the noise texture to both the near-distance display texture and the far-distance display texture can ensure that the adjacent laid textures will not align with each other.

[0007] Preferably, the anime scene rendering unit includes a scene interaction rendering module. The scene interaction rendering module is electrically connected to the texture rendering network construction module. When the same type of vegetation in the anime natural scene needs to be rendered, the scene interaction rendering module designs a rendering strategy. The specific operation is as follows: The hierarchical static mesh instance is used to render the same type of vegetation in the anime natural scene. The LOD technology and the distance culling technology are applied to the vegetation individuals through the hierarchical static mesh instance to realize the corresponding reduction of the number of polygons of the mesh according to the distance between the mesh and the camera. And since the number of draw calls is an important factor affecting the time spent on rendering 1 frame of the picture, each material of each static mesh in the scene corresponds to 1 draw call to the GPU. When there are multiple identical objects in the scene, the draw call only needs to save 1 mesh instance in the memory, and all the same materials of the objects in the scene can be rendered through 1 draw call.

[0008] Preferably, the anime scene rendering unit further includes a scene lighting rendering module, which is electrically connected to the scene interaction rendering module. In order to improve the lighting rendering effect of the anime natural scene during the day-night alternation process, the scene lighting rendering module designs an illuminance correction strategy. The specific operation is to adjust the Pitch of the rotating body attribute of the directional light source in the scene, thereby changing the incident angle of the light. Assume that a minutes in the real scene represents 24 hours in the anime natural scene. When the directional light source rotates at a constant speed, D per second in the real scene corresponds to 2 / 5a hours in the anime natural scene. Update the time in the anime natural scene through the time axis, and then calculate the rotation angle of the directional light source at the current moment through the directional light source angle algorithm. The specific calculation formula is as follows:

[0009] In the formula, represents the time in the anime natural scene. In the initial state, the rotation angle of the directional light source is 90°. As time changes, the number of light rays from the directional light source hitting the upper surface of the anime natural scene increases first and then slowly decreases. After a minutes, the directional light source returns to the initial state. Judge the lighting rendering effect by observing the brightness information of the texture in the anime natural scene within a minutes, so as to achieve the correction of the lighting angle.

[0010] Preferably, the anime scene optimization unit includes a neural renderer design module. With the aim of optimizing the rendering effect of the anime natural scene, by combining the traditional renderer and the neural renderer, a multi-scale rendering strategy is provided to efficiently process the scene image. The specific operation is to design a hybrid rendering framework, which includes a traditional renderer and a neural renderer. And the neural renderer contains two sub-networks, namely the adaptive rendering network ARNet and the iterative upsampling network IUNet. For high-resolution scene images, ARNet calculates the adaptive downsampling ratio according to the blur parameter of the scene image and the preset maximum blur amount of the network, and downsamples the input scene image, and then renders. And to achieve a smooth transition when fusing the results of the two renderers, the scene image is restored to the original resolution size through bilinear upsampling. For low-resolution scene images, IUNet is used for iterative upsampling.

[0011] Preferably, the anime scene optimization unit further includes a multi-plane representation rendering optimization module. The multi-plane representation rendering optimization module is electrically connected to the neural renderer design module. In response to the problem that when the foreground should be focused, the foreground appears blurred and semi-transparent, causing some originally occluded background areas to be revealed, the multi-plane representation rendering optimization module uses the multi-plane image representation method to design a rendering framework based on background filling for rendering optimization. The specific operation is to use an encoder-decoder network MPI-Net to obtain the initial multi-plane image representation of the revealed background image. The encoder uses a pre-trained ResNet-18. For each scene image, the encoder runs only once, while the decoder needs to run once for each disparity plane of the scene image, so as to generate scene representations of multiple planes. Combining with the existing image filling method, fill the background image area where the foreground is occluded and revealed, and perform rendering optimization in the way of depth stratification and layer-by-layer rendering.

[0012] Preferably, the rendering quality management unit includes a renderability definition module and a rendering quality calculation module. The renderability definition module defines the renderability parameter based on the sources of rendering errors in the anime scene image. The sources of errors specifically include that the pixel resolution of the source view is lower than that of the target view and the incorrect reprojection caused by the 3D reconstruction error. The calculation is performed by pixel-by-pixel evaluation and then summation. The specific calculation formula is as follows:

[0013] In the formula, measures the quality of the image pixel color of point at the selected source view that can be restored at the target view , measures the accuracy of the color synthesis of the background light corresponding to the pixel, measures the integrity of the overall rendering result, represents the set of rays projected by the light center of the directional light source, represents the set of intersection points of the rays intersecting with the proxy geometry in represents the set of rays intersecting with the proxy geometry in

[0014] Compared with the prior art, the beneficial effects of the present invention at least include: The present invention proposes an animation scene rendering system based on three-dimensional computer-aided design. An animation scene rendering system based on three-dimensional computer-aided design is applied to the rendering of animation natural scenes. The three-dimensional modeling unit of the animation scene is set to perform three-dimensional modeling on animation characters and animation scenes. The rendering unit of the animation scene is set to differentiate the texture rendering characteristics of the animation natural scene, thereby optimizing the rendering effect of the terrain in the animation natural scene. And in order to improve the lighting rendering effect of the animation natural scene during the day-night alternation process, an illuminance correction strategy is designed. The optimization unit of the animation scene is set with the purpose of optimizing the rendering effect of the animation natural scene. By combining the traditional renderer and the neural renderer, a multi-scale rendering strategy is provided. And for the problem that when the foreground should be focused, the foreground appears blurred and semi-transparent, resulting in the exposure of some originally occluded background areas, a multi-plane image representation method is adopted, and a rendering framework based on background filling is designed for rendering optimization. The rendering quality management unit is set to define the renderability parameter, and then determine the rendering quality of the animation scene. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a schematic diagram of the overall process of an embodiment of the present invention; Figure 2 It is a schematic diagram of the process of the three-dimensional modeling unit of the animation scene in an embodiment of the present invention; Figure 3 It is a schematic diagram of the process of the rendering unit of the animation scene in an embodiment of the present invention; Figure 4 It is a schematic diagram of the process of the optimization unit of the animation scene in an embodiment of the present invention; Figure 5 It is a schematic diagram of the process of the rendering quality management unit in an embodiment of the present invention.

[0016] In the figure: 100, three-dimensional modeling unit of the animation scene; 101, module for obtaining animation scene requirements; 102, module for constructing the character scene model; 200, rendering unit of the animation scene; 201, module for constructing the texture rendering network; 202, module for interactive rendering of the scene; 203, module for lighting rendering of the scene; 300, optimization unit of the animation scene; 301, module for designing the neural renderer; 302, module for rendering optimization of multi-plane representation; 400, rendering quality management unit; 401, module for defining renderability; 402, module for calculating rendering quality. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0018] Please refer to Figures 1-5 , the present invention provides a technical solution: a 3D computer-aided design-based animation scene rendering system, including an animation scene 3D modeling unit 100. A 3D computer-aided design-based animation scene rendering system is applied to the rendering of animation natural scenes. The animation scene 3D modeling unit 100 obtains the design requirements of the animation scene and conducts the preliminary preparation work for the production of the animation scene. It uses 3D modeling software to perform 3D modeling on the animation characters and the animation scene, and gradually adds details to make the animation character scene more exquisite. An animation scene rendering unit 200. The animation scene rendering unit 200 differentiates the texture rendering characteristics of the animation natural scene by constructing a texture rendering network, thereby optimizing the rendering effect of the terrain in the animation natural scene. When rendering the same type of vegetation in the animation natural scene, it designs a rendering strategy. In order to improve the lighting rendering effect during the day-night alternation process of the animation natural scene, it designs an illuminance correction strategy. An animation scene optimization unit 300. The animation scene optimization unit 300 aims to optimize the rendering effect of the animation natural scene. By combining a traditional renderer and a neural renderer, it provides a multi-scale rendering strategy to efficiently process the scene image. For the problem that when the foreground should be focused, the foreground appears blurred and semi-transparent, resulting in the exposure of some originally occluded background areas, it adopts a multi-plane image representation method and designs a rendering framework based on background filling for rendering optimization. A rendering quality management unit 400. The rendering quality management unit 400 defines a renderability parameter based on the source of the rendering error of the animation scene image, and then determines the rendering quality of the animation scene through this parameter of renderability.

[0019] The animation scene 3D modeling unit 100 includes an animation scene requirement acquisition module 101 and a character scene model construction module 102. The animation scene requirement acquisition module 101 obtains the design requirements of the animation scene and conducts the preliminary preparation work for the production of the animation scene, including drawing a sketch of the scene, determining the overall style layout of the scene, and collecting relevant reference images. The character scene model construction module 102 is electrically connected to the animation scene requirement acquisition module 101. The character scene model construction module 102 uses 3D modeling software to perform 3D modeling on the animation characters and the animation scene, and gradually adds details, including buildings, vegetation, and props in the scene, to make the animation character scene more exquisite. The anime scene rendering unit 200 includes a texture rendering network construction module 201. The texture rendering network construction module 201 constructs a texture rendering network to differentiate the texture rendering features of the anime natural scene for the rendering of the terrain in the anime natural scene. The specific operation is as follows: First, a composite terrain layer is constructed by adding terrain attributes. The natural environment includes various terrains such as soil, rocks, and vegetation. The texture rendering network uses multiple different types of layers to cover the surface of the terrain model in the anime scene. Second, to ensure the texture quality, PBR-standardized texture materials are selected from the 3D scan resource library Quixel Megascans. During the material rendering process, considering that the shader takes a lot of time to sample each texture map, to optimize the size of the texture space, the number of sampled textures in the material is kept to a minimum, and only two texture maps, namely the base color and the normal, are retained. Finally, the texture features are differentiated from two aspects: texture coordinate scaling and noise mixing. The operation of differentiating texture features from the aspect of texture coordinate scaling is that the texture forms a two-dimensional coordinate system with the UV channel, and the material assigns a default coordinate system to the input texture to map it to the bound mesh. When the texture density does not correctly match the size of the mesh, a spare texture coordinate system is applied to the texture sample through the TexCoord node to avoid occupying additional texture memory. For the input texture, the texture is scaled twice according to the near-distance UV coefficient and the far-distance UV coefficient respectively to obtain the near-distance display texture and the far-distance display texture. The operation of differentiating texture features from the aspect of noise mixing is that since the noise texture is also a grayscale texture, as the pixel value ranges from 0 to 1, the pixel color transitions from black 0sRGB to white 255sRGB, and the black, white, and gray parts are randomly distributed in the texture area. Therefore, applying the noise texture to both the near-distance display texture and the far-distance display texture can ensure that the adjacent laid textures will not align with each other; The anime scene rendering unit 200 includes a scene interaction rendering module 202. The scene interaction rendering module 202 is electrically connected to the texture rendering network construction module 201. When the same type of vegetation in the anime natural scene needs to be rendered, the scene interaction rendering module 202 designs a rendering strategy. The specific operation is as follows: The hierarchical static mesh instance is used to render the same type of vegetation in the anime natural scene. The LOD technology and the distance culling technology are applied to the vegetation individuals through the hierarchical static mesh instance to realize the corresponding reduction of the number of polygons of the mesh according to the distance between the mesh and the camera. And since the number of draw calls is an important factor affecting the time spent on rendering 1 frame of the picture, each material of each static mesh in the scene corresponds to 1 draw call to the GPU. When there are multiple identical objects in the scene, only 1 mesh instance needs to be saved in the memory for the draw call, and all the same materials of the objects in the scene can be rendered through 1 draw call; The anime scene rendering unit 200 further includes a scene lighting rendering module 203. The scene lighting rendering module 203 is electrically connected to the scene interaction rendering module 202. In order to improve the lighting rendering effect of the anime natural scene during the day-night alternation process, the scene lighting rendering module 203 designs an illuminance correction strategy. The specific operation is to adjust the Pitch of the rotating body attribute of the directional light source in the scene, thereby changing the incident angle of the light. Assume that a minutes in the real scene represents 24 hours in the anime natural scene. When the directional light source rotates at a constant speed, D per second in the real scene corresponds to 2 / 5a hours in the anime natural scene. Update the time in the anime natural scene through the time axis, and then calculate the rotation angle of the directional light source at the current moment through the directional light source angle algorithm. The specific calculation formula is as follows:

[0020] In the formula, represents the time in the anime natural scene. In the initial state, the rotation angle of the directional light source is 90°. As time changes, the number of light rays emitted by the directional light source onto the upper surface of the anime natural scene increases first and then slowly decreases. After a minutes, the directional light source returns to the initial state. Judge the lighting rendering effect by observing the brightness information of the texture in the anime natural scene within a minutes, so as to achieve the correction of the lighting angle; The anime scene optimization unit 300 includes a neural renderer design module 301. The neural renderer design module 301 aims to optimize the rendering effect of the anime natural scene. By combining the traditional renderer and the neural renderer, it provides a multi-scale rendering strategy for efficiently processing scene images. The specific operation is to design a hybrid rendering framework, which includes a traditional renderer and a neural renderer. The neural renderer contains two sub-networks, namely the adaptive rendering network ARNet and the iterative upsampling network IUNet. For high-resolution scene images, ARNet calculates the adaptive downsampling ratio according to the blur parameter of the scene image and the preset maximum blur amount of the network, and downsamples the input scene image, and then renders. In order to achieve a smooth transition when fusing the results of the two renderers, the scene image is restored to the original resolution size through bilinear upsampling. For low-resolution scene images, IUNet is used for iterative upsampling; The anime scene optimization unit 300 further includes a multi-plane representation rendering optimization module 302. The multi-plane representation rendering optimization module 302 is electrically connected to the neural renderer design module 301. For the problem that when the foreground should be focused, the foreground appears blurred and semi-transparent, resulting in some originally occluded background areas being revealed, the multi-plane representation rendering optimization module 302 adopts the multi-plane image representation method and designs a rendering framework based on background filling for rendering optimization. The specific operation is to use an encoder-decoder network MPI-Net to obtain the initial multi-plane image representation of the revealed background image. The encoder uses the pre-trained ResNet-18. For each scene image, the encoder runs only once, while the decoder needs to run once for each disparity plane of the scene image, so as to generate the scene representation of multiple planes. Combining with the existing image filling method, fill the background image area where the foreground is occluded and revealed, and perform rendering optimization in the way of depth stratification and layer-by-layer rendering; The rendering quality management unit 400 includes a renderability definition module 401 and a rendering quality calculation module 402. The renderability definition module 401 defines the renderability parameter based on the sources of rendering errors in the anime scene image. The sources of errors specifically include that the pixel resolution of the source view is lower than that of the target view and the incorrect reprojection caused by the 3D reconstruction error, and calculates it by the method of pixel-by-pixel evaluation and then summation. The specific calculation formula is as follows:

[0021] In the formula, measures the quality of the image pixel color of point at the selected source view that can be restored at the target view , measures the accuracy of the color synthesis of the background light corresponding to the pixel, measures the integrity of the overall rendering result, represents the set of rays projected by the light center of the directional light source, represents the intersection set of the rays intersecting with the proxy geometry in represents the set of rays intersecting with the proxy geometry in . The rendering quality calculation module 402 is electrically connected to the renderability definition module 401. The rendering quality calculation module 402 calculates the renderability parameter for the rendered anime natural scene and compares it with the preset threshold to determine the rendering quality.

[0022] Working principle: The anime scene requirement acquisition module 101 obtains the anime scene design requirements and conducts the preliminary preparation work for anime scene production. The character and scene model construction module 102 performs 3D modeling on anime characters and anime scenes and gradually adds details. The texture rendering network construction module 201 constructs a texture rendering network to differentiate the texture rendering features of the anime natural scene. When the same type of vegetation in the anime natural scene needs to be rendered, the scene interaction rendering module 202 designs a rendering strategy. The scene lighting rendering module 203 designs an illuminance correction strategy to improve the lighting rendering effect of the anime natural scene during the day-night alternation process. The neural renderer design module 301 combines the traditional renderer and the neural renderer to provide a multi-scale rendering strategy for efficiently processing scene images. The multi-plane representation rendering optimization module 302 adopts the multi-plane image representation method to design a rendering framework based on background filling for rendering optimization. The renderability definition module 401 defines the renderability parameters based on the sources of rendering errors in the anime scene images. The rendering quality calculation module 402 calculates the renderability parameters for the rendered anime natural scene and compares them with the preset threshold to determine the rendering quality.

[0023] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or device.

[0024] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principle and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An animation scene rendering system based on three-dimensional computer-aided design, characterized in that: It includes a 3D modeling unit for anime scenes (100). An anime scene rendering system based on 3D computer-aided design is applied to the rendering of anime natural scenes. The 3D modeling unit for anime scenes (100) obtains the design requirements of the anime scene and conducts the preliminary preparation work for the production of the anime scene. It uses 3D modeling software to perform 3D modeling on anime characters and anime scenes, and gradually adds details to make the anime character scene more exquisite. Anime scene rendering unit (200). The anime scene rendering unit (200) differentiates the texture rendering characteristics of the anime natural scene by constructing a texture rendering network, thereby optimizing the rendering effect of the terrain in the anime natural scene. When rendering the same type of vegetation in the anime natural scene, it designs a rendering strategy. In order to improve the lighting rendering effect during the day-night alternation of the anime natural scene, it designs an illuminance correction strategy. Anime scene optimization unit (300). The anime scene optimization unit (300) aims to optimize the rendering effect of the anime natural scene. By combining a traditional renderer and a neural renderer, it provides a multi-scale rendering strategy to efficiently process scene images. For the problem that when the foreground should be focused, the foreground appears blurred and semi-transparent, resulting in some originally occluded background areas being revealed, it adopts a multi-plane image representation method and designs a rendering framework based on background filling for rendering optimization. Rendering quality management unit (400). The rendering quality management unit (400) defines a renderability parameter based on the source of the rendering error of the anime scene image, and then determines the rendering quality of the anime scene through this parameter of renderability.

2. The 3D computer-aided design-based animation scene rendering system according to claim 1, wherein: The 3D modeling unit for anime scenes (100) includes an anime scene requirement acquisition module (101) and a character scene model construction module (102). The anime scene requirement acquisition module (101) obtains the design requirements of the anime scene and conducts the preliminary preparation work for the production of the anime scene, including drawing a sketch of the scene, determining the overall style layout of the scene, and collecting relevant reference images. The character scene model construction module (102) is electrically connected to the anime scene requirement acquisition module (101). The character scene model construction module (102) uses 3D modeling software to perform 3D modeling on anime characters and anime scenes, and gradually adds details, including buildings, vegetation, and props in the scene, to make the anime character scene more exquisite.

3. A 3D computer-aided design-based anime scene rendering system according to claim 1, characterized in that: The anime scene rendering unit (200) includes a texture rendering network construction module (201). The texture rendering network construction module (201) constructs a texture rendering network to differentiate the texture rendering features of the anime natural scene for the rendering of the terrain in the anime natural scene. The specific operation is to first increase the terrain attributes to construct a composite terrain layer. The natural environment includes various terrains such as soil, rocks, and vegetation. The texture rendering network uses multiple different types of layers to cover the surface of the terrain model in the anime scene. Secondly, to ensure the texture quality, PBR-standardized texture materials are selected from the 3D scanning resource library Quixel Megascans. During the material rendering process, considering that the shader needs to spend a lot of time sampling each texture map, in order to optimize the size of the texture space, the number of textures sampled in the material is kept to a minimum, and only two texture maps, namely the base color and the normal, are retained. Finally, the texture features are differentiated from two aspects: texture coordinate scaling and noise mixing.

4. The 3D computer-aided design-based animation scene rendering system according to claim 3, wherein: The anime scene rendering unit (200) includes a scene interaction rendering module (202). The scene interaction rendering module (202) is electrically connected to the texture rendering network construction module (201). When the same type of vegetation in the anime natural scene needs to be rendered, the scene interaction rendering module (202) designs a rendering strategy. The specific operation is to use hierarchical static mesh instances to render the same type of vegetation in the anime natural scene, and apply LOD technology and distance culling technology to the vegetation individuals through the hierarchical static mesh instances, so as to reduce the number of polygons of the mesh according to the distance between the mesh and the camera. And since the number of draw calls is an important factor affecting the time spent on rendering 1 frame of the screen, each material of each static mesh in the scene corresponds to 1 draw call to the GPU. When there are multiple identical objects in the scene, only 1 mesh instance needs to be saved in the memory for the draw call, and all the same materials of the objects in the scene can be rendered through 1 draw call.

5. The anime scene rendering system based on 3D computer-aided design according to claim 4, characterized in that: The anime scene rendering unit (200) further includes a scene lighting rendering module (203). The scene lighting rendering module (203) is electrically connected to the scene interaction rendering module (202). In order to improve the lighting rendering effect of the anime natural scene during the day-night alternation process, the scene lighting rendering module (203) designs an illuminance correction strategy. The specific operation is to adjust the Pitch of the rotating body attribute of the directional light source in the scene, thereby realizing the change of the incident angle of the light. Assuming that a minutes in the real scene represents 24 hours in the anime natural scene, when the directional light source rotates at a constant speed, D per second in the real scene corresponds to 2 / 5a hours in the anime natural scene. Update the time in the anime natural scene through the time axis, and then calculate the rotation angle of the directional light source at the current moment through the directional light source angle algorithm. In the initial state, the rotation angle of the directional light source is 90°. As time changes, the number of light rays irradiated by the directional light source on the upper surface of the anime natural scene increases first and then slowly decreases. After a minutes, the directional light source returns to the initial state. Judge the lighting rendering effect by observing the brightness information of the texture in the anime natural scene within a minutes, so as to realize the lighting angle correction.

6. The 3D computer-aided design-based animation scene rendering system according to claim 1, wherein: The anime scene optimization unit (300) includes a neural renderer design module (301). The neural renderer design module (301) aims to optimize the rendering effect of the anime natural scene. By combining the traditional renderer and the neural renderer, it provides a multi-scale rendering strategy for efficiently processing scene images. The specific operation is to design a hybrid rendering framework, which includes a traditional renderer and a neural renderer. The neural renderer contains two sub-networks, namely the adaptive rendering network ARNet and the iterative upsampling network IUNet. For high-resolution scene images, ARNet calculates the adaptive downsampling ratio according to the blur parameter of the scene image and the preset maximum blur amount of the network, and downsamples the input scene image, and then renders. And to achieve a smooth transition when fusing the results of the two renderers, the scene image is restored to the original resolution size through bilinear upsampling. For low-resolution scene images, IUNet is used for iterative upsampling.

7. An anime scene rendering system based on 3D computer-aided design according to claim 6, characterized in that: The anime scene optimization unit (300) further includes a multi-plane representation rendering optimization module (302). The multi-plane representation rendering optimization module (302) is electrically connected to the neural renderer design module (301). When the foreground should be focused and appears blurred and semi-transparent, resulting in some originally occluded background areas being revealed, the multi-plane representation rendering optimization module (302) uses the multi-plane image representation method to design a rendering framework based on background filling for rendering optimization. The specific operation is to use an encoder-decoder network MPI-Net to obtain the initial multi-plane image representation of the revealed background image. The encoder uses a pre-trained ResNet-18. For each scene image, the encoder runs only once, while the decoder needs to run once for each disparity plane of the scene image, so as to generate scene representations of multiple planes. Combining with the existing image filling method, fill the occluded foreground and the revealed background image area, and perform rendering optimization in the way of depth stratification and layer-by-layer rendering.

8. An animation scene rendering system based on three-dimensional computer-aided design according to claim 1, characterized in that: The rendering quality management unit (400) includes a renderability definition module (401) and a rendering quality calculation module (402). The renderability definition module (401) defines the renderability parameter based on the sources of rendering errors in the anime scene image. The sources of errors specifically include that the pixel resolution of the source view is lower than that of the target view and the incorrect reprojection caused by the 3D reconstruction error, and calculates it by means of pixel-by-pixel evaluation and then summation. The rendering quality calculation module (402) is electrically connected to the renderability definition module (401). The rendering quality calculation module (402) calculates the renderability parameter for the rendered anime natural scene and compares it with a preset threshold to determine the rendering quality.

Images