An efficient graphics rendering system based on artificial intelligence core computing power

Abstract

The present invention relates to the field of graphics rendering, and in particular to an efficient graphics rendering system based on artificial intelligence core computing power, comprising: an intelligent task decoupling module: performing task decoupling on graphics rendering tasks; an AI enhanced texture generation optimization module: learning the changing rules of dynamic scenes to generate texture maps in real time, using a reuse algorithm to reduce computing overhead in repeated texture areas, and using a deep convolutional neural network to perform super-resolution compensation on low-resolution textures; an AI driven lighting and shadow calculation module: using AI to predict the lighting transmission path in the scene and reconstruct dynamic shadow effects to achieve high global lighting and shadow simulation; a task result synchronization module: synchronizing the rendering data output by each subtask; the present invention dynamically generates high-quality texture maps and performs super-resolution compensation on low-resolution textures, avoids repeated calculations through a texture reuse algorithm, significantly reduces computing overhead and improves the real-time rendering efficiency of large-scale complex scenes.

Description

Technical Field

[0001] The present invention relates to the field of graphics rendering technology, and in particular to an efficient graphics rendering system based on artificial intelligence core computing power. Background Art

[0002] With the rapid development of virtual reality (VR), augmented reality (AR), film and television special effects and 3D games, graphics rendering technology is facing increasingly complex scenes and high-quality real-time rendering requirements. Traditional graphics rendering technology relies on methods such as ray tracing and rasterization. These methods often have huge computational requirements when dealing with complex light and shadow effects and high-dynamic scenes, and it is difficult to balance the requirements of rendering quality and real-time performance.

[0003] There are still the following problems in effectively applying AI technology to graphics rendering tasks:

[0004] Challenges in processing dynamic scenes: In dynamic scenes, lighting, shadows, and textures change frequently over time and from different viewpoints. Traditional methods have difficulty achieving real-time synchronization and efficient computing, especially in large-scale scenes, where rendering delays and errors are particularly prominent. Summary of the invention

[0005] The present invention provides a high-efficiency graphics rendering system based on artificial intelligence core computing power.

[0006] An efficient graphics rendering system based on artificial intelligence core computing power, including:

[0007] Intelligent task decoupling module:

[0008] Decouple the graphics rendering task, which includes the lighting and shadow processing subtasks and the texture generation subtasks.

[0009] AI enhanced texture generation optimization module:

[0010] Generative adversarial networks (GANs) are used to process the texture generation subtask, learn the changing rules of dynamic scenes to generate high-quality texture maps in real time, use reuse algorithms in repeated texture areas to reduce computational overhead, and use deep convolutional neural networks (DCNNs) to perform super-resolution compensation for low-resolution textures.

[0011] AI-driven lighting and shadow calculation module:

[0012] Based on the neural radiance field model, the lighting and shadow processing subtasks are processed, and AI is used to predict the light transmission path in the scene and reconstruct the dynamic shadow effect to achieve high global illumination and shadow simulation;

[0013] Task result synchronization module:

[0014] The rendering data output by each subtask is synchronized to eliminate rendering errors and time delays caused by task decoupling.

[0015] Optionally, the intelligent task decoupling module specifically includes:

[0016] Scene feature extraction: Analyze the input graphics rendering task to extract the lighting features, shadow distribution features, and texture requirement features in the scene;

[0017] Task decomposition: Based on the extracted scene features, the graphics rendering task is decomposed into lighting and shadow processing subtasks and texture generation subtasks, where:

[0018] The lighting and shadow processing subtasks include global illumination calculation and dynamic shadow generation;

[0019] The texture generation subtask includes high-quality texture map generation and optimization of repeated texture areas.

[0020] Optionally, the AI ​​enhanced texture generation optimization module includes a dynamic texture generation unit, a repeated texture multiplexing unit, and a low-resolution texture super-resolution compensation unit, wherein:

[0021] The dynamic texture generation unit generates texture maps that meet the scene requirements in real time;

[0022] The repeated texture multiplexing unit performs mapping processing on the repeated texture area;

[0023] The low-resolution texture super-resolution compensation unit performs super-resolution compensation on the low-resolution texture to repair the blurred area in the texture.

[0024] Optionally, the dynamic texture generation unit processes the texture generation subtask using a generative adversarial network, and generates a texture map that meets the scene requirements in real time by learning the dynamic change rules of the scene, wherein the generator of the generative adversarial network is responsible for generating detailed textures, and the discriminator of the generative adversarial network is used to optimize texture authenticity and visual consistency;

[0025] The generative adversarial network includes a generator and the discriminator , and its optimization objective is expressed as:

[0026] ,in, is the generator model, with noise Generate a pseudo texture as input , is the discriminator model, taking input Determine whether it is a real texture. is the objective function, representing the adversarial loss between the generator and the discriminator, is the distribution of the real texture, is the noise distribution, which is a Gaussian distribution, is the expectation operator, which is used to calculate the expected value of real and fake textures.

[0027] Optionally, the repeated texture reuse unit identifies a repeated texture area in the scene, calculates the similarity of feature vectors of two texture blocks by a feature matching algorithm, sets a similarity threshold, and identifies a repeated texture area if the similarity is greater than the similarity threshold, extracts the generated texture blocks, and uses a texture reuse algorithm to map the repeated texture area, thereby avoiding repeated calculations and significantly reducing calculation overhead;

[0028] The feature matching algorithm is calculated as follows: the cosine similarity of the texture block feature vector Determine whether a texture block is repeated: ,in, are the feature vectors of two texture patches, extracted from a pre-trained convolutional neural network, represents the dot product operation of vectors, is the Euclidean norm of the vector, is the cosine similarity, with a value range of [-1,1], The higher the similarity between the two texture blocks, the Exceeding the set threshold , indicating that the texture block is a repeated block, directly maps the generated texture block to avoid repeated calculation.

[0029] Optionally, the low-resolution texture super-resolution compensation unit performs super-resolution compensation on the low-resolution texture based on a deep convolutional neural network, reconstructs the high-resolution texture details through a multi-layer convolution structure, and repairs the blurred texture area through a residual network (ResNet) to ensure the clarity and detail fidelity of the output texture.

[0030] Optionally, the AI-driven illumination and shadow calculation module specifically includes:

[0031] Light transmission path prediction unit: Based on the neural radiance field model, the light transmission path is predicted using scene geometry data and light source parameters. Mapping input locations and direction To light intensity value ,in, Represents the neural radiance field model, the parameter set obtained through training , is the sampling point position in space, Indicates the direction of light, is the predicted light intensity;

[0032] Dynamic shadow generation unit: Analyzes the occlusion relationship between the light source and the geometric shape of the object in the scene, estimates the degree of illumination occlusion in the shadow area through the volume rendering integral in the neural radiance field model, and generates a dynamic shadow effect. The volume rendering integral is expressed as:

[0033] ,in, is the transmittance of light passing through, indicating the proportion of light that is not blocked. is the density function, indicating the position The light blocking intensity at is the color value function, indicating The color of the light at Represents the near and far boundaries of the ray in the scene, respectively.

[0034] Optionally, the AI-driven illumination and shadow calculation module also includes a global illumination optimization unit, which globally optimizes the illumination distribution results based on the light transmission path to enhance the realism of the global illumination. The global illumination distribution optimization needs to balance the illumination intensity in the scene to make it more visually consistent. The optimization formula is as follows:

[0035] ,in, It's location The global illumination distribution, is the direction predicted from the light transmission path The light intensity on is a weight function that adjusts the lighting distribution according to the directionality of the light source in the scene (for example, the main light source direction has a higher weight). Is the set of all light directions in the scene.

[0036] Optionally, the task result synchronization module includes a time synchronization unit: aligning the rendering data output by each subtask based on the timestamp, and using a time series interpolation method to synchronize the dynamic light and shadow results of the lighting and shadow processing subtask with the output texture of the texture generation subtask to ensure consistency in the time dimension.

[0037] Optionally, the time series interpolation method specifically includes:

[0038] Data sampling and time alignment: Sample output data from the lighting and shadow processing subtask and the texture generation subtask, and add a timestamp to each data to indicate the specific time point when it was generated. The goal is to align the output data of the two tasks in time so that they can be synchronized and matched at the same time point;

[0039] Interpolation synchronization: Based on the data of adjacent time frames, the ungenerated intermediate frame data is estimated by interpolation method;

[0040] Error detection and correction: After completing interpolation synchronization, check whether there is a significant time deviation or rendering artifact between lighting, shadows, and textures. If there is an error, correct the timestamp of the task output again or regenerate the required intermediate frame data.

[0041] Beneficial effects of the present invention:

[0042] The present invention decomposes complex graphics rendering tasks into lighting and shadow processing subtasks and texture generation subtasks through an intelligent task decoupling module, combines generative adversarial networks (GANs) and deep convolutional neural networks (DCNNs), dynamically generates high-quality texture maps and performs super-resolution compensation for low-resolution textures, and avoids repeated calculations through a texture reuse algorithm, significantly reducing computing overhead and improving the real-time rendering efficiency of large-scale complex scenes.

[0043] The present invention utilizes the neural radiosity field model in the AI-driven lighting and shadow calculation module to accurately predict and generate the lighting transmission path and dynamic shadow effects in the scene, and further balances the lighting distribution through the global lighting optimization unit to improve the realism and consistency of global lighting and dynamic shadows. At the same time, in the texture generation process, the generator and the discriminator collaborate to optimize the texture details to ensure the visual authenticity and consistency of the texture, and combine the residual network to repair the blurred texture area, thereby improving the accuracy and detail expression of the final rendering result.

[0044] The present invention, through the time synchronization unit in the task result synchronization module, adopts the time series interpolation method to time align and synchronize the dynamic light and shadow results of the illumination and shadow processing subtask with the output texture of the texture generation subtask, thereby ensuring the consistency of the time dimension after task decoupling in the dynamically changing scene and solving the rendering error and delay problems that are easily caused by the decoupling method. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0046] Figure 1 A schematic diagram of system function modules according to an embodiment of the present invention;

[0047] Figure 2 Schematic diagram of an AI enhanced texture generation optimization module according to an embodiment of the present invention;

[0048] Figure 3 Schematic diagram of an AI-driven lighting and shadow calculation module according to an embodiment of the present invention. DETAILED DESCRIPTION

[0049] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. At the same time, it is explained here that in order to make the embodiments more detailed, the following embodiments are the best and preferred embodiments, and those skilled in the art may also adopt other alternatives to implement some known technologies; and the accompanying drawings are only for more specific description of the embodiments, and are not intended to specifically limit the present invention.

[0050] It should be noted that the references to "one embodiment", "embodiment", "exemplary embodiments", "some embodiments" and the like in the specification indicate that the embodiments described may include specific features, structures or characteristics, but not every embodiment may include the specific features, structures or characteristics. In addition, when a specific feature, structure or characteristic is described in conjunction with an embodiment, it should be within the knowledge of a person skilled in the art to implement such feature, structure or characteristic in conjunction with other embodiments (whether or not explicitly described).

[0051] In general, a term can be understood, at least in part, from its use in context. For example, depending, at least in part, on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in the singular sense, or can be used to describe a combination of features, structures, or characteristics in the plural sense. Additionally, the term "based on" can be understood as not necessarily intended to convey an exclusive set of factors, but can instead, depending, at least in part, on the context, allow for the presence of other factors that are not necessarily explicitly described.

[0052] like Figure 1-Figure 3 As shown, an efficient graphics rendering system based on artificial intelligence core computing power includes:

[0053] Intelligent task decoupling module:

[0054] Decouple the graphics rendering task, which includes the lighting and shadow processing subtasks and the texture generation subtasks.

[0055] AI enhanced texture generation optimization module:

[0056] Generative adversarial networks (GANs) are used to process the texture generation subtask, learn the changing rules of dynamic scenes to generate high-quality texture maps in real time, use reuse algorithms in repeated texture areas to reduce computational overhead, and use deep convolutional neural networks (DCNNs) to perform super-resolution compensation for low-resolution textures.

[0057] AI-driven lighting and shadow calculation module:

[0058] Based on the neural radiance field model, the lighting and shadow processing subtasks are processed, and AI is used to predict the light transmission path in the scene and reconstruct the dynamic shadow effect, achieving high global illumination and shadow simulation while reducing the computational cost of traditional ray tracing methods;

[0059] Task result synchronization module:

[0060] The rendering data output by each subtask is synchronized to eliminate rendering errors and time delays caused by task decoupling, ensuring the consistency of the final rendering results.

[0061] The intelligent task decoupling module specifically includes:

[0062] Scene feature extraction: Analyze the input graphics rendering task to extract the lighting features, shadow distribution features, and texture requirement features in the scene;

[0063] Task decomposition: Based on the extracted scene features, the graphics rendering task is decomposed into lighting and shadow processing subtasks and texture generation subtasks, where:

[0064] The lighting and shadow processing subtasks include global illumination calculation and dynamic shadow generation;

[0065] The texture generation subtask includes high-quality texture map generation and optimization of repeated texture areas.

[0066] Light feature extraction: Use a convolutional neural network model to analyze the number, intensity, and direction parameters of light sources in the scene;

[0067] Shadow distribution feature extraction technology: Apply occlusion detection based on image analysis, determine the occlusion relationship between objects through geometric models and light source positions, generate shadow areas in the scene, use edge detection to extract the boundaries of the shadow areas, and analyze the depth and layers of the shadows in combination with the depth map;

[0068] Texture requirement feature extraction technology: Based on the convolutional neural network, the resolution, detail level and repeated area features of the texture are extracted. Through feature clustering, the distribution of texture areas in the scene is analyzed to identify high-resolution and low-resolution texture requirement areas.

[0069] The AI ​​enhanced texture generation optimization module includes a dynamic texture generation unit, a repeated texture reuse unit, and a low-resolution texture super-resolution compensation unit, among which:

[0070] The dynamic texture generation unit generates texture maps that meet the scene requirements in real time;

[0071] The repeated texture multiplexing unit performs mapping processing on the repeated texture area;

[0072] The low-resolution texture super-resolution compensation unit performs super-resolution compensation on the low-resolution texture to repair the blurred area in the texture.

[0073] The dynamic texture generation unit uses a generative adversarial network to process the texture generation subtask. By learning the dynamic change rules of the scene, it generates texture maps that meet the scene requirements in real time. The generator of the generative adversarial network is responsible for generating detailed textures, and the discriminator of the generative adversarial network is used to optimize texture authenticity and visual consistency.

[0074] Generative adversarial networks include generators and the discriminator , and its optimization objective is expressed as:

[0075] ,in, is the generator model, with noise Generate a pseudo texture as input , is the discriminator model, taking input Determine whether it is a real texture. is the objective function, representing the adversarial loss between the generator and the discriminator, is the distribution of the real texture, is the noise distribution, which is a Gaussian distribution, is the expectation operator, used to calculate the expected value of real and pseudo textures;

[0076] The generator update rule is expressed as: ,in, are the parameters of the generator model, is the learning rate, is the gradient of the generator parameters;

[0077] The discriminator update rule is expressed as:

[0078] ,in, are the parameters of the discriminator model.

[0079] The repeated texture reuse unit identifies repeated texture areas in the scene, calculates the similarity of the feature vectors of two texture blocks through a feature matching algorithm, sets a similarity threshold, and identifies the area as a repeated texture area if it is greater than the similarity threshold. It extracts the generated texture blocks and uses a texture reuse algorithm to map the repeated texture area, avoiding repeated calculations and significantly reducing computational overhead.

[0080] The feature matching algorithm is calculated as follows: the cosine similarity of the texture block feature vector Determine whether a texture block is repeated: ,in, are the feature vectors of two texture patches, extracted from a pre-trained convolutional neural network, represents the dot product operation of vectors, is the Euclidean norm of the vector, is the cosine similarity, with a value range of [-1,1], The higher the similarity between the two texture blocks, the Exceeding the set threshold , indicating that the texture block is a repeated block, directly maps the generated texture block to avoid repeated calculation.

[0081] The low-resolution texture super-resolution compensation unit performs super-resolution compensation on low-resolution textures based on a deep convolutional neural network. It reconstructs high-resolution texture details through a multi-layer convolutional structure and repairs blurred texture areas through a residual network (ResNet) to ensure the clarity and detail fidelity of the output texture.

[0082] Deep convolutional neural network achieves low-resolution texture through multi-layer convolution operations To high resolution textures The mapping is expressed as:

[0083] ,in, is the output high-resolution texture, is the input low-resolution texture, is the convolution mapping function from input to output, with parameters , It is The convolution kernel weights of the layer, It is The bias term of the layer, * is the convolution operator, is the activation function, is the number of network layers;

[0084] The optimization goal of the residual network is to optimize the residual To compensate for blurry areas of the texture:

[0085] ,in, It is the texture detail information learned by the residual network;

[0086] Loss function: ,in, is the loss function, which represents the error between the predicted texture and the real high-resolution texture. is the number of texture samples, is a high-resolution texture generated by the network, It's a true high-resolution texture.

[0087] The AI-driven lighting and shadow calculation module specifically includes:

[0088] Light transmission path prediction unit: Based on the neural radiance field model, the light transmission path is predicted using scene geometry data and light source parameters. Mapping input locations and direction To light intensity value , to achieve efficient calculation of global illumination, where Represents the neural radiance field model, the parameter set obtained through training , is the sampling point position in space, Indicates the direction of light, is the predicted light intensity;

[0089] Dynamic shadow generation unit: Analyzes the occlusion relationship between the light source and the geometric shape of the object in the scene, estimates the degree of illumination occlusion in the shadow area through the volume rendering integral in the neural radiance field model, and generates a dynamic shadow effect. The volume rendering integral is expressed as:

[0090] ,in, is the transmittance of light passing through, indicating the proportion of light that is not blocked. is the density function, indicating the position The light blocking intensity at is the color value function, indicating The color of the light at Represents the near and far boundaries of the ray in the scene, respectively.

[0091] The AI-driven lighting and shadow calculation module also includes a global illumination optimization unit, which globally optimizes the illumination distribution results based on the light transmission path to improve the realism of global illumination. Global illumination distribution optimization needs to balance the illumination intensity in the scene to make it more visually consistent. The optimization formula is as follows:

[0092] ,in, It's location The global illumination distribution, is the direction predicted from the light transmission path The light intensity on is a weight function that adjusts the lighting distribution according to the directionality of the light source in the scene (for example, the main light source direction has a higher weight). Is the set of all light directions in the scene.

[0093] The task result synchronization module includes a time synchronization unit: it aligns the rendering data output by each subtask based on the timestamp, and uses the time series interpolation method to synchronize the dynamic light and shadow results of the lighting and shadow processing subtask with the output texture of the texture generation subtask to ensure consistency in the time dimension.

[0094] The time series interpolation methods include:

[0095] Data sampling and time alignment: Sample output data from the lighting and shadow processing subtask and the texture generation subtask, and add a timestamp to each data to indicate the specific time point when it was generated. The goal is to align the output data of the two tasks in time so that they can be synchronized and matched at the same time point;

[0096] Interpolation synchronization: Based on the data of adjacent time frames, the ungenerated intermediate frame data is estimated through interpolation. This process is performed separately between the light and shadow processing results and the texture map to fill the gaps on the time axis and ensure that the dynamically changing scenes are consistent in the time dimension.

[0097] Error detection and correction: After completing interpolation synchronization, check whether there is a significant time deviation or rendering artifact between lighting, shadows, and textures. If there is an error, correct the timestamp of the task output again or regenerate the required intermediate frame data.

[0098] Combine the aligned and corrected lighting and shadow results with the texture map data to generate the final rendering, ensuring temporal coherence and visual consistency.

[0099] The present invention covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present invention. In order to make the public have a thorough understanding of the present invention, specific details are described in detail in the following preferred embodiments of the present invention, but those skilled in the art can fully understand the present invention without the description of these details. In addition, in order to avoid unnecessary confusion about the essence of the present invention, well-known methods, processes, procedures, components and circuits are not described in detail.

[0100] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. An efficient graphics rendering system based on artificial intelligence core computing power, characterized in that: include: Intelligent task decoupling module: The graphics rendering task is decoupled, and the task decoupling includes the lighting and shadow processing subtask and the texture generation subtask. The intelligent task decoupling module specifically includes: Scene feature extraction: Analyze the input graphics rendering task to extract the lighting features, shadow distribution features, and texture requirement features in the scene; Task decomposition: Based on the extracted scene features, the graphics rendering task is decomposed into lighting and shadow processing subtasks and texture generation subtasks, where: The lighting and shadow processing subtasks include global illumination calculation and dynamic shadow generation; The texture generation subtask includes high-quality texture map generation and repeated texture area optimization; AI enhanced texture generation optimization module: Generative adversarial networks are used to process texture generation subtasks, learn the changing rules of dynamic scenes to generate texture maps in real time, use reuse algorithms in repeated texture areas to reduce computational overhead, and use deep convolutional neural networks to perform super-resolution compensation for low-resolution textures. AI-driven lighting and shadow calculation module: Based on the neural radiance field model, the lighting and shadow processing subtasks are processed, and AI is used to predict the light transmission path in the scene and reconstruct the dynamic shadow effect to achieve high global illumination and shadow simulation; Task result synchronization module: The rendering data output by each subtask is synchronized to eliminate rendering errors and time delays caused by task decoupling.

2. According to claim 1, a high-efficiency graphics rendering system based on artificial intelligence core computing power is characterized in that: The AI ​​enhanced texture generation optimization module includes a dynamic texture generation unit, a repeated texture multiplexing unit, and a low-resolution texture super-resolution compensation unit, wherein: The dynamic texture generation unit generates texture maps that meet the scene requirements in real time; The repeated texture multiplexing unit performs mapping processing on the repeated texture area; The low-resolution texture super-resolution compensation unit performs super-resolution compensation on the low-resolution texture to repair the blurred area in the texture.

3. The efficient graphics rendering system based on artificial intelligence core computing power according to claim 2, characterized in that: The dynamic texture generation unit uses a generative adversarial network to process the texture generation subtask, and generates a texture map that meets the scene requirements in real time by learning the dynamic change rules of the scene, wherein the generator of the generative adversarial network is responsible for generating detailed textures, and the discriminator of the generative adversarial network is used to optimize texture authenticity and visual consistency; The generative adversarial network includes a generator and the discriminator , and its optimization objective is expressed as: ,in, is the generator model, with noise Generate a pseudo texture as input , is the discriminator model, taking input Determine whether it is a real texture. is the objective function, representing the adversarial loss between the generator and the discriminator, is the distribution of the real texture, is the noise distribution, which is Gaussian distribution, is the expectation operator, which is used to calculate the expected value of real and fake textures.

4. The efficient graphics rendering system based on artificial intelligence core computing power according to claim 3, characterized in that: The repeated texture reuse unit identifies a repeated texture area in the scene, calculates the similarity of feature vectors of two texture blocks by a feature matching algorithm, sets a similarity threshold, identifies a repeated texture area if the similarity is greater than the similarity threshold, extracts the generated texture blocks, and uses a texture reuse algorithm to map the repeated texture area; The feature matching algorithm is calculated as follows: the cosine similarity of the texture block feature vector Determine whether a texture block is repeated: ,in, are the feature vectors of the two texture blocks, represents the dot product operation of vectors, is the Euclidean norm of the vector, is the cosine similarity, with a value range of [-1,1], The higher the similarity between the two texture blocks, the higher the similarity between the two texture blocks.

5. The efficient graphics rendering system based on artificial intelligence core computing power according to claim 4, characterized in that: The low-resolution texture super-resolution compensation unit performs super-resolution compensation on low-resolution textures based on a deep convolutional neural network, reconstructs high-resolution texture details through a multi-layer convolution structure, and repairs blurred texture areas through a residual network.

6. The efficient graphics rendering system based on artificial intelligence core computing power according to claim 1, characterized in that: The AI-driven illumination and shadow calculation module specifically includes: Light transmission path prediction unit: Based on the neural radiance field model, the light transmission path is predicted using scene geometry data and light source parameters. Mapping input locations and direction To light intensity value ,in, Represents the neural radiance field model, the parameter set obtained through training , is the sampling point position in space, Indicates the direction of light, is the predicted light intensity; Dynamic shadow generation unit: Analyzes the occlusion relationship between the light source and the geometric shape of the object in the scene, estimates the degree of illumination occlusion in the shadow area through the volume rendering integral in the neural radiance field model, and generates a dynamic shadow effect. The volume rendering integral is expressed as: ,in, is the transmittance of light passing through, indicating the proportion of light that is not blocked. is the density function, indicating the position The light blocking intensity at is the color value function, indicating The color of the light at Represents the near and far boundaries of the ray in the scene, respectively.

7. The efficient graphics rendering system based on artificial intelligence core computing power according to claim 6, characterized in that: The AI-driven illumination and shadow calculation module also includes a global illumination optimization unit. Global illumination distribution optimization requires balancing the illumination intensity in the scene. The optimization formula is as follows: ,in, It's location The global illumination distribution, is the direction predicted from the light transmission path The light intensity on is a weight function that adjusts the illumination distribution according to the directionality of the light source in the scene. Is the set of all light directions in the scene.

8. The efficient graphics rendering system based on artificial intelligence core computing power according to claim 1, characterized in that: The task result synchronization module includes a time synchronization unit: the rendering data output by each subtask is aligned based on the timestamp, and the dynamic light and shadow results of the illumination and shadow processing subtask and the output texture of the texture generation subtask are synchronized and adjusted using the time series interpolation method to ensure the consistency of the time dimension.

9. The efficient graphics rendering system based on artificial intelligence core computing power according to claim 8, characterized in that: The time series interpolation method specifically includes: Data sampling and time alignment: Sample output data from the lighting and shadow processing subtask and the texture generation subtask, and add a timestamp to each data to indicate the specific time point when it was generated. The goal is to align the output data of the two tasks in time so that they can be synchronized and matched at the same time point; Interpolation synchronization: Based on the data of adjacent time frames, the ungenerated intermediate frame data is estimated by interpolation method; Error detection and correction: After completing interpolation synchronization, check whether there is time deviation or rendering artifacts between lighting, shadows and textures. If there is an error, correct the timestamp of the task output again or regenerate the required intermediate frame data.

Images