High-efficiency three-dimensional animation production method based on artificial intelligence

Abstract

The invention relates to the technical field of computer vision, and discloses a high-efficiency three-dimensional animation production method based on artificial intelligence, which comprises the following steps: step 1, constructing a light propagation calculation model, and establishing the light propagation calculation model for describing the distribution relationship of illumination in different media based on the propagation characteristics of photons in a three-dimensional space, the light propagation calculation model comprises description of a photon motion state, and illumination intensity distribution is calculated in combination with scene geometric information and material characteristics; and step 2, calculating illumination distribution based on a light propagation calculation model, and determining the distribution condition of illumination in the three-dimensional scene by solving the photon motion state. Light path sampling is optimized through a Markov chain Monte Carlo method, adaptive photon propagation path selection based on the state transition probability is realized, the effects of reducing invalid light path calculation and improving the convergence speed of illumination calculation are achieved, and global illumination calculation in a complex illumination environment is efficiently carried out.

Description

Technical Field

[0001] The present invention relates to the technical field of computer vision, and particularly to a high-efficiency three-dimensional animation production method based on artificial intelligence. Background Technique

[0002] The three-dimensional animation production technology has developed rapidly in recent years and is widely used in the technical fields of film and television, games, and virtual reality. In the process of three-dimensional animation production, light calculation is a key factor affecting the final image quality and rendering efficiency. High-quality light simulation can enhance the realism of the image, making the light and shadow performance of characters and scenes more natural. However, there are still technical problems in the existing three-dimensional animation production methods in terms of light calculation, and the technical problems are as follows:

[0003] Traditional light calculation methods mainly rely on path tracing and photon mapping. The core idea is to simulate global illumination by sampling the light propagation path. However, this method requires a large number of samples to approximate the real light distribution, resulting in extremely high computational costs, especially in complex lighting environments.

[0004] In the process of three-dimensional animation production, light calculation usually needs to handle the global illumination of large-scale scenes and the local illumination of complex geometric structures. Traditional global illumination calculation methods, such as radiosity method and path tracing, can better simulate the multiple reflections of light, but the computational complexity increases sharply in large-scale scenes and cannot be processed in real time.

[0005] Shadows are an important part of three-dimensional animation rendering, which can enhance the layering of the image and make the spatial relationship between characters and the environment clearer. The current shadow calculation methods mainly include shadow mapping and ray tracing shadows. Shadow mapping has high computational efficiency, but it is easy to produce jagged artifacts at the shadow edges, resulting in unnatural shadows; ray tracing shadows can provide high-quality dynamic shadows, but the computational cost is relatively high, especially in the case of dynamic light sources and complex geometric structures, and the shadow boundaries are prone to discontinuity or blurring problems.

[0006] In three-dimensional animation production, the materials in the scene usually have different optical properties. Traditional light calculation methods use a fixed reflectivity model to calculate the light response of materials, and cannot dynamically adjust the light adaptability of materials under different light source conditions, resulting in low light calculation accuracy and affecting the consistency and authenticity of the image.

[0007] High dynamic range rendering technology is used to simulate the light intensity changes in the real world, making the bright and dark details in the three-dimensional animation scene clearer. However, traditional light calculation methods are prone to overexposure in the bright part and loss of dark details in the high dynamic range environment, resulting in insufficient layering of the image.

[0008] Therefore, those skilled in the art provide a high-efficiency three-dimensional animation production method based on artificial intelligence to solve the above-mentioned problems. Summary of the Invention

[0009] Aiming at the deficiencies of the prior art, the present invention provides a high-efficiency three-dimensional animation production method based on artificial intelligence to solve the problems raised in the above background art.

[0010] To achieve the above objectives, the present invention is realized through the following technical solutions: A high-efficiency three-dimensional animation production method based on artificial intelligence, including:

[0011] Step 1: Construct a light propagation calculation model. Based on the propagation characteristics of photons in three-dimensional space, establish a light propagation calculation model that describes the distribution relationship of light in different media. Among them, the light propagation calculation model includes a description of the motion state of photons, and calculates the light intensity distribution by combining scene geometric information and material characteristics;

[0012] Step 2: Calculate the light distribution based on the light propagation calculation model. By solving the motion state of photons, determine the distribution of light in the three-dimensional scene, and adjust the attenuation and scattering characteristics of light by combining the material parameters of the scene to generate preliminary light data;

[0013] Step 3: Optimize the preliminary light data through a neural network. Based on the constructed light propagation calculation model and the light distribution situation, use a neural network to perform feature learning on the light data, and adjust the calculated light distribution to make the light data adapt to the characteristics of different scene light sources. Among them, the input of the neural network is the light distribution data, and the output is the optimized light distribution information;

[0014] Step 4: Optimize the light path based on the optimized light distribution. According to the light distribution information optimized by the neural network, use the path optimization method to adjust the propagation path of light in three-dimensional space, and optimize the directivity of light by combining the photon propagation state to make the propagation of light on the surfaces of different media conform to optical characteristics;

[0015] Step 5: Calculate the global illumination based on the light path optimization result. According to the propagation information obtained by the light path optimization, combined with the geometric structure of the three-dimensional scene, use the illumination calculation method to solve the global illumination and generate global illumination data. The global illumination data is used as the input information for the final light rendering;

[0016] Step 6: Render the anime scene based on the global illumination data. According to the calculated global illumination data, combined with the material properties and illumination direction in the 3D anime scene, perform illumination mapping calculation on the objects in the scene to generate a 3D anime image that conforms to the illumination characteristics. Among them, the illumination mapping calculation adjusts the light and shadow in combination with the illumination distribution information in the global illumination data to ensure that the rendering result conforms to the illumination changes in the 3D scene.

[0017] Preferably, the light propagation calculation model in Step 1 calculates the illumination intensity distribution using the light propagation formula based on wave optics. The calculation formula is as follows:

[0018] E(r, t) = E 0 e i(k·r-ωt) ,

[0019] where E(r, t) represents the amplitude of the light wave at the 3D spatial position r;

[0020] E 0 is the initial amplitude of the light, representing the energy intensity of the light source;

[0021] e i(k·r-ωt) is the phase factor of the light wave, describing the propagation of the light wave in space and time;

[0022] k is the wave vector of the light, describing the propagation direction of the light in different media;

[0023] ω is the angular frequency of the light, reflecting the time evolution characteristics of the light wave;

[0024] t is the time variable.

[0025] Preferably, the calculation of the illumination distribution in Step 2 uses the Monte Carlo integration method to calculate the photon state. The calculation formula is as follows:

[0026]

[0027] where I(r) is the illumination intensity at the 3D spatial position r;

[0028] N is the number of Monte Carlo samplings, representing the number of samples used when calculating the illumination;

[0029] f(r i ) is the contribution factor of the photon at the position r i ;

[0030] P(r i ) is the probability distribution of the photon at the position r i .

[0031] Preferably, in step 3, the neural network optimizes the illumination data by using a convolutional neural network to extract illumination features and calculate the optimization error of the illumination data. The loss function is as follows:

[0032]

[0033] where L cnn is the loss function for neural network optimized illumination calculation;

[0034] M is the number of training samples;

[0035] I cnn (s j ) is the illumination intensity calculated by the neural network;

[0036] I gt (s j ) is the true illumination intensity;

[0037] s j is the spatial position of the training sample.

[0038] Preferably, in step 4, the light path is optimized by using the Markov chain Monte Carlo method to adjust the propagation path of light. The probability calculation of the photon path is as follows:

[0039] P(x k+1 |x k ) = A(x k , x k+1 )P(x k+1 ),

[0040] where P(x k+1 |x k ) is the probability that the photon propagates from x k to x k+1 ;

[0041] x k is the current photon position, representing the propagation point of the photon in space;

[0042] x k+1 is the candidate for the next photon propagation position;

[0043] A(x k , x k+1 ) is the acceptance probability, calculated as follows:

[0044]

[0045] where P(x k+1 ) is the probability of the candidate new path point x k+1 ;

[0046] P(x k ) is the current path point xk probability

[0047] Preferably, in step 5, the global illumination calculation uses the finite element method to solve the illumination distribution, and the calculation formula is as follows:

[0048]

[0049] where K lm is the finite element matrix, representing the mutual influence of light in different spatial units;

[0050] I m is the illumination intensity of the m-th unit;

[0051] F l is the contribution of the external light source to the l-th unit;

[0052] F l is the total number of units divided by the finite element.

[0053] Preferably, in step 6, the boundary element method is used to optimize the local illumination calculation in the anime scene rendering, and the calculation formula is as follows:

[0054] K·I = F,

[0055] where K is the boundary element matrix, representing the propagation relationship of the boundary illumination;

[0056] I is the boundary illumination intensity distribution vector;

[0057] F is the boundary light source contribution vector.

[0058] Preferably, in step 6, when performing the anime scene rendering, the adaptive illumination mapping method is used to adjust the illumination reflectivity of the material, and the calculation formula is as follows:

[0059] R mat (λ) = R 0 +α(λ - λ 0 ),

[0060] where R mat (λ) is the material illumination reflectivity at wavelength λ;

[0061] R 0 is the reflectivity at the reference wavelength λ 0 ;

[0062] α is the reflectivity change coefficient of the material;

[0063] λ is the wavelength of the current light; λ 0 is the reference wavelength.

[0064] Preferably, in step 6, during the animation scene rendering process, the shadow distribution of the object is adjusted based on the illumination direction, and the shadow generation function is calculated as follows:

[0065] S(p,θ)=max(0,cosθ),

[0066] Where S(p, θ) is the shadow intensity at position p in three-dimensional space;

[0067] p is the spatial position in the three-dimensional scene, indicating the coordinates of the point where the shadow needs to be calculated;

[0068] θ is the incident angle of light; cosθ is the angle between the object surface and the light source direction.

[0069] Preferably, in step 6, in order to enhance light and shadow details during the animation scene rendering process, the light brightness is adjusted using a high dynamic range, and the calculation formula is as follows:

[0070]

[0071] Among them, L hdr (x) is the light brightness after high dynamic range adjustment;

[0072] L raw (x) is the original calculated light intensity.

[0073] The present invention provides a high-efficiency three-dimensional animation production method based on artificial intelligence. It has the following beneficial effects:

[0074] 1. The present invention optimizes light path sampling through the Markov chain Monte Carlo method, realizes adaptive photon propagation path selection based on state transition probability, reduces invalid light path calculations, and improves the convergence speed of illumination calculations, so that global illumination calculations in complex illumination environments can be performed efficiently.

[0075] 2. The present invention optimizes illumination calculation by combining the finite element method with the boundary element method to implement a hierarchical calculation strategy for global illumination and local illumination, so that global illumination is used for large-scale scene calculations and local illumination optimizes the light and shadow transition of complex geometric structures, thereby improving rendering accuracy, reducing illumination boundary errors, and enhancing the sense of light and shadow layering of three-dimensional animation images.

[0076] 3. The present invention performs dynamic shadow calculation through an angle-based light and shadow calculation model, realizes the calculation of light and shadow distribution according to the light source direction and the object normal vector, so that the shadow can be dynamically adjusted according to the lighting direction, thereby avoiding the problems of boundary discontinuity, blur and unreality in traditional shadow calculation methods, and making the shadows in three-dimensional animation conform to the real physical lighting characteristics. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] Figure 1This is the flowchart of the present invention. Detailed implementation manners

[0078] To enable those skilled in the art to understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Based on the embodiments of the present invention, other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0079] The present invention will be described in detail below with reference to the accompanying drawings:

[0080] Embodiment:

[0081] Please refer to the attached Figure 1 , the embodiment of the present invention provides a high-efficiency three-dimensional animation production method based on artificial intelligence, including:

[0082] Step 1: Construct a light propagation calculation model. Based on the propagation characteristics of photons in three-dimensional space, establish a light propagation calculation model that describes the distribution relationship of light in different media. Among them, the light propagation calculation model includes a description of the motion state of photons, and calculates the light intensity distribution by combining scene geometric information and material characteristics;

[0083] Step 2: Calculate the light distribution based on the light propagation calculation model. By solving the motion state of photons, determine the distribution of light in the three-dimensional scene, and adjust the attenuation and scattering characteristics of light by combining the material parameters of the scene to generate preliminary light data;

[0084] Step 3: Optimize the preliminary light data with a neural network. Based on the constructed light propagation calculation model and the light distribution situation, use a neural network to perform feature learning on the light data, and adjust the calculated light distribution to make the light data adapt to the characteristics of different scene light sources. Among them, the input of the neural network is the light distribution data, and the output is the optimized light distribution information;

[0085] Step 4: Optimize the light path based on the optimized light distribution. According to the light distribution information optimized by the neural network, use a path optimization method to adjust the propagation path of light in three-dimensional space, and optimize the directivity of light by combining the photon propagation state to make the propagation of light on the surface of different media conform to optical characteristics;

[0086] Step 5: Calculate the global illumination based on the light path optimization result. According to the propagation information obtained by the light path optimization, combine the geometric structure of the three-dimensional scene, and use a light calculation method to solve the global illumination to generate global illumination data, and the global illumination data is used as the input information for the final light rendering;

[0087] Step 6: Render the anime scene based on the global illumination data. According to the calculated global illumination data, combined with the material properties and illumination direction in the 3D anime scene, perform illumination mapping calculation on the objects in the scene to generate a 3D anime image that conforms to the illumination characteristics. Among them, the illumination mapping calculation adjusts the light and shadow according to the illumination distribution information in the global illumination data to ensure that the rendering result conforms to the illumination changes in the 3D scene.

[0088] Benefits of Step 1: By establishing a propagation model of photons in 3D space, the illumination calculation can accurately simulate the reflection, refraction, and scattering behaviors of light in different media, improving the physical realism of the illumination calculation. Combining the scene geometric information and material characteristics, calculate the spatial distribution of the illumination intensity, enabling the illumination to have an accurate performance in a complex 3D environment and reducing the illumination distortion problem in traditional methods.

[0089] Benefits of Step 2: By solving the motion state of photons, optimize the propagation trajectory of light in the 3D scene, making the illumination calculation conform to the actual optical laws and reducing non-physical illumination effects. Combining the scene material parameters, dynamically adjust the illumination attenuation and scattering characteristics to ensure the natural illumination distribution on different material surfaces and improve the light and shadow realism of the anime scene.

[0090] Benefits of Step 3: Use a neural network to perform feature learning on the illumination data, enabling the illumination calculation to automatically adapt to different light source conditions and improving the generalization ability of the illumination calculation. Optimize the illumination data through deep learning methods, making the illumination calculation smooth under low sampling conditions, effectively reducing the high-frequency noise problem in the Monte Carlo sampling method, and improving the rendering quality.

[0091] Benefits of Step 4: Based on the illumination data optimized by the neural network, adjust the photon propagation trajectory through a path optimization method, improving the calculation efficiency of ray tracing, reducing redundant ray calculations, and reducing the consumption of computing resources. Combine the photon propagation state to optimize the illumination directionality, making the propagation of light on different material surfaces conform to the optical characteristics, avoiding the problem of inconsistent illumination directionality in traditional methods, and improving the light and shadow quality.

[0092] Benefits of Step 5: Through the global illumination calculation method, utilize the optimized light path information to make the illumination calculation of the 3D scene efficient, reduce the sampling error in the global illumination calculation, and improve the illumination distribution accuracy. Combine the scene geometric structure to solve the illumination, enabling the illumination calculation to fully consider environmental factors, improve the layering and realism of the illumination, and enhance the visual performance of the 3D anime scene.

[0093] Benefits of step 6: Combine global illumination data with the material information of the 3D animation scene to perform accurate illumination mapping calculations on the surface of the object, so that the illumination distribution is uniform and the light and shadow realism of the final rendering result is improved. By combining the illumination distribution information of the global illumination data to adjust the light and shadow, the rendering result conforms to the scene illumination changes, and a dynamic light and shadow effect is achieved, which improves the visual impact of the 3D animation picture.

[0094] The light propagation calculation model in step 1 uses the light propagation formula based on wave optics to calculate the light intensity distribution. The calculation formula is as follows:

[0095] E(r, t) = E 0 e i(k·r-ωt) ,

[0096] Where E(r, t) represents the amplitude of the light wave at the position r in three-dimensional space;

[0097] E 0 is the initial amplitude of the light, indicating the energy intensity of the light source;

[0098] e i(k·r-ωt) is the phase factor of the light wave, describing the propagation of the light wave in space and time;

[0099] k is the wave vector of light, which describes the propagation direction of light in different media;

[0100] ω is the angular frequency of light, reflecting the time evolution characteristics of light waves;

[0101] t is the time variable.

[0102] The wave optics model is used to calculate the light intensity distribution, so that the light simulation conforms to the laws of physics and can truly reproduce the interference, diffraction and phase change characteristics of light.

[0103] Traditional geometric optics methods are mainly based on ray tracing and ignore the wave characteristics of light. The method of the present invention can accurately describe the propagation characteristics of light in different media and improve the accuracy of illumination calculation.

[0104] By calculating the propagation direction of light in different media through the wave vector k of light, the refraction and reflection behavior of light on transparent, translucent and highly reflective materials can be dynamically adjusted to avoid the errors that occur in refraction calculations using traditional methods.

[0105] The time variation of illumination is calculated in combination with the angular frequency ω, so that the illumination variation of dynamic light sources is smooth and the problem of illumination jumps during rendering is reduced.

[0106] The light propagation formula of wave optics is used to calculate the light intensity, making the light and shadow effects of shadows, diffuse reflections, and specular reflections natural, reducing the problems of harsh shadows and excessive uneven light and shadows in traditional methods.

[0107] By calculating the light wave amplitude E(r, t) and its changes, the light intensity distribution in the highlight area and the shadow area is simulated to make the light and shadow contrast of the 3D animation picture realistic.

[0108] The calculation of the illumination distribution in step 2 uses the Monte Carlo integration method to calculate the photon state, and the calculation formula is as follows:

[0109]

[0110] Where I(r) is the light intensity at position r in three-dimensional space;

[0111] N is the number of Monte Carlo sampling, which indicates the number of samples used when calculating illumination;

[0112] f(r i ) is the photon at position r i Contribution factors at

[0113] P(r i ) is the photon at position r i The probability distribution at .

[0114] The Monte Carlo integration method is used to calculate the illumination distribution, which makes the estimation of the photon state accurate and can effectively approximate the actual illumination distribution.

[0115] Through sampling technology, the calculation error caused by the fixed sampling method can be reduced, making the lighting calculation stable and suitable for global illumination calculation of complex three-dimensional scenes.

[0116] Traditional lighting calculation methods have high computational costs when dealing with complex lighting scenes, while the Monte Carlo method can achieve high-precision lighting estimation with limited computing resources.

[0117] By adjusting the sampling times N, the calculation accuracy and efficiency can be dynamically balanced, which is suitable for different rendering requirements and improves the calculation flexibility of 3D animation production.

[0118] By introducing the probability distribution P(r i ) Model the photon state so that the lighting calculation can conform to the laws of physics and reduce the generation of non-physical lighting artifacts.

[0119] Combined with the photon contribution factor f(r i ) Calculates the effect of photons on lighting at different positions, which can optimize the transition effect of lighting on complex geometric structures, reduce the sharp transition problem of light and shadow edges, and make the lighting distribution natural.

[0120] In step 3, the neural network optimizes the illumination data using a convolutional neural network to extract illumination features and calculate the optimization error of the illumination data. The loss function is as follows:

[0121]

[0122] Among them, L cnn is the loss function for optimizing the lighting calculation by the neural network;

[0123] M is the number of training samples;

[0124] I cnn (s j ) is the lighting intensity calculated by the neural network;

[0125] I gt (s j ) is the true lighting intensity;

[0126] s j is the spatial position of the training sample.

[0127] The convolutional neural network is used to extract the lighting features, enabling the neural network to learn the spatial correlation in the lighting data and accurately predict the lighting distribution.

[0128] By optimizing the error calculation, the output of the neural network is made to approach the true lighting intensity, reducing the lighting distortion problem caused by insufficient sampling and calculation deviation in the traditional lighting calculation method.

[0129] The traditional lighting calculation method is prone to generating high-frequency noise under low sampling, while the neural network optimized lighting data can effectively remove this noise and improve the rendering quality.

[0130] By calculating the loss function L optimized by the neural network cnn , the parameters of the neural network can be dynamically adjusted, making the lighting calculation robust and improving the consistency of lighting estimation under different scenarios.

[0131] By the neural network learning the features of a large amount of lighting data, in a new 3D anime scene, the lighting distribution can be quickly predicted without relying on a large number of ray tracing calculations in the traditional method.

[0132] The convolutional neural network structure has a weight sharing mechanism, which can efficiently process the lighting data, reduce redundant calculations, improve the real-time performance of lighting rendering, and enable the efficient production of high-quality 3D anime.

[0133] In step 4, the light path optimization uses the Markov chain Monte Carlo method to adjust the propagation path of light, and the probability calculation of the photon path is as follows:

[0134] P(x k+1 |x k ) = A(x k , x k+1 )P(x k+1 ),

[0135] Among them, P(xk+1 |x k ) is the probability that a photon propagates from x k to x k+1 ;

[0136] x k is the current photon position, representing the propagation point of the photon in space;

[0137] x k+1 is the candidate for the next photon propagation position;

[0138] A(x k , x k+1 ) is the acceptance probability, calculated as follows:

[0139]

[0140] where P(x k+1 ) is the probability of the candidate new path point x k+1 ;

[0141] P(x k ) is the probability of the current path point x k .

[0142] Using the Markov chain Monte Carlo method to optimize the light path sampling can find the optimal photon propagation path within fewer calculation steps, enabling the light illumination calculation to converge quickly.

[0143] By calculating the photon state transition probability P(x k+1 |x k ), the propagation of light in three-dimensional space is calculated, enabling the light illumination calculation to be adaptively optimized, rather than relying on random ray tracing, and significantly reducing the calculation of invalid light paths.

[0144] Traditional ray tracing methods usually adopt uniform sampling, resulting in a large amount of computing resources being wasted on invalid paths. However, this method uses the acceptance probability A(x k , x k+1 ) of the Markov chain, which can dynamically adjust the light propagation direction, prioritize the calculation of important light paths, and improve the calculation efficiency.

[0145] By adaptively adjusting the sampling probability of the photon path, unnecessary light illumination calculations are reduced, enabling complex light illumination scenarios to obtain calculation results quickly.

[0146] Since the Markov chain Monte Carlo uses probability-based optimal path search, it can effectively reduce the sampling error in traditional ray tracing methods and avoid the problem of uneven light illumination caused by insufficient sampling.

[0147] By calculating the acceptance probability It can ensure smooth light propagation during the light calculation process, reduce the appearance of artifacts, and make the light and shadow effects in the 3D animation scene natural.

[0148] In step 5, the global illumination calculation uses the finite element method to solve the light distribution, and the calculation formula is as follows:

[0149]

[0150] Among them, K lm is the finite element matrix, representing the mutual influence of light in different spatial units;

[0151] I m is the light intensity of the m-th unit;

[0152] F l is the contribution of the external light source to the l-th unit;

[0153] F l is the total number of units divided by the finite element method.

[0154] The finite element method is used to divide the three-dimensional space into multiple groups of finite elements, and the light intensity of each unit is calculated by solving the system of equations, making the light calculation accurate and especially suitable for the global illumination calculation of complex geometric structures.

[0155] Through the finite element matrix K lm The mutual influence relationship of light between different units is established, so that the light calculation can consider multiple reflections and indirect illumination, and improve the physical authenticity of the global illumination.

[0156] The traditional path tracing method requires a large number of samples of the scene, resulting in a high calculation cost. However, this method uses finite element mesh division and calculates on key units, reducing unnecessary calculations and improving the calculation efficiency.

[0157] Through the matrix solution method, the calculation time is greatly reduced, making the global illumination calculation fast and stable in large 3D animation scenes.

[0158] Since the finite element method can perform light interpolation calculation in space, it can effectively reduce the light unevenness problem caused by discrete sampling, making the light and shadow transition natural in the 3D animation scene.

[0159] Using the inter-unit light influence matrix K lm Calculate the light interaction in different spatial regions, reduce the errors at the light boundaries, make the light calculation smooth, and improve the overall light quality of the 3D animation picture.

[0160] In step 6, the boundary element method is used to optimize the local illumination calculation in the animation scene, and the calculation formula is as follows:

[0161] K·I = F,

[0162] Among them, K is the boundary element matrix, representing the propagation relationship of boundary illumination;

[0163] I is the boundary illumination intensity distribution vector;

[0164] F is the boundary light source contribution vector.

[0165] In step 6, when performing anime scene rendering, an adaptive illumination mapping method is used to adjust the illumination reflectivity of the material, and the calculation formula is as follows:

[0166] R mat (λ) = R 0 + α(λ - λ 0 ),

[0167] Among them, R mat (λ) is the material illumination reflectivity at wavelength λ;

[0168] R 0 is the reflectivity at the reference wavelength λ 0 ;

[0169] α is the reflectivity change coefficient of the material;

[0170] λ is the wavelength of the current light; λ 0 is the reference wavelength.

[0171] In step 6, during the anime scene rendering process, the shadow distribution of the object is adjusted based on the illumination direction, and the shadow generation function is calculated as follows:

[0172] S(p, θ) = max(0, cosθ),

[0173] Among them, S(p, θ) is the shadow intensity at the three-dimensional space position p;

[0174] p is the space position in the three-dimensional scene, representing the point coordinates where the shadow needs to be calculated;

[0175] θ is the illumination incident angle; cosθ is the influence of calculating the angle between the object surface and the light source direction.

[0176] In step 6, during the anime scene rendering process, to enhance the light and shadow details, high dynamic range is used to adjust the illumination brightness, and the calculation formula is as follows:

[0177]

[0178] Among them, L hdr (x) is the illumination brightness after high dynamic range adjustment;

[0179] L raw (x) is the originally calculated illumination brightness.

[0180] Benefits of optimizing local illumination calculation using the boundary element method:

[0181] By establishing the illumination propagation relationship on the boundary through the boundary element matrix K, the local illumination calculation is made accurate, especially suitable for scenes with complex material surfaces, fine geometric structures, and uneven illumination.

[0182] Traditional global illumination calculation methods are difficult to provide high-precision illumination in local detail areas, while this method calculates the local illumination intensity I through boundary element calculation, optimizes the local light and shadow transition, and makes the illumination effect natural.

[0183] Using the boundary element method only requires calculation on the boundary, without the need to calculate the illumination distribution in a three-dimensional scene. Therefore, the calculation amount is reduced, the rendering efficiency is improved, and it is suitable for high-precision animation production.

[0184] Benefits of adjusting the material illumination reflectivity using the adaptive illumination mapping method:

[0185] By calculating the material reflectivity R mat (λ) = R 0 +α(λ - λ 0 ), the illumination adaptability of the material under different light sources can be dynamically adjusted.

[0186] Traditional methods use a fixed reflectivity model and cannot adapt to multi-light source and dynamic illumination environments, while this method can be adaptively adjusted by spectral changes to ensure the authenticity of material illumination in three-dimensional animation scenes.

[0187] Benefits of dynamic shadow adjustment based on the illumination direction:

[0188] Through the shadow calculation formula S(p, θ) = max(0, cosθ), it is ensured that the shadow is adjusted in real time according to the change of the light source direction, making the shadow natural and avoiding the problem of static shadow distortion.

[0189] Traditional shadow mapping methods are prone to hard shadow edges and aliasing artifacts, while this method calculates based on the illumination angle and can effectively optimize the shadow transition, making the light and shadow in three-dimensional animation scenes smooth.

[0190] Benefits of adjusting the illumination brightness using high dynamic range:

[0191] Through the high dynamic range calculation formula, overexposure can be avoided in the highlight area, and details can be retained in the dark area, making the picture clear and vivid.

[0192] Traditional illumination calculation methods are prone to problems such as highlight overflow and detail loss in the dark area, while this method automatically balances the illumination intensity through non-linear high dynamic range mapping, optimizes the illumination level, and makes the animation picture more natural.

[0193] Although embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A high-efficiency 3D animation production method based on artificial intelligence, characterized in that: include: Step 1: Construct a light propagation calculation model. Based on the propagation characteristics of photons in three-dimensional space, a light propagation calculation model is established to describe the distribution relationship of light in different media. The light propagation calculation model includes a description of the motion state of photons and calculates the light intensity distribution in combination with scene geometry information and material properties. Step 2: Calculate the illumination distribution based on the light propagation calculation model, determine the distribution of illumination in the three-dimensional scene by solving the motion state of photons, adjust the attenuation and scattering characteristics of illumination in combination with the material parameters of the scene, and generate preliminary illumination data; Step 3: Perform neural network optimization on the preliminary illumination data. Based on the constructed light propagation calculation model and illumination distribution, a neural network is used to learn the characteristics of illumination data, and the calculated illumination distribution is adjusted to adapt the illumination data to the characteristics of different scene light sources. The neural network input is illumination distribution data, and the output is optimized illumination distribution information. Step 4: Optimize the light path based on the optimized light distribution. According to the light distribution information obtained by the neural network optimization, the light propagation path in the three-dimensional space is adjusted by the path optimization method. The directionality of the light is optimized in combination with the photon propagation state, so that the propagation of light on the surface of different media conforms to the optical characteristics. Step 5: Calculate global illumination based on the light path optimization result. According to the propagation information obtained by light path optimization and the geometric structure of the three-dimensional scene, use the illumination calculation method to solve the global illumination and generate global illumination data. The global illumination data is used as the input information for the final illumination rendering. Step 6: Render the animation scene based on the global illumination data. According to the calculated global illumination data, combined with the material properties and lighting direction in the three-dimensional animation scene, perform lighting mapping calculations on the objects in the scene to generate a three-dimensional animation image that conforms to the lighting characteristics. The lighting mapping calculation is combined with the lighting distribution information in the global illumination data to adjust light and shadow to ensure that the rendering result conforms to the lighting changes in the three-dimensional scene.

2. The high-efficiency 3D animation production method based on artificial intelligence according to claim 1 is characterized in that: The light propagation calculation model in step 1 uses the light propagation formula based on wave optics to calculate the light intensity distribution. The calculation formula is as follows: E(r,t)=E0e i(k·r-ωt) , Where E(r, t) represents the amplitude of the light wave at the position r in three-dimensional space; E0 is the initial amplitude of the light, indicating the energy intensity of the light source; e i(k·r-ωt) is the phase factor of the light wave, describing the propagation of the light wave in space and time; k is the wave vector of light, which describes the propagation direction of light in different media; ω is the angular frequency of light, reflecting the time evolution characteristics of light waves; t is the time variable.

3. The high-efficiency 3D animation production method based on artificial intelligence according to claim 1 is characterized in that: The calculation of the illumination distribution in step 2 uses the Monte Carlo integration method to calculate the photon state, and the calculation formula is as follows: Where I(r) is the light intensity at position r in three-dimensional space; N is the number of Monte Carlo sampling, which indicates the number of samples used when calculating illumination; f(r i ) is the photon at position r i Contribution factors at P(r i ) is the photon at position r i The probability distribution at .

4. The high-efficiency 3D animation production method based on artificial intelligence according to claim 1 is characterized in that: In step 3, the neural network optimizes the illumination data by using a convolutional neural network to extract illumination features and calculate the optimization error of the illumination data. The loss function is as follows: Among them, L cnn Optimizing the loss function for lighting calculations for neural networks; M is the number of training samples; I cnn (s j ) is the light intensity calculated by the neural network; I gt (s j ) is the real light intensity; s j is the spatial position of the training sample.

5. The high-efficiency 3D animation production method based on artificial intelligence according to claim 1 is characterized in that: In step 4, the light path optimization uses the Markov chain Monte Carlo method to adjust the light propagation path, and the probability of the photon path is calculated as follows: P(x k+1 |x k )=A(x k ,x k+1 )P(x k+1 ), Among them, P(x k+1 |x k ) is the photon from x k Propagate to x k+1 The probability of x k is the current photon position, indicating the propagation point of the photon in space; x k+1 The next photon propagation position for the candidate; A(x k , x k+1 ) is the acceptance probability, calculated as follows: Among them, P(x k+1 ) is a candidate new path point x k+1 probability; P(x k ) is the current path point x k probability.

6. The high-efficiency 3D animation production method based on artificial intelligence according to claim 1 is characterized in that: In step 5, the global illumination calculation uses the finite element method to solve the illumination distribution, and the calculation formula is as follows: Among them, K lm is a finite element matrix, which represents the mutual influence of light in different spatial units; I m is the light intensity of the mth unit; F l Contribution to the external light source of unit l; F l The total number of elements in the finite element partition.

7. The high-efficiency 3D animation production method based on artificial intelligence according to claim 1 is characterized in that: In step 6, the animation scene rendering uses the boundary element method to optimize the local illumination calculation, and the calculation formula is as follows: K·I=F, Among them, K is the boundary element matrix, which represents the propagation relationship of boundary illumination; I is the boundary illumination intensity distribution vector; F is the boundary light source contribution vector.

8. The high-efficiency 3D animation production method based on artificial intelligence according to claim 7 is characterized in that: In step 6, when rendering the animation scene, an adaptive light mapping method is used to adjust the light reflectivity of the material, and the calculation formula is as follows: R mat (λ)=R0+α(λ-λ0), Among them, R mat (λ) is the material light reflectance at wavelength λ; R0 is the reflectivity at the reference wavelength λ0; α is the reflectivity variation coefficient of the material; λ is the wavelength of the current light; λ0 is the reference wavelength.

9. The high-efficiency 3D animation production method based on artificial intelligence according to claim 8 is characterized in that: In the animation scene rendering process, step 6 adjusts the shadow distribution of the object based on the lighting direction, and the shadow generation function is calculated as follows: S(p,θ)=max(0,cosθ), Where S(p, θ) is the shadow intensity at position p in three-dimensional space; p is the spatial position in the three-dimensional scene, indicating the coordinates of the point where the shadow needs to be calculated; θ is the incident angle of light; cosθ is the angle between the object surface and the light source direction.

10. The high-efficiency 3D animation production method based on artificial intelligence according to claim 9, characterized in that: In the animation scene rendering process, in step 6, in order to enhance the light and shadow details, the light brightness is adjusted using a high dynamic range. The calculation formula is as follows: Among them, L hdr (x) is the light brightness after high dynamic range adjustment; L raw (x) is the original calculated light intensity.

Images