Development Method and System of Graphics Processing Engine Based on Virtual Simulation

Abstract

The application discloses a kind of based on virtual simulation graphics processing engine development method and system, it is related to computer graphics and virtual reality technical field, including acquisition three-dimensional point cloud data, convolution operation is carried out by convolution neural network algorithm, extract three-dimensional space feature matrix, construct pattern recognition model, the original data of the graphics to be processed is input to pattern recognition model, output graphics parameter set, in the matching virtual scene model in preestablished scene library, and the parameter of model is adjusted;According to the model after adjustment establishes physical coordinate system, by grid division algorithm the graphics to be processed is converted into grid unit, calculates the displacement data of grid unit;Read displacement data, generate and send graphics processing instruction sequence to graphics processor, by to grid unit superimposed texture information and ray tracing effect, output imaging result.The application establishes complete processing chain from data acquisition to rendering output, improves the automation level and accuracy of graphics processing.

Description

Technical Field

[0001] This invention relates to the fields of computer graphics and virtual reality technology, and in particular to a method and system for developing a graphics processing engine based on virtual simulation. Background Technology

[0002] Traditional graphics processing engines primarily employ geometric modeling to construct 3D scenes and generate images through a rasterization rendering pipeline. In recent years, with the rise of deep learning technology, neural network-based 3D feature extraction and scene understanding methods have gradually demonstrated their superiority. Simultaneously, the introduction of physical simulation technology has enabled virtual scenes to possess more realistic dynamic effects, while the development of real-time ray tracing technology has provided new solutions for high-quality rendering. Currently, mainstream graphics processing engines such as Unity3D and Unreal Engine have begun to integrate technologies such as deep learning, physical simulation, and ray tracing, but there is still considerable room for optimization in system architecture and processing workflows.

[0003] However, existing graphics processing engines still face numerous challenges when handling complex scenes: First, traditional feature extraction methods struggle to accurately capture deep semantic information in 3D space, leading to insufficient accuracy in scene understanding and reconstruction; second, during virtual scene matching, the lack of effective parameter optimization mechanisms often results in model deformation and loss of detail; third, physical simulation computation is inefficient, failing to meet the demands of real-time interaction; and finally, texture mapping and lighting calculations during rendering often employ simplified models, resulting in insufficient image realism. These technical bottlenecks severely restrict the further development of virtual simulation applications.

[0004] To address the aforementioned problems, this invention proposes a method and system for developing a graphics processing engine based on virtual simulation. This method enhances 3D feature extraction capabilities through deep learning technology, improves model reconstruction quality by employing improved scene matching and parameter optimization algorithms, introduces efficient physical simulation calculation methods, and combines advanced rendering technology to enhance image realism. This effectively solves the technical problems of existing graphics processing engines in terms of feature extraction accuracy, scene reconstruction quality, physical simulation efficiency, and rendering effects. Summary of the Invention

[0005] In view of the problems of insufficient accuracy in 3D feature extraction, poor quality of virtual scene matching, low efficiency of physical simulation calculation and insufficient rendering realism of existing graphics processing engines, this invention is proposed.

[0006] Therefore, the problem to be solved by this invention is how to improve the modeling accuracy and rendering realism of virtual scenes through deep learning and physical simulation technology.

[0007] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0008] In a first aspect, embodiments of the present invention provide a method for developing a graphics processing engine based on virtual simulation, comprising: acquiring 3D point cloud data output by a depth camera of a training scene, and performing convolution operations on the 3D point cloud data using a convolutional neural network algorithm to extract a 3D spatial feature matrix; constructing a graphics recognition model based on the 3D spatial feature matrix, and inputting the original data of the graphics to be processed into the graphics recognition model to output a graphics parameter set; matching a virtual scene model in a preset scene library based on the graphics parameter set, and adjusting the parameters of the virtual scene model; establishing a physical coordinate system based on the adjusted virtual scene model, converting the graphics to be processed into grid cells using a mesh partitioning algorithm, and calculating the displacement data of the grid cells based on physical kinematics formulas; reading the displacement data, generating and sending a sequence of graphics processing instructions to a graphics processor, and outputting an imaging result by superimposing texture information and ray tracing effects on the grid cells.

[0009] As a preferred embodiment of the virtual simulation-based graphics processing engine development method of the present invention, the method for generating the imaging result is as follows: Displacement data is sorted by timestamp using a temporal data stream processing method and entered into a graphics processing buffer. The displacement data in the graphics processing buffer is smoothed by interpolation and then mapped to the graphics rendering pipeline to generate a graphics processing instruction sequence. Based on the graphics processing instruction sequence, texture coordinates are assigned to mesh units according to the physical coordinate system, and anisotropic filtering is used to improve texture sampling quality. Simultaneously, surface details are enhanced through normal mapping and displacement mapping, and a preset material template is applied to the mesh units to perform bidirectional reflection distribution function calculation, simulating the light scattering characteristics of the material surface. A hierarchical bounding box structure is used to accelerate ray-scene intersection operations, calculate the composite effect of direct lighting, indirect lighting, and global illumination, and add ambient occlusion and soft shadow effects to achieve ray tracing effects. Based on the ray tracing effect, the rendering result is output to the display buffer through tone mapping and anti-aliasing processing in the post-processing pipeline to generate the imaging result.

[0010] As a preferred embodiment of the virtual simulation-based graphics processing engine development method of the present invention, the displacement data is formed as follows: A physical coordinate system is established based on the right-handed Cartesian coordinate system and the adjusted virtual scene model; an adaptive quadrilateral meshing algorithm is used to mesh the graphics to be processed, and the mesh density is dynamically adjusted according to the local curvature and boundary features of the graphics to generate a set of mesh cells; based on the set of mesh cells, using Newton's second law as the basis of the physical kinematics formula and combining it with the Lagrange equations, a force is applied to the mesh cells; simultaneously, the Runge-Kutta integral algorithm is used to solve a system of differential equations to calculate the velocity vector and displacement vector of the mesh cells within the discrete time step; the spatial position of the mesh cells in the physical coordinate system is updated according to the displacement vector to form displacement data.

[0011] As a preferred embodiment of the virtual simulation-based graphics processing engine development method of the present invention, the following steps are included: matching the corresponding virtual scene model in a preset scene library according to the graphics parameter set, and adjusting the parameters of the virtual scene model, including: constructing a hierarchical scene index structure and generating standardized feature vectors by standardizing the graphics parameter set through feature vector normalization; mapping the graphics parameter set to the category space of the scene library based on a fast retrieval algorithm using a KD tree, and constructing a feature vector index in the category space; calculating the similarity between the standardized feature vectors and the feature vector index to determine the matching virtual scene model; extracting the geometric parameters from the graphics parameter set and performing geometric transformations on the virtual scene model according to the affine transformation principle, wherein the geometric transformations include scaling, rotation, translation, and projection mapping; adjusting the transformed virtual scene model through a bilinear interpolation algorithm, and simultaneously optimizing the material parameters, lighting parameters, and shadow parameters of the virtual scene model.

[0012] As a preferred embodiment of the virtual simulation-based graphics processing engine development method of the present invention, the specific formula for the similarity is as follows:

[0013]

[0014] Where S is the similarity score, V1 is the standardized feature vector to be matched, V2 is the feature vector index in the scene library, and v 1i and v 2i Let be the i-th components in V1 and V2 respectively, σ be the bandwidth parameter of the Gaussian kernel function, λ be the attenuation coefficient, and w be the ith component in V1 and V2 respectively. j f is the weight coefficient of the j-th feature. 1j and f 2j Let be the j-th global feature of V1 and V2 respectively, n be the dimension of the feature vector, and m be the number of global features.

[0015] As a preferred embodiment of the virtual simulation-based graphics processing engine development method of the present invention, the method for obtaining the graphics parameter set is as follows: Based on a three-dimensional spatial feature matrix, a graphics recognition model is established through a neural network architecture, wherein the neural network architecture includes an encoder and a decoder; the encoder includes a fully connected layer, a normalization layer, a ReLU activation function, and a bidirectional long short-term memory network unit; the decoder includes a parallel branch structure; the parallel branch structure includes a first branch structure and a second branch structure; the first branch structure includes a deconvolution layer; the second branch structure includes an attention mechanism; the fully connected layer is used to perform dimensionality reduction mapping on the three-dimensional spatial feature matrix, and the dimensionality-reduced feature vector is input to the bidirectional long short-term memory network unit for temporal feature encoding; the first branch structure uses the deconvolution layer to perform spatial restoration of the temporal feature encoding, while the second branch structure uses the attention mechanism to extract local key features; the spatially restored temporal feature encoding and the local key features are fused through a channel-dimensional connection operation; the original data of the graphics to be processed is input into the graphics recognition model, and through the sequential processing of the encoder and the decoder, the geometric parameters, position parameters, and orientation parameters of the graphics are extracted to form a graphics parameter set.

[0016] As a preferred embodiment of the virtual simulation-based graphics processing engine development method of the present invention, the depth camera includes an original point cloud data stream that outputs spatial coordinate information of the X-axis, Y-axis, and Z-axis and reflection intensity values; the method for extracting the three-dimensional spatial feature matrix is ​​as follows: the original point cloud data stream is preprocessed with spatial noise reduction and density equalization to generate a point cloud dataset; the point cloud dataset is divided into three-dimensional voxel grids of equal size, and the point cloud features of the three-dimensional voxel grids are statistically encoded to generate local feature descriptors; based on the convolutional neural network algorithm, multi-scale three-dimensional convolutional kernels are used to extract features layer by layer from the local feature descriptors; the feature maps extracted at different scales are adaptively weighted and combined through a cross-layer feature fusion module to output a three-dimensional feature matrix including multi-scale spatial information, wherein the row vectors of the three-dimensional feature matrix represent the number of sampling points, and the column vectors represent the feature dimensions.

[0017] Secondly, embodiments of the present invention provide a graphics processing engine development system based on virtual simulation, comprising: a data acquisition module, used to acquire three-dimensional point cloud data output by a depth camera of a training scene, and to perform convolution operations on the three-dimensional point cloud data using a convolutional neural network algorithm to extract a three-dimensional spatial feature matrix; a feature extraction module, used to construct a graphics recognition model based on the three-dimensional spatial feature matrix, and to input the original data of the graphics to be processed into the graphics recognition model, and output a graphics parameter set; a parameter adjustment module, used to match a virtual scene model in a preset scene library based on the graphics parameter set, and to adjust the parameters of the virtual scene model; a meshing processing module, used to establish a physical coordinate system based on the adjusted virtual scene model, to convert the graphics to be processed into mesh cells using a meshing algorithm, and to calculate the displacement data of the mesh cells based on physical kinematics formulas; and a graphics imaging module, used to read the displacement data, generate and send a sequence of graphics processing instructions to a graphics processor, and output an imaging result by superimposing texture information and ray tracing effects on the mesh cells.

[0018] Thirdly, embodiments of the present invention provide a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program instructions are executed by the processor, they implement the steps of the virtual simulation-based graphics processing engine development method described in the first aspect of the present invention.

[0019] Fourthly, embodiments of the present invention provide a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program instructions are executed by a processor, they implement the steps of the virtual simulation-based graphics processing engine development method described in the first aspect of the present invention.

[0020] The beneficial effects of this invention are as follows: by acquiring 3D point cloud data through a depth camera and combining it with a convolutional neural network for feature extraction, the accuracy of spatial feature representation is improved; by constructing a graphic recognition model to achieve feature dimensionality reduction and parameterization, computational complexity is reduced; by utilizing scene library matching and dynamic parameter adjustment mechanisms, the accuracy of virtual scene reconstruction is ensured; by combining mesh partitioning algorithms and physical kinematics calculations, the computational efficiency of physical simulation is improved; and by employing optimized rendering techniques for texture overlay and ray tracing, the realism of the image is enhanced. Attached Figure Description

[0021] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein:

[0022] Figure 1 This is a flowchart of the development method for a graphics processing engine based on virtual simulation, as described in Example 1. Detailed Implementation

[0023] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0024] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0025] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0026] Example 1

[0027] Reference Figure 1 This is the first embodiment of the present invention, which provides a method for developing a graphics processing engine based on virtual simulation, including:

[0028] S1: Collect 3D point cloud data output by the depth camera of the training scene, and perform convolution operation on the 3D point cloud data through the convolution neural network algorithm to extract the 3D spatial feature matrix.

[0029] Specifically, the depth camera includes a raw point cloud data stream that outputs spatial coordinate information of the X, Y, and Z axes and reflection intensity values; the raw point cloud data stream is preprocessed with spatial noise reduction and density equalization to generate a point cloud dataset; the point cloud dataset is divided into a three-dimensional voxel grid of equal size, and the point cloud features of the three-dimensional voxel grid are statistically encoded to generate local feature descriptors.

[0030] Furthermore, based on the convolutional neural network algorithm, multi-scale three-dimensional convolutional kernels are used to extract features layer by layer from the local feature descriptors; the feature maps extracted at different scales are adaptively weighted and combined through a cross-layer feature fusion module to output a three-dimensional feature matrix including multi-scale spatial information, wherein the row vectors of the three-dimensional feature matrix represent the number of sampling points and the column vectors represent the feature dimensions.

[0031] It should be noted that the multi-scale 3D convolution kernel includes a first-layer convolution kernel, a second-layer convolution kernel, and a third-layer convolution kernel; the first-layer convolution kernel has a size of 3×3×3 and is used to capture local geometric structure features, the second-layer convolution kernel has a size of 5×5×5 and is used to extract medium-scale spatial relationship features, and the third-layer convolution kernel has a size of 7×7×7 and is used to obtain large-scale global context features.

[0032] S2: Construct a graphic recognition model based on the three-dimensional spatial feature matrix, input the original data of the graphic to be processed into the graphic recognition model, and output a graphic parameter set.

[0033] Specifically, based on a three-dimensional spatial feature matrix, an image recognition model is established through a neural network architecture, which includes an encoder and a decoder.

[0034] It should be noted that the encoder includes fully connected layers, normalization layers, ReLU activation functions, and bidirectional long short-term memory network units; the decoder includes a parallel branch structure; the parallel branch structure includes a first branch structure and a second branch structure; the first branch structure includes deconvolution layers; and the second branch structure includes an attention mechanism.

[0035] Furthermore, fully connected layers are used to perform dimensionality reduction mapping on the three-dimensional spatial feature matrix, and the dimensionality-reduced feature vectors are input into bidirectional long short-term memory network units for temporal feature encoding; the output dimensions of each fully connected layer are 1024, 512, and 256, respectively.

[0036] Furthermore, the first branch structure uses deconvolutional layers to spatially restore the temporal feature encoding, while the second branch structure uses an attention mechanism to extract local key features. The spatially restored temporal feature encoding and local key features are then fused through channel-dimensional concatenation. The original data of the image to be processed is input into the image recognition model, and through sequential processing by the encoder and decoder, the geometric parameters, position parameters, and orientation parameters of the image are extracted to form an image parameter set.

[0037] S3: Based on the set of graphics parameters, match a virtual scene model in the preset scene library and adjust the parameters of the virtual scene model.

[0038] Specifically, a hierarchical scene index structure is constructed, and the graphics parameter set is processed through feature vector standardization to generate standardized feature vectors; a fast retrieval algorithm based on KD-tree maps the graphics parameter set to the category space of the scene library, and a feature vector index is constructed within the category space; the similarity between the standardized feature vectors and the feature vector index is calculated to determine the matching virtual scene model.

[0039] Furthermore, the specific formula for similarity is as follows:

[0040]

[0041] Where S is the similarity score, V1 is the standardized feature vector to be matched, V2 is the feature vector index in the scene library, and v 1i and v 2i Let be the i-th components in V1 and V2 respectively, σ be the bandwidth parameter of the Gaussian kernel function, λ be the attenuation coefficient, and w be the ith component in V1 and V2 respectively. j f is the weight coefficient of the j-th feature. 1j and f 2j Let be the j-th global feature of V1 and V2 respectively, n be the dimension of the feature vector, and m be the number of global features.

[0042] It should be noted that when the similarity score is greater than a preset threshold, the corresponding virtual scene model is marked as a candidate matching model. If the similarity score of a candidate matching model is higher than the score of other candidate models by a preset difference, this model is directly determined as the final matching model. If the difference in similarity scores among multiple candidate matching models is less than a preset difference, the similarity scores are re-ranked based on the usage frequency weight of the candidate matching models, and the model with the highest weighted score is selected as the final matching model. When the similarity score is less than or equal to the preset threshold, an empty matching result is returned, and a prompt message indicating that there is no suitable matching model is sent to the user. The preset threshold is the interval boundary value obtained by K-means clustering analysis based on the similarity distribution of the historical matching dataset.

[0043] Furthermore, geometric parameters are extracted from the set of graphic parameters, and geometric transformations are performed on the virtual scene model according to the principle of affine transformation. The transformed virtual scene model is then adjusted using a bilinear interpolation algorithm, while the material parameters, lighting parameters, and shadow parameters of the virtual scene model are optimized.

[0044] It should be noted that geometric transformations include scaling, rotation, translation, and projection mapping.

[0045] S4: Establish a physical coordinate system based on the adjusted virtual scene model, convert the graphic to be processed into grid cells using a grid division algorithm, and calculate the displacement data of the grid cells based on the physical kinematics formula.

[0046] Specifically, a physical coordinate system is established based on the right-handed Cartesian coordinate system and the adjusted virtual scene model.

[0047] It should be noted that in the physical coordinate system, the origin is set to coincide with the center of the virtual scene, and the X-axis, Y-axis and Z-axis correspond to the length, height and depth of the scene, respectively. The coordinate units are scaled to the actual physical scale.

[0048] Furthermore, an adaptive quadrilateral meshing algorithm is used to mesh the graphic to be processed, and the mesh density is dynamically adjusted according to the local curvature and boundary features of the graphic to generate a set of mesh cells, where each mesh cell contains node coordinates, cell mass and boundary constraints.

[0049] Furthermore, based on the set of grid cells, and using Newton's second law as the foundation of the physical kinematics formula, combined with the Lagrange equations, forces are applied to the grid cells. At the same time, the Runge-Kutta integral algorithm is used to solve the system of differential equations to calculate the velocity vector and displacement vector of the grid cells within the discrete time step. The spatial position of the grid cells in the physical coordinate system is updated according to the displacement vector to form displacement data.

[0050] S5: Read the displacement data, generate and send a sequence of graphics processing instructions to the graphics processor, and output the imaging result by superimposing texture information and ray tracing effect on the grid cells.

[0051] Specifically, a time-series data stream processing method is used to sort the displacement data by timestamp and enter it into the graphics processing buffer. The displacement data in the graphics processing buffer is then mapped to the graphics rendering pipeline after interpolation and smoothing to generate a sequence of graphics processing instructions.

[0052] It should be noted that the graphics processing instruction sequence includes vertex shading instructions, fragment shading instructions, and post-processing effects instructions. Vertex shading instructions control the spatial transformation and geometric deformation of mesh units; fragment shading instructions are responsible for texture mapping and material rendering; and post-processing effects instructions are used for ray tracing and environmental effects.

[0053] Furthermore, based on the graphics processing instruction sequence, texture coordinates are assigned to mesh units according to the physical coordinate system, and anisotropic filtering is used to improve texture sampling quality. At the same time, surface details are enhanced by normal mapping and displacement mapping, and a preset material template is applied to the mesh units to perform bidirectional reflection distribution function calculation to simulate the light scattering characteristics of the material surface.

[0054] It should be noted that the preset material template defines the diffuse reflection coefficient, specular reflection coefficient, and bump map parameters.

[0055] Furthermore, a hierarchical bounding box structure is used to accelerate the ray-scene intersection operation, calculate the composite effect of direct lighting, indirect lighting and global illumination, and add ambient occlusion and soft shadow effects to achieve the effect of ray tracing; based on the effect of ray tracing, the rendering result is output to the display buffer through tone mapping and anti-aliasing processing of the post-processing pipeline to generate the imaging result.

[0056] Furthermore, this embodiment also provides a virtual simulation-based graphics processing engine development system, including: a data acquisition module, used to acquire 3D point cloud data output by a depth camera in a training scene, and perform convolution operations on the 3D point cloud data using a convolutional neural network algorithm to extract a 3D spatial feature matrix; a feature extraction module, used to construct a graphics recognition model based on the 3D spatial feature matrix, input the original data of the graphics to be processed into the graphics recognition model, and output a graphics parameter set; a parameter adjustment module, used to match a virtual scene model in a preset scene library based on the graphics parameter set, and adjust the parameters of the virtual scene model; a meshing processing module, used to establish a physical coordinate system based on the adjusted virtual scene model, convert the graphics to be processed into mesh cells using a meshing algorithm, and calculate the displacement data of the mesh cells based on physical kinematics formulas; and a graphics imaging module, used to read the displacement data, generate and send a sequence of graphics processing instructions to the graphics processor, and output an imaging result by superimposing texture information and ray tracing effects on the mesh cells.

[0057] This embodiment also provides a computer device applicable to the development method of a graphics processing engine based on virtual simulation, including a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the graphics processing engine development method based on virtual simulation as proposed in the above embodiment.

[0058] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0059] In summary, this invention improves the accuracy of spatial feature representation by acquiring 3D point cloud data through a depth camera and combining it with a convolutional neural network for feature extraction; it reduces computational complexity by constructing an image recognition model to achieve feature dimensionality reduction and parameterization; it ensures the accuracy of virtual scene reconstruction by utilizing scene library matching and dynamic parameter adjustment mechanisms; it improves the computational efficiency of physical simulation by combining mesh partitioning algorithms and physical kinematics calculations; and it enhances the realism of images by employing optimized rendering techniques for texture overlay and ray tracing.

[0060] Example 2

[0061] Referring to Table 1, which is the second embodiment of the present invention, this embodiment provides a method for developing a graphics processing engine based on virtual simulation. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0062] Specifically, an industrial parts design scenario was selected as the test object. An Intel RealSense D455 depth camera was used to collect 3D point cloud data of six mechanical parts models with varying complexity. The depth camera's sampling frequency was set to 30Hz, spatial resolution to 1280×720 pixels, and depth accuracy to ±0.5mm. During the data acquisition process, the parts under test were placed on a calibration platform, and a robotic arm drove the depth camera to perform a 360-degree surround scan. Each part was acquired for 60 seconds, resulting in approximately 1800 frames of raw point cloud data.

[0063] Furthermore, the acquired point cloud data was denoised using voxel filtering, with voxel sizes set to 0.5mm × 0.5mm × 0.5mm. An improved PointNet++ network architecture was used for feature extraction, comprising four point cloud convolutional layers with kernel sizes of 16, 32, 64, and 128, a stride of 1, and LeakyReLU activation. After feature extraction, a 256-dimensional 3D spatial feature matrix was obtained. Based on the extracted 3D spatial feature matrix, a deep neural network containing an encoder and a decoder was constructed as an image recognition model. The encoder consisted of three fully connected layers (512, 256, and 128 nodes), and the decoder employed a parallel branching structure. The first branch used transposed convolutions for feature reconstruction, while the second branch extracted key feature points through an attention mechanism. The outputs of the two branches were fused through a 1×1 convolutional layer, ultimately outputting a set of image parameters including geometric, positional, and orientation parameters.

[0064] Specifically, model matching is performed within a pre-defined library of 10,000 industrial parts scenarios. An improved KD-tree algorithm is used to establish an index structure, with a feature vector similarity threshold set to 0.85. Parameter adjustments are made to the matched virtual scene models, including geometric transformations (scaling factor 0.8-1.2, rotation angle ±15°) and material optimization (diffuse reflection coefficient 0.2-0.8, roughness 0.1-0.9). In the physical simulation stage, an adaptive quadrilateral mesh generation algorithm is used, with a mesh size ranging from 0.1mm to 2mm and a curvature adaptive threshold of 0.05. The fourth-order Runge-Kutta method is used to solve the motion equations, with a time step of 0.001s. In the rendering stage, a physically based material system is used, with ray tracing sampling number set to 1024 and recursion depth to 4. Ambient occlusion and soft shadow effects are also enabled.

[0065] Furthermore, as shown in Table 1, the present invention achieves a point cloud acquisition accuracy of 0.48mm, a 49.5% improvement over the traditional method's 0.95mm, thus enhancing the quality of 3D data acquisition. Feature extraction time is reduced from 195ms to 58ms, a 70.3% improvement in processing efficiency, thanks to the improved convolutional neural network algorithm and optimized parallel computing architecture. Feature dimension compression reaches 85.5%, an 89.2% improvement over the traditional method's 45.2%, demonstrating the significant advantages of the present invention in data dimensionality reduction and feature representation. The model matching accuracy reaches 95.8%, a 16.1% improvement over the traditional method's 82.5%, while scene reconstruction error is reduced from 1.85mm to 0.62mm, a 66.5% improvement. This indicates that the present invention has made significant progress in virtual scene model matching and parameter optimization, providing more reliable basic data for subsequent physical simulation and rendering processing. In terms of physical simulation, the present invention achieves a simulation frame rate of 105fps, a 133.3% improvement over the traditional method's 45fps. The mesh generation accuracy was improved to 0.15 mm, a 66.7% improvement over the traditional method's 0.45 mm. These improvements are attributed to the adaptive mesh generation algorithm and the efficient physical kinematics calculation framework, which significantly enhances the accuracy and real-time performance of the simulation process.

[0066] Table 1. Comparison between the method of the present invention and traditional methods

[0067]

[0068]

[0069] Furthermore, the rendering quality score reached 9.2 out of 10, a 24.3% improvement over the traditional method's 7.4. Material reproduction accuracy increased from 75.8% to 92.5%, a 22.0% improvement. Ray tracing sampling efficiency reached 960 samples / ms, a 152.6% improvement over the traditional method's 380 samples / ms. These metrics fully demonstrate the advantages of this invention in terms of graphics rendering effects and processing efficiency. System response time decreased from 125ms in the traditional method to 32ms, a 74.4% improvement, while memory usage decreased from 1250MB to 485MB, saving 61.2% of system resources. This indicates that this invention not only improves processing performance but also optimizes resource utilization efficiency, making it more practical.

[0070] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for developing a graphics processing engine based on virtual simulation, the method comprising: The application relates to a method for generating an imaging result, comprising the following steps: ​ collecting three-dimensional point cloud data output by a depth camera of a training scene, and performing convolution operation on the three-dimensional point cloud data through a convolution neural network algorithm to extract a three-dimensional space feature matrix; constructing a pattern recognition model according to the three-dimensional space feature matrix, inputting original data of a to-be-processed pattern into the pattern recognition model, and outputting a pattern parameter set; matching a virtual scene model in a preset scene library based on the pattern parameter set, and adjusting parameters of the virtual scene model; establishing a physical coordinate system according to the adjusted virtual scene model, converting the to-be-processed pattern into a grid unit through a grid division algorithm, and calculating displacement data of the grid unit based on a physical kinematics formula; reading the displacement data, generating and sending a pattern processing instruction sequence to a graphic processor, superimposing texture information and ray tracing effects on the grid unit, and outputting an imaging result; the imaging result is generated by, adopting a time sequence data flow processing method to sort the displacement data according to timestamps, and entering the displacement data into a pattern processing buffer area, wherein the displacement data in the pattern processing buffer area is mapped to a pattern rendering pipeline after being processed through interpolation smoothing, and a pattern processing instruction sequence is generated; based on the pattern processing instruction sequence, texture coordinates are allocated to the grid unit according to the physical coordinate system, and an anisotropic filtering method is adopted to improve the texture sampling quality; at the same time, the surface details are enhanced through normal mapping and displacement mapping, a preset material template is applied to the grid unit, bidirectional reflectance distribution function calculation is performed, and the light scattering characteristics of the material surface are simulated; a hierarchical bounding volume structure is adopted to accelerate the intersection operation of rays and scenes, the synthesis effects of direct light, indirect light and global light are calculated, environmental light shielding and soft shadow effects are added, and the effect of ray tracing is realized; based on the effect of ray tracing, the rendering result is output to a display buffer area through tone mapping and anti-aliasing processing of a post-processing pipeline, and an imaging result is generated.

2. The virtual simulation-based graphic processing engine development method of claim 1, wherein: the formation method of the displacement data is, based on a right-hand Cartesian coordinate system, a physical coordinate system is established according to the adjusted virtual scene model; an adaptive quadrilateral grid division algorithm is adopted to divide the to-be-processed pattern into grids, and the grid density is dynamically adjusted according to the local curvature and boundary characteristics of the pattern to generate a grid unit set; based on the grid unit set, a force is applied to the grid unit based on Newton's second law as a physical kinematics formula and in combination with Lagrange's mechanical equation; at the same time, the velocity vector and displacement vector of the grid unit in a discrete time step are calculated by solving a differential equation set through a Runge-Kutta integral algorithm; the space position of the grid unit in the physical coordinate system is updated according to the displacement vector to form displacement data.

3. The virtual simulation-based graphics processing engine development method of claim 2, wherein: matching a corresponding virtual scene model in a preset scene library according to the pattern parameter set, and adjusting parameters of the virtual scene model, comprising: constructing a hierarchical scene index structure, and generating a standardized feature vector through feature vector standardization processing of the pattern parameter set; mapping the pattern parameter set to a category space of the scene library based on a KD tree fast retrieval algorithm, and constructing a feature vector index in the category space; Calculate the similarity of the standardized feature vector and the feature vector index to determine a matched virtual scene model; Extract geometric parameters in the set of graphic parameters and perform geometric transformation on the virtual scene model according to the principle of affine transformation, wherein the geometric transformation includes size scaling, rotation translation and projection mapping; Adjust the transformed virtual scene model through a bilinear interpolation algorithm, and simultaneously optimize the configuration of material parameters, lighting parameters and shadow parameters of the virtual scene model.

4. The virtual simulation-based graphic processing engine development method of claim 3, wherein: The specific formula of the similarity is as follows: ； wherein, is a similarity score, is a normalized feature vector to be matched, is a feature vector index in the scene library, and are respectively and is the i-th component in is a bandwidth parameter of the Gaussian kernel function, is a decay coefficient, is a weight coefficient of the j-th feature, and are respectively and is the j-th global feature in is a feature vector dimension, is a global feature number.

5. The virtual simulation-based graphic processing engine development method of claim 3, wherein: The method for obtaining the set of graphic parameters is, Based on the three-dimensional space feature matrix, a graphic recognition model is established through a neural network architecture, wherein the neural network architecture includes an encoder and a decoder; the encoder includes a fully connected layer, a normalization layer, a ReLU activation function and a bidirectional long short-term memory network unit; the decoder includes a parallel branch structure; the parallel branch structure includes a first branch structure and a second branch structure; the first branch structure includes a deconvolution layer; the second branch structure includes an attention mechanism; The fully connected layer is used to perform dimension reduction mapping on the three-dimensional space feature matrix, and the dimension-reduced feature vector is input into the bidirectional long short-term memory network unit for time sequence feature coding; The first branch structure uses the deconvolution layer to perform spatial restoration on the time sequence feature coding, and the second branch structure uses the attention mechanism to extract local key features; The spatially restored time sequence feature coding and the local key features are connected through channel dimensions and then fused; The original data of the to-be-processed graphic are input into the graphic recognition model, and the geometric parameters, position parameters and direction parameters of the graphic are extracted through sequential processing of the encoder and the decoder to form a set of graphic parameters.

6. The virtual simulation-based graphics processing engine development method of claim 1, wherein: The depth camera includes an original point cloud data stream outputting spatial coordinate information and reflection intensity values of X, Y and Z axes; The method for extracting the three-dimensional space feature matrix is, The original point cloud data stream is preprocessed through spatial noise reduction and density equalization to generate a point cloud data set; The point cloud data set is divided into three-dimensional voxel grids of equal size, and the point cloud features of the three-dimensional voxel grids are statistically coded to generate local feature descriptors; Based on a convolutional neural network algorithm, multi-scale three-dimensional convolution kernels are used to extract features layer by layer from the local feature descriptors; Through a cross-layer feature fusion module, the feature maps extracted under different scales are adaptively weighted and combined to output a three-dimensional feature matrix including multi-scale spatial information, wherein the row vector of the three-dimensional feature matrix represents the number of sampling points, and the column vector represents the feature dimension.

7. A virtual simulation-based graphics processing engine development system based on the virtual simulation-based graphics processing engine development method according to any one of claims 1 to 6, characterized by: Further comprising, a data acquisition module configured to acquire three-dimensional point cloud data output by a depth camera of a training scene, and perform convolution operation on the three-dimensional point cloud data through a convolutional neural network algorithm to extract a three-dimensional space feature matrix; a feature extraction module configured to construct a graphic recognition model according to the three-dimensional space feature matrix, and input original data of a to-be-processed graphic into the graphic recognition model to output a set of graphic parameters; The parameter adjustment module is configured to match a virtual scene model in a preset scene library based on the graphics parameter set, and adjust parameters of the virtual scene model. The gridding processing module is configured to establish a physical coordinate system according to the adjusted virtual scene model, convert the to-be-processed graphics into grid cells through a grid division algorithm, and calculate displacement data of the grid cells based on a physical kinematics formula. The image forming module is configured to read the displacement data, generate and send a graphics processing instruction sequence to a graphics processor, superimpose texture information and ray tracing effects on the grid cells, and output an imaging result.

8. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that: The processor executes the computer program to implement the steps of the virtual simulation-based graphics processing engine development method according to any one of claims 1-6.

9. A computer readable storage medium having stored thereon a computer program, characterized in that: The computer program is executed by the processor to implement the steps of the virtual simulation-based graphics processing engine development method according to any one of claims 1-6.

Images