Virtual reality rapid presentation method and system based on real physical attributes
By obtaining and analyzing the physical attribute information of the simulation object, establishing a mechanism model and performing finite element analysis, and combining deep learning technology to perform frame rate complementation, the problem of insufficient interaction between multiple physics in virtual reality is solved, and high-realistic Virtual reality scene presentation improves user immersion and simulation efficiency.
Patent Information
- Application Number
- CN202411119139.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-15
- Publication Date
- 2024-11-19
AI Technical Summary
Existing virtual reality technologies are difficult to accurately simulate changes in physical attributes caused by the interaction of multiple physics fields in the real world in a virtual environment, resulting in insufficient sense of reality in virtual scenes.
By obtaining the physical attribute information of the simulation object, establishing a mechanism model, performing finite element analysis, generating state change animations, and using deep learning technology to perform frame rate complementation, ensuring the smoothness and authenticity of the animation.
It realizes the presentation of changes in physical states in virtual reality, improves the immersion of users, and significantly saves the numerical simulation time of physical objects, and is suitable for high-risk training and other fields.
Smart Images

Figure CN118981253A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of virtual reality, and in particular to a method and system for rapid presentation of virtual reality based on real physical properties. Background Art
[0002] Virtual reality (VR) technology can create a three-dimensional, interactive virtual environment by rendering high-quality graphics and sound effects. With the expansion of the application scope of virtual reality technology, it has been widely used in digital visualization and virtual simulation in the industrial field, which enables users to observe and experience highly realistic real physical scenes and environmental changes in a virtual environment.
[0003] In the current field of virtual reality modeling, although the technology has made significant progress, there is still a key limitation, that is, the lack of physical property change data associated with the interaction of multiple physical fields that are commonly present in the real environment; in the real world, various physical phenomena such as electromagnetic fields, fluid dynamics, heat conduction, etc. often interact with each other and jointly affect the change of physical objects in the environment; however, in the virtual environment, due to the lack of sufficiently accurate and comprehensive physical object property change laws and information data, the simulated scenes are still insufficient in terms of realism;
[0004] Therefore, how to better integrate the effects of multiple physical fields and real-time changes in object properties and improve the ability of virtual reality technology in reasoning about changes in physical states is an urgent problem to be solved in this field. Summary of the invention
[0005] The present invention provides a virtual reality rapid presentation method based on real physical properties, comprising: obtaining physical property information of a simulation object; establishing a mechanism model corresponding to the simulation object based on the physical property information corresponding to the simulation object; obtaining property change information of the simulation object based on the mechanism model corresponding to the simulation object; generating a state change animation of the simulation object based on the property change information of the simulation object; judging whether to perform frame interpolation on the state change animation of the simulation object based on frame rate requirements; if not, importing the state change animation of the simulation object into a virtual reality engine for presentation; if so, interpolating two adjacent frames of a simulation state diagram of the state change animation of the simulation object through deep learning technology to generate a state change animation of the simulation object after frame interpolation, and importing the state change animation of the simulation object after frame interpolation into a virtual reality engine for presentation.
[0006] Furthermore, the obtaining of the physical property information of the simulation object includes: obtaining the geometric shape information and size information of the simulation object; using a data acquisition tool to collect the property data of the simulation object, wherein the data acquisition tool includes an experimental test tool and / or a simulation tool; processing and analyzing the property data of the simulation object to determine the physical property value of the simulation object, wherein the physical property information of the simulation object includes at least the geometric shape information, size information and physical property value of the simulation object.
[0007] Furthermore, based on the physical property information corresponding to the simulation object, a mechanism model corresponding to the simulation object is established, including: based on the geometric shape information and size information of the simulation object, a geometric model of the simulation object is established; based on the geometric model of the simulation object, a mechanism model corresponding to the simulation object is established.
[0008] Furthermore, based on the geometric model of the simulation object, a mechanism model of the simulation object is established, including: obtaining the characteristics and research objectives of the physical field; based on the geometric model of the simulation object, establishing at least one mathematical equation for describing the behavior of the physical field, wherein the at least one mathematical equation includes at least a partial differential equation and / or a differential equation.
[0009] Furthermore, based on the mechanism model corresponding to the simulation object, the property change information of the simulation object is obtained, including: performing finite element analysis based on the mechanism model corresponding to the simulation object to obtain the property change information of the simulation object.
[0010] Furthermore, finite element analysis is performed based on the mechanism model of the simulation object to obtain property change information of the simulation object, including: applying boundary conditions and initial conditions; using finite element analysis software to solve at least one established mathematical equation through finite element numerical method to obtain property change information of the simulation object.
[0011] Furthermore, judging whether to add frames to the state change animation of the simulated object based on the frame rate requirement includes: determining whether the state image change rate of the state change animation of the simulated object meets preset requirements based on the frame rate requirement, dynamic image quality requirements and virtual reality presentation hardware performance.
[0012] Furthermore, the method uses deep learning technology to interpolate two adjacent frames of the simulation state graph of the state change animation of the simulation object to generate the state change animation of the simulation object after interpolation, including: for two adjacent frames of the simulation state graph of the state change animation of the simulation object, calculating the pure noise hidden variables of the simulation state graph of each frame, interpolating the pure noise hidden variables of the simulation state graph of two adjacent frames to obtain multiple pure noise hidden variables generated by interpolation, training the low-rank adaptive parameter matrix of the simulation state graph of each frame, interpolating the low-rank adaptive parameter matrix of the simulation state graph of two adjacent frames to obtain multiple low-rank adaptive parameter matrices generated by interpolation, calculating the self-attention mechanism matrix of the simulation state graph of each frame, interpolating the self-attention mechanism matrix of the simulation state graph of two adjacent frames to obtain multiple self-attention mechanism matrices generated by interpolation, and using the multiple low-rank adaptive parameter matrices generated by interpolation and the multiple self-attention mechanism matrices generated by interpolation to cyclically denoise the multiple pure noise hidden variables generated by interpolation to obtain multiple smooth transition images.
[0013] Furthermore, the low-rank adaptive parameter matrix of the simulation state diagram of two adjacent frames is interpolated based on the following formula: ((t,t+1),n) =α n W t +(1-α n )W t+1 ,n=1,2……N, where N ((t,t+1),n) The low-rank adaptive parameter matrix of the simulation state diagram of the t-th frame and the low-rank adaptive parameter matrix of the simulation state diagram of the t+1-th frame are interpolated to obtain the n-th interpolated low-rank adaptive parameter matrix, W t is the low-rank adaptive parameter matrix of the simulation state diagram of the t-th frame, W t+1 is the low-rank adaptive parameter matrix of the simulation state diagram of the t+1th frame, α n W ((t,t+1),n) The corresponding interpolation coefficient, N is the total number of low-rank adaptive parameter matrices generated by interpolating the low-rank adaptive parameter matrix of the t-th frame simulation state diagram and the low-rank adaptive parameter matrix of the t+1-th frame simulation state diagram.
[0014] The present invention provides a virtual reality rapid presentation system based on real physical properties, comprising: an attribute determination module, used for obtaining physical attribute information of a simulation object; a model establishment module, used for establishing a mechanism model corresponding to the simulation object based on the physical attribute information corresponding to the simulation object; a change acquisition module, used for obtaining attribute change information of the simulation object based on the mechanism model corresponding to the simulation object; an animation generation module, used for generating a state change animation of the simulation object based on the attribute change information of the simulation object; a frame supplement determination module, used for judging whether to supplement the state change animation of the simulation object based on a frame rate requirement; a virtual reality presentation module, used for importing the state change animation of the simulation object into a virtual reality engine for presentation after the frame supplement determination model determines that the state change animation of the simulation object is not supplemented, and also used for supplementing two adjacent frames of the simulation state diagram of the state change animation of the simulation object through a deep learning technology after the frame supplement determination module determines that the state change animation of the simulation object is supplemented, generating the state change animation of the simulation object after supplementing the frames, and importing the state change animation of the simulation object after supplementing the frames into a virtual reality engine for presentation.
[0015] Compared with the prior art, the virtual reality rapid presentation method and system based on real physical properties provided by the present invention has at least the following beneficial effects:
[0016] The physical property changes of simulated objects in the real environment are quickly mapped to the virtual reality scene. The established virtual reality scene and the dynamic real-time changes of virtual objects in the virtual environment are presented according to the real physical property change rules, which can achieve high-fidelity virtual scene reproduction and enhance the immersion of virtual reality users. It has important practical significance in a wide range of application fields (such as emergency drills, high-risk training, etc.). The deep learning technology is used to quickly realize the image interpolation of simulation results, which can significantly save the numerical simulation time of physical objects and provide a rapid implementation technology for high-fidelity and high-efficiency presentation of virtual scenes. BRIEF DESCRIPTION OF THE DRAWINGS
[0017] This specification will be further described in the form of exemplary embodiments, which will be described in detail by the accompanying drawings. These embodiments are not restrictive, and in these embodiments, the same number represents the same structure, wherein:
[0018] Figure 1 It is a flowchart of a method for rapid virtual reality presentation based on real physical properties according to some embodiments of this specification;
[0019] Figure 2 It is a schematic diagram of a process of interpolating two adjacent frames of a simulation state diagram of a state change animation of a simulation object according to some embodiments of this specification;
[0020] Figure 3 It is a schematic diagram of frame interpolation using an algorithm model integrated with LoRA and UNet according to some embodiments of this specification;
[0021] Figure 4 It is a schematic diagram of the result after interpolating two adjacent frames of the simulation state diagram of the state change animation of the simulation object according to some embodiments of this specification;
[0022] Figure 5 It is a module diagram of a virtual reality rapid presentation system based on real physical properties according to some embodiments of this specification. DETAILED DESCRIPTION
[0023] In order to more clearly illustrate the technical solutions of the embodiments of this specification, the following is a brief introduction to the drawings required for the description of the embodiments. Obviously, the drawings described below are only some examples or embodiments of this specification. For ordinary technicians in this field, this specification can also be applied to other similar scenarios based on these drawings without creative work. Unless it is obvious from the language environment or otherwise explained, the same reference numerals in the figures represent the same structure or operation.
[0024] Figure 1 is a flowchart of a method for rapid virtual reality presentation based on real physical properties according to some embodiments of this specification, such as Figure 1 As shown, a method for rapid virtual reality presentation based on real physical properties may include the following steps.
[0025] Step 110, obtaining physical property information of the simulation object.
[0026] In some embodiments, step 110 specifically includes:
[0027] Obtaining geometric shape information and size information of the simulation object;
[0028] Using a data acquisition tool to collect attribute data of the simulation object, wherein the data acquisition tool includes an experimental test tool and / or a simulation tool;
[0029] The property data of the simulation object is processed and analyzed to determine the physical property value of the simulation object, wherein the physical property information of the simulation object at least includes geometric shape information, size information and physical property value of the simulation object.
[0030] Specifically, experimental testing tools collect the property data of simulation objects through actual physical tests or experiments. These tools can provide the most direct and realistic data feedback, which helps to verify the accuracy and reliability of the simulation model. Experimental testing tools include, but are not limited to, sensors: used to measure various physical quantities, such as temperature, pressure, speed, acceleration, etc., which are important components of the property data of simulation objects; measuring instruments: such as multimeters, oscilloscopes, spectrometers, etc., used to accurately measure electrical, optical and other physical parameters; experimental equipment: used to test the performance of simulation objects under specific conditions.
[0031] Simulation tools use mathematical models and algorithms to simulate the behavior and properties of simulation objects to obtain relevant data. This type of tool has the advantages of low cost, high efficiency, and strong repeatability. Simulation tools include but are not limited to: Commercial simulation software: such as ANSYS, MATLAB / Simulink, etc. These software provide a wealth of simulation modules and tools to support users to build simulation models according to specific needs and perform various complex simulation analyses; open source simulation platforms: such as OpenFOAM, SciPy, etc. These platforms provide open source simulation algorithms and tool sets, and users can customize and expand simulation functions as needed; customized simulation tools: customized simulation tools developed for specific industries or application scenarios, such as aircraft simulation software in the aerospace field, vehicle dynamics simulation software in the automotive industry, etc.
[0032] Before using data acquisition tools to collect attribute data of simulation objects, you can first determine the simulation objects and requirements, clarify which attribute data needs to be collected, and the role and importance of these data in the simulation model. According to the characteristics and requirements of the simulation objects, select appropriate experimental test tools or simulation tools. Use the selected tools to perform actual tests or simulation analysis to collect attribute data of simulation objects. Process and analyze the collected data to extract useful information.
[0033] Taking the simulation of droplets freezing on wings as an example, based on high-precision cameras, thermal imagers, laser rangefinders and other equipment, the physical property information of droplets changing during the de-icing process is recorded, including the geometric shape, size, speed and temperature of the droplets; the external conditions under airport experimental conditions are measured, including air flow rate, temperature and humidity; the temperature distribution on the wing surface and the heat transfer when the droplets interact with the wing surface are measured through contact and non-contact sensors.
[0034] Step 120: establishing a mechanism model corresponding to the simulation object based on the physical property information corresponding to the simulation object.
[0035] In some embodiments, step 120 specifically includes:
[0036] Based on the geometric shape information and size information of the simulation object, a geometric model of the simulation object is established;
[0037] Based on the geometric model of the simulation object, a mechanism model corresponding to the simulation object is established.
[0038] Specifically, a modeling tool can be used to establish a geometric model of the simulation object based on the geometric shape information and size information of the simulation object. As an example only, the modeling tool can be a commercial CAD software (such as SolidWorks, AutoCAD, CATIA, etc.), or a software specifically used for simulation (such as ANSYS Workbench, Simulink 3DAnimation, etc.), or an open source modeling platform (such as Blender, FreeCAD, etc.).
[0039] Taking the simulation of the freezing process of droplets on the wing as an example, CAD software can be used to establish a droplet geometric model with a volume of 5μL and a droplet radius of 0.001m.
[0040] In some embodiments, establishing a mechanism model of the simulation object based on the geometric model of the simulation object includes:
[0041] Obtain the characteristics of the physical field and the research objectives;
[0042] Based on the geometric model of the simulation object, at least one mathematical equation is established to describe the behavior of the physical field, wherein the at least one mathematical equation includes at least a partial differential equation and / or a differential equation.
[0043] Specifically, the essential characteristics of physical fields involve an in-depth understanding of physical phenomena (such as mechanics, electromagnetism, thermodynamics, fluid mechanics, etc.). The characteristics of physical fields may include their spatial distribution, time evolution, interaction mechanisms, and relationships with external conditions (such as boundary conditions, initial conditions, external field sources, etc.). The research goal is to quantitatively analyze or predict these physical field characteristics in order to solve specific scientific problems or engineering needs. For example, the research goal may be to predict the flow state of fluids in pipes, calculate the effects of electromagnetic fields on electronic devices, analyze the deformation behavior of materials under thermal stress, etc.
[0044] The geometric model of the simulation object is the basis for describing the spatial distribution of the physical field. This model can be two-dimensional (such as a plane figure) or three-dimensional (such as a three-dimensional structure), depending on the complexity of the physical field and the research requirements. The geometric model defines the scope and boundary conditions of the physical field and is an important reference for establishing mathematical equations. Based on the characteristics of the physical field and the geometric model of the simulation object, at least one mathematical equation needs to be established to describe the behavior of the physical field. This equation can be a partial differential equation, a differential equation, or a combination of them, depending on the nature of the physical field and the research objectives. When the behavior of the physical field changes not only with time but also with the spatial position, a partial differential equation is needed to describe it. Partial differential equations can simultaneously consider the evolution of the physical field in time and space. If the behavior of the physical field only changes with time (or changes in one-dimensional space), differential equations can be used to describe it. The establishment of mathematical equations is usually based on physical laws (such as Newton's second law, Maxwell's equations, Fourier's heat conduction equations, etc.) and conservation principles (such as conservation of mass, conservation of energy, conservation of momentum, etc.). These equations need to accurately reflect the internal mechanism of the physical field and the influence of external conditions on the physical field.
[0045] Taking the simulation of the freezing process of a droplet on a wing as an example, in the process of applying the melting-solidification model to solve the temperature, the solution of the temperature is an iteration between the energy equation and the liquid fraction equation;
[0046]
[0047] Where t is the solidification time, is the speed; k is the thermal coefficient, T is the temperature at any time, and H is the enthalpy, which is the sum of the latent enthalpy and the sensible enthalpy;
[0048]
[0049] In the formula, h ref is the reference enthalpy value; T ref is the reference temperature; C p is the constant pressure specific heat capacity of the fluid; β is the liquid fraction; L is the latent heat of phase change; h is the sensible enthalpy.
[0050] Step 130: Acquire property change information of the simulation object based on the mechanism model corresponding to the simulation object.
[0051] In some embodiments, step 130 specifically includes:
[0052] Finite element analysis is performed based on the mechanism model corresponding to the simulation object to obtain the property change information of the simulation object.
[0053] Specifically, the geometric model of the simulation object is divided into multiple small units (such as triangles, quadrilaterals, tetrahedrons, etc.), which are connected to each other through nodes. The density and quality of the grid have an important influence on the accuracy of the analysis results and the calculation efficiency. The material properties of the simulation object (such as elastic modulus, density, thermal conductivity, etc.) are assigned to the corresponding units in the finite element model. According to the actual working environment of the simulation object, define appropriate boundary conditions (such as fixed constraints, sliding constraints, symmetric boundaries, etc.) and loads (such as force, pressure, temperature, etc.). These conditions will affect the property changes of the simulation object. Select the solver (such as static analysis, dynamic analysis, thermal analysis, fluid-solid coupling analysis, etc.) and solution options (such as convergence criteria, number of iterations, time step, etc.). These settings will affect the stability of the solution process and the accuracy of the results. Run the FEA (FiniteElementAnalysis) software to execute the solution process. The software will automatically calculate the property changes of each unit and node and generate the corresponding result file.
[0054] Preferably, the distribution of physical quantities such as stress, strain, temperature, velocity, and Reynolds coefficient of the simulation object in the entire geometric range can be obtained through matrix operations and numerical solution techniques.
[0055] Taking the simulation of the freezing process of droplets on the wing as an example, the established mechanism model is imported into the finite element analysis software, and the mesh is divided by combining structured and unstructured grids; the initial conditions and boundary conditions of droplet freezing are set, including parameters such as ambient temperature, initial temperature of droplets, thermal conductivity and latent heat of phase change; based on the finite element analysis software, the heat conduction equation and phase change equation are defined and solved to obtain the quantitative distribution and changes of various physical properties in the formation, growth and freezing of ice, and generate simulation animations of the formation and growth of ice on the wing and the heating and removal process of ice by the de-icing system; the boundary conditions involved in the calculation domain are air boundary and solid wall boundary. The width L of the simulation calculation domain is 0.009m, the height H is 0.012m, the air inlet boundary is the pressure inlet boundary (Pressure inlet), and the outlet boundary is the pressure outlet boundary; the solid wall boundary is a no-slip boundary, that is, the normal velocity along the wall is regarded as zero; the solid wall boundary is a no-slip boundary, that is, the normal velocity along the wall is zero; the boundary conditions use structured quadrilateral grids.
[0056] Step 140: Generate a state change animation of the simulation object based on the attribute change information of the simulation object.
[0057] Taking the simulation of droplet freezing on the wing as an example, the state change animation of the heat transfer process of the droplet freezing process is obtained.
[0058] Step 150: Based on the frame rate requirement, determine whether to add frames to the state change animation of the simulation object.
[0059] In some embodiments, step 150 specifically includes:
[0060] Based on the frame rate requirements, dynamic image quality requirements and virtual reality presentation hardware performance, determine whether the state image change rate of the state change animation of the simulation object meets the preset requirements.
[0061] Specifically, based on the frame rate requirements, dynamic image quality requirements and virtual reality presentation hardware performance, it is determined whether the state image change rate of the state change animation generated by the numerical simulation meets the smoothness required by the engineering project. If not, it is necessary to add frames to the state change animation of the simulation object.
[0062] Step 160: If not, the state change animation of the simulation object is imported into the virtual reality engine for presentation.
[0063] Specifically, relying on data interaction methods such as particle systems, the state change animation of the simulated object is imported into the virtual reality engine to generate dynamic vision, and then presented through virtual reality equipment.
[0064] Step 170, if yes, interpolate two adjacent frames of the simulation state diagram of the state change animation of the simulation object through deep learning technology to generate the state change animation of the simulation object after interpolation, and import the state change animation of the simulation object after interpolation into the virtual reality engine for presentation.
[0065] Figure 2 is a flow chart of interpolating two adjacent frames of a simulation state diagram of a state change animation of a simulation object according to some embodiments of this specification, such as Figure 2 As shown, in some embodiments, two adjacent frames of the simulation state diagram of the state change animation of the simulation object are interpolated by deep learning technology to generate the state change animation of the simulation object after interpolation, including:
[0066] For two adjacent frames of the simulation state graph of the state change animation of the simulation object, the pure noise hidden variables of each frame of the simulation state graph are calculated, the pure noise hidden variables of the two adjacent frames of the simulation state graph are interpolated to obtain multiple pure noise hidden variables generated by interpolation, the low-rank adaptive parameter matrix of each frame of the simulation state graph is trained, the low-rank adaptive parameter matrix of the two adjacent frames of the simulation state graph is interpolated to obtain multiple low-rank adaptive parameter matrices generated by interpolation, the self-attention mechanism matrix of each frame of the simulation state graph is calculated, the self-attention mechanism matrix of the two adjacent frames of the simulation state graph is interpolated to obtain multiple self-attention mechanism matrices generated by interpolation, and the multiple pure noise hidden variables generated by interpolation are cyclically denoised using the multiple low-rank adaptive parameter matrices generated by interpolation and the multiple self-attention mechanism matrices generated by interpolation to obtain multiple smooth transition images.
[0067] Figure 3is a schematic diagram of frame interpolation using an algorithm model integrated with LoRA and UNet according to some embodiments of this specification, such as Figure 3 As shown, as a preferred method, DDIM (Denoision Diffusion Implicit Models) can be used to invert and obtain pure noise hidden variables of each frame of the simulation state diagram. The pure noise hidden variables of the simulation state diagrams of two adjacent frames can be smoothly interpolated to obtain multiple pure noise hidden variables generated by interpolation, and the LoRA (Low-Rank Adaption) fine-tuning model can be used to encapsulate high-level image semantics into a low-rank parameter space, and generate semantically consistent samples when traversing potential noise. The algorithm model integrated by LoRA and UNet can be used to utilize multiple low-rank adaptive parameter matrices generated by interpolation and multiple self-attention mechanism matrices generated by interpolation, and multiple pure noise hidden variables generated by interpolation can be cyclically denoised to obtain multiple smooth transition images, thereby completing the interpolation of two adjacent frames of the simulation state diagram.
[0068] In some embodiments, the low-rank adaptive parameter matrix of two adjacent frames of simulation state diagrams is interpolated based on the following formula:
[0069] W ((t,t+1),n) =α n W t +(1-α n )W t+1 ,n=1,2……N
[0070] Among them, W ((t,t+1),n) The low-rank adaptive parameter matrix of the simulation state diagram of the t-th frame and the low-rank adaptive parameter matrix of the simulation state diagram of the t+1-th frame are interpolated to obtain the n-th interpolated low-rank adaptive parameter matrix, W t is the low-rank adaptive parameter matrix of the simulation state diagram of the t-th frame, W t+1 is the low-rank adaptive parameter matrix of the simulation state diagram of the t+1th frame, α n W ((t,t+1),n) The corresponding interpolation coefficient, N is the total number of low-rank adaptive parameter matrices generated by interpolation of the low-rank adaptive parameter matrix of the simulation state diagram of the t-th frame and the low-rank adaptive parameter matrix of the simulation state diagram of the t+1-th frame. The interpolation coefficients corresponding to any two low-rank adaptive parameter matrices generated by interpolation are different.
[0071] In some embodiments, the self-attention mechanism matrix of two adjacent frames of simulation state diagram is interpolated based on the following formula:
[0072] S ((t,t+1),n) =α n S t +(1-α n )S t+1,n=1,2……N
[0073] Among them, S ((t,t+1),n) The self-attention mechanism matrix of the simulation state diagram of the t-th frame and the self-attention mechanism matrix of the simulation state diagram of the t+1-th frame are interpolated to obtain the n-th interpolated self-attention mechanism matrix, S t is the self-attention mechanism matrix of the simulation state diagram of the t-th frame, S t+1 is the self-attention mechanism matrix of the simulation state diagram of the t+1th frame, α n For S ((t,t+1),n) The corresponding interpolation coefficient, N is the total number of self-attention mechanism matrices generated by interpolation of the self-attention mechanism matrix of the simulation state diagram of the t-th frame and the self-attention mechanism matrix of the simulation state diagram of the t+1-th frame. The interpolation coefficients corresponding to any two interpolated self-attention mechanism matrices are different.
[0074] In order to meet the requirements of fast presentation, the numerical simulation step size needs to be set as large as possible; because the use of numerical simulation to simulate the real physical property changes of objects will inevitably take a lot of computing time when the simulation step size is small or the simulation time is long, which will greatly reduce the efficiency of virtual reality scene modeling. When the numerical simulation step size is large, the rendering of the state change animation in the simulation results is not smooth, so it is necessary to perform frame interpolation on the simulation results with a low refresh rate. A deep learning algorithm is used to generate continuous frames of physical property changes for virtual visualization, that is, data-driven visualization presentation; among them, the deep learning algorithm improves the smoothness of semantic conversion and continuous images by fitting the low-rank adaptation of two images, introducing attention mechanism interpolation and injection technology, adaptive normalization adjustment method and new sampling scheme. Virtual reality engine dynamic visual presentation, according to high-fidelity virtual reality presentation, the physical object changes in the virtual engine not only conform to the real physical property change law, but also the animation is displayed in no less than the specified number of frames.
[0075] Preferably, the state change animation of the simulated object after frame complementation is imported into the virtual reality engine to generate dynamic vision by relying on data interaction methods such as particle system, and then presented through virtual reality equipment.
[0076] Taking the simulation of the freezing process of droplets on the wing as an example, the animation of the freezing process of droplets on the wing after frame filling is used to capture the corresponding color rendering values in the form of discrete points, and then imported into the virtual reality engine for rendering in the form of particle swarms, and finally presented to the user through the virtual reality device; the UE-5 engine is known for its highly realistic graphics rendering capabilities, especially in the fields of virtual textures and global illumination. The animation of the freezing process of droplets on the wing after frame filling can be imported into the high-level realm and highly realistic virtual environment rendered by the UE-5 engine. Figure 4is a schematic diagram of the result after interpolating two adjacent frames of the simulation state diagram of the state change animation of the simulation object according to some embodiments of this specification, such as Figure 4 As shown, image interpolation is performed on the simulation state diagrams of 7-8 seconds, 8-9 seconds, and 9-10 seconds, respectively, and the four discrete simulation pictures in the middle are transformed into continuous simulation results with a high refresh rate; the results of the animation frame filling reflect the dynamic changes in the physical properties of the droplets, which can realistically reflect the changes in the physical properties of the actual droplet freezing.
[0077] Figure 5 is a module diagram of a virtual reality rapid presentation system based on real physical properties according to some embodiments of this specification, such as Figure 5 As shown, a virtual reality rapid presentation system based on real physical properties may include a property determination module, a model building module, a change acquisition module, an animation generation module, a frame supplement determination module and a virtual reality presentation module.
[0078] The property determination module can be used to obtain the physical property information of the simulation object.
[0079] The model building module can be used to build a mechanism model corresponding to the simulation object based on the physical property information corresponding to the simulation object.
[0080] The change acquisition module can be used to obtain the property change information of the simulation object based on the mechanism model corresponding to the simulation object.
[0081] The animation generation module can be used to generate state change animation of the simulation object based on the property change information of the simulation object.
[0082] The interpolation frame determination module can be used to determine whether to interpolate frames for the state change animation of the simulation object based on the frame rate requirement.
[0083] The virtual reality presentation module can be used to import the state change animation of the simulation object into the virtual reality engine for presentation after the frame interpolation judgment model determines that the state change animation of the simulation object does not need to be interpolated. It can also be used to interpolate two adjacent frames of the simulation state diagram of the state change animation of the simulation object through deep learning technology after the frame interpolation judgment module determines that the state change animation of the simulation object needs to be interpolated, generate the state change animation of the simulation object after interpolation, and import the state change animation of the simulation object after interpolation into the virtual reality engine for presentation.
[0084] A virtual reality rapid presentation system based on real physical properties can be used to execute a virtual reality rapid presentation method based on real physical properties. For more descriptions of a virtual reality rapid presentation system based on real physical properties, please refer to the relevant descriptions of a virtual reality rapid presentation method based on real physical properties, which will not be repeated here.
[0085] Finally, it should be understood that the embodiments described in this specification are only used to illustrate the principles of the embodiments of this specification. Other variations may also fall within the scope of this specification. Therefore, as an example and not a limitation, alternative configurations of the embodiments of this specification may be considered consistent with the teachings of this specification. Accordingly, the embodiments of this specification are not limited to the embodiments explicitly introduced and described in this specification.
Claims
1. A virtual reality rapid presentation method based on real physical properties, characterized in that: include: Obtain physical property information of simulation objects; Based on the physical property information corresponding to the simulation object, establishing a mechanism model corresponding to the simulation object; Based on the mechanism model corresponding to the simulation object, obtaining the property change information of the simulation object; Based on the property change information of the simulation object, generating a state change animation of the simulation object; Based on the frame rate requirement, determining whether to add frames to the state change animation of the simulation object; If not, the state change animation of the simulation object is imported into a virtual reality engine for presentation; If so, two adjacent frames of the simulation state diagram of the state change animation of the simulation object are interpolated by deep learning technology to generate the state change animation of the simulation object after interpolation, and the state change animation of the simulation object after interpolation is imported into the virtual reality engine for presentation.
2. The method for rapid virtual reality presentation based on real physical properties according to claim 1, characterized in that: The obtaining of physical property information of the simulation object includes: Acquiring geometric shape information and size information of the simulation object; Using a data acquisition tool to collect the attribute data of the simulation object, wherein the data acquisition tool includes an experimental test tool and / or a simulation tool; The property data of the simulation object is processed and analyzed to determine the physical property value of the simulation object, wherein the physical property information of the simulation object at least includes geometric shape information, size information and physical property value of the simulation object.
3. The method for rapid virtual reality presentation based on real physical properties according to claim 2, characterized in that: Based on the physical property information corresponding to the simulation object, a mechanism model corresponding to the simulation object is established, including: Based on the geometric shape information and size information of the simulation object, a geometric model of the simulation object is established; Based on the geometric model of the simulation object, a mechanism model corresponding to the simulation object is established.
4. The method for rapid virtual reality presentation based on real physical properties according to claim 3, characterized in that: Based on the geometric model of the simulation object, a mechanism model of the simulation object is established, including: Obtain the characteristics of the physical field and the research objectives; Based on the geometric model of the simulation object, at least one mathematical equation is established to describe the behavior of the physical field, wherein the at least one mathematical equation includes at least a partial differential equation and / or a differential equation.
5. The method for rapid virtual reality presentation based on real physical properties according to claim 4, characterized in that: Based on the mechanism model corresponding to the simulation object, obtaining the property change information of the simulation object includes: Finite element analysis is performed based on a mechanism model corresponding to the simulation object to obtain property change information of the simulation object.
6. The method for rapid virtual reality presentation based on real physical properties according to claim 5, characterized in that: Performing finite element analysis based on the mechanism model of the simulation object to obtain property change information of the simulation object includes: Imposing boundary conditions and initial conditions; Finite element analysis software is used to solve at least one established mathematical equation through a finite element numerical method to obtain the property change information of the simulation object.
7. A virtual reality rapid presentation method based on real physical properties according to any one of claims 1 to 6, characterized in that: The step of determining whether to add frames to the state change animation of the simulation object based on the frame rate requirement includes: Based on the frame rate requirement, dynamic image quality requirement and virtual reality presentation hardware performance, it is determined whether the state image change rate of the state change animation of the simulation object meets the preset requirement.
8. A virtual reality rapid presentation method based on real physical properties according to any one of claims 1 to 6, characterized in that: The method of interpolating two adjacent frames of the state change animation of the simulation object by deep learning technology to generate the state change animation of the simulation object after interpolation includes: For two adjacent frames of the simulation state graph of the state change animation of the simulation object, the pure noise hidden variables of each frame of the simulation state graph are calculated, the pure noise hidden variables of the two adjacent frames of the simulation state graph are interpolated to obtain multiple pure noise hidden variables generated by interpolation, the low-rank adaptive parameter matrix of each frame of the simulation state graph is trained, the low-rank adaptive parameter matrix of the two adjacent frames of the simulation state graph is interpolated to obtain multiple low-rank adaptive parameter matrices generated by interpolation, the self-attention mechanism matrix of each frame of the simulation state graph is calculated, the self-attention mechanism matrix of the two adjacent frames of the simulation state graph is interpolated to obtain multiple self-attention mechanism matrices generated by interpolation, and the multiple low-rank adaptive parameter matrices generated by interpolation and the multiple self-attention mechanism matrices generated by interpolation are used to perform cyclic denoising on the multiple pure noise hidden variables generated by interpolation to obtain multiple smooth transition images.
9. The method for rapid virtual reality presentation based on real physical properties according to claim 8, characterized in that: The low-rank adaptive parameter matrix of the simulation state diagram of two adjacent frames is interpolated based on the following formula: W ((t,t+1),n) =a n W t +(1-a n )W t+1 ,n=1,2……N Among them, w ((t,t+1),n) The low-rank adaptive parameter matrix of the simulation state diagram of the t-th frame and the low-rank adaptive parameter matrix of the simulation state diagram of the t+1-th frame are interpolated to obtain the n-th interpolated low-rank adaptive parameter matrix, W t is the low-rank adaptive parameter matrix of the simulation state diagram of the t-th frame, W t+1 is the low-rank adaptive parameter matrix of the simulation state diagram of the t+1th frame, α n W ((t,t+1),n) The corresponding interpolation coefficient, N is the total number of low-rank adaptive parameter matrices generated by interpolating the low-rank adaptive parameter matrix of the t-th frame simulation state diagram and the low-rank adaptive parameter matrix of the t+1-th frame simulation state diagram.
10. A virtual reality rapid presentation system based on real physical properties, characterized in that: include: The property determination module is used to obtain the physical property information of the simulation object; A model building module, used to build a mechanism model corresponding to the simulation object based on the physical property information corresponding to the simulation object; A change acquisition module, used for acquiring attribute change information of the simulation object based on a mechanism model corresponding to the simulation object; An animation generation module, used for generating a state change animation of the simulation object based on the property change information of the simulation object; A frame interpolation determination module, used for determining whether to interpolate frames for the state change animation of the simulation object based on a frame rate requirement; A virtual reality presentation module is used to import the state change animation of the simulation object into a virtual reality engine for presentation after the frame interpolation determination model determines that the state change animation of the simulation object does not need to be interpolated, and is also used to interpolate two adjacent frames of the simulation state diagram of the state change animation of the simulation object through deep learning technology after the frame interpolation determination module determines that the state change animation of the simulation object is to be interpolated, to generate the state change animation of the simulation object after the interpolation, and to import the state change animation of the simulation object after the interpolation into the virtual reality engine for presentation.
