Fast gauss splash physics simulation method and related apparatus
By combining segmentation, field occupancy, and Monte Carlo sampling with a large language model to optimize physical parameters, the problem of long simulation time and inaccurate parameters in existing four-dimensional physics simulations is solved, achieving fast and efficient physics simulation, which is applicable to fields such as virtual reality and games.
Patent Information
- Application Number
- CN202610828083.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-06-09
- Publication Date
- 2026-07-14
AI Technical Summary
Existing four-dimensional physics simulation methods suffer from problems such as long processing time, complex and inaccurate physical parameter design when generating realistic physical dynamic content, especially when simulating in three-dimensional Gaussian scenes, which can easily lead to collapsed and unrealistic results.
The fast Gaussian splash physics simulation method is adopted. By segmenting objects in a static 3D Gaussian scene, an occupation field is constructed and Monte Carlo importance probability sampling is used to fill points. Combined with a large language model to optimize physical parameters, the material point method simulation is realized.
It enables the rapid and accurate generation of physical models, improving the stability and realism of simulation results, and is suitable for applications such as virtual reality, game and movie scene production.
Smart Images
Figure CN122389654A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of physical model simulation generation technology, and in particular to a fast Gaussian splash physics simulation method and related equipment. Background Technology
[0002] Text-based 4D content generation methods: Unlike 3D content, 4D content requires more consideration in terms of spatiotemporal consistency and dynamic realism. Recently, video generation models have developed rapidly, such as SVD and Sora, and many works are based on video generation models to create 4D dynamic content. MAV3D first utilizes temporal SDS in the text-to-video diffusion model to optimize the HexPlane spatial representation. 4D-fy introduces hybrid fractional distillation HSD during training to achieve high-quality text-to-4D generation. Consistent4D uses SDS for geometric optimization and combines interpolation loss to improve the spatiotemporal consistency of 4D content generated from monocular videos. 4DGen uses pseudo-labels generated by a multi-view diffusion model for supervised training. Animate124 proposes a graph-based 4D framework using a coarse-to-fine strategy. DreamGaussian4D generates 4D content from a single image without using temporal SDS, but instead optimizes it based on reference videos generated by a video diffusion model. However, these methods often face challenges when simulating more realistic physical dynamics because video diffusion models lack fully reasonable physical understanding and reasoning capabilities, and such methods require a lengthy optimization process.
[0003] Four-dimensional Gaussian simulation methods based on physical modeling: Recently, some methods have proposed using the Matter Point Method (MPM) for more realistic dynamic simulations. MPM discretizes particles into Lagrange points carrying physical states while solving the momentum equation on an Eulerian background grid. MPM can naturally handle large deformations and complex interactions, exhibiting high robustness. A representative work is PhysGaussian proposed by Xie et al., which achieves realistic four-dimensional Gaussian content simulation by combining the MPM method with 3DGS. However, it requires users to design complex physical parameters, which is inconvenient. Some methods use fractional distillation sampling (SDS) to extract dynamic priors from video diffusion models; for example, DreamPhysics and Physics3D utilize SDS to optimize physical parameters. However, this approach is time-consuming and cannot guarantee the reasonableness of the dynamic priors in the video. To address this, methods such as PhysSplat and Phys4DGen use Large Language Modeling (LLM) to predict physical parameters, but this introduces perceptual errors with the actual physical space, leading to unreliable predictions. Furthermore, existing methods often ignore the inherent characteristics of the original 3D representation. For example, 3D Gaussian is a surface reconstruction algorithm with a hollow internal structure. Using such a hollow structure for simulation will lead to collapsed and unrealistic results. Summary of the Invention
[0004] The main objective of this application is to propose a fast Gaussian splash physics simulation method and related equipment to generate physical models efficiently and accurately.
[0005] To achieve the above objectives, one aspect of this application proposes a fast Gaussian splash physics simulation method, the method comprising the following steps: Segment the objects in a static 3D Gaussian scene; An occupancy field is constructed, and then corresponding points are assigned to the interior of the object based on the occupancy probability; The points inside the object are filled using the Monte Carlo importance probability sampling method; The physical parameters of the object for material point method simulation are obtained based on the text description using a large language model; Simulation results for each of the objects are generated based on the physical parameters.
[0006] In some embodiments, segmenting the objects in a static 3D Gaussian scene includes the following steps: The static 3D Gaussian scene is clustered and segmented using the DBSCAN algorithm. The calculation formula includes: ; in, Let be the cluster and label of the k-th cluster, respectively. Represents a three-dimensional Gaussian point. The cluster radius; Calculate the convex hull using the QuickHull algorithm. , to compute the enclosing set of each cluster; Calculate the AABB bounding box to obtain the points around the boundary of the object. The calculation formula includes: ; in, Let b be the AABB bounding box constructed based on the convex hull, where b is the bounding box boundary value. Candidate points around the AABB bounding box, Representative from The j-th point of the internal sampling, This represents random sampling from the bounding box.
[0007] In some embodiments, constructing an occupancy field and then assigning corresponding points to the interior of the object based on occupancy probabilities includes the following steps: Construct an occupancy field (OCC), and then calculate the occupancy probability of each point; Points with an occupancy probability greater than a set threshold are assigned to the interior of the corresponding object.
[0008] In some embodiments, filling the points inside the object using the Monte Carlo importance probability sampling method includes the following steps: Calculate the distance between each point and the nearest point. : ; in, c Represents the coordinates of the nearest point; Based on the Gaussian distribution properties, the importance weights are... Defined as a probability distribution, the calculation formula includes: ; in, Represents the Gaussian standard deviation; Importance weight Normalization: ; Based on the probability function obtained after normalization Sampling candidate points yields the final filled Gaussian points.
[0009] In some embodiments, obtaining the physical parameters of the object for material point method simulation based on the text description using a large language model includes the following steps: Predict the Young's modulus E and potential energy model of the object based on the text description using a large language model. The material density ρ is used as the physical parameter; a fixed Poisson's ratio is used to define the neutral deformation behavior of the material, including elastic and plastic properties, thereby determining the material density ρ; the potential energy model is determined using a predefined functional form. ; The Young's modulus E is optimized by using a gradient-guided stress-strain τ that reflects stress changes.
[0010] In some embodiments, optimizing the Young's modulus E by using a gradient-guided stress-strain τ that reflects stress changes includes the following steps: The gradient reflecting the stress change is calculated as follows: ; in, This refers to the gradient that reflects stress changes; The total number of images, The image number is the sequence number. For the first Stress at points in the image, For the first Young's modulus of a point in the image; Amplitude control is performed using log operations, and the log gradient value is multiplied by the CLIP coding score to guide the optimization of the amplitude. The calculation includes: ; in, The optimization range is described above; The image-text similarity loss is obtained by encoding using the CLIP model; The optimized Young's modulus E obtained using the gradient is used for forward material point method simulation. The calculation formula includes: .
[0011] In some embodiments, the method further includes the following steps: The simulation results of each of the objects are applied to at least one of virtual reality, games, movie scene creation, or embodied intelligent simulation.
[0012] To achieve the above objectives, another aspect of the embodiments of this application proposes a fast Gaussian splash physics simulation device, the device comprising: The object segmentation unit is used to segment out individual objects in a static 3D Gaussian scene; A point partitioning unit is used to construct an occupancy field, and then partition the corresponding points into the interior of the object according to the occupancy probability; Point filling units are used to fill points inside the object using the Monte Carlo importance probability sampling method; The physical parameter determination unit is used to obtain the physical parameters of the object for material point method simulation based on the text description using a large language model; The simulation generation unit is used to generate simulation results for each of the objects based on the physical parameters.
[0013] To achieve the above objectives, another aspect of this application provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the above-described method.
[0014] To achieve the above objectives, another aspect of the embodiments of this application proposes a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method.
[0015] To achieve the above objectives, another aspect of this application provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.
[0016] The embodiments of this application include at least the following beneficial effects: This application provides a fast Gaussian splash physics simulation method and related equipment. The scheme involves segmenting objects in a static 3D Gaussian scene; constructing an occupancy field; then assigning corresponding points to the interior of the objects based on occupancy probabilities; filling the points inside the objects using a Monte Carlo importance probability sampling method; obtaining the physical parameters of the objects for material point method simulation based on text descriptions using a large language model; and generating simulation results for each object based on the physical parameters. This application first segments the input static 3D Gaussian scene, then obtains preliminary filling results using the occupancy field, and then uses the proposed Monte Carlo importance probability sampling method to fill complex structures based on Gaussian importance. The text-based physics optimization based on a large language model proposed in this application can quickly optimize and determine the physical parameters of the objects, obtaining stable simulation results. Attached Figure Description
[0017] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0018] Figure 1 A flowchart illustrating the fast Gaussian splash physics simulation method provided in this application embodiment; Figure 2 Example flowchart of the fast Gaussian splash physics simulation method provided in the embodiments of this application; Figure 3 A schematic diagram of the structure of the fast Gaussian splash physics simulation device provided in the embodiments of this application; Figure 4 This is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of this application. Detailed Implementation
[0019] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit it. In the following description, when referring to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with those of this application; they are merely examples of apparatuses and methods consistent with some aspects of the embodiments of this application as detailed in the appended claims.
[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.
[0021] This application provides a fast Gaussian splash physics simulation method and related equipment, relating to the field of physical model simulation and generation technology. The fast Gaussian splash physics simulation method and related equipment provided in this application can be applied to a terminal, a server, or software running on a terminal or server. In some embodiments, the terminal can be a smartphone, tablet, laptop, desktop computer, smart speaker, smartwatch, or vehicle terminal, but is not limited thereto; the server can be configured as an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The server can also be a node server in a blockchain network; the software can be an application implementing the fast Gaussian splash physics simulation method, but is not limited to the above forms.
[0022] This application can be used in a wide variety of general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics devices, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.
[0023] Reference Figure 1 This application provides a fast Gaussian splash physics simulation method, which may include, but is not limited to, steps S100 to S140, as follows: S100: Segment the objects in a static 3D Gaussian scene; S110: Construct an occupancy field, and then assign the corresponding points to the interior of the object according to the occupancy probability; S120: Fill the points inside the object using the Monte Carlo importance probability sampling method; S130: Use a large language model to obtain the physical parameters of the object for material point method simulation based on the text description; S140: Generate simulation results for each of the objects based on the physical parameters.
[0024] In some embodiments, segmenting the objects in a static 3D Gaussian scene includes the following steps: The static 3D Gaussian scene is clustered and segmented using the DBSCAN algorithm. The calculation formula includes: ; in, Let be the cluster and label of the k-th cluster, respectively. Represents a three-dimensional Gaussian point. The cluster radius; Calculate the convex hull using the QuickHull algorithm. , to compute the enclosing set of each cluster; Calculate the AABB bounding box to obtain the points around the boundary of the object. The calculation formula includes: ; in, Let b be the AABB bounding box constructed based on the convex hull, where b is the bounding box boundary value. Candidate points around the AABB bounding box, Representative from The j-th point of the internal sampling, This represents random sampling from the bounding box.
[0025] In some embodiments, constructing an occupancy field and then assigning corresponding points to the interior of the object based on occupancy probabilities includes the following steps: Construct an occupancy field (OCC), and then calculate the occupancy probability of each point; Points with an occupancy probability greater than a set threshold are assigned to the interior of the corresponding object.
[0026] In some embodiments, filling the points inside the object using the Monte Carlo importance probability sampling method includes the following steps: Calculate the distance between each point and the nearest point. : ; in, c Represents the coordinates of the nearest point; Based on the Gaussian distribution properties, the importance weights are... Defined as a probability distribution, the calculation formula includes: ; in, Represents the Gaussian standard deviation; Importance weight Normalization: ; Based on the probability function obtained after normalization Sampling candidate points yields the final filled Gaussian points.
[0027] In some embodiments, obtaining the physical parameters of the object for material point method simulation based on the text description using a large language model includes the following steps: Predict the Young's modulus E and potential energy model of the object based on the text description using a large language model. The material density ρ is used as the physical parameter; a fixed Poisson's ratio is used to define the neutral deformation behavior of the material, including elastic and plastic properties, thereby determining the material density ρ; the potential energy model is determined using a predefined functional form. ; The Young's modulus E is optimized by using a gradient-guided stress-strain τ that reflects stress changes.
[0028] In some embodiments, optimizing the Young's modulus E by using a gradient-guided stress-strain τ that reflects stress changes includes the following steps: The gradient reflecting the stress change is calculated as follows: ; in, This refers to the gradient that reflects stress changes; The total number of images, The image number is the sequence number. For the first Stress at points in the image, For the first Young's modulus of a point in the image; Amplitude control is performed using log operations, and the log gradient value is multiplied by the CLIP coding score to guide the optimization of the amplitude. The calculation includes: ; in, The optimization range is described above; The image-text similarity loss is obtained by encoding using the CLIP model; The optimized Young's modulus E obtained using the gradient is used for forward material point method simulation. The calculation formula includes: .
[0029] In some embodiments, the method further includes the following steps: The simulation results of each of the objects are applied to at least one of virtual reality, games, movie scene creation, or embodied intelligent simulation.
[0030] The following sections will provide a detailed description and explanation of some optional embodiments of this application, using specific application examples.
[0031] This embodiment proposes a fast four-dimensional Gaussian splash physics simulation method, FastPhysGS, based on the matter point method. To obtain a stable spatial structure, the internal particles are first roughly filled using 3D segmentation and occupancy field, and Monte Carlo importance sampling is proposed to refine the filling. To achieve fast and robust simulation, a text-based bidirectional graph optimization algorithm is proposed. The forward graph performs only matter point method simulation, while the backward graph focuses on the rapid optimization of physical parameters, resulting in a low-time-consuming and low-memory simulation process. Experiments show that each strategy and module is necessary; their organic combination enables fast and robust four-dimensional Gaussian splash physics simulation, including common materials (elastoplastic, liquid, soft, etc.) and physical motions (movement, rotation, collision, oscillation, flow, etc.).
[0032] The process of the method in this embodiment is as follows: Figure 2 As shown, this implementation mainly consists of two stages: Hull-based Fast Particle Filling (HFPF) and Text-Physics Optimization (TPO). In HFPF, a static 3D Gaussian scene is first given, and the DBSCAN algorithm is used to segment and derive instances of each object. Next, an occupancy field is constructed, and the occupancy probability is used to determine whether a point is inside an object. Then, Monte Carlo Importance Sampling (MCIS) is designed to handle the filling of complex structures. In TPO, a bidirectional graph decoupling algorithm is proposed, which achieves rapid optimization and dynamic simulation of physical parameters through forward simulation and backward optimization. The details of each stage will be described in the following sections.
[0033] Specifically, this embodiment includes the following technical solutions: 1. Fast particle filling HFPF based on convex hull.
[0034] After inputting a static 3D Gaussian scene, the DBSCAN algorithm is first used for clustering and segmentation: ; in, For the k-th cluster and its label, Represents a three-dimensional Gaussian point. is the cluster radius.
[0035] To compute the bounding set of each cluster, the QuickHull algorithm is used to calculate the convex hull. Then, the AABB bounding box was calculated, and the points around the boundary were initially obtained: ; in, Let b be the AABB bounding box constructed based on the convex hull, where b is the bounding box boundary value. Candidate points around the bounding box, Representative from The j-th point of the internal sampling, This represents random sampling from the bounding box.
[0036] Then, to determine whether a point is inside an object, an occupancy field (OCC) is constructed, and an occupancy threshold is used to roughly determine whether a point is inside an object. ; It is understood that the threshold value set in this embodiment is 0.6.
[0037] However, for some complex structures, such as irregular curved surfaces or concave objects, the above method can lead to incorrect spatial filling. Therefore, Monte Carlo Importance Sampling (MCIS) is introduced to address this issue. First, to calculate the importance of each point, its distance to the nearest point is calculated: ; Where c represents the coordinates of the nearest point. Then, based on the Gaussian distribution property, the importance weights are defined as a probability distribution: ; in, This represents the Gaussian standard deviation. Then, the probability weights are normalized for easier calculation: ; Finally, candidate points can be sampled according to this probability function to obtain the final filled Gaussian points.
[0038] 2. Text Physical Optimization (TPO).
[0039] Optimize the target analysis. To facilitate the acquisition of physical parameters for MPM simulation, a text description is provided, such as "This is an object with a relatively soft rubber material. Please predict its physical parameters for material point method simulation." Large Language Model (LLM) is used to predict Young's modulus E and the potential energy model. And the material density ρ. By fixing the Poisson's ratio ν to 0.3, this value effectively defines the neutral deformation behavior of the material, including both elastic and plastic properties. Simultaneously, the potential energy model... Using predefined functional forms (e.g., fixed rotational elastic model, StVK elastic model, Drucker-Prager plasticity model, and fluid plasticity model), LLM is directly used to predict the category to which the material belongs. Therefore, given the crucial role of Young's modulus E in determining material stiffness, it is used as the primary optimization objective. However, a perceptual gap exists between the LLM inference output and the native 3DGS space, resulting in generated parameters that are not suitable for simulation. To address this, a fast and robust optimization method is explored, using stress-strain τ from the MPM simulation process to guide the optimization.
[0040] Two-way graph decoupling. To address the above issues, a two-way graph decoupling method is proposed to decouple the forward simulation from the backward graph optimization process. For example... Figure 2 As shown, a forward MPM simulation without gradient calculation is first performed. The forward MPM simulation process utilizes a Lagrange-Eulerian mesh to combine the deformation gradient and positional changes of the 3D Gaussian grid, achieving continuous dynamic simulation. This process is temporal and causal, meaning that subsequent frames are only affected by previous states. Therefore, only three rendered keyframes (initial frames) are recorded. Intermediate frames and the final frame The properties of the particle are: particle position x, particle velocity v, velocity gradient C, and deformation gradient F. This provides effective prior information for subsequent optimization while minimizing computational and storage overhead. Subsequently, gradient calculation is enabled in the inverse plot, and a pseudo-simulation process is executed—this process only preserves the gradient propagation path and does not involve real physical motion. To achieve the goal of optimizing Young's modulus E using stress and strain τ, a gradient reflecting stress changes is used to guide the optimization process: ; Then, the magnitude is controlled using a log operation, and the log gradient value is multiplied by the CLIP encoding score (obtained by calculating embedding similarity from CLIP) to guide the optimization magnitude: ; In this way, the gradient can be used to obtain the optimized result of Young's modulus E, thereby enabling a more stable forward MPM simulation. The calculation formula includes: ; In summary, the method of this embodiment has potential application value for downstream tasks such as virtual reality, games, movie scene production, and embodied intelligent simulation.
[0041] For example, some application scenarios of this embodiment are as follows: 1. User-interactive four-dimensional Gaussian splashing and physical dynamic simulation.
[0042] 2. A resource-friendly paradigm for rapid space filling and optimization.
[0043] 3. Batch production of high-quality static-dynamic pairing datasets.
[0044] This embodiment constructs a four-dimensional Gaussian splash physics simulation method, FastPhysGS, based on the material point method. For static Gaussian scene input, users can describe motion and materials using text. First, the input Gaussian scene is segmented, and then the occupancy field is used to obtain preliminary filling results. Subsequently, the proposed Monte Carlo Importance Sampling (MCIS) can fill complex structures according to Gaussian importance. The proposed text physics optimization uses a convenient paradigm of bidirectional graph decoupling to quickly optimize physical parameters and obtain stable simulation results. Compared with other four-dimensional physics generation and simulation methods, the method of this embodiment is 1.3% and 6% higher than the current best methods in terms of text matching accuracy and image fidelity, reaching 0.292 and 0.87 (the top two indicators are approximately 0.279 and 0.81). Furthermore, the method of this embodiment has significant advantages in visual effects, which is sufficient to prove the effectiveness of the method in this embodiment.
[0045] Reference Figure 3 This application also provides a fast Gaussian splash physics simulation device, which can implement the above-mentioned fast Gaussian splash physics simulation method. The device includes: The object segmentation unit is used to segment out individual objects in a static 3D Gaussian scene; A point partitioning unit is used to construct an occupancy field, and then partition the corresponding points into the interior of the object according to the occupancy probability; Point filling units are used to fill points inside the object using the Monte Carlo importance probability sampling method; The physical parameter determination unit is used to obtain the physical parameters of the object for material point method simulation based on the text description using a large language model; The simulation generation unit is used to generate simulation results for each of the objects based on the physical parameters.
[0046] It is understood that the content of the above method embodiments is applicable to the present device embodiments. The specific functions implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.
[0047] This application also provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the method of this application. This electronic device can be any smart terminal, including tablet computers, in-vehicle computers, etc.
[0048] It is understood that the content of the above method embodiments is applicable to the device embodiments. The specific functions implemented by the device embodiments are the same as those of the methods of this application, and the beneficial effects achieved are the same as those achieved by the methods of this application.
[0049] Figure 4 The hardware structure of an electronic device according to another embodiment is illustrated. The electronic device includes: The processor 101 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application. The memory 102 can be implemented as a read-only memory (ROM), static storage device, dynamic storage device, or random access memory (RAM). The memory 102 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 102 and is called and executed by the processor 101. Input / output interface 103 is used to implement information input and output; The communication interface 104 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.). Bus 105 transmits information between various components of the device (e.g., processor 101, memory 102, input / output interface 103, and communication interface 104); The processor 101, memory 102, input / output interface 103 and communication interface 104 are connected to each other within the device via bus 105.
[0050] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method of this application.
[0051] It is understood that the content of the above method embodiments is applicable to this storage medium embodiment. The specific functions implemented in this storage medium embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.
[0052] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.
[0053] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.
[0054] The embodiments described in this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. As those skilled in the art will know, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.
[0055] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of this application, and may include more or fewer steps than shown, or combine certain steps, or different steps.
[0056] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.
[0057] Those skilled in the art will understand that all or some of the steps in the methods disclosed above, as well as the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, or suitable combinations thereof.
[0058] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.
[0059] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.
[0060] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of the units described above is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.
[0061] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.
[0062] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.
[0063] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes multiple instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing programs, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0064] The preferred embodiments of the present application have been described above with reference to the accompanying drawings, but this does not limit the scope of the claims of the present application. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and substance of the embodiments of the present application shall be within the scope of the claims of the present application.
Claims
1. A fast Gaussian splash physics simulation method, characterized in that, The method includes the following steps: Segment the objects in a static 3D Gaussian scene; An occupancy field is constructed, and then corresponding points are assigned to the interior of the object based on the occupancy probability; The points inside the object are filled using the Monte Carlo importance probability sampling method; The physical parameters of the object for material point method simulation are obtained based on the text description using a large language model; Simulation results for each of the objects are generated based on the physical parameters.
2. The fast Gaussian splash physics simulation method according to claim 1, characterized in that, The process of segmenting objects in a static 3D Gaussian scene includes the following steps: The static 3D Gaussian scene is clustered and segmented using the DBSCAN algorithm. The calculation formula includes: ； in, Let be the cluster and label of the k-th cluster, respectively. Represents a three-dimensional Gaussian point. The cluster radius; Calculate the convex hull using the QuickHull algorithm. , to compute the enclosing set of each cluster; Calculate the AABB bounding box to obtain the points around the boundary of the object. The calculation formula includes: ； in, Let b be the AABB bounding box constructed based on the convex hull, where b is the bounding box boundary value. Candidate points around the AABB bounding box, Representative from The j-th point of the internal sampling, This represents random sampling from the bounding box.
3. The fast Gaussian splash physics simulation method according to claim 2, characterized in that, The process of constructing an occupancy field and then assigning corresponding points to the interior of the object based on the occupancy probability includes the following steps: Construct an occupancy field (OCC), and then calculate the occupancy probability of each point; Points with an occupancy probability greater than a set threshold are assigned to the interior of the corresponding object.
4. The fast Gaussian splash physics simulation method according to claim 3, characterized in that, The method of filling the points inside the object using the Monte Carlo importance probability sampling method includes the following steps: Calculate the distance between each point and the nearest point. : ； in, c Represents the coordinates of the nearest point; Based on the Gaussian distribution properties, the importance weights are... Defined as a probability distribution, the calculation formula includes: ； in, Represents the Gaussian standard deviation; Importance weight Normalization: ； Based on the probability function obtained after normalization Sampling candidate points yields the final filled Gaussian points.
5. The fast Gaussian splash physics simulation method according to claim 1, characterized in that, The process of obtaining the physical parameters of the object for material point method simulation based on the text description using a large language model includes the following steps: Predict the Young's modulus E and potential energy model of the object based on the text description using a large language model. The material density ρ is used as the physical parameter; a fixed Poisson's ratio is used to define the neutral deformation behavior of the material, including elastic and plastic properties, thereby determining the material density ρ; the potential energy model is determined using a predefined functional form. ; The Young's modulus E is optimized by using a gradient-guided stress-strain τ that reflects stress changes.
6. The fast Gaussian splash physics simulation method according to claim 5, characterized in that, The optimization of Young's modulus E using gradient-guided stress-strain τ that reflects stress changes includes the following steps: The gradient reflecting the stress change is calculated as follows: ； in, This refers to the gradient that reflects stress changes; The total number of images, The image number is the sequence number. For the first Stress at points in the image, For the first Young's modulus of a point in the image; Amplitude control is performed using log operations, and the log gradient value is multiplied by the CLIP coding score to guide the optimization of the amplitude. The calculation includes: ； in, The optimization range is described above. The image-text similarity loss is obtained by encoding using the CLIP model; The optimized Young's modulus E obtained using the gradient is used for forward material point method simulation. The calculation formula includes: 。 7. The fast Gaussian splash physics simulation method according to any one of claims 1 to 6, characterized in that, The method further includes the following steps: The simulation results of each of the objects are applied to at least one of virtual reality, games, movie scene creation, or embodied intelligent simulation.
8. A rapid Gaussian splashing physics simulation device, characterized in that, The device includes: The object segmentation unit is used to segment out individual objects in a static 3D Gaussian scene; A point partitioning unit is used to construct an occupancy field, and then partition the corresponding points into the interior of the object according to the occupancy probability; Point filling units are used to fill points inside the object using the Monte Carlo importance probability sampling method; The physical parameter determination unit is used to obtain the physical parameters of the object for material point method simulation based on the text description using a large language model; The simulation generation unit is used to generate simulation results for each of the objects based on the physical parameters.
9. An electronic device, characterized in that, The electronic device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method as described in any one of claims 1 to 7.
10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1 to 7.