PhysGaussian-based digital twin ancient building construction method and system

Abstract

The application belongs to the field of digital twinning, and provides a digital twinning ancient building construction method and system based on PhysGaussian. A differential optimization algorithm is used to fit geometric primitives, a PhysGaussian modeling framework is used to construct a scene model of the target ancient building, and a scene model fused with physical parameters is dynamically simulated and predicted to render visual effects and generate a digital twinning model. The application fuses ancient building geometry, material properties and environmental interaction data through a Gaussian kernel, realizes "twinning is real", is suitable for complex structures and fragile materials of ancient buildings, supports preventive protection and repair decisions, and supports real-time updating of urban-level ancient building groups.

Description

Technical Field

[0001] This invention belongs to the field of digital twins, specifically relating to a method and system for constructing digital twin ancient buildings based on PhysGaussian. Background Technology

[0002] Digital twin technology is an advanced modeling method that integrates multi-source data acquisition, simulation calculation, and intelligent analysis. Its core objective is to construct a "twin" in virtual space that is highly consistent with the real-world physical object. This twin not only replicates the geometric shape and material properties of the physical object but can also dynamically respond to its operational status, environmental changes, and even predict future behavior through real-time data. Ancient buildings, as the core carriers of cultural heritage, face enormous challenges in digital preservation and dynamic monitoring. While traditional digital twin technology is widely used in the industrial sector, it struggles to meet the stringent requirements of ancient buildings for high-precision geometric details, accurate material physical properties, and dynamic behavior simulation.

[0003] Ancient architecture is characterized by complex geometric forms and intricate structural relationships, often containing numerous irregular components such as curved tiles, brackets, openwork carvings, and mortise and tenon joints. Its structural features are highly dependent on historical craftsmanship and cultural symbols. Traditional geometric modeling methods often approximate these non-standard geometric forms, making it difficult to accurately reproduce their true forms and resulting in a loss of detail. Furthermore, the materials used in ancient architecture are primarily wood, brick, stone, and glazed tiles, which, under the influence of long-term wind and rain erosion and temperature and humidity changes, exhibit complex physical behaviors such as weathering, aging, and cracking. These nonlinear and time-varying characteristics are difficult to accurately capture and represent within current modeling frameworks. The optical properties of materials (such as the high gloss reflection of glazed tiles and the diffuse reflection roughness of wood) and their mechanical responses (such as the elastic-plastic transition of mortise and tenon joints and the post-earthquake residual deformation of beam-column structures) lack effective analytical methods.

[0004] Furthermore, existing digital twin methods suffer from significant limitations in dynamic simulation capabilities. When faced with complex boundary conditions such as wind loads and seismic impacts, traditional methods can mostly only perform static or simplified linear response analyses, failing to capture the multi-stage evolution of real buildings under disaster conditions. For ancient buildings, which are exposed to the natural environment for extended periods, simulating the coupling process of chronic deterioration and sudden damage remains a significant obstacle for digital twin systems. Moreover, in real urban environments, ancient buildings often interact with environmental elements such as vegetation, water bodies, and topography. For example, humidity affects foundation stability through soil conduction, and tree cover affects ventilation and corrosion rates. These complex multiphysics interactions lack unified modeling and simulation support in current digital twin systems. Summary of the Invention

[0005] This invention provides a method and system for constructing digital twin ancient buildings based on PhysGaussian, in order to solve the problem that traditional digital twin technology cannot meet the stringent requirements of ancient buildings for high-precision geometric details, authenticity of material physical properties, and simulation of dynamic behavior. To address the aforementioned technical problems, the present invention discloses the following technical solutions: One aspect of the present invention provides a method for constructing digital twin ancient buildings based on PhysGaussian, comprising: Acquire point cloud data and building exterior feature images within the target ancient building area; A pre-defined geometric segmentation network is used to identify geometric primitives from point cloud data; Differentiable optimization algorithms are used to fit geometric primitives, and the PhysGaussian modeling framework is used to construct a scene model of the target ancient building. The building's exterior feature image and the data collected by the mechanical sensors on the target ancient building are input into a preset differentiable physical sensing network to obtain the physical parameters of each geometric unit. The physical parameters include at least optical parameters and mechanical parameters. This integrates each geometric primitive in the scene model with its corresponding physical parameters. The PhysGaussian modeling framework is used to perform dynamic simulation and prediction on scene models that incorporate physical parameters, and then renders visual effects to generate digital twin models.

[0006] Optionally, acquiring point cloud data and building exterior feature images within the target ancient building area includes: The drone equipped with high-precision lidar was used to scan the target ancient building to obtain basic point cloud data of the basic structure and dense point cloud data of the complex structure. RGB-D depth images of the target ancient building surface are acquired using a high-definition camera. The images record at least the wood surface texture, carving patterns, and coating cracks of the target ancient building. The microscopic data of the target ancient building is detected by a multispectral imaging system. The microscopic data includes at least the characteristics of wood aging, internal insect infestation, and water stain erosion.

[0007] Optionally, the step of identifying geometric primitives from point cloud data using a preset geometric segmentation network includes: Point cloud data is processed using PointNet++ or DGCNN networks to automatically identify the geometric primitives that constitute the target ancient building.

[0008] Optionally, the fitting of the geometric primitives using a differentiable optimization algorithm includes: A differentiable optimization algorithm is used to adjust the parameters of each geometric primitive with the goal of minimizing the geometric error between the original point cloud data and the constructed model. All geometric primitives are fitted to an initial structural model based on Boolean operations; The initial structural model is smoothed using deformation field interpolation technology to obtain the overall structural model of the target ancient building.

[0009] Optionally, the construction of the scene model of the target ancient building using the PhysGaussian modeling framework includes: The incremental Gaussian evolution algorithm in the PhysGaussian modeling framework is used to discretize each geometric primitive in the overall structural model of the target ancient building into Gaussian particles. Specifically, the key geometric primitives are discretized into fine-grained Gaussian particles, while other geometric primitives are discretized into coarse-grained Gaussian particles. In dynamic scenes, deformation is simulated by adjusting the Gaussian particle parameters of geometric primitives to generate dynamic scene models.

[0010] Optionally, the step of inputting the building exterior feature image and the data collected by the mechanical sensors on the target ancient building into a preset differentiable physical sensing network to obtain the physical parameters of each geometric primitive includes: Mechanical data are collected from each mechanical sensor, which is installed on the target ancient building at each corresponding geometric element; A sub-image of each geometric primitive is obtained based on the building exterior feature image; The mechanical data and sub-image corresponding to each geometric primitive are input into a preset differentiable physical sensing network to obtain the physical parameters of each geometric primitive. The physical parameters include at least Young's modulus, Poisson's ratio, reflectivity, density, and yield stress.

[0011] Optionally, the process of fusing each geometric primitive in the scene model with its corresponding physical parameters includes: Based on the PhysGaussian particle-mesh bidirectional coupling mechanism, the physical parameters corresponding to each geometric primitive are matched with Gaussian particles. The mechanical and optical characteristics of each Gaussian particle are determined based on physical parameters.

[0012] Optionally, the step of using the PhysGaussian modeling framework to perform dynamic simulation prediction on the scene model with fused physical parameters includes: Based on the explicit time integration in the Material Point Method physical simulation method and the local affine transformation assumption in the PhysGaussian modeling framework, each Gaussian particle in the scene model is dynamically evolved. A differentiable simulation framework is adopted, and the mechanical and optical characteristics of each Gaussian particle are dynamically adjusted by backpropagating simulation errors. Dynamic simulation prediction is performed based on the updated Gaussian particles to obtain a simulation model.

[0013] Optionally, the rendering of the visual effects to generate a digital twin model includes: PhysGaussian Splatting technology is used to render the visual effects of the simulation model and highlight key geometric primitives to generate a digital twin model of the target ancient building.

[0014] Optionally, the method further includes: Based on the simulation parameters input by the user, the mechanical and optical characteristics of each Gaussian particle are adjusted, and the simulation parameters include at least wind speed and seismic intensity. Based on the adjusted Gaussian particle generation simulation model; Extract the mechanical properties of each structure in the simulation model, and generate alarms and protection schemes when the mechanical properties exceed the preset strength threshold.

[0015] Optionally, the scene model includes at least a model of the target ancient building, and a model of the vegetation and water system within the target ancient building area.

[0016] Another aspect of the present invention provides a PhysGaussian-based digital twin ancient building construction system, the system performing the PhysGaussian-based digital twin ancient building construction method described in the foregoing aspect.

[0017] This invention discloses a method and system for constructing digital twin ancient buildings based on PhysGaussian. It utilizes a differentiable optimization algorithm to fit geometric primitives, employs the PhysGaussian modeling framework to construct a scene model of the target ancient building, and performs dynamic simulation prediction on the scene model incorporating physical parameters to render visual effects and generate a digital twin model. This invention achieves "twin equals reality" by fusing geometric, material properties, and environmental interaction data of the ancient building using Gaussian kernels. It is applicable to complex structures and fragile materials in ancient buildings, supports preventative protection and restoration decisions, and supports real-time updates for city-level ancient building complexes.

[0018] The summary section is provided to present the chosen concepts in a simplified form, which will be further described in the detailed description below. The summary section is not intended to identify essential or necessary features of this disclosure, nor is it intended to limit the scope of this disclosure. Attached Figure Description

[0019] The above and other objects, features and advantages of this disclosure will become more apparent from the accompanying drawings, in which like reference numerals generally denote like parts.

[0020] Figure 1 A flowchart illustrating a method for constructing a digital twin ancient building based on PhysGaussian provided in this embodiment of the invention; Figure 2 An implementation provided by an embodiment of the present invention Figure 1 A flowchart illustrating step S100; Figure 3 An implementation provided by an embodiment of the present invention Figure 1 A flowchart illustrating step S300; Figure 4 An implementation provided by an embodiment of the present invention Figure 1 A flowchart illustrating step S400; Figure 5 An implementation provided by an embodiment of the present invention Figure 1 A flowchart of step S500; Figure 6 An implementation provided by an embodiment of the present invention Figure 1 A flowchart of step S600; Figure 7 A flowchart illustrating another method for constructing a digital twin ancient building based on PhysGaussian, provided as an embodiment of the present invention. Detailed Implementation

[0021] Embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that the present disclosure will be thorough and complete, and will fully convey the scope of the present disclosure to those skilled in the art.

[0022] The term "comprising" and its variations as used herein signify open inclusion, i.e., "including but not limited to". Unless otherwise stated, the term "or" means "and / or". The term "based on" means "at least partially based on". The terms "one example embodiment" and "one embodiment" mean "at least one example embodiment". The term "another embodiment" means "at least one additional embodiment". The terms "first", "second", etc., may refer to different or the same objects. Other explicit and implicit definitions may also be included below.

[0023] Figure 1A flowchart illustrating a method for constructing a digital twin of ancient buildings based on PhysGaussian, as provided in this embodiment of the invention, is shown below. Figure 1 As shown, the method includes the following steps: Step S100: Obtain point cloud data and building exterior feature images within the target ancient building area.

[0024] The embodiments of the present invention acquire multi-source data of the target ancient building area through various acquisition methods, including at least point cloud data, RGB-D depth images and multispectral microscopic detection data.

[0025] In one embodiment of the present invention, such as Figure 2 As shown, step S100 includes the following sub-steps: Step S101: Use a drone equipped with a high-precision lidar to scan the target ancient building to obtain basic point cloud data of the basic structure and dense point cloud data of the complex structure.

[0026] A drone platform equipped with a high-precision lidar system was used to conduct a comprehensive aerial scan of the target ancient building to obtain three-dimensional spatial distribution information of the building's external structure. The lidar system used has centimeter-level spatial resolution and rapid point cloud imaging capabilities, and can continuously scan the building and surrounding objects from multiple angles and with multiple flight paths to generate broad-coverage basic point cloud data.

[0027] To acquire high-density 3D information of complex architectural structures (such as eaves, brackets, and window lattices), a specific area can be scanned multiple times in low-altitude hovering mode, and dense point cloud data can be constructed by combining structured light or laser triangulation modules. This dense point cloud data is then spatially registered and fused with the base point cloud data, resulting in a comprehensive point cloud data that possesses both global structural coverage and detailed precision.

[0028] During data acquisition, the system uses a combination of inertial navigation unit and GPS for positioning to ensure spatial consistency of the point cloud data. To improve scanning accuracy, ground control points can also be used for post-processing point cloud calibration. This method can completely record the overall form of the target ancient building, providing a precise 3D foundation for subsequent geometric segmentation and modeling.

[0029] Step S102: Use a high-definition camera to acquire RGB-D depth images of the surface of the target ancient building.

[0030] A high-resolution RGB-D camera system was used to closely capture images of the surface of the target ancient building. The RGB-D camera system can simultaneously acquire visible light images (RGB) and depth maps. The RGB images record the true texture and color information of the building surface, while the depth map records the distance information of each pixel from the sensor, forming a color dot matrix surface data.

[0031] In actual data acquisition, operators can handhold or mount RGB-D devices on a ground-based mobile platform or lifting frame to perform omnidirectional scanning of areas such as building facades, eaves, and roof edges in a close-up manner. To ensure image quality, multi-angle and multi-time-period acquisition methods can be used, combined with SLAM (Simultaneous Localization and Mapping) technology for image sequence registration.

[0032] The RGB-D images acquired in this step not only record the true texture and color of the wooden structure surface, but also accurately reflect its surface microstructure features such as carved patterns, paint peeling, and weathering cracks. This information will be used in subsequent physical perception modeling steps to help determine the material properties and surface state of each geometric unit, thereby improving the realism of physical simulation and rendering.

[0033] Step S103: Detect the microscopic data of the target ancient building based on the multispectral imaging system.

[0034] Microscopic data include at least wood aging, internal insect infestation, and water stain erosion.

[0035] A multispectral imaging system was used to perform detailed scanning of the target ancient building to detect its material degradation and the distribution of defects. The imaging system integrates infrared, ultraviolet, near-infrared, and visible light bands, and multi-band response analysis is used to identify different materials and defects.

[0036] Multispectral imaging can reveal microscopic changes in the surface of wood that cannot be directly observed. For example, by using the light absorption characteristics in the ultraviolet band, the degree of lignin degradation in wood can be determined; by using the reflection characteristics in the near-infrared band, structural anomalies caused by internal cavities or insect infestation can be identified; and by using the humidity response imaging in the infrared and short-wave infrared bands, water absorption areas in wood can be identified, and water seepage paths and potential structural hazards can be analyzed.

[0037] To improve acquisition accuracy, an automatic calibration lens and a stable gimbal can be equipped, and real-time positioning can be combined with point cloud data to ensure that the multispectral image, RGB-D image and point cloud are precisely aligned in space to form a complete "geometry-texture-material" three-dimensional information fusion.

[0038] Step S200: Use a preset geometric segmentation network to identify geometric primitives from point cloud data.

[0039] By employing deep learning-based geometric segmentation methods and differentiable optimization techniques, intelligent structural analysis is performed on point cloud data of ancient buildings. Geometric primitives are identified and an accurate initial structural model is constructed, laying a geometric foundation for subsequent physical modeling and simulation.

[0040] In one embodiment of the present invention, this step can be implemented in the following manner: The input point cloud data can be processed using PointNet++ or DGCNN network architectures. Among them, PointNet++ can extract multi-scale local features from the point cloud by constructing hierarchical local regions, which is suitable for capturing local changes in architectural components such as tiles and brackets. DGCNN (Dynamic Graph CNN) can dynamically construct k-nearest neighbor graphs between points and perform convolution operations on the graph structure, making it suitable for representing complex local relationships between points in a point cloud.

[0041] The network input is the multi-source fused point cloud obtained in the preceding steps, and the output is the geometric primitive label to which each point belongs. Post-processing (such as K-means clustering, region growing, etc.) aggregates points of the same category to extract complete geometric primitives.

[0042] Meanwhile, a multi-scale feature fusion mechanism is introduced during the network training phase to jointly model local scales (such as the bending features of the corners of the brackets) with global scales (such as the roof ridge and beams), thereby improving the segmentation accuracy in complex ancient architectural environments.

[0043] Step S300: Fit the geometric primitives using a differentiable optimization algorithm, and construct a scene model of the target ancient building using the PhysGaussian modeling framework.

[0044] In one embodiment of the present invention, such as Figure 3 As shown, fitting geometric primitives using a differentiable optimization algorithm includes the following sub-steps: Step S301: Use a differentiable optimization algorithm to adjust the parameters of each geometric primitive with the goal of minimizing the geometric error between the original point cloud data and the constructed model.

[0045] For the segmented geometric primitive regions, a differentiable optimization algorithm is further used for accurate fitting. A parameterized geometric template (such as a cylinder with parameters including radius, height, and orientation vector) is introduced, and the objective function is to minimize the geometric error (such as Euclidean distance) between the template surface and the original point cloud.

[0046] For example, the distance from each point to the geometric primitive model is defined as a loss term, and the parameters of each primitive are dynamically optimized through the backpropagation algorithm, including the center coordinate position, size parameters (such as length, width, height, radius, etc.), spatial orientation (such as rotation angle), and optional offset and local deformation tensor.

[0047] Differentiable optimization algorithms can be seamlessly integrated with deep learning segmentation networks, allowing the construction of end-to-end structural reconstruction systems and enabling fully automated modeling from point clouds to analytical geometric models.

[0048] Step S302: Fit all geometric primitives to the initial structural model based on Boolean operations.

[0049] After fitting all the basic geometries, Boolean geometric operations are used to stitch them together into a continuous initial structural model.

[0050] Boolean operations include: 1. Union: connecting adjacent components into a complete roof, corridor, or other overall structure; 2. Intersection: extracting overlapping areas to determine the nesting or joining of structures; 3. Difference: removing interfering data or trimming unnecessary component fragments.

[0051] This process ensures that the initial structural model is continuous and consistent in spatial structure, avoiding problems such as gaps, overlaps, or unreasonable interweaving between components.

[0052] Boolean combinations also introduce a geometric topology consistency verification mechanism to ensure that the model does not produce topological vulnerabilities (such as "dangling surfaces" and "self-intersecting edges"), thereby improving the availability of the model for subsequent physical simulation and rendering.

[0053] Step S303: Use deformation field interpolation technology to smooth the initial structural model to obtain the overall structural model of the target ancient building.

[0054] By constructing a continuous deformation field, the initial structural model is calibrated and smoothed in detail, so that it can maintain the accuracy of the geometric structure while having good visual continuity and physical simulability.

[0055] The methods for constructing deformation fields include the following: 1. Local deformation interpolation based on sparse control points: Select several key control points on the initial model, calculate the offset vectors of these points from the original point cloud, and propagate the deformation to the overall model through RBF or MLS interpolation techniques.

[0056] 2. Physics-driven detail fitting: By constructing a local rigid / flexible transformation field, specific components such as tile edges and arched brackets can be elastically adjusted.

[0057] 3. Deep residual correction network: Introducing a lightweight neural network to learn and correct the model residuals, improving the accuracy and naturalness of the fit.

[0058] The final output is a complete structural model with smooth details and minimal error, which serves as the input for subsequent PhysGaussian modeling.

[0059] In one embodiment of the present invention, after completing the point cloud geometric modeling of the target ancient building, in order to support the structural evolution modeling of the building in dynamic environments (such as wind, vibration, aging, earthquakes, etc.), the PhysGaussian modeling framework is further introduced. The geometric primitives are discretized into multi-scale Gaussian particles, and the geometric state is dynamically updated through an incremental Gaussian evolution algorithm, thereby constructing a digital twin scene model that adapts to real-time changes.

[0060] A scene model of the target ancient building is constructed using the PhysGaussian modeling framework, such as... Figure 3 As shown, this can be achieved using the following sub-steps: Step S304: Using the incremental Gaussian evolution algorithm in the PhysGaussian modeling framework, each geometric primitive in the overall structural model of the target ancient building is discretized into Gaussian particles.

[0061] Specifically, the key geometric primitives are discretized into fine-grained Gaussian particles, while other geometric primitives are discretized into coarse-grained Gaussian particles.

[0062] The overall structural model of the target ancient building output from the preceding steps is used as input. The PhysGaussian modeling framework is then introduced, and its incremental Gaussian evolution algorithm is employed to convert all geometric primitives in the structural model into dynamically updatable Gaussian particle representations. The main process is as follows: (I) Particle-based Discretization of Geometric Elements 1. For each geometric element in the structural model (such as beams, columns, roof ridges, tiles, etc.), construct its three-dimensional geometric parameters (center position, orientation, and dimensions).

[0063] 2. Classify the granularity according to the importance and complexity of the primitives. For example, key geometric primitives (such as load-bearing columns, bracket nodes, and roof interfaces) are discretized into fine-grained Gaussian particles (Gaussian Mixture with small covariance) to capture small structural changes; minor geometric primitives (such as walls, platforms, and window sills) are discretized into coarse-grained Gaussian particles to reduce computational redundancy.

[0064] 3. Each Gaussian particle is represented as a three-dimensional position vector and a covariance matrix, with additional weights indicating the particle's physical importance.

[0065] (ii) Incremental Gaussian Evolution Initialization 1. Based on the initial state of the structural model and environmental response, a particle set is established as the initial scene in the static state.

[0066] 2. Spatial constraint diagrams (such as particle connection tension and damping) can be introduced between particles to model their interactions.

[0067] 3. Introduce parametric physical models (such as Young's modulus, density, and material properties) to support physical simulation.

[0068] Step S305: In a dynamic scene, the deformation is simulated by adjusting the Gaussian particle parameters of the geometric primitives to generate a dynamic scene model.

[0069] To support the continuous evolution of ancient building structures under changes in the external environment or human intervention, this step is based on an incremental Gaussian evolution algorithm to update the state of each Gaussian particle in real time in dynamic scenarios, thus constructing a complete dynamic simulation model. Specific technical details are as follows: 1. External input sensing mechanism Multiple external dynamic input sources are introduced, such as wind speed / direction sensors, temperature and humidity changes, earthquake simulation data, structural strain sensors, or time-series video inference. These external inputs are mapped to physical perturbation functions acting on the particle ensemble, driving particle deformation.

[0070] 2. Gaussian particle state update For each time step, the particle position and covariance are updated based on the perturbation function. The update can be based on the particle's elastic response, inertial term, external force influence, etc., to simulate real physical changes such as deformation, shaking, and warping.

[0071] 3. Continuous modeling and backtracking mechanism All particle states are stored in a time-series cache, forming a time-series model. The current scene state can be visualized in real time, and it also supports backtracking analysis or evolution prediction for a specific time period. Intermediate states can also be output for applications such as structural safety early warning and material aging assessment.

[0072] 4. The re-conversion from particle to structural model At any given time step, the 3D model after the current structural deformation can be restored using Gaussian field-to-mesh reconstruction algorithms (such as iso-surface extraction or voxel reconstruction).

[0073] Step S400: Input the building exterior feature image and the data collected by the mechanical sensors on the target ancient building into the preset differentiable physical sensing network to obtain the physical parameters of each geometric unit.

[0074] To improve the physical simulability and simulation accuracy of the digital twin model of the target ancient building in real-world scenarios, this embodiment of the invention pre-constructs a differentiable physical perception network, which is used to automatically infer the physical attribute parameters of each geometric element of the target ancient building from multimodal input data (including building exterior images, RGB-D images, mechanical sensor data, etc.). The physical parameters include at least optical parameters and mechanical parameters.

[0075] In one embodiment of the present invention, such as Figure 4 As shown, step S400 includes the following sub-steps: Step S401: Collect mechanical data from each mechanical sensor. The mechanical sensors are set on the target ancient building at each corresponding geometric element.

[0076] Mechanical sensor modules are installed on each key geometric element (such as columns, beams, and bracket nodes). For example, stress gauges are placed on beams to monitor bending stress; axial stress / compressive stress gauges are placed on columns; seismic response accelerometers are placed at column bases; and temperature and humidity / wet expansion strain gauges can be added to some nodes to assist in modeling.

[0077] The data sampling frequency is no less than 100Hz, and it can continuously record mechanical parameters such as compressive stress, tensile stress, and displacement response. The acquired data includes long-term stress data under static load and short-term dynamic impact response. Each sensor data point records the physical response values ​​in three axes. The acquired data must be synchronized with image data through spatiotemporal alignment; that is, all sensor data are aligned with image data via timestamps, and the system automatically marks the geometric primitive number corresponding to each data point.

[0078] Step S402: Obtain sub-images of each geometric primitive based on the building exterior feature image.

[0079] The building exterior images are correlated with the mechanical data acquired in step S401, and local image fragments corresponding to each geometric primitive are extracted as part of the network input. See the following steps for details: 1. Image segmentation and annotation Semantic segmentation networks (such as DeepLabV3+) are used to identify geometric primitive regions in the original RGB-D image. Sub-image regions in the image are automatically generated and associated with primitives. Geometric contour information is supplemented using fusionable 3D depth maps.

[0080] 2. Feature enhancement processing: Each sub-image is subjected to illumination normalization, edge enhancement, and crack texture high-pass filtering. Auxiliary labels (such as material type, historical repair traces, etc.) are embedded in the input image as auxiliary input items for the network.

[0081] Step S403: Input the mechanical data and sub-image corresponding to each geometric primitive into a preset differentiable physical sensing network to obtain the physical parameters of each geometric primitive.

[0082] Physical parameters include at least Young's modulus, Poisson's ratio, reflectivity, density, and yield stress. Among these, Young's modulus measures the stiffness of the material; Poisson's ratio reflects the lateral deformation rate; optical reflectivity can be used for realistic rendering; and density and yield stress can be used for physical simulation.

[0083] The mechanical data and sub-images from the preceding steps are input into a pre-defined differentiable physical sensing network (DPPN) to estimate the physical properties of each geometric primitive.

[0084] Each primitive outputs a set of physical parameter vectors, where Young's modulus and Poisson's ratio are used for structural finite element modeling, reflectivity is used for realistic rendering, and density / yield stress is used for dynamic response simulation.

[0085] All output physical properties are synchronously mapped back to the physical simulation model to support real-time simulation based on physical properties (such as simulating the response of ancient buildings under earthquakes and wind loads), and can also be used for applications such as structural safety assessment and disaster early warning.

[0086] Step S500: Merge each geometric primitive in the scene model with its corresponding physical parameters.

[0087] This invention introduces a PhysGaussian-based particle-mesh bidirectional coupling mechanism. This mechanism can precisely bind the physical parameters of each geometric primitive obtained through a differentiable physics sensing network to Gaussian particles represented by PhysGaussian, achieving a unified expression of geometric structure and mechanical behavior. Ultimately, it can support the realistic simulation of the response of ancient building structures under wind loads and seismic actions, including plastic deformation, stress distribution, and even local fracture prediction.

[0088] In one embodiment of the present invention, such as Figure 5 As shown, step S500 may include the following sub-steps. Step S501: Based on the PhysGaussian particle-mesh bidirectional coupling mechanism, the physical parameters corresponding to each geometric primitive are matched with Gaussian particles.

[0089] For the aforementioned target ancient building structure model constructed using the Gaussian evolution algorithm (where each geometric primitive has been discretized into Gaussian particles of a certain density), this embodiment of the invention introduces a particle-mesh bidirectional coupling mechanism to achieve the following functions: 1. The input data is a set of Gaussian particles, each containing its spatial position, covariance matrix, etc.; the set of physical parameters of the geometric primitive to which each particle belongs; and the external environmental load field (such as wind pressure distribution function, earthquake ground acceleration time history).

[0090] 2. Binding Mechanism The physical parameters are assigned to the corresponding Gaussian particles, where Young's modulus and Poisson's ratio determine the displacement stiffness of the particles during the elastic response process; density determines the mass matrix terms; yield stress is used for plastic deformation threshold control; and optical reflectivity is used for image rendering consistency.

[0091] Using the weighted projection function in the PhysGaussian framework, particle distribution information is mapped onto a continuous physical simulation grid (Grid-based FEM or MPM grid). The projection uses a Gaussian kernel function (such as RBF or SPH kernel), and the local stiffness, mass, and damping matrices of the grid nodes can be obtained by weighted averaging of particles.

[0092] Particle-to-mesh mapping (P2G) is used to initialize the physics field; and mesh-to-particle feedback (G2P) is used to transmit the physics response results after the load is applied and update the particle states (position, deformation, and destruction markers).

[0093] 3. During the binding process, introduce boundary constraints (such as fixing the building root) and contact stiffness constraints between adjacent particles to avoid particle penetration or non-physical movement during the simulation process.

[0094] Step S502: Determine the mechanical and optical characteristics of each Gaussian particle based on the physical parameters.

[0095] Based on the physical parameters corresponding to each Gaussian particle, determine its mechanical characteristic parameters (e.g., mass density, inertia tensor, connection stiffness) and optical characteristic parameters (e.g., color, transparency, luminescence factor).

[0096] Step S600: Use the PhysGaussian modeling framework to perform dynamic simulation prediction on the scene model that integrates physical parameters, and render the visual effects to generate a digital twin model.

[0097] In one embodiment of the present invention, such as Figure 6 This step, as shown, includes the following sub-steps: Step S601: Based on the explicit time integration in the Material Point Method physical simulation method and the local affine transformation assumption in the PhysGaussian modeling framework, dynamically evolve each Gaussian particle in the scene model.

[0098] The dynamic simulation prediction first uses the PhysGaussian modeling framework as a foundation, integrating the explicit time integration strategy in Material Point Method (MPM) to progressively evolve each Gaussian particle representing the ancient building structure. This step treats each Gaussian particle not only as a geometric representation of its three-dimensional shape, but also as endows it with physical properties (including mass, stress tensor, deformation gradient, etc.), achieving a unified representation of geometry and physics.

[0099] In each frame of time progression, the system processes the particles according to the MPM method as follows: 1. Transfer the physical quantities (mass, momentum) of the particles to the background mesh nodes.

[0100] 2. Perform mechanical calculations on the grid nodes and update the speed.

[0101] 3. Then map the results back to the particles and update their positions and deformation gradients.

[0102] 4. Synchronously update the covariance matrix of particles to support subsequent rendering.

[0103] This step supports modeling complex physical behaviors such as elasticity, plasticity, and fracture. For example, in fracture simulation, the Drucker-Prager plasticity model is used to describe the nonlinear mechanical behavior of masonry materials, and a yield surface and plastic return mapping are defined. When the cumulative plastic strain difference between particles exceeds a set threshold, a particle separation mechanism is triggered to simulate brick breakage.

[0104] To improve computational efficiency, a multi-resolution particle strategy is adopted: high-resolution particles (such as those with resolution r) are placed in beam-column connections and high-stress contact areas. p <0.01m), while using coarse-resolution particles (such as r) in static background areas. p >0.1m).

[0105] The entire simulation is executed in parallel in a GPU-accelerated environment. CUDA is used to achieve parallel computation of key MPM modules (including particle-mesh mapping, stress update, backpropagation, etc.). Combined with an explicit time-progression strategy, it achieves real-time simulation updates of 30 frames per second, supports user interaction, and maintains stable graphics response.

[0106] Step S602: Using a differentiable simulation framework, the mechanical and optical characteristics of each Gaussian particle are dynamically adjusted by backpropagating simulation errors.

[0107] To further improve the simulation accuracy of digital twin ancient buildings, a differentiable physical simulation framework is introduced, incorporating mechanical and optical parameters into the joint optimization process to achieve dynamic matching between simulation behavior and real data.

[0108] In terms of mechanical parameters, the system supports adaptive calibration of elastic modulus (such as Young's modulus of wood E=8.5GPa), density, yield stress, friction angle and cohesion.

[0109] 1. Apply external loads or interactive forces and use motion capture to obtain real response data (such as beam and column deformation, tile slip trajectory).

[0110] 2. Construct a loss function to evaluate the deviation between simulation and reality.

[0111] 3. Use the automatic differentiation tool to backpropagate the loss function and update the physical parameters so that the simulation results gradually approach historical observations.

[0112] 4. In the fracture simulation scenario, the crack paths and fragment distribution generated in the simulation are compared with the actual post-earthquake photos, and the Drucker-Prager parameter set is finely tuned through differentiable optimization.

[0113] In terms of optical characteristics, the parameters of the object's surface are inverted using photometric stereo vision technology: 1. After inputting RGB images from multiple angles, inversely deduce the reflectivity and normal direction.

[0114] 2. The optical inversion results are used to optimize the coefficients of the spherical harmonic basis functions and reconstruct the illumination response of the Gaussian elements.

[0115] 3. By using differentiable rendering technology, the difference between the real image and the rendered image is used as the source of loss, and the optical properties of Gaussian particles, such as roughness and metallicity, are optimized through backpropagation.

[0116] This enables a closed-loop optimization process that combines mechanical simulation and visual reconstruction, achieving high-fidelity digital twin modeling driven by historical data.

[0117] Step S603: Perform dynamic simulation prediction based on the updated Gaussian particles to obtain the simulation model.

[0118] The system constructs a highly realistic simulation model based on an optimized Gaussian particle ensemble, and performs dynamic prediction and visualization.

[0119] In one embodiment of the present invention, the generation of a digital twin model by rendering visual effects can be achieved in the following way: PhysGaussian Splatting technology is used to render the visual effects of the simulation model and highlight key geometric primitives to generate a digital twin model of the target ancient building.

[0120] After the dynamic simulation is completed, PhysGaussian Splatting technology is used to perform high-quality visual rendering of the simulation results, and to build and present the digital twin model of the target ancient building.

[0121] PhysGaussian Splatting technology can accurately present the complex optical effects of ancient buildings, such as high-reflection and subsurface scattering. Combined with spherical harmonic lighting and real-time shadow functions, it vividly restores the historical appearance of ancient buildings. At the same time, through multi-dimensional data mapping, physical information such as stress field and displacement field is encoded into the transparency and color gradient of Gaussian kernel, thereby intuitively displaying the mechanical state of key structures of ancient buildings (such as beams and columns), which facilitates the timely detection of potential risks.

[0122] PhysGaussian Splatting is a real-time differentiable rendering method based on three-dimensional Gaussian primitives. Its core is to directly project each physically simulated Gaussian particle onto the image plane, and to accumulate and synthesize color and depth in the form of Gaussian kernels, thereby generating a continuous, smooth, and mesh-artifact-free visual image.

[0123] During this rendering process, each Gaussian particle carries attributes such as its position, covariance matrix, color, normal, and roughness, which are encoded into a rendering parameter vector of more than six dimensions. The system performs perspective projection on the particles according to the camera's viewpoint and calculates the screen space shape (elliptical outline) and transparency weight of each particle.

[0124] The GPU-accelerated Alpha Blending algorithm is used for forward synthesis to generate realistic simulation images. At the same time, physical information such as the stress state and deformation gradient of particles is visually mapped. For example, high-stress areas are highlighted in red, and fracture edges are faded with transparency, allowing users to easily identify weak points in the structure.

[0125] For geometrically significant components (such as brackets, eaves, and load-bearing columns), the system employs a visual tagging and high-resolution particle highlighting mechanism, setting their Gaussian elements to a fine granularity (e.g., resolution r). p <0.01m), enhancing rendering resolution and interactive experience.

[0126] Through the above rendering and display process, the system ultimately generates a digital twin model with realistic geometric appearance, dynamic physical response, and visual interactive control. This model can not only be used to display the structure and form of ancient buildings, but also support multiple application scenarios such as structural stability analysis, disaster early warning assessment, historical state restoration, and virtual conservation teaching.

[0127] Digital twin models possess the following key characteristics: 1. Real-time response and interactive control Users can directly apply disturbances (such as wind load, seismic intensity adjustment, dragging objects, etc.) through interface controls, which will be converted into force terms in the simulation domain in real time, triggering physical response simulation. The updated particle position and deformation parameters will be updated every frame to ensure that the feedback delay is controlled within 33ms (above 30FPS).

[0128] 2. Faults and Disaster Prediction By incorporating fracture mechanisms, the simulation model can model the slippage of roof tiles during earthquakes, the plastic flow of beams and columns, and the fragmentation of bricks under impact. By comparing the predicted results with historical damage patterns, a risk analysis report is automatically generated, and structural reinforcement schemes are recommended.

[0129] 3. Multimodal data fusion and display By using a Gaussian kernel to store geometric, optical, and mechanical information in a unified manner, and mapping it into visual variables such as color, transparency, and normal maps in the rendering pipeline, a seamless transition from physical simulation to visual expression is achieved.

[0130] 4. Smart Protection Recommendations By combining dynamic prediction results, the system can automatically analyze the safety threshold of ancient buildings under different disaster conditions and propose protection and restoration suggestions (such as reinforcement of specific nodes), providing technical support for cultural relic restoration work.

[0131] By using a digital twin model centered on Gaussian particles, not only was dynamic evolution and predictive simulation achieved, but high-precision rendering and interactive feedback were also integrated to construct a digital twin system for ancient buildings that can be operated in real time, continuously optimized, and visually analyzed.

[0132] The following is a specific example in the simulation process: 1. Determine the elastic modulus of wood The elastic modulus of wood is E=8.5GPa. The sliding of roof tiles and plastic deformation of beams and columns under earthquake action are simulated to optimize the structural reinforcement scheme.

[0133] 2. Optical Parameter Analysis Using photometric stereoscopic vision technology, the surface reflectance of an object can be inferred from multi-light source RGB images. and normal direction The spherical harmonic coefficients of the Gaussian kernel are optimized by combining differentiable rendering.

[0134] 3. Mechanical parameter calibration: Apply a known external force to an elastic body and invert Young's modulus using motion capture data. The strain-stress relationship was calculated using incremental deformation calculations of PhysGaussian and iteratively optimized until the error was minimized.

[0135] The reflectivity calculation is performed again here to further verify and optimize the existing calculation results, and to ensure that it is highly consistent with the surface characteristics of the building in actual simulation. In addition, comparing the reflectivity output by manual calibration and the physical sensing network can effectively reduce the error of the model when dealing with complex surface reflections, thereby improving the overall simulation accuracy.

[0136] 4. Parameter Prediction and Simulation (1) Fracture simulation example: Application of the Drucker-Prager plasticity model Technical principles and processes: a. Model initialization The bricks of the ancient building are modeled as a dense collection of Gaussian particles, with each particle bound to a mass. Initial deformation gradient (Initially an identity matrix) and Drucker-Prager plasticity parameters (friction angles) Cohesion ).

[0137] Plasticity returns a mapping by definition. The evolution of plastic strain of particles under yield conditions is calculated.

[0138] b. Application of impact load Dynamic boundary conditions are set in the simulation domain to apply instantaneous impact force to the surface of the brick. Simulates external shocks (such as impacts or seismic loads).

[0139] The velocity update of the grid nodes is calculated by explicit time integration of MPM. And transmit it to the particle to update its position. and deformation gradient .

[0140] c. Simulation of fracture process When a particle undergoes large deformation, the Drucker-Prager yield condition is triggered, updating the elastic deformation gradient. And mark the accumulation of plastic strain.

[0141] By using a particle separation algorithm, when the difference in plastic strain between adjacent particles exceeds a threshold, a fracture effect is triggered, splitting the particle cluster into independent groups to simulate the effect of brick breaking.

[0142] d. Parameter optimization and historical data matching Based on the actual fracture patterns (such as crack paths and fragment distribution), a loss function is constructed. Simulation Results - Historical Data .

[0143] Using a differentiable simulation framework, the gradient is backpropagated to the plasticity parameter. and The gradient descent method is used for iterative optimization until the error between the simulation results and historical data is minimized.

[0144] (2) Real-time interactive simulation: GPU-accelerated MPM and Gaussian dynamic update Technical implementation steps: a. GPU Parallelization Architecture Design The mesh computation and particle state update of MPM are decomposed into CUDA kernel functions, and the parallel computing capabilities of GPUs are used to accelerate key steps (such as mass / momentum transfer from particles to the mesh and stress calculation).

[0145] By optimizing memory (such as shared memory caching and asynchronous data transfer), the latency between GPU cores is reduced, ensuring that the simulation frame rate remains stable at 30 FPS.

[0146] b. Real-time deformation update of Gaussian elements At each time step Based on the deformation gradient updated by the particles Dynamically calculate the covariance matrix of the Gaussian kernel. .

[0147] Incremental evolution avoids the accumulation of errors in global deformation gradients, supporting numerical stability in long-term simulations.

[0148] c. User interaction and dynamic response Interactive force input: Capturing external forces applied by the user through UI controls (such as mouse dragging, touch gestures). Map it to an external force term in the simulation domain. .

[0149] Real-time feedback: In the GPU pipeline, particle positions are... With deformation parameters It is delivered to the rendering pipeline in real time, and the current frame image is generated through 3D Gaussian Splatting to ensure that the latency of user operation and visual feedback is less than 33ms (30 FPS).

[0150] PhysGaussian plays a key role in simulation in the following aspects: 1. Unified physical-geometric representation: The Gaussian kernel simultaneously carries the geometric shape (covariance matrix) ) and physical properties (stress) Deformation gradient Eliminating "simulated mesh" in traditional pipelines Data transformation for "rendering mesh".

[0151] 2. Efficient dynamic evolution: Real-time simulation of complex physical phenomena (fracture, plastic flow) is achieved through incremental deformation updates and GPU-accelerated MPM.

[0152] 3. Closed-loop parameter optimization: The differentiable framework supports end-to-end optimization from data to parameters, ensuring the high fidelity and predictive ability of the digital twin model.

[0153] In one embodiment of the present invention, such as Figure 7 As shown, this method for constructing digital twin ancient buildings based on PhysGaussian also includes the following steps: Step S700: Adjust the mechanical and optical characteristics of each Gaussian particle according to the simulation parameters input by the user.

[0154] Simulation parameters should include at least wind speed and seismic intensity.

[0155] This invention provides a parameter input module based on a human-computer interaction interface, allowing users to set disaster simulation conditions in real time. A graphical interactive panel is used, providing controls such as sliders, numerical input boxes, and selectors, supporting user input of parameters such as wind speed level, seismic intensity, and flood level. The input parameters are mapped to corresponding physical simulation boundary conditions in real time. For example, wind speed parameters are converted into wind pressure distribution varying with height; seismic intensity is mapped to a ground acceleration time history function and applied to the foundation and bottom boundary of the structure; flood level affects the buoyancy field of particles and the impact surface pressure distribution.

[0156] After loading the physical field, the system dynamically adjusts the mechanical characteristic parameters (including mass density, inertia tensor, and connection stiffness) and optical characteristic parameters (such as color, transparency, and luminescence factor) of each Gaussian particle according to the intensity of the disaster. For example, when the structure enters a disaster state, the color of the corresponding particle gradually changes from green to yellow or even red, so as to provide real-time feedback on the stress state in the visualization rendering; at the same time, the transparency will increase according to the local strain to help identify the critical damage area.

[0157] Step S800: Based on the adjusted Gaussian particle generation simulation model.

[0158] The embodiments of the present invention are based on the incremental evolution simulation engine of PhysGaussian, which performs real-time dynamic simulation and constructs a complete digital simulation model.

[0159] A fusion algorithm combining the Material Point Method (MPM) and Gaussian Splatting is employed to construct particle field evolution equations, and GPU parallel architecture is used to calculate the motion and interaction of particles under external forces. This allows for real-time simulation of structural deformation, crack propagation, and node detachment in particle systems, with a simulation frame rate exceeding 30 FPS, ensuring interactive responsiveness.

[0160] For key structural areas (such as brackets, beams, columns, and eaves), the simulation accuracy is enhanced by increasing particle density (reducing particle radius) and precision parameters to reflect the development of damage at the detail level.

[0161] During the physical simulation, the optical characteristics (such as color) of particles change in real time, reflecting the evolution of their mechanical state. Simultaneously, PhysGaussian Splatting is used for continuous real-time rendering, enabling dynamic disaster demonstration and interactive analysis in a 3D view.

[0162] Step S900: Extract the mechanical properties of each structure in the simulation model, and generate alarms and protection schemes after the mechanical properties exceed the preset strength threshold.

[0163] 1. Extraction of mechanical indicators and determination of safety threshold: Key mechanical parameters for each structural region in the simulation model are continuously extracted, including maximum stress, maximum displacement, local strain, and deformation gradient tensor.

[0164] Based on traditional material mechanics standards and ancient building protection codes, failure criteria for different materials (such as wood, brick, stone, and concrete) are preset. For example, the Tresca yield criterion is used for wooden components; the Mohr-Coulomb fracture model is used for stone components; and the separation conditions for joint connections are set according to the limit displacement and friction slip model.

[0165] When the structural indicators reach 80% of the elastic limit, the system generates a potential risk warning; when the indicators exceed the limit strength, the system immediately triggers a red alarm and automatically enters the protection scheme recommendation process.

[0166] A visual feedback mechanism is adopted. For example, in the 3D view, high-stress areas are marked with a color gradient (green → yellow → red) and overlaid with a V-shaped hazard label to help users intuitively judge the risk areas.

[0167] (2) Automatically generate protection schemes: Data clustering and deformation pattern recognition algorithms are used to classify and determine the typical damage types that occur in the structure, such as bending buckling, shear cracks, tenon and mortise loosening, beam and column displacement, etc.

[0168] The system utilizes a knowledge base and rule-based reasoning engine for ancient building conservation to automatically match the optimal conservation recommendations under the current damage pattern. For example, in cases where earthquakes cause mortise and tenon joints to detach, it recommends installing custom-made iron fittings or carbon fiber cladding; in cases where wind pressure causes tiles to slide, it recommends repairing with traditional mortar or antique-style polymer adhesive; and in cases where floods erode wooden pillars, it recommends installing a waterproof isolation layer and a wood sealing and reinforcement agent.

[0169] The recommended measures are loaded back into the simulation model in a parametric form to automatically simulate the response effect of the reinforced building under the same disaster scenario. The scheme is iteratively optimized by comparing indicators, such as reducing the amount of reinforcement material and adjusting the reinforcement location.

[0170] The final output includes a comprehensive report with illustrations, 3D models, and budget estimates, including: reinforcement locations and material specifications, repair construction steps and precautions, project cost estimates, and a "post-disaster performance preview" that supports interactive browsing on VR devices.

[0171] The following is a specific example of generating early warning and protection schemes: 1. Implementation mechanism for real-time prediction of safety thresholds (1) Dynamic parameter input and simulation triggering a. User interface control design: Provide a visual interactive panel (such as sliders and numerical input boxes) to allow users to dynamically adjust disaster parameters (such as wind speed level, earthquake intensity, and flood level) and trigger simulation calculations in real time.

[0172] b. Parameter-Physics Mapping: This transforms input parameters into boundary conditions for the physics simulation engine. For example, wind speed is mapped to dynamic wind pressure distribution, and seismic intensity is converted into ground acceleration time history curves, which are then loaded onto the building structure using PhysGaussian's particle-field coupling model.

[0173] (2) Real-time physical simulation and index extraction a. GPU-accelerated computing: Based on the Material Point Method (MPM) and PhysGaussian incremental evolution algorithm, and utilizing the GPU parallel computing architecture, real-time dynamic simulation (over 30 FPS) of large-scale particle systems (such as ancient building bricks and tiles, beam and column nodes) is realized.

[0174] b. Monitoring of key safety indicators: Mechanical properties: Real-time extraction of maximum structural stress Displacement Local strain Data such as...

[0175] Stability threshold: Preset failure criteria (such as Tresca yield criterion, Mohr-Coulomb fracture criterion) for ancient building materials (such as wood and stone) and dynamically compare them with simulation data.

[0176] (3) Safety threshold determination and early warning Multi-level early warning mechanism: Yellow alert: When key indicators reach the material's elastic limit. (like This indicates potential risks.

[0177] Red alert: When the index exceeds the material strength threshold (e.g., ... The system determines that the structure is about to fail and triggers an emergency protection recommendation.

[0178] Visual feedback: In the 3D Gaussian rendering interface, through color gradient (green... yellow (Red) High-stress areas are marked in real time and overlaid. Straight label.

[0179] 2. Technical Logic for Automatically Generating Protection Schemes (1) Knowledge base and rule engine a. Engineering Conservation Knowledge Base: Integrates cultural heritage protection standards (such as the "Technical Specifications for Maintenance and Reinforcement of Ancient Wooden Structures"), material mechanics databases (such as wood aging models and stone weathering coefficients), and historical restoration case studies.

[0180] b. Rule-based reasoning engine: Based on expert systems and machine learning models, it establishes mapping rules for "disaster type - structural damage - protective measures". For example: The earthquake caused the mortise and tenon joints to come loose. It is recommended to add iron reinforcement or wrap with carbon fiber cloth.

[0181] Wind load caused tile slippage It is recommended to use traditional mortar bonding or new antique-style waterproof adhesive for fixing.

[0182] (2) Protection scheme generation process a. Damage pattern identification: Through cluster analysis of simulation data, the main damage types (such as bending deformation, shear cracks, and node loosening) are identified.

[0183] b. Solution matching and optimization: Preliminary solution: Match protective measures that match the current damage mode from the knowledge base and sort them by priority (e.g., lowest cost, least impact on the original appearance).

[0184] Parametric optimization: Combining the geometric model and physical parameters of the ancient building, the effectiveness of the solution is verified through differentiable simulation. For example, the stress distribution of the reinforced structure under the same disaster is simulated, and the reinforcement location and material usage are iteratively optimized.

[0185] c. Solution Output: Generate graphic reports and 3D visualization guides, including: Construction drawings: indicate the location of reinforcement points and material specifications (such as the thickness of carbon fiber cloth and the type of anchor bolt).

[0186] Budget assessment: Costs are automatically estimated based on material unit prices and project quantities.

[0187] VR simulation: Allows users to preview the performance of reinforced buildings in disaster scenarios using virtual reality devices.

[0188] In one embodiment of the present invention, the scene model includes at least a model of the target ancient building and a model of the vegetation and water system within the target ancient building area.

[0189] This invention, through the deep integration of multi-source data and physical simulation technology, constructs a highly collaborative digital twin platform to achieve unified modeling, dynamic visualization, and interactive analysis of all urban elements—including ancient buildings, vegetation, and water systems. First, using multimodal data acquisition methods such as laser scanning, UAV photography, and hydrological sensors, millimeter-level geometric details of ancient buildings, three-dimensional root structures of vegetation, and dynamic flow information of water systems are precisely acquired. These heterogeneous data are then uniformly transformed into a Gaussian-based primitive model. Each Gaussian primitive not only encodes spatial distribution and geometric features but also embeds material properties (such as the elastic modulus of timber in ancient buildings and the optical reflectivity of glazed tiles) and physical interaction parameters (such as the adhesion of vegetation roots to the soil and the erosion intensity of floodwaters on building foundations), thereby achieving deep integration of physics and geometry within a unified framework.

[0190] At the visualization level, a multi-resolution rendering strategy is employed: the carved brackets and weathered bricks of ancient buildings are realistically presented using high-precision Gaussian particles (resolution less than 0.01 meters), while the vegetation canopy and water surface are rendered using coarse-grained particles (resolution greater than 0.1 meters) to balance efficiency and realism. Through layered overlay technology, users can freely switch between geometric detail layers, physical field data layers (such as stress heat maps of ancient buildings, stress distribution of vegetation roots, and flood velocity fields), and dynamic behavior layers (such as the swaying of ancient trees in the wind and building vibrations caused by earthquakes), intuitively observing the complex interactions between multiple elements. For example, in flood simulation, water flow particles collide with Gaussian elements of ancient bridge piers in real time, the system dynamically calculates the impact of scouring force on the bridge foundation, and maps the erosion risk level from blue to red through a color gradient, while also marking the additional stress value of the foundation from the roots of nearby ancient trees, forming a comprehensive risk warning.

[0191] Interactive analysis capabilities empower users with dynamic control and decision support. By dragging and dropping controls, users can adjust wind speed, water level, or earthquake intensity in real time. The system, based on GPU-accelerated Material Point Method and PhysGaussian incremental evolution algorithms, updates the physical state of the entire scene in millisecond-level responses. For example, when a user raises the flood level to 2 meters, the system not only simulates the impact of water flow on the ancient bridge piers but also simultaneously calculates the attenuation of root grip on ancient trees along the riverbank due to soil saturation, thereby predicting the risk of the bridge tilting. To address these risks, the platform's built-in intelligent decision engine can automatically generate multiple protection schemes—such as reinforcing the bridge piers, transplanting some ancient trees, or constructing diversion channels—and verify the effectiveness of each scheme through differentiable simulation. Users can immerse themselves in virtual reality to preview the scenarios after different solutions are implemented: the stress peak of the bridge piers reinforced with carbon fiber is reduced by 30% in floods, the stress distribution of the transplanted ancient tree roots tends to be more balanced, and the diversion channel diverts the flood away from the core area of ​​the building. All data is fed back in real time with dynamic charts and three-dimensional heat maps, supplemented by cost estimation and original appearance protection assessment, providing managers with decision-making basis that is both scientific and culturally sensitive.

[0192] This platform not only provides precise tools for the preventative protection of cultural heritage but also reveals the deep connection between the natural and human environments through collaborative simulation of all urban elements. For example, the millennia-long stability of a certain ancient bridge depends not only on its own structure but also on the soil-stabilizing effect of ancient trees along the riverbanks and the history of river dredging. Through spatiotemporal comparison, the impact of vegetation changes and waterway rerouting on ancient buildings over a century can be traced, injecting historical wisdom into future urban planning. Ultimately, this platform transforms the complexity of the physical world into a calculable, interactive, and predictable digital mirror.

[0193] This invention also provides a PhysGaussian-based digital twin ancient building construction system, which is used to execute the PhysGaussian-based digital twin ancient building construction method disclosed in the above embodiments.

[0194] The various embodiments of this disclosure have been described above. These descriptions are exemplary and not exhaustive, and are not limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical applications, or technical improvements to the technology in the market, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. A method for constructing digital twin ancient buildings based on PhysGaussian, characterized in that, include: Acquire point cloud data and building exterior feature images within the target ancient building area; A pre-defined geometric segmentation network is used to identify geometric primitives from point cloud data; Differentiable optimization algorithms are used to fit geometric primitives, and the PhysGaussian modeling framework is used to construct a scene model of the target ancient building. The building's exterior feature image and the data collected by the mechanical sensors on the target ancient building are input into a preset differentiable physical sensing network to obtain the physical parameters of each geometric unit. The physical parameters include at least optical parameters and mechanical parameters. This involves fusing each geometric primitive in the scene model with its corresponding physical parameters; specifically, it includes matching the physical parameters of each geometric primitive with Gaussian particles based on a particle-mesh bidirectional coupling mechanism using PhysGaussian; and determining the mechanical and optical characteristics of each Gaussian particle based on the physical parameters. The Gaussian particles are obtained based on geometric primitive discretization. The PhysGaussian modeling framework is used to perform dynamic simulation and prediction on scene models that incorporate physical parameters, and to render visual effects to generate digital twin models. The method of using the PhysGaussian modeling framework to perform dynamic simulation prediction on a scene model that incorporates physical parameters includes: Based on the explicit time integration in the Material Point Method physical simulation method and the local affine transformation assumption in the PhysGaussian modeling framework, each Gaussian particle in the scene model is dynamically evolved. A differentiable simulation framework is adopted, and the mechanical and optical characteristics of each Gaussian particle are dynamically adjusted by backpropagating simulation errors. A simulation model is obtained by performing dynamic simulation prediction based on the updated Gaussian particles. The method, based on the explicit time integration in the Material Point Method and the local affine transformation assumption in the PhysGaussian modeling framework, dynamically evolves each Gaussian particle in the scene model. This includes: using the PhysGaussian modeling framework as a foundation and integrating the explicit time integration strategy from the Material Point Method, progressively evolving each Gaussian particle representing the ancient building structure. Each Gaussian particle possesses geometric features of a three-dimensional shape and is associated with physical properties, including mass, stress tensor, and deformation gradient. In each frame of time progression, the particles are processed according to the Material Point Method as follows: 1) the physical quantities on the particles are transferred to the background mesh nodes; 2) mechanical calculations are performed on the mesh nodes to update the velocity; 3) the results are mapped back to the particles to update their position and deformation gradient; 4) the covariance matrix of the particles is updated synchronously to support subsequent rendering.

2. The method according to claim 1, characterized in that, The acquisition of point cloud data and building exterior feature images within the target ancient building area includes: The drone equipped with high-precision lidar was used to scan the target ancient building to obtain basic point cloud data of the basic structure and dense point cloud data of the complex structure. RGB-D depth images of the target ancient building surface are acquired using a high-definition camera. The images record at least the wood surface texture, carving patterns, and coating cracks of the target ancient building. The microscopic data of the target ancient building is detected by a multispectral imaging system. The microscopic data includes at least the characteristics of wood aging, internal insect infestation, and water stain erosion.

3. The method according to claim 1, characterized in that, The method of identifying geometric primitives from point cloud data using a preset geometric segmentation network includes: Point cloud data is processed using PointNet++ or DGCNN networks to automatically identify the geometric primitives that constitute the target ancient building.

4. The method according to claim 1, characterized in that, The fitting of geometric primitives using a differentiable optimization algorithm includes: A differentiable optimization algorithm is used to adjust the parameters of each geometric primitive with the goal of minimizing the geometric error between the original point cloud data and the constructed model. All geometric primitives are fitted to an initial structural model based on Boolean operations; The initial structural model is smoothed using deformation field interpolation technology to obtain the overall structural model of the target ancient building.

5. The method according to claim 4, characterized in that, The process of constructing a scene model of the target ancient building using the PhysGaussian modeling framework includes: The incremental Gaussian evolution algorithm in the PhysGaussian modeling framework is used to discretize each geometric primitive in the overall structural model of the target ancient building into Gaussian particles. Specifically, the key geometric primitives are discretized into fine-grained Gaussian particles, while other geometric primitives are discretized into coarse-grained Gaussian particles. In dynamic scenes, deformation is simulated by adjusting the Gaussian particle parameters of geometric primitives to generate dynamic scene models.

6. The method according to claim 5, characterized in that, The process involves inputting images of the building's exterior features and data collected by mechanical sensors on the target ancient building into a pre-defined differentiable physical sensing network to obtain the physical parameters of each geometric unit, including: Mechanical data are collected from each mechanical sensor, which is installed on the target ancient building at each corresponding geometric element; A sub-image of each geometric primitive is obtained based on the building exterior feature image; The mechanical data and sub-image corresponding to each geometric primitive are input into a preset differentiable physical sensing network to obtain the physical parameters of each geometric primitive. The physical parameters include at least Young's modulus, Poisson's ratio, reflectivity, density, and yield stress.

7. The method according to claim 6, characterized in that, The rendering of visual effects generates a digital twin model, including: PhysGaussian Splatting technology is used to render the visual effects of the simulation model and highlight key geometric primitives to generate a digital twin model of the target ancient building.

8. The method according to claim 7, characterized in that, The method further includes: Based on the simulation parameters input by the user, the mechanical and optical characteristics of each Gaussian particle are adjusted, and the simulation parameters include at least wind speed and seismic intensity. Based on the adjusted Gaussian particle generation simulation model; Extract the mechanical properties of each structure in the simulation model, and generate alarms and protection schemes when the mechanical properties exceed the preset strength threshold.

9. The method according to claim 1, characterized in that, The scene model includes at least a model of the target ancient building, as well as a model of the vegetation and water system within the target ancient building area.

10. A digital twin ancient building construction system based on PhysGaussian, characterized in that, The system executes the PhysGaussian-based digital twin ancient building construction method according to any one of claims 1-9.

Images