Spatial intelligent world modeling method and system based on global NURBS parameter domain

Abstract

The invention belongs to the technical field of space intelligent modeling, and relates to a space intelligent world modeling method based on a global NURBS parameter domain. Consistent mapping and management are carried out on the multi-Patch geometric objects in a global NURBS parameter domain; generating and optimizing a control point matrix and a weight matrix for geometric expression based on a global NURBS parameter domain; realizing automatic smooth transition of the multi-Patch geometry in the splicing area based on a continuity keeping mechanism of curvature and normal constraint; executing NURBS curved surface subdivision operation according to the curvature change rate and the geometric error threshold value; and performing global solution on the control point matrixes of all the Patches, and outputting a space intelligent world model which is continuous and differentiable in a global range and has consistent parameters. According to the method, the continuity and controllability of geometric modeling are remarkably improved. The invention further provides a spatial intelligent world modeling system based on the global NURBS parameter domain.

Description

Technical Field

[0001] This invention belongs to the field of spatial intelligence and 3D spatial intelligent modeling technology, and is applied to industries such as digital twins, virtual reality, smart cities and generative spatial understanding. Specifically, it relates to a spatial intelligent world modeling method based on the global NURBS parameter domain. Background Technology

[0002] Currently, the 3D world modeling of spatial intelligent systems mainly relies on discrete geometric representations such as meshes, voxels, or implicit radiation fields (NeRF). While these methods have made some progress in local scene reconstruction and visual rendering, they still have significant limitations in representing large-scale, city-level, and continuous spaces.

[0003] In recent years, generative artificial intelligence models (such as NeRF and the Marble model from the LiFei-Fei team) have been able to learn the implicit geometric structure of the world through neural networks, enabling the generation of 3D spatial representations from multiple views or text inputs. However, the underlying geometric primitives of these models (such as voxels, point clouds, or Gaussian spheres) are essentially still discrete or probabilistically distributed representations, and they suffer from the following major problems in terms of continuous differentiability, geometric controllability, and global consistency: 1. Insufficient geometric continuity: Current representations based on volume density or Gough units can only achieve surface smoothness in a probabilistic sense, lacking strict geometric continuity constraints and making it difficult to achieve controllable continuity at the G0 / G1 / G2 level, thus limiting the application of the model in precise scenarios such as urban modeling.

[0004] 2. Lack of global parameter consistency: Existing neural fields and Gaussian representations mostly rely on local coordinates or independent surface patches, which cannot establish a globally unified parameter domain mapping, resulting in scale drift and deformation inconsistencies when stitching large scenes or across regions.

[0005] 3. The learnable geometric parameters are unclear: In Mesh or Gaussian Splatting representations, geometric shapes are often implicit in the network weights, lacking explicit learnable geometric control parameters, which is not conducive to AI models performing interpretable and adjustable spatial learning.

[0006] 4. Insufficient dynamic adaptive capability: Existing models cannot achieve adaptive subdivision and reconstruction under different resolutions, viewpoints, or task conditions. Rendering and geometry updates still require manual resampling, and there is a lack of intelligent subdivision mechanisms based on error tolerance.

[0007] 5. Difficulty in smoothly stitching together multi-patch scenes: For city-level, complex building complexes or natural terrain scenes, existing methods struggle to maintain boundary continuity between multiple local models, often resulting in inconsistencies in normals, surface tears, or occlusion gaps at the splicing points.

[0008] Therefore, while existing spatial intelligent world modeling techniques have made significant progress in visual perception and generative rendering, they still face obvious bottlenecks in continuous geometric representation, global parameter consistency, learnable controllability, and adaptive generation capabilities. There is an urgent need for a geometric modeling method based on a globally continuous parameter space, capable of using a unified NURBS parameter domain as the underlying logic to achieve continuous representation of multi-patch worlds, curvature preservation, adaptive subdivision, and AI-driven generative optimization, thereby constructing a truly spatial intelligent world model. Summary of the Invention

[0009] This invention provides a spatial intelligent world modeling method and system based on the global NURBS parameter domain, which solves the problems of discrete geometric representation, insufficient continuity, and inconsistent parameter domains in existing spatial intelligent modeling technologies.

[0010] This invention is achieved using the following technical solution: A spatial intelligent world modeling method based on the global NURBS parameter domain includes: Data input and data preprocessing; Construct a global NURBS parameter domain for modeling the intelligent world in space, and perform consistent mapping and management of multi-patch geometric objects under the global NURBS parameter domain; Generate and optimize the control point matrix for geometric representation based on the global NURBS parameter domain; Based on the continuity preservation mechanism of curvature and normal constraints, automatic smooth transition of multi-patch geometry in the stitching region is achieved. Based on an adaptive subdivision and error control mechanism, NURBS surface subdivision operation is performed according to the rate of curvature change and geometric error threshold. Based on the global consistency optimization solution mechanism, the global solution is performed on the control point matrix of all patches, and the output is a spatial intelligent world model that is continuously differentiable and has consistent parameters in the global scope.

[0011] Preferably, the method further includes: Based on a multi-scale hierarchical modeling and fusion mechanism, curvature and topological continuity are maintained between different levels through nested parameter domain mapping relationships.

[0012] Preferably, the global NURBS parameter domain construction process includes: The point cloud data is normalized in the world coordinate system. A two-dimensional parameter domain is established in the normalized space, and the point cloud coordinates are converted into parameter domain coordinates through an affine mapping function. Automatically generate non-uniform node vectors based on point cloud density and curvature characteristics; Based on the point cloud features, the parameter domain is divided into multiple patch regions, and a global parameter index table is established.

[0013] Preferably, generating and optimizing the control point matrix for geometric representation based on the global NURBS parameter domain includes: In the global NURBS parameter domain, a control point parameter grid is generated according to a preset resolution or adaptive rules. Based on the mapping relationship between the parameter domain and spatial coordinates, the control points in the parameter domain are mapped to the corresponding spatial control point coordinates, thereby forming an initial control point matrix. Perform global collaborative optimization on the initial control point matrix.

[0014] Preferably, the control point matrix, as an optimization variable, is subject to the combined effects of the following constraints: Parameter domain continuity constraints are used to ensure that the boundary control point parameter positions of adjacent patches remain consistent in the global parameter domain. Geometric smoothing constraints are used to suppress geometric abrupt changes caused by excessively dense or sparse distribution of local control points; Error tolerance constraints are used to limit the adjustment range of control points so that the geometric accuracy meets the preset requirements.

[0015] Preferably, the mechanism for maintaining the continuity of curvature and normal constraints includes: At the junction of adjacent patches, the curvature and normal information are calculated based on the generated control point matrix, and the curvature difference and normal deviation at the junction of adjacent patches are used as the continuity evaluation index; when the curvature difference or normal deviation exceeds the preset threshold, the continuity optimization process is automatically triggered. In the continuous optimization process, the coordinates of the control points in the control point matrix are used as optimization variables. The positions of the control points are iteratively adjusted by simultaneously constraining the curvature and normal changes at the boundaries of adjacent patches.

[0016] Preferably, performing NURBS surface subdivision operation based on the rate of curvature change and a geometric error threshold includes: When the local curvature exceeds the threshold or the accumulated geometric error exceeds the threshold, a new control point layer is automatically inserted to refine the NURBS surface hierarchically.

[0017] Preferably, the multi-scale hierarchical modeling and fusion mechanism includes: A hierarchical NURBS representation system is adopted: the upper-level patch is responsible for representing the macro structure of the city; the middle-level patch is responsible for the terrain and building outlines; the lower-level patch describes local geometric details; and the curvature and topological continuity between the layers are maintained through nested parameter domain mapping relationships.

[0018] A spatial intelligent world modeling system based on global NURBS parameter domain includes: Parameter domain management module: responsible for establishing and maintaining the mapping relationship of global NURBS parameter domains; Control point generation and optimization module: used to automatically generate initial control point matrix and weight matrix based on input point cloud, curvature and semantic features; Continuity constraint module: This module is responsible for detecting and constraining the differences in normal and curvature between adjacent patches to achieve automatic continuity at geometric boundaries. Subdivision and Error Control Module: Used to perform an adaptive subdivision process based on error tolerance; when the local curvature or error exceeds the set threshold, the system automatically inserts new control points and regenerates the node vector; Global Optimization and Model Output Module: Used to uniformly solve the control point matrix of all patches, coordinate parameters, and output the model.

[0019] Preferably, the system further includes: Multi-scale synchronization module: used to maintain parameter consistency and geometric smoothness between different levels, so that the models at each scale remain nested and continuous in terms of topology and curvature.

[0020] 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.

[0021] Compared with existing spatial intelligent modeling technologies, this invention has significant technical advantages and comprehensive effects in terms of geometric continuity, parameter consistency, generation efficiency, and system scalability, including: 1. Regarding geometric continuity, this invention defines a multi-patch geometric structure within a unified global NURBS parameter domain and employs a continuity optimization mechanism based on curvature and normal constraints to achieve automatic boundary stitching, ensuring smooth connections between adjacent surfaces in both the normal and curvature directions. This mechanism enables surface continuity at G1 to G2 levels without manual intervention, effectively eliminating polylines, cracks, and surface irregularities at the stitching points of traditional mesh models, thereby achieving globally consistent geometric representation in large scenes.

[0022] 2. Regarding modeling and optimization efficiency, this invention employs a parameterized control point matrix as the core geometric representation, reducing reliance on high-density point clouds or voxel data and making the modeling process more lightweight. Since control points and weight parameters can be optimized using analytical gradients, this invention significantly outperforms sampling-based implicit field methods in terms of iterative convergence speed and computational stability, improving overall modeling efficiency by approximately 30%. It can complete continuous geometric generation and global optimization of city-level or complex scenes in a shorter time.

[0023] 3. In terms of geometric accuracy and structural representation capability, the global parameter domain structure of this invention can precisely control the surface morphology and directly constrain the curvature distribution through the control point matrix, making the generated results superior to traditional methods in terms of surface smoothness, structural alignment, and morphological rationality. Compared with mesh-based discrete models, the world geometry generated by this invention is more consistent in terms of continuity and differentiability, making it suitable for spatial intelligent scenarios requiring high-precision modeling and subsequent physical simulation.

[0024] 4. In terms of system structure and engineering applications, the parametric representation system of this invention is highly compatible with existing BIM modeling systems. The generated NURBS control point matrix can be directly exported as a standard industrial format file, facilitating integration with architectural design, structural analysis, and digital twin platforms. This feature makes the generation and optimization process of spatial models more automated, eliminating the need for complex post-processing operations and significantly reducing manual modeling costs and time.

[0025] The method of this invention not only significantly improves the continuity and controllability of geometric modeling, but also has significant advantages in efficiency, compatibility, and learnability. Compared with existing technologies, this invention demonstrates significant technological progress and engineering application value in multiple aspects, including structural simplification, computational acceleration, automated modeling, and resource conservation. Attached Figure Description

[0026] 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.

[0027] Figure 1 This is a schematic diagram of a spatial intelligent world modeling method based on a global NURBS parameter domain according to an embodiment of the present invention; Figure 2 This is a schematic diagram of a city-level 3D scene modeling process based on a global NURBS parameter domain according to an embodiment of the present invention; Figure 3 This is a schematic diagram illustrating the global parameter domain definition and node vector construction process provided in one embodiment of the present invention; Figure 4A schematic diagram of a NURBS surface generation and geometry export module provided in one embodiment of the present invention; Figure 5 This is a block diagram of a spatial intelligent world modeling system based on a global NURBS parameter domain, according to an embodiment of the present invention. Detailed Implementation

[0028] 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.

[0029] The term "comprising" and its variations, as used in this invention, signify an open-ended inclusion, i.e., "including but not limited to". Unless specifically stated otherwise, 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.

[0030] This invention uses a global NURBS parametric domain as the underlying logic for geometric expression. It organizes multi-patch geometric structures within a unified global parameter space, achieving continuous world-class geometric generation and optimization. The core idea of ​​the entire method is to form an end-to-end spatial intelligent geometric modeling framework through parameter domain mapping, control point matrix constraints, curvature continuity preservation, adaptive subdivision optimization, and globally consistent solution. This enables spatial intelligent systems to perform geometric optimization and generation within a unified parameter space, thereby establishing a world modeling framework that combines geometric continuity, AI learnability, and dynamic adaptability.

[0031] To facilitate understanding, the technical terms involved in the embodiments of this invention will be briefly introduced first: NURBS is an abbreviation for Non-Uniform Rational B-Spline. NURBS is a method that uses mathematical formulas (through control points and weights) to accurately describe smooth curves and surfaces.

[0032] In geometric modeling, a patch is usually translated as a surface patch or facet. A patch is the basic building block of a complex surface. It is a smooth, continuous surface segment defined by a mathematical formula.

[0033] Control points are the core parameters defining the geometry of NURBS. They constitute a set of points in space, which, through a linear combination of weighted B-spline basis functions, generate an accurate mathematical surface. Each control point contains three-dimensional coordinates and a weight value: the coordinates determine its spatial position, and the weight, as a rational factor, adjusts the strength of the point's "attraction" to the surface. This structure allows NURBS to accurately express freeform surfaces and conic sections (such as circular arcs) using a unified mathematical model. Control points are typically not located on the surface but form external control polygons or control meshes. Moving control points can smoothly and predictably drive local or global deformation of the surface.

[0034] The control point matrix is ​​the core data structure in the mathematical definition of NURBS surfaces. It systematically organizes all control points of the surface in the u and v parametric directions as a two-dimensional array. The topology of this matrix directly determines the control grid shape of the NURBS surface: each row of the matrix corresponds to a sequence of control points along an isoparametric line in the u direction, and each column corresponds to a sequence of control points in the v direction. This matrix arrangement not only provides a clear parametric coordinate framework for the surface, allowing any surface point to be calculated using the basis functions corresponding to the (u,v) parameter index and the subset of control points, but also forms the geometric foundation for advanced surface operations (such as trimming, blending, and continuity matching). In industrial CAD systems, manipulating the control point matrix—that is, adjusting the position and weight of its elements (control points) through interactive tools—has become an efficient operational paradigm for editing surface shapes. It ensures that even with local modifications, the entire surface can strictly maintain its mathematical smoothness and geometric continuity.

[0035] Example 1: The following section introduces a spatial intelligent world modeling method based on the global NURBS parameter domain, as described in this invention.

[0036] A spatial intelligent world modeling method based on the global NURBS parameter domain, such as Figure 1 As shown, it includes: S100, Data Input and Data Preprocessing.

[0037] In one embodiment, the input data includes LiDAR point clouds, photogrammetric point clouds, or other discrete spatial sampling data. Data preprocessing includes automatically performing noise removal, normal estimation, and downsampling operations on the point cloud data. Neighborhood query efficiency is improved through spatial indexing, resulting in a standardized point cloud file containing coordinate, normal, and semantic information.

[0038] S200. Construct a global NURBS parameter domain for modeling the intelligent spatial world, and perform consistent mapping and management of multi-patch geometric objects under this global NURBS parameter domain.

[0039] S200 is a global parameter domain establishment and mapping mechanism. During the initialization phase, the system defines a unified two-dimensional parameter domain as the basic coordinate system for the global geometric space. The original point cloud coordinates are transformed to the parameter domain space through affine mapping, and multiple logical patches are created based on the point cloud distribution. All patches are generated under the same parameterization rules, achieving a globally geometrically continuous and parameter-consistent expression structure.

[0040] Unlike existing technologies that perform conventional parametric processing on a single NURBS surface, this invention constructs a global NURBS parameter domain for modeling a spatial intelligent world, and performs consistent mapping and management of multi-patch geometric objects under this global NURBS parameter domain.

[0041] The global NURBS parameter domain is a continuous, unified, and scalable two-dimensional parameter space. Its parameter coordinate system remains globally unique during the modeling process, serving to carry the parameter mapping relationships of multiple geometric patches in the space. Unlike traditional NURBS surface methods where "each surface has its own custom local (u,v) parameter domain," this invention pre-establishes a global parameter domain, ensuring that the parameter coordinates of different geometric patches are no longer independent but are subject to unified parameter scale, orientation, and boundary constraints.

[0042] After the global parameter domain is established, this invention maps each geometric patch in space to the corresponding parameter sub-region in the global parameter domain. During the mapping process, parameter boundary alignment and scale normalization mechanisms ensure the continuity of boundaries and consistency of orientation of adjacent patches in the parameter domain, thereby avoiding parameter drift, scale inconsistency, and cross-region stitching distortion problems caused by the independent definition of local parameter domains in the prior art.

[0043] Furthermore, the global NURBS parameter domain of this invention not only serves as the parameter basis for geometry generation, but also as a unified parameter coordinate system for subsequent control point matrix generation, curvature continuity constraints, adaptive subdivision, and global consistency optimization. By organizing geometric information within a unified parameter domain, the spatial modeling process possesses excellent continuity, controllability, and scalability, providing underlying support for constructing city-level or large-scale continuous spatial models.

[0044] Specifically, during the initialization phase, a unified two-dimensional parameter domain is first defined in the global coordinate system as the global NURBS parameter domain. This global NURBS parameter domain serves as the fundamental parameter coordinate system for the entire spatial intelligent world model. The global NURBS parameter domain remains globally unique throughout the modeling process, with its parameter scale, orientation, and boundary rules remaining fixed throughout the entire modeling workflow. This provides a unified parameter basis for subsequent geometry generation, optimization, and subdivision.

[0045] For city-level or complex world scenarios, the system divides the overall space into multiple logical patch regions, each patch corresponding to a parameter sub-interval in the global parameter domain. Unlike existing technologies where each patch independently establishes its local parameter domain, in this invention, all patches share the same global parameter domain coordinate system, differing only in the range of parameter values. The system uses preset parameter mapping rules to map the spatial coordinates of each patch to the corresponding sub-interval in the global parameter domain, ensuring natural connection between adjacent patches within the parameter domain.

[0046] During the mapping process, the boundary parameters of the patches are uniformly aligned to ensure that the boundary parameter ranges of adjacent patches remain continuous and consistent in the parameter domain. This allows for parameter continuity during geometric stitching without the need for additional manual stitching. This mapping mechanism effectively avoids the scale drift, orientation inconsistency, and cross-regional stitching distortion problems caused by the independent definition of local parameter domains in existing technologies.

[0047] The global NURBS parameter domain established by S200 provides a unified parameter basis for subsequent control point matrix generation, continuity preservation, adaptive subdivision, and global consistency optimization. This enables multi-patch geometries to be continuously expressed and collaboratively optimized within the same parameter space, thereby constructing a spatial intelligent world model with global consistency.

[0048] S300: Generates and optimizes the control point matrix for geometric representation based on the global NURBS parameter domain.

[0049] In this invention, S300 is a control point matrix generation and global collaborative optimization mechanism, including: automatically generating a control point matrix and a weight matrix based on the global NURBS parameter domain partitioning results to describe the geometric features of the surface. The control point positions and weights are optimized through algorithms or learning models to make the surface shape smooth, continuous, and differentiable. This mechanism realizes the parameterized reconstruction from discrete samples to a continuous geometric model. Specifically: After establishing the global NURBS parameter domain and mapping the patches, a control point matrix for geometric representation is generated and optimized based on the global parameter domain. The control point matrix, as an explicit parameterized representation of NURBS geometry, is a computable and optimizable core geometric variable in the spatial intelligent modeling process of this invention.

[0050] In the spatial intelligent world modeling method based on the global NURBS parameter domain of this invention, the geometric generation and optimization of three-dimensional space are both implemented based on the NURBS surface model, and its mathematical expression is as follows: (1) in, Represents parameter coordinates The corresponding three-dimensional space point; These are the coordinates of the control points in the control point matrix; The weights of the control points; and They are respectively towards and The B-spline basis functions are given by p and q, where p and q are the spline orders, and n and m are the number of control point matrices in the two directions, respectively.

[0051] Formula (1) reflects the basic logic of geometric generation in this invention: any point in three-dimensional space All of them are composed of all control points in the control point matrix. The weighted combination is generated. The influence weight of each control point is determined by the basis function. With weight The values ​​are determined jointly. The numerator represents the combined contribution of each control point to the current parameter point, while the denominator is a normalization term used to ensure the smoothness and numerical stability of the surface shape. The resulting surface is geometrically continuous and differentiable, locally controllable and globally consistent, achieving strict geometric continuity and smooth transitions within the global parameter domain.

[0052] In the system of this invention, formula (1) is not only used for surface generation, but also serves as the core geometric representation mechanism for intelligent spatial modeling.

[0053] Specifically, geometric continuity is achieved through the smoothing properties of basis functions and control point position constraints, ensuring G0, G1, and G2 continuity between patches; global parameter consistency is ensured by defining the parameter coordinates (u,v) within a globally shared NURBS parameter domain, enabling all geometric segments to be generated in a unified coordinate system; and learnability and optimizability are achieved through the control point matrix. and weight It can be used as an optimization variable, updated directly through gradient descent or AI models, to achieve automatic generation and adaptive optimization of geometry; hierarchical adaptive capability: when the curvature of a local region changes significantly or the error exceeds a threshold, new control points can be automatically interpolated and generated within the parameter domain, achieving hierarchical subdivision. This formula describes the generation method of geometric points, forming the mathematical foundation of the globally continuous optimizable geometric system in this invention. Through the control point matrix optimization framework based on this, this invention can achieve consistent modeling and accurate optimization from local geometric fragments to the global scene in a unified parameter space, providing a computable and differentiable geometric foundation for the high-precision generation and adaptive updating of spatial intelligent world models.

[0054] It is important to note that, unlike existing technologies that directly construct control point matrices based on a single surface or local patch, the control point matrix generation process of this invention is performed under the constraints of the global NURBS parameter domain. Specifically, this invention constructs a regular or adaptive control point distribution structure in the parameter domain based on the mapping relationship between the positions of each parameter in the global NURBS parameter domain and spatial geometry, and maps this control point distribution to the corresponding spatial control point matrix, thereby ensuring global consistency of the control point matrix at the parameter domain level. Specifically: First, the system generates a control point parameter mesh in the global NURBS parameter domain according to a preset resolution or adaptive rules. Then, based on the mapping relationship from the parameter domain to spatial coordinates, it maps the control points in the parameter domain to their corresponding spatial control point coordinates, thus forming an initial control point matrix. This control point matrix does not only serve a single patch but is generated under the constraints of the global parameter domain, ensuring that the control points in different patches have a unified distribution rule and index relationship at the parameter level.

[0055] Secondly, after the control point matrix is ​​generated, the system performs global collaborative optimization on the control point matrix. Unlike existing technologies that independently adjust control points within a single surface or local patch, this invention treats the control point matrices of multiple patches as a unified optimization object, performing joint optimization under global parameter domain constraints. Specifically, the control point matrix, as an optimization variable, is subject to the combined effects of the following constraints: Parameter domain continuity constraints are used to ensure that the boundary control point parameter positions of adjacent patches remain consistent in the global parameter domain. Geometric smoothing constraints are used to suppress geometric abrupt changes caused by excessively dense or sparse distribution of local control points; Error tolerance constraints are used to limit the adjustment range of control points so that the geometric accuracy meets the preset requirements.

[0056] By applying the aforementioned constraints, the system maintains its local geometric representation capabilities while achieving a consistent distribution and collaborative optimization of the control point matrix at the global scale. The optimized control point matrix is ​​not only used for current geometry generation but also serves as the foundational input for subsequent curvature continuity preservation, adaptive subdivision, and globally consistent solution, thereby enabling the geometric model to possess iterative update and adaptive optimization capabilities.

[0057] Therefore, although the NURBS geometric representation used in this invention is a well-known technology in the field, by introducing global parameter domain constraints and a collaborative optimization mechanism for multi-patch control point matrices, the control point matrix is ​​transformed from the traditional static modeling result into a globally optimizable set of geometric parameters, which is significantly different from the conventional control point generation method in the prior art.

[0058] S400, based on the continuity preservation mechanism of curvature and normal constraints, realizes automatic smooth transition of multi-patch geometry in the stitching region.

[0059] To achieve automatic and smooth transition of multi-patch geometry in the stitching region, this invention establishes a continuity maintenance mechanism based on curvature and normal constraints in S400. This mechanism is used to maintain stable geometric continuity between adjacent patches in the global parameter domain without relying on manual repair. The mechanism includes: the system calculating the curvature and normal differences of the boundaries of adjacent patches to detect geometric discontinuities; and automatically achieving continuity at the normal (G1) and curvature (G2) levels through control point position adjustment and boundary constraint optimization. This mechanism ensures smooth stitching between different patches, with no visible seams or abrupt morphological changes.

[0060] Specifically, at the junction of adjacent patches, the system calculates the corresponding curvature and normal information based on the generated control point matrix, and uses the curvature difference and normal deviation at the junction of adjacent patches as continuity evaluation indicators. When the curvature difference or normal deviation exceeds a preset threshold, the system automatically triggers a continuity optimization process.

[0061] During the continuity optimization process, the system uses the coordinates of control points in the control point matrix as optimization variables. By simultaneously constraining the curvature and normal changes at the boundaries of adjacent patches, the positions of the control points are iteratively adjusted. In this way, while ensuring the reasonable local geometry, the normal direction at the patch boundary can be made continuous, and curvature abrupt changes can be further reduced, thereby achieving geometric continuity at the G1 to G2 level.

[0062] This continuity-preserving mechanism for curvature and normal constraints does not only apply to a single patch boundary, but also uniformly processes the boundaries of multiple patches under global parameter domain constraints, ensuring that the entire spatial model maintains a continuous and smooth geometric structure globally. This mechanism effectively avoids common problems in existing multi-patch modeling methods, such as boundary polylines, normal jumps, and surface tearing, thereby constructing a stable and controllable continuous world geometric model.

[0063] The core optimization equation for this continuity preservation mechanism of curvature and normal constraints is defined as follows: (2) in: The coordinates of the control points in the control point matrix of the NURBS surface are used as optimization variables. This represents the curvature difference at the boundary of adjacent patches; Indicates the normal deviation of the boundary between adjacent patches; It is a balance coefficient used to adjust the weight between curvature continuity (G2 continuity) and normal continuity (G1 continuity).

[0064] Automatic smooth connection at the geometric stitching boundary is achieved by simultaneously minimizing the curvature difference and normal deviation between adjacent patches.

[0065] When Δn=0, the surface is continuous in the boundary normal direction (G1 continuity). When Δk=0, the curvature direction is continuous (G2 continuous).

[0066] By adjusting the coefficient λ, the system can dynamically balance normal smoothing and curvature smoothing according to the needs of the scenario.

[0067] During the optimization process, the system uses a control point matrix. and their weights Using the values ​​of the independent variable, the deviations in curvature and normal of adjacent patch boundaries are calculated, and the control point positions are iteratively updated using numerical methods such as gradient descent or least squares. The goal of this optimization process is to ensure the continuity of the energy gradient at the patching boundary of the surface, avoiding broken lines, depressions, or irregular abrupt changes.

[0068] Through the above optimization mechanism, the spatial distribution of the control point matrix can be automatically adjusted, so that the multi-patch NURBS surface can maintain the geometric continuity of G1 / G2 level in the global parameter domain, thereby forming a smooth and stable continuous world geometry.

[0069] The S500, based on an adaptive subdivision and error control mechanism, automatically performs NURBS surface subdivision operations according to the rate of curvature change and geometric error threshold.

[0070] This invention achieves adaptive subdivision based on geometric complexity by setting an error tolerance ε. Control point density is increased in complex or high-curvature regions, while a simplified structure is maintained in flat regions. This mechanism reduces data volume while ensuring geometric accuracy, achieving a balance between modeling accuracy and computational efficiency.

[0071] When the local curvature exceeds a threshold or the geometric error accumulates excessively, the system automatically inserts a new layer of control points to perform hierarchical NURBS refinement on the NURBS surface. The refinement process is completed entirely within the global NURBS parameter domain. The system can dynamically allocate the control point density according to the rate of curvature change, thereby effectively reducing data and computational load while ensuring geometric accuracy, achieving an adaptive balance between accuracy and efficiency.

[0072] S600, based on a global consistency optimization solution mechanism, performs a global solution on the control point matrix of all patches, and outputs a spatial intelligent world model that is continuously differentiable and has consistent parameters in the global scope.

[0073] After optimizing the control points within a local patch, a global solution is performed on the control point matrices of all patches. Optimization objectives include geometric continuity, parameter smoothness, and global energy constraints. Through global convergence updates, consistent coordination and smooth transition of the entire world model across the parameter domain are achieved.

[0074] Specifically, the control point matrix of all patches is solved by a unified global optimizer, and the objective function comprehensively considers geometric continuity, parameter smoothness, and energy constraints. (3) in: Represents the geometric continuity error term. Indicates the parameter domain smoothing term. To prevent excessive control point density, the optimizer employs a multi-level solution strategy. It first locally optimizes the control points within a patch, and then performs a global synchronous update to achieve globally convergent geometrically consistent modeling.

[0075] Preferably, based on the completed S600 spatial intelligent world model, further multi-scale expression, format export, system integration, or application layer processing are performed on the spatial intelligent world model according to specific application requirements to improve the model's applicability.

[0076] This invention provides a spatial intelligent world modeling method based on the global NURBS parameter domain, which further includes: S700, based on a multi-scale hierarchical modeling and fusion mechanism, maintains curvature and topological continuity between different levels through nested parameter domain mapping relationships, realizing cross-scale seamless fusion of spatial intelligent world modeling.

[0077] The multi-scale hierarchical modeling and fusion mechanism adopts a multi-level NURBS representation to achieve continuous modeling from the city level to the building level and then to the local structure. The upper-level patch represents the macro structure, while the middle and lower-level patches progressively refine the geometric details. The parameter domain and curvature are maintained continuously between different levels, realizing a spatial intelligent modeling system that seamlessly integrates across scales.

[0078] To adapt to the needs of spatial intelligent modeling at different scales, this invention adopts a hierarchical NURBS representation system: the upper-layer patch is responsible for representing the macro-structure at the city level; the middle-layer patch is responsible for the terrain and building outlines; and the lower-layer patch describes local geometric details. The layers maintain curvature and topological continuity through nested parameter domain mapping relationships, achieving seamless cross-scale fusion modeling.

[0079] The following example, using city-level 3D scene modeling under the global NURBS parameter domain, further illustrates the spatial intelligent world modeling method of this invention: This embodiment is used to realize continuous geometric representation of city-level 3D scenes under the global NURBS parameter domain. Using NURBS (Non-Uniform Rational B-Splines) as the underlying geometric logic, it achieves parametric reconstruction from discrete point cloud data to a continuous differentiable world model. For example... Figure 2 As shown, it includes: S101. Data Preparation and Preprocessing: Input city-level LiDAR point cloud data is automatically processed by noise removal, normal estimation, and downsampling. Spatial indexing is used to improve neighborhood query efficiency, resulting in standardized point cloud files containing coordinate, normal, and semantic information.

[0080] The input data in this embodiment includes LiDAR point clouds, photogrammetric point clouds, or other discrete spatial sampling data. Point cloud data serves only as raw spatial sampling input, providing information on scene geometry, normal direction, and local curvature. This invention does not directly perform geometric modeling on the point cloud; instead, it uses it as a geometric constraint, transforming discrete data into a continuous geometric representation through parametric mapping and control point fitting.

[0081] Specifically, the system establishes a two-dimensional parameter domain in the global coordinate system. The point cloud data is projected onto this parameter domain using an affine mapping function, and a control point matrix is ​​generated based on the point cloud distribution density and curvature characteristics. With weight matrix Therefore, LiDAR point clouds serve only as the source of input samples, while the NURBS parameter domain serves as the underlying geometric representation structure.

[0082] The input data is a city-level LiDAR point cloud (LAS format), with an average density of approximately 10 points / m². 2 The system first performs the following preprocessing steps: noise filtering: a statistical outlier removal algorithm is used (radius 1.0m, number of neighborhood points 20); normal estimation: the normal of each point is estimated using PCA; downsampling: voxel filtering (voxel size 0.2m) is used for uniform sampling; spatial index construction: an index is built using KD-Tree to accelerate subsequent neighborhood curvature calculation. The preprocessed point cloud file contains spatial coordinates, normals, and semantic label information.

[0083] S201. Global Parameter Domain Definition and Node Vector Construction: A unified two-dimensional NURBS parameter domain is established in the global coordinate system, mapping the point cloud into a parameterized spatial representation. The system performs coordinate normalization, parameter mapping, node vector generation, and patch partitioning to ensure global continuity and differentiability.

[0084] A unified NURBS parameter domain is established in the global coordinate system to map discrete point cloud spatial data into a continuous two-dimensional parameter representation. For example... Figure 3 Step S201 includes: S2011, Global Coordinate Normalization: The input LiDAR point cloud is normalized in the world coordinate system to eliminate scale differences between different scenes and provide a unified computational basis for subsequent parametric mapping.

[0085] S2012, Parameter Domain Mapping: A two-dimensional parameter domain (u,v) is established in the normalized space, and the point cloud coordinates are converted into two-dimensional parameter domain (u,v) coordinates through an affine mapping function. Each point cloud sample obtains a unique parameter representation, realizing a unified mapping from geographic coordinates to parameter space.

[0086] S2013, Node Vector Generation: Automatically generates non-uniform node vectors based on point cloud density and curvature characteristics to control surface resolution. Nodes are more densely distributed in high-curvature regions and spaced further apart in flat regions, thus improving efficiency while maintaining smoothness.

[0087] S2014, Patch Region Division and Stitching Relationship: Based on point cloud features, the parameter domain is divided into multiple patch regions, and a global parameter index table is established. Adjacent patches share boundary nodes, achieving seamless connection at parameter continuity and geometric stitching points, ensuring global consistency.

[0088] Specifically: First, the input LiDAR point cloud is represented in WGS-84 or UTM coordinates in the world coordinate system, with a large numerical range and inconsistent magnitudes. To ensure the stability of subsequent parameter calculations, the system first performs global normalization, normalizing the original spatial coordinates to [0,1]. 3 The normalized space is calculated using the following formula: Where: X, Y, Z: represent the three-dimensional spatial coordinate components of any sampling point in the LiDAR point cloud in the original world coordinate system (WGS-84 or UTM coordinate system), corresponding to the original coordinate values ​​of the point in the X-axis, Y-axis, and Z-axis directions, respectively. X′, Y′, Z′: represent the normalized coordinate components obtained after global normalization of the original spatial coordinates X, Y, and Z, with their values ​​mapped to [0,1] for subsequent parameter domain mapping and numerical calculation. Xmax, Ymax, Zmax: represent the maximum coordinate values ​​in the X-axis, Y-axis, and Z-axis directions of the current input point cloud dataset, used to determine the upper bound of the spatial range of the point cloud in the corresponding coordinate axis directions. Xmin, Ymin, Zmin: represent the minimum coordinate values ​​in the X-axis, Y-axis, and Z-axis directions of the current input point cloud dataset, used to determine the lower bound of the spatial range of the point cloud in the corresponding coordinate axis directions.

[0089] This step maps point cloud data to a uniform spatial scale, providing a computational basis for subsequent parametric mapping.

[0090] Secondly, the system defines a two-dimensional parameter domain in the normalized space. And establish the correspondence between point cloud coordinates and parameter coordinates through affine mapping functions: ,in, These are linear transformation coefficients, automatically calculated from the size range of the global scene. Through this mapping function, each point cloud sample obtains unique (u,v) coordinates in the two-dimensional parameter domain, achieving a unified transformation from geographic coordinates to parameter space.

[0091] Subsequently, the system adaptively divides the parameter domain based on the point cloud distribution density and curvature gradient, forming several rectangular subdomains (i.e., patch regions). Each patch corresponds to a local geometric unit in the urban space (such as a building facade, road, plot, or vegetation area). The patches are arranged continuously in the parameter space, with seamless connection of boundary parameter ranges, thus maintaining global continuity and differentiability throughout the entire parameter domain. For each patch, node vectors U and V are automatically generated based on the number of control points and the selected NURBS order (cubic NURBS, i.e., p=q=3, is used in this embodiment), as shown below: The non-uniform distribution of node vectors reflects the differences in the geometric complexity of point clouds. In regions with high curvature or dense data, the node spacing is automatically reduced to support high-resolution local surface representation; while in flat or sparse regions, the node spacing is increased to reduce computational redundancy and improve overall computational efficiency.

[0092] The automatic generation of node vectors employs a "curvature-weighted allocation method." The system calculates the local curvature average value k(u,v) in each parameter direction and applies it according to the weighting function: , The curvature adjustment coefficient dynamically adjusts the node spacing, enabling the surface resolution to adaptively match the scene's geometric complexity, thereby significantly improving modeling efficiency and accuracy while maintaining geometric continuity.

[0093] After the node vectors are generated, the system establishes a global parameter index table to record the parameter range for each patch. When the boundary parameter ranges of adjacent patches overlap, the system automatically binds the corresponding node index to ensure parameter continuity and node consistency at the stitching boundary. Through this "parameter domain sharing" mechanism, the entire city-level scene achieves a globally consistent expression in the parameter space, and even when stitching across regions, there will be no abrupt changes in normals or parameter breaks, thus providing a unified geometric basis for subsequent curvature-driven continuity optimization and adaptive subdivision.

[0094] S301. Control Point Matrix and Weight Matrix Generation: Control point matrices and weight matrices are automatically generated within each patch to define the initial surface structure. The number of control points is adaptively adjusted based on curvature characteristics. The generated matrix files are used for subsequent surface stitching and optimization calculations.

[0095] Specifically: within each parameter patch, the system automatically generates a control point matrix. With weight matrix The control point mesh size is 40×40 by default (can be increased to 50×50 in high curvature regions). Each control point records its coordinates and local normal. Weight initialization formula: .in This represents the curvature variance of the control point neighborhood. The control point matrix and weight matrix are stored as files for subsequent continuous optimization and AI training module calls.

[0096] S401, NURBS Surface Generation and Assembly: Surface solving and assembly are performed by calling geometric computation modules (such as libnurbs or OpenNURBS). Local support and global smoothness are achieved through basis function interpolation and node management, maintaining G1 / G2 level continuous connectivity between adjacent patches.

[0097] In the spatial intelligent world modeling system of this invention, the NURBS (non-uniform rational B-spline) geometric calculations for NURBS surface generation and stitching are implemented by the NURBS geometric operation module at the system's bottom layer, such as... Figure 4 The NURBS geometry calculation module includes: Basis function evaluation submodule: used to calculate the numerical results of B-spline basis functions and generate continuously differentiable shape functions in the parameter domain for surface weighted interpolation.

[0098] The surface evaluation and derivative calculation submodule is used to calculate the coordinates of surface points based on the control point matrix and weights, and at the same time to obtain the first derivative, second derivative and normal vector of the surface, providing input for subsequent curvature-driven optimization and continuity constraints.

[0099] The node vector management submodule is used to implement node vector interpolation and adaptive subdivision operations. When the local curvature gradient exceeds the threshold, new nodes are automatically inserted to improve local geometric accuracy.

[0100] Curvature and Continuity Calculation Submodule: Used to calculate the curvature difference and normal deviation of the surface at the splicing boundary, supports G0, G1 and G2 condition detection, and outputs constraint terms for control point optimization.

[0101] The geometry export submodule is used to export the generated NURBS model for seamless data exchange with BIM and digital twin systems.

[0102] Built upon general-purpose mathematical libraries (such as libnurbs, OpenNURBS, or equivalent geometric kernels), the system numerically solves for the control point matrix, node vectors, and weight parameters, and generates surfaces. Each surface patch is defined as a continuously differentiable function in the parameter domain, and local support and global smoothness are guaranteed through the interpolation properties of the node vectors. At adjacent patches, the system automatically adjusts the boundary nodes and weights to ensure that the merged boundary satisfies the G1 (normal continuity) and G2 (curvature continuity) conditions, thereby forming a globally continuous world geometric model.

[0103] S501. Continuity Optimization of Curvature and Normal Constraints: Calculate the curvature difference and normal deviation of adjacent patch boundaries, and minimize energy using control point positions as optimization variables. The system automatically updates boundary control points to achieve seamless splicing and smooth surface transitions, enhancing global geometric stability.

[0104] Specifically, after patch assembly, to further improve overall smoothness and numerical stability, the system calculates the curvature difference Δk and normal deviation Δn of adjacent patch boundaries, and performs energy minimization with control point positions as optimization variables: Where λ is the balance coefficient (values ​​from 0.5 to 1.0). Optimization is achieved through gradient descent with a learning rate of 0.01 and a convergence threshold. The optimization results automatically update the control point matrix and boundary parameters, enabling seamless stitching between multiple patches.

[0105] S601, Adaptive Subdivision and Error Control: When the local curvature change exceeds a threshold, the patch is automatically subdivided and new control point rows and columns are generated through interpolation. Node vectors are updated synchronously using B-spline interpolation rules to ensure that high curvature regions maintain curvature continuity and consistent accuracy.

[0106] Specifically, when the local curvature change exceeds a preset threshold, the system automatically interpolates to generate new rows and columns of control points and updates the basis function node vectors to maintain surface parameterization consistency. Subdivision employs B-spline interpolation rules and synchronizes boundary control points in real time to ensure G2 continuity is maintained even in high curvature regions.

[0107] S701, Global Consistency Solution and Model Output: After all patches are optimized, the system performs a global consistency solution to smooth the distribution of control points and parameter coordinates. This ultimately achieves continuously differentiable, globally consistent spatial intelligent city-level modeling results.

[0108] Specifically: After optimization at all patch levels, the system performs a globally consistent solution to smooth the control point distribution and renormalize the parameter coordinates. The output model includes the following formats: STEP / IGES file: for direct loading by the BIM platform; JSON file: containing parameter domains, node vectors, control point and weight information, which can be used for training digital twin systems, simulation engines, or AI. The generated model is globally continuous and differentiable with consistent parameters, enabling smooth and seamless spatial intelligent modeling in city-level scenarios.

[0109] Example 2: Corresponding to the spatial intelligent world modeling method based on the global NURBS parameter domain provided in Embodiment 1 of this application, Embodiment 2 of this application also provides a spatial intelligent world modeling system based on the global NURBS parameter domain, such as... Figure 5 As shown, it mainly includes the following functional modules: The parameter domain management module is responsible for establishing and maintaining the mapping relationship of the global NURBS parameter domain. This module defines a unified two-dimensional parameter coordinate system, mapping the input point cloud or spatial sampling data to the global parameter domain, providing a unified geometric coordinate basis for subsequent control point generation and surface stitching.

[0110] Control point generation and optimization module: This module automatically generates an initial control point matrix and weight matrix based on the input point cloud, curvature, and semantic features. The system performs control point position optimization within this module, enabling the surface to have a continuously differentiable and locally controllable geometric expression in the global parameter domain.

[0111] The continuity constraint module is responsible for detecting and constraining the differences in normal and curvature between adjacent patches, achieving automatic continuity preservation at geometric boundaries. This module ensures smooth connection of the model at the G1 (normal) and G2 (curvature) levels by iteratively adjusting the boundary control point parameters.

[0112] Subdivision and Error Control Module: Used to perform an adaptive subdivision process based on the error tolerance ε. When the local curvature or error exceeds a set threshold, the system automatically inserts new control points and regenerates node vectors to achieve local refinement and smoothing compensation in high-precision areas.

[0113] Global Optimization and Model Output Module: This module is used to uniformly solve the control point matrices of all patches, coordinate parameters, and output the model. It comprehensively considers geometric continuity, parameter smoothness, and energy constraints to achieve globally consistent optimization of the overall structure. The output model is continuously differentiable and has consistent parameters globally.

[0114] Preferably, it further includes: Multi-scale synchronization module: Used to maintain parameter consistency and geometric smoothness across different levels. This module uses a multi-level NURBS representation mechanism to ensure that models at each scale maintain nested continuity in topology and curvature, thereby achieving unified modeling and seamless fusion across scales.

[0115] The various embodiments of this disclosure have been described above. These descriptions are exemplary and not exhaustive, nor are they 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 in this invention is chosen to best explain the principles, practical applications, or technological improvements to the embodiments in the market, or to enable others skilled in the art to understand the disclosed embodiments.

Claims

1. A spatial intelligent world modeling method based on a global NURBS parameter domain, characterized in that, include: Data input and data preprocessing; Construct a global NURBS parameter domain for modeling the intelligent world in space, and perform consistent mapping and management of multi-patch geometric objects under the global NURBS parameter domain; Generate and optimize the control point matrix for geometric representation based on the global NURBS parameter domain; Based on the continuity preservation mechanism of curvature and normal constraints, automatic smooth transition of multi-patch geometry in the stitching region is achieved. Based on an adaptive subdivision and error control mechanism, NURBS surface subdivision operation is performed according to the rate of curvature change and geometric error threshold. Based on the global consistency optimization solution mechanism, the global solution is performed on the control point matrix of all patches, and the output is a spatial intelligent world model that is continuously differentiable and has consistent parameters in the global scope.

2. The spatial intelligent world modeling method according to claim 1, characterized in that, The method further includes: Based on a multi-scale hierarchical modeling and fusion mechanism, curvature and topological continuity are maintained between different levels through nested parameter domain mapping relationships.

3. The spatial intelligent world modeling method according to claim 1, characterized in that, The global NURBS parameter domain construction process includes: The point cloud data is normalized in the world coordinate system. A two-dimensional parameter domain is established in the normalized space, and the point cloud coordinates are converted into parameter domain coordinates through an affine mapping function. Automatically generate non-uniform node vectors based on point cloud density and curvature characteristics; Based on the point cloud features, the parameter domain is divided into multiple patch regions, and a global parameter index table is established.

4. The spatial intelligent world modeling method according to claim 1, characterized in that, The generation and optimization of the control point matrix for geometric representation based on the global NURBS parameter domain includes: In the global NURBS parameter domain, a control point parameter grid is generated according to a preset resolution or adaptive rules. Based on the mapping relationship between the parameter domain and spatial coordinates, the control points in the parameter domain are mapped to the corresponding spatial control point coordinates, thereby forming an initial control point matrix. Perform global collaborative optimization on the initial control point matrix.

5. The spatial intelligent world modeling method according to claim 1 or 4, characterized in that, The control point matrix, as an optimization variable, is subject to the combined effects of the following constraints: Parameter domain continuity constraints are used to ensure that the boundary control point parameter positions of adjacent patches remain consistent in the global parameter domain. Geometric smoothing constraints are used to suppress geometric abrupt changes caused by excessively dense or sparse distribution of local control points; Error tolerance constraints are used to limit the adjustment range of control points so that the geometric accuracy meets the preset requirements.

6. The spatial intelligent world modeling method according to claim 1, characterized in that, The mechanisms for maintaining the continuity of curvature and normal constraints include: At the junction of adjacent patches, the curvature and normal information are calculated based on the generated control point matrix, and the curvature difference and normal deviation at the junction of adjacent patches are used as the continuity evaluation index; when the curvature difference or normal deviation exceeds the preset threshold, the continuity optimization process is automatically triggered. In the continuous optimization process, the coordinates of the control points in the control point matrix are used as optimization variables. The positions of the control points are iteratively adjusted by simultaneously constraining the curvature and normal changes at the boundaries of adjacent patches.

7. The spatial intelligent world modeling method according to claim 1, characterized in that, Performing NURBS surface subdivision operations based on the rate of curvature change and a geometric error threshold includes: When the local curvature exceeds the threshold or the accumulated geometric error exceeds the threshold, a new control point layer is automatically inserted to refine the NURBS surface hierarchically.

8. The spatial intelligent world modeling method according to claim 2, characterized in that, Multi-scale hierarchical modeling and fusion mechanisms include: A hierarchical NURBS representation system is adopted: the upper-level patch is responsible for representing the macro structure of the city; the middle-level patch is responsible for the terrain and building outlines; the lower-level patch describes local geometric details; and the curvature and topological continuity between the layers are maintained through nested parameter domain mapping relationships.

9. A spatial intelligent world modeling system based on a global NURBS parameter domain, characterized in that, include: Parameter domain management module: responsible for establishing and maintaining the mapping relationship of global NURBS parameter domains; Control point generation and optimization module: used to automatically generate initial control point matrix and weight matrix based on input point cloud, curvature and semantic features; Continuity constraint module: This module is responsible for detecting and constraining the differences in normal and curvature between adjacent patches to achieve automatic continuity at geometric boundaries. Subdivision and Error Control Module: Used to perform an adaptive subdivision process based on error tolerance; when the local curvature or error exceeds the set threshold, the system automatically inserts new control points and regenerates the node vector; Global Optimization and Model Output Module: Used to uniformly solve the control point matrix of all patches, coordinate parameters, and output the model.

10. The spatial intelligent world modeling system according to claim 9, characterized in that, The system also includes: Multi-scale synchronization module: used to maintain parameter consistency and geometric smoothness between different levels, so that the models at each scale remain nested and continuous in terms of topology and curvature.

Images