A space grid adaptive partitioning method, device, equipment and medium

Abstract

This application discloses a spatial grid adaptive partitioning method, apparatus, device, and medium. The embodiment acquires multi-source heterogeneous 3D data of a target region and preprocesses the data to obtain the vertex density of each vertex within the target region. Based on the vertex coordinates, a pre-trained continuous environmental factor expression model predicts the multivariate environmental factors corresponding to each vertex. For any grid within the target region, based on the vertex density and multivariate environmental factors, it is determined whether the grid is to be partitioned. The grid to be partitioned is further partitioned based on the longest axis bisection strategy. Thus, by incorporating vertex density and environmental factors into the determination mechanism of the grid to be partitioned, and recursively partitioning through this mechanism, complex regions or high-risk environments obtain finer grids, while simple regions or low-risk environments maintain coarser grids, achieving non-uniform and differentiated adaptive partitioning.

Description

Technical Field

[0001] This application relates to the field of spatial information processing technology, specifically to a spatial grid adaptive partitioning method, apparatus, device, and medium. Background Technology

[0002] With the rapid development of smart cities, digital twins, and low-altitude economy, the Discrete Global Grid System (DGGS), as a framework that discretizes the Earth's surface and near-Earth space into a multi-resolution hierarchical grid, provides a unified digital foundation for the organization, indexing, computation, and fusion of global spatial data.

[0003] In related technologies, typical DGGS frameworks such as GeoSOT-3D usually divide the 3D space using pre-defined fixed-level rules (e.g., a 32-level partitioning system). Each level corresponds to a defined mesh size and encoding length, thereby ensuring the consistency of spatial reference globally. In this method, the mesh fineness depends on the pre-defined fixed-level partitioning, resulting in a lack of ability to dynamically adjust according to the actual geometric size and complexity of the data elements. Summary of the Invention

[0004] This application provides a spatial grid adaptive partitioning method, apparatus, device, and medium that can realize dynamic adaptive partitioning of spatial grids.

[0005] This application provides a spatial grid adaptive partitioning method, including:

[0006] Acquire multi-source heterogeneous 3D data of the target region, and preprocess the multi-source heterogeneous 3D data to obtain the vertex density of each vertex in the target region;

[0007] Based on the vertex coordinates of each vertex, the multivariate environmental factors corresponding to each vertex are predicted using a pre-trained continuous expression model of environmental factors.

[0008] For any grid within the target area, based on the vertex density of each vertex within the grid and multiple environmental factors, determine whether the grid is a grid to be divided;

[0009] The grid to be divided is further divided based on the longest axis bisection strategy.

[0010] Optionally, for any grid within the target area, determining whether the grid is a grid to be divided is based on the vertex density of each vertex within the grid and multiple environmental factors, including:

[0011] For any grid within the target area, the importance of the grid is determined based on the vertex density of each vertex within the grid and multiple environmental factors;

[0012] If the importance of the grid is greater than or equal to the partitioning threshold, then the grid is determined to be a grid to be partitioned.

[0013] Optionally, for any grid within the target region, the importance of the grid is determined based on the vertex density of each vertex within the grid and multiple environmental factors, including:

[0014] For any grid within the target region, the geometric complexity of the grid is determined based on a preset geometric weight and the sum of the vertex densities of all vertices within the grid.

[0015] The environmental complexity of the grid is determined based on the multivariate environmental factors of each vertex in the grid and the environmental weights corresponding to each multivariate environmental factor.

[0016] The importance of the mesh is determined based on the geometric complexity and the environmental complexity.

[0017] Optionally, the mesh to be divided is further divided based on the longest axis bisection strategy, including:

[0018] The grid is divided into two sub-grids along the target axis, where the target axis refers to the direction of the longest axis among the three axes of length, width, and height of the grid.

[0019] The method further includes:

[0020] Based on the axis position and sub-block position of the sub-mesh, the local code of the sub-mesh is determined. The axis position refers to the code representation corresponding to the target axis, and the sub-block position refers to the code representation corresponding to the partition position of the sub-mesh on the target axis. The partition position can be one of left-right partition, top-bottom partition, or front-back partition.

[0021] The local code of the sub-mesh and the local code of the parent mesh node to which the sub-mesh belongs are concatenated hierarchically to obtain the global code of the sub-mesh. The parent mesh node refers to the mesh related to the sub-mesh division process.

[0022] Optionally, the continuous expression model of environmental factors includes at least a location encoding module and a multilayer perceptron regression module;

[0023] Based on the vertex coordinates of each vertex, the multivariate environmental factors corresponding to each vertex are predicted using a pre-trained continuous representation model of environmental factors, including:

[0024] Input the vertex coordinates of each vertex into the position encoding module;

[0025] In the position encoding module, the vertex coordinates of each vertex are mapped to a high-dimensional space through a random Gaussian matrix, and then transformed by sine and cosine functions to obtain a high-dimensional feature vector;

[0026] The high-dimensional feature vector is input into the multilayer perceptron regression module, and then undergoes nonlinear transformation through multiple hidden layers. Finally, the multivariate environmental factors corresponding to the vertex coordinates of each vertex are obtained through regression through the output layer.

[0027] Output the multivariate environmental factors corresponding to the vertex coordinates of each vertex.

[0028] Optionally, the multi-source heterogeneous 3D data is preprocessed to obtain the vertex density of each vertex within the target region, including:

[0029] The multi-source heterogeneous 3D data is projected onto the same coordinate system to obtain the initial data of the target region;

[0030] The initial data is cleaned to obtain valid data;

[0031] Based on the valid data, the vertex density of each vertex within the target region is determined.

[0032] Optionally, the effective data includes point feature data, surface model data, and 3D model data;

[0033] Based on the valid data, the vertex density of each vertex within the target region is determined, including:

[0034] For the surface model data and the 3D model data, determine the volume of the axis-aligned outer bounding box of the surface model data and the 3D model data;

[0035] The number of original vertices contained in the surface model data and the 3D model data is counted.

[0036] The original vertex density is determined based on the volume of the axis-aligned outer bounding box and the original number of vertices;

[0037] The corner points of the outer bounding box aligned with the axis are used as representative vertices of the face model data and the 3D model data, and the vertex density of the representative vertices is determined based on the original vertex density.

[0038] For the point feature data, the vertex density of the point feature data is determined to be one;

[0039] Based on the vertex density of the representative vertices of all the surface model data and the 3D model data within the target area, and the vertex density of all the point feature data, the vertex density of each vertex within the target area is obtained.

[0040] This application embodiment also provides a spatial grid adaptive partitioning device, including:

[0041] The vertex density acquisition module is used to acquire multi-source heterogeneous 3D data of the target region and preprocess the multi-source heterogeneous 3D data to obtain the vertex density of each vertex in the target region.

[0042] The multivariate environmental factor prediction module is used to predict the multivariate environmental factors corresponding to each vertex based on the vertex coordinates of each vertex and a pre-trained continuous expression model of environmental factors.

[0043] The grid to be divided module is used to determine whether a grid is a grid to be divided for any grid within the target area, based on the vertex density of each vertex in the grid and multiple environmental factors.

[0044] The mesh generation module is used to further divide the mesh to be divided based on the longest axis bisection strategy.

[0045] This application also provides an electronic device, including a memory storing multiple instructions; the processor loads the instructions from the memory to execute the steps in any of the spatial grid adaptive partitioning methods provided in this application.

[0046] This application also provides a computer program product, including a computer program / instructions, which, when executed by a processor, implement the steps in any of the spatial grid adaptive partitioning methods provided in this application.

[0047] This application also provides a computer-readable storage medium storing a plurality of instructions adapted for loading by a processor to execute steps in any of the spatial grid adaptive partitioning methods provided in this application.

[0048] This application embodiment can acquire multi-source heterogeneous 3D data of a target region, and preprocess the multi-source heterogeneous 3D data to obtain the vertex density of each vertex in the target region; based on the vertex coordinates of each vertex, predict the multivariate environmental factors corresponding to each vertex based on a pre-trained continuous expression model of environmental factors; for any grid in the target region, determine whether the grid is a grid to be divided based on the vertex density of each vertex in the grid and the multivariate environmental factors; and continue to divide the grid to be divided based on the longest axis bisection strategy.

[0049] Therefore, by incorporating vertex density and environmental factors into the decision-making mechanism for mesh generation, the importance of both geometric space and environmental factors is considered. Furthermore, recursive generation through the decision-making mechanism allows for finer grids in complex or high-risk environments, while maintaining coarser grids in simple or low-risk environments. This achieves non-uniform, differentiated, adaptive generation, thus meeting the needs of specific application scenarios and realistically reflecting the complex and ever-changing low-altitude environment. Attached Figure Description

[0050] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0051] Figure 1 This is a flowchart illustrating an adaptive spatial grid partitioning method provided in an embodiment of this application;

[0052] Figure 2 This is a schematic diagram of the structure of the continuous expression model of environmental factors in the embodiments of this application;

[0053] Figure 3 This is a schematic diagram of the global encoding structure of the grid in an embodiment of this application;

[0054] Figure 4 This is a schematic diagram of the structure of a spatial grid adaptive partitioning device provided in an embodiment of this application;

[0055] Figure 5 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0056] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0057] Existing DGGS frameworks based on rule-based hierarchical grid partitioning exhibit the following two major limitations in practical applications, especially in scenarios involving significant scale differences and complex environmental factors, such as citywide or low-altitude airspace:

[0058] (1) The regular hierarchical grid division is difficult to adapt to the scale differences of data elements, resulting in an imbalance between computational efficiency and accuracy.

[0059] Existing DGGS frameworks rely on pre-defined fixed hierarchical divisions to determine grid fineness, lacking the ability to dynamically adjust based on the actual geometric size and complexity of data elements. For example, in low-altitude airspace management scenarios, spatial data simultaneously contains large-scale sparse airspace blocks and small-scale complex buildings or obstacles. Using a uniform hierarchical division presents two problems: firstly, for simple, sparse, large-scale airspace data, overly fine grid divisions generate a large number of invalid redundant grids, resulting in a huge waste of storage space and computing resources; secondly, for complex, detailed small-scale obstacles (such as utility poles and small buildings), overly coarse grid divisions fail to accurately capture their geometric features, leading to severely insufficient accuracy in spatial computations (such as collision detection and proximity analysis), increasing the safety risks of aircraft route planning. Therefore, existing technologies cannot achieve adaptive, unified, and efficient encoding and computation of "multi-source heterogeneous data of different sizes and finenesses" while ensuring global consistency.

[0060] (2) Existing grid coding mechanisms are difficult to effectively support the fusion expression of dynamic environmental factors, which limits the ability to analyze in complex environments.

[0061] Existing DGGS (Discrete Geometric Grids) solutions (such as collision detection methods based on GeoSOT-3D) primarily focus on spatial location encoding and set operations (such as intersection and union) of geometric elements like points, lines, surfaces, and volumes to determine conflicts related to UAV positions, obstacle ranges, and trajectory paths. However, in real-world low-altitude flight path planning, in addition to geometric spatial relationships, environmental factors (such as sudden changes in wind speed / direction, variations in light intensity, geomagnetic anomalies, and GPS signal obstruction) have a decisive impact on flight safety and path optimization. Since the core encoding mechanism of DGGS uses spatial location as the sole primary key, existing technologies cannot directly embed these continuous or discrete environmental factor attributes into the mesh generation, subdivision, and encoding process. Environmental factors are typically only indirectly associated and expressed as external attribute tables or independent layers. This "internal-external separation" model results in the mesh system itself lacking the inherent ability to represent environmental states and their dynamic changes (such as the intensity of airflow disturbances evolving over time). The mesh generation results in a poor ability to accurately reflect the complex and ever-changing low-altitude environment, leading to limited adaptability.

[0062] In view of this, this application proposes a spatial grid adaptive partitioning method that can deeply integrate environmental factors into the dynamic partitioning process of the grid, thereby achieving a non-uniform and differentiated adaptive partitioning effect on a global scale, and can meet the needs of specific application scenarios, truly reflecting the complex and ever-changing low-altitude environment.

[0063] For details, please refer to Figure 1 The specific process of this adaptive spatial grid partitioning method can be summarized as follows:

[0064] S110. Obtain multi-source heterogeneous 3D data of the target region, and preprocess the multi-source heterogeneous 3D data to obtain the vertex density of each vertex in the target region.

[0065] In this embodiment, the target area refers to the region of interest selected by the user, that is, the specific geographic space to be subdivided into a refined 3D grid. For example, the low-altitude airspace in a city plan (such as 0-200 meters above the ground) or the entire 3D space of a digital twin city.

[0066] Multi-source heterogeneous 3D data refers to a collection of various 3D data from different sensors, in different formats, with different levels of precision, describing different objects within a target area. For example, it can include the following data elements: POIs (Points of Interest), such as cafes, convenience stores, and charging stations within a park; building facades, such as the two-dimensional building outlines of a park; oblique photogrammetry 3D models, realistic 3D models generated through drone aerial photography, used to reflect terrain and building appearance; and BIM models (Building Information Modeling), detailed internal models of key buildings, including information such as pipes and structures.

[0067] Vertex density is a metric for measuring the geometric complexity of a 3D model. For example, a simple cube has a low vertex density, while a finely crafted sculpture has a high vertex density.

[0068] Specifically, after acquiring multi-source heterogeneous 3D data of the target region, preprocessing is performed first, including but not limited to: coordinate system I, data format conversion, geometric cleaning (removal of invalid faces), and most importantly, vertex density calculation. The preprocessing process mainly includes:

[0069] S110-1. Project the multi-source heterogeneous 3D data onto the same coordinate system to obtain the initial data of the target area.

[0070] In this embodiment, a suitable projection coordinate system can be selected based on the project scope and requirements. For example, CGCS2000 / 3-degree Gauss-Kruger (a 3-degree zone Gauss-Kruger projection coordinate system under the China Geodetic Coordinate System 2000) or UTM Zone 50N (the 50th zone of the Universal Transverse Mercator Projection). Then, multi-source heterogeneous 3D data from different coordinate systems are uniformly projected onto this projection coordinate system, ensuring that all data are correctly superimposed and aligned, thereby obtaining the initial data for the target area, which is then used for subsequent density calculations.

[0071] S110-2. Clean the initial data to obtain valid data.

[0072] During coordinate projection transformation, the model automatically generated by oblique photogrammetry may contain a small number of broken or coplanar triangular faces. The BIM model may also generate tiny invalid faces after conversion. These invalid geometries can lead to rendering errors (such as broken surfaces or black spots) and analysis calculation failures (such as volume calculation errors). Therefore, it is necessary to clean the initial data to remove geometric errors, improve data quality, and ensure the correctness of subsequent calculations and visualization effects.

[0073] For example, data cleaning mainly includes the following aspects: 1. Invalid faces: These are faces that cannot form valid polygons. For example, a face with three collinear points has zero area and needs to be automatically detected and deleted. 2. Anomalies in normals: The direction of the normals of a triangle determines the reflection of light (which side is the front and which is the back). Anomalies in normals may cause faces to be invisible or flickering during rendering, so it is necessary to unify the normal directions. 3. Duplicate triangles: Two completely overlapping triangles are redundant data and should be deleted directly.

[0074] S110-3. Based on valid data, determine the vertex density of each vertex within the target area.

[0075] The cleaned valid data includes point feature data, polygon model data, and 3D model data. The vertex density calculation process includes: determining the volume of the axis-aligned bounding box of the polygon model data and 3D model data; counting the original number of vertices contained in the polygon model data and 3D model data; determining the original vertex density based on the volume of the axis-aligned bounding box and the original number of vertices; using the corner points of the axis-aligned bounding box as representative vertices of the polygon model data and 3D model data, and determining the vertex density of the representative vertices based on the original vertex density; determining the vertex density of the point feature data as one; and obtaining the vertex density of each vertex in the target area based on the vertex densities of the representative vertices of all polygon model data and 3D model data, and the vertex densities of all point feature data.

[0076] In this embodiment, for surface model data and 3D model data, such as oblique photogrammetry modularized models and BIM models, the smallest cuboid bounding box that can completely enclose the entire model is first calculated, i.e., the axis-aligned outer bounding box, and its volume is calculated. Then, the number of original vertices contained in the model is counted. For example, a building contains 500,000 original vertices. Next, the original vertex density of the model can be calculated:

[0077]

[0078] in, This represents the original vertex density of the model. This indicates the volume of the outer bounding box aligned with the axis. This represents the original number of vertices.

[0079] The original vertex density is easy to understand. The number of vertices contained within a unit volume is a core metric for measuring the geometric complexity and precision of a model. A higher value indicates a more detailed model, but also consumes more computer resources. Therefore, to alleviate computational pressure, this embodiment uses the corner points of the axis-aligned bounding box as representative vertices of the model and assigns a density to each representative vertex: for example, for a building, the eight corner points of its axis-aligned bounding box are taken as its representative vertices, and the vertex density of each representative vertex is: Thus, each complex element has eight representative points, which can be used for rapid spatial queries and analysis without having to call up the huge complete model every time.

[0080] For point feature data, such as points of interest (like a charging station), simply set its vertex density to 1.

[0081] Finally, the vertex density of each element within the target area can be calculated. For example, the vertex density of the representative vertex of a simple factory building is 10 vertices per cubic meter, while the vertex density of the representative vertex of the R&D center BIM model is 1000 vertices per cubic meter.

[0082] S120. Based on the vertex coordinates of each vertex, the multivariate environmental factors corresponding to each vertex are predicted using a pre-trained continuous expression model of environmental factors.

[0083] In this embodiment, vertex coordinates are used to represent the position of the vertex in three-dimensional space, which is the basis for spatial positioning and querying environmental factors. For example, a triplet can be used. To indicate, among which, Indicates longitude. Indicates latitude, Indicates altitude.

[0084] Multivariate environmental factors refer to a variety of physical or environmental parameters within the target area that have a significant impact on the application. These environmental factors collectively describe the environmental conditions in three-dimensional space. For example, multivariate environmental factors may include: wind speed, temperature, light intensity, magnetic field strength, GPS signal-to-noise ratio, air pollution index, etc.

[0085] The pre-trained continuous representation model of environmental factors is a pre-trained implicit neural representation (INR) network. It learns an expression of environmental factors from spatial coordinates. A continuous mapping function to multiple environmental factor values; that is, the input to this continuous expression model of environmental factors is arbitrary spatial coordinates. The output is the values ​​of various environmental factors at that coordinate point.

[0086] Specifically, the training process of this continuous expression model of environmental factors includes: firstly, acquiring the original discrete sampling data of environmental factors within the target area (such as weather station records, sensor readings, etc.), and then expanding each sampling point into a quadruple. ,in, That is, representing the coordinate point The model incorporates multiple environmental factors, such as wind speed and temperature. The coordinates are then normalized to a uniform [0, 1] interval and divided into training and testing sets at an 80%:20% ratio. Simultaneously, the multiple environmental factors are preprocessed to handle extreme outliers (such as outliers caused by sensor malfunctions) to prevent excessive interference with model training. Then, based on the sample data formed by the coordinates and their corresponding multiple environmental factors, the initial implicit neural representation network is trained and tested: the coordinates are input into the initial implicit neural representation network for feature extraction and prediction to obtain the predicted multiple environmental factors corresponding to those coordinates. The loss function value is calculated based on the predicted multiple environmental factors and their actual multiple environmental factors, and the network is updated using a backward gradient based on the loss function value. The AdamW optimizer is used to update the network weights until the loss function converges or the preset number of training iterations is reached, at which point the update stops, resulting in a trained continuous representation model of environmental factors. Through training, the implicit neural representation network can learn a continuous mapping relationship from arbitrary spatial coordinates to the intensity of environmental factors, ultimately achieving fast and high-precision prediction of environmental factor values ​​at any location (including unsampled points) within the target area. Furthermore, the continuous representation model of environmental factors can also be trained based on gated recurrent units (GRUs), Transformer / Informer series models, spatiotemporal convolutional neural networks (3D-CNNs), or temporal graphical neural networks (T-GCNs), etc., and this application does not impose specific limitations on these methods.

[0087] Further, please refer to Figure 2 The continuous representation model of environmental factors includes at least a location encoding module and a multilayer perceptron (MLP) regression module. Step S120 above predicts the multivariate environmental factors corresponding to each vertex based on the vertex coordinates and the pre-trained continuous representation model of environmental factors, including:

[0088] The vertex coordinates of each vertex are input into the position encoding module. In the position encoding module, the vertex coordinates of each vertex are mapped to a high-dimensional space through a random Gaussian matrix, and then transformed by sine and cosine functions to obtain a high-dimensional feature vector. The high-dimensional feature vector is input into the multilayer perceptron regression module, and then nonlinearly transformed through multiple hidden layers. Finally, the multivariate environmental factors corresponding to the vertex coordinates of each vertex are obtained through regression through the output layer. The multivariate environmental factors corresponding to the vertex coordinates of each vertex are output.

[0089] In this embodiment, during the position encoding stage, the input three-dimensional spatial coordinates A high-dimensional feature vector of 512 dimensions is obtained by mapping to a random Gaussian matrix B and then applying sine and cosine function transformations in parallel along each dimension. This vector contains rich frequency information to capture complex spatial variations. This high-dimensional feature vector is then fed into an MLP regression architecture. This multilayer perceptron regression module consists of one input layer, four hidden layers, and one output layer. The input layer maintains the same dimension as the location encoding output (512 dimensions). Each hidden layer contains 256 neurons and uses the ReLU activation function to enhance nonlinear representation and accelerate model convergence. The output layer is a single neuron that directly regresses the intensity values ​​of environmental factors at the corresponding spatial location, ultimately outputting the multivariate environmental factors for that location. This design ensures that the INR network can capture high-frequency details in the spatial field while maintaining the stability and continuity of the regression process.

[0090] S130. For any grid within the target area, determine whether the grid is a grid to be divided based on the vertex density of each vertex and multiple environmental factors within the grid.

[0091] S140. The mesh to be divided is further divided based on the longest axis bisection strategy.

[0092] In this embodiment, after completing the vertex density calculation and environmental factor prediction, for any grid within the target area, the importance of the grid is determined based on the vertex density of each vertex and multiple environmental factors. If the importance of the grid is greater than or equal to the partitioning threshold, the grid is determined to be a grid to be partitioned.

[0093] Among them, grid importance is a comprehensive quantitative indicator used to measure the criticality and priority of a 3D spatial grid in the entire analysis scenario. It is calculated based on the vertex density of each vertex within the grid and multiple environmental factors, and is a dimensionless score that integrates geometric complexity and environmental risk. That is, the importance of a grid depends not only on the complexity of its internal physical world (how many complex buildings and facilities there are), but also on the dynamic risk level of its environment (how strong the wind is, how poor the signal is, etc.).

[0094] For any grid within the target area, a decision is made on whether to further subdivide it based on its calculated importance index. When the importance of a grid is greater than or equal to a set subdivision threshold, the grid is determined to be a grid to be subdivided and requires further subdivision; when the importance of a grid is less than the set subdivision threshold, the grid is considered sufficiently "simple" or "safe" and does not require further subdivision.

[0095] For example, an initial large grid A covers an open grassland with low vertex density and gentle wind speed. Its overall importance is lower than the partitioning threshold, so it is not partitioned. On the other hand, another initial grid B of the same size covers the R&D building and its surroundings. It has high vertex density and high wind speed on the roof. Its overall importance is much higher than the partitioning threshold, so it is marked as a grid to be partitioned.

[0096] For the mesh to be divided, the longest axis bisection strategy is used to further divide the mesh. For example, the mesh is divided into two sub-mesh along the target axis, which refers to the direction of the longest axis among the three axes of length, width, and height. That is, by calculating the side lengths of the mesh in the x, y, and z directions, the longest axis is selected as the dividing direction, and the mesh is divided into two sub-mesh along this direction.

[0097] After the subdivision is completed, the importance of each subgrid is recalculated, and the subdivision is determined again. Subgrids that can be further subdivided are then further subdivided. In this way, the above steps S130-S140 are repeated to subdivide the grid layer by layer until the importance of a certain subgrid is less than the set subdivision threshold, at which point the subdivision stops.

[0098] Therefore, by recursively partitioning the grid through a decision-making mechanism, a non-uniform grid that conforms to actual characteristics is dynamically formed in spatial distribution. This takes into account both the importance of space and the balance between computational overhead and storage efficiency, transforming grid partitioning from the traditional "uniform partitioning" to "on-demand focusing," enabling precise allocation of computing resources. Ultimately, large grids are used in simple, safe areas (such as open skies) to save resources, while the grid is automatically densified in complex, high-risk areas (such as densely built-up areas and windy areas) to ensure accuracy.

[0099] Furthermore, the process of determining the importance of a grid based on the vertex density and multivariate environmental factors includes:

[0100] S130-1. For any grid within the target area, determine the geometric complexity of the grid based on the preset geometric weights and the sum of the vertex densities of each vertex within the grid.

[0101] S130-2. Determine the environmental complexity of the grid based on the multivariate environmental factors of each vertex and the environmental weights corresponding to each multivariate environmental factor.

[0102] S130-3. Determine the importance of the mesh based on geometric complexity and environmental complexity.

[0103] Specifically, the formula for calculating grid importance is as follows:

[0104]

[0105] in, Indicates the importance of the grid. This indicates the number of vertices in the grid. This represents the preset geometric weights. This represents the first normalization function. Indicates the first Vertex density of vertices, Indicates the first Environmental weights of each environmental factor This represents the second normalization function. Indicates the first The vertex of the first vertex The values ​​of environmental factors, This indicates the number of environmental factors.

[0106] importance of the grid It consists of two parts, the first part For the geometric complexity of the mesh, the second part The environmental complexity of the grid.

[0107] Specifically, in the calculation of geometric complexity, firstly, for any given mesh, all vertices falling within that mesh are traversed, and their vertex densities are summed to obtain the total vertex density. This value intuitively reflects the total amount of geometric detail contained in the mesh. A mesh containing multiple corner points of a highly detailed landmark building will have a much higher sum of vertex densities than a small mesh containing only simple face or lawn vertices. It should be noted that within this mesh, for both face model data and 3D model data, the vertex density of representative vertices (i.e., the vertex density of the 8 corner points aligned with the outer bounding box) is taken. The calculation is performed using a vertex density of 1 for point feature data.

[0108] Then, the sum of vertex densities is normalized and mapped to the [0, 1] interval to eliminate dimensional differences. It is then compared with a preset geometric weight. Multiplying these gives the final geometric complexity. The geometric weights are then calculated. This setting can be adjusted based on the specific application. For example, if the current application focuses more on the geometry of the physical world (e.g., primarily for 3D visualization rendering load assessment), the setting can be increased. This allows the mesh to better represent complex models. If the geometry is relatively minor, the mesh size can be reduced. .

[0109] In the process of calculating environmental complexity, for each vertex within the grid, a pre-trained continuous representation model of environmental factors is used to query the values ​​of various environmental factors (such as wind speed, illumination, GPS intensity, etc.) at its location, resulting in an environmental factor vector. Then, for each environmental factor... Normalization is performed to eliminate the influence of extreme outliers, and a weighted sum is calculated based on the predefined environmental weights of each environmental factor to obtain the final environmental complexity of the grid. The sum of the environmental weights of all environmental factors is 1; for example, wind speed has a weight of 0.4, GPS signal quality has a weight of 0.3, and illumination has a weight of 0.3.

[0110] In summary, the embodiments of this application have the following effects:

[0111] 1. Adaptability and Efficiency: Building upon traditional spatial range-based rule-based partitioning criteria, this model further incorporates weighted evaluations of spatial geometric complexity and dynamic environmental factors, forming a joint weighted model of geometry and environmental factors to guide the splitting and termination of spatial grid cells. This approach considers both the importance of geometric space and environmental factors. Simultaneously, by setting partitioning thresholds, it controls the further subdivision or termination of the grid, allowing for finer grids in complex or high-risk environments, while maintaining coarser grids in simple or low-risk environments. This achieves non-uniform, differentiated spatial partitioning, thus meeting the needs of specific application scenarios and realistically reflecting the complex and ever-changing low-altitude environment.

[0112] 2. Accuracy and flexibility of the continuous expression model of environmental factors: Based on the implicit neural network (INR), the continuous expression of environmental factors in space is realized, and environmental factors can be queried at any fine scale, providing extremely accurate data support for grid division.

[0113] Furthermore, this application also provides an encoding method for binary encoding the grid during the adaptive partitioning of the grid using the above-described spatial grid adaptive partitioning method.

[0114] Specifically, the encoding process may include:

[0115] The grid is divided into two sub-grids along the target axis, which refers to the direction of the longest of the three axes (length, width, and height) of the grid. Based on the axis and sub-block positions of the sub-grid, the local code of the sub-grid is determined. The axis refers to the code representation corresponding to the target axis, and the sub-block position refers to the code representation corresponding to the partition position of the sub-grid along the target axis. The partition position can be one of left-right partition, top-bottom partition, or front-back partition. The local code of the sub-grid and the local code of the parent grid node to which the sub-grid belongs are concatenated hierarchically to obtain the global code of the sub-grid. The parent grid node refers to the grid related to the sub-grid division process.

[0116] In this embodiment, for a three-dimensional mesh (e.g., a cuboid), its dimensions in the X, Y, and Z directions are first calculated. Then, the lengths of these three dimensions are compared, and the longest direction is selected as the target axis for this partitioning. For example, if a mesh has dimensions of (100m, 50m, 20m), the X-axis is the longest and is selected as the target axis. After determining the target axis, the entire mesh is precisely divided into two sub-mesh of equal volume along the midpoint of this axis using a plane perpendicular to the axis. If the target axis is the X-axis, it is partitioned left and right, resulting in a left sub-mesh and a right sub-mesh. If the target axis is the Y-axis, it is partitioned front and back, resulting in a front sub-mesh and a back sub-mesh. If the target axis is the Z-axis, it is partitioned top and bottom, resulting in an upper sub-mesh and a lower sub-mesh. This partitioning along the longest side most effectively reduces the size of the mesh's maximum bounding rectangle, making the mesh geometrically closer to a cube, and can adapt to regions of different shapes, ensuring that the partitioning process remains balanced.

[0117] Then, local codes are assigned to the two sub-grids generated in this meshing operation. A local code is a relative identifier assigned to a newly generated sub-grid in a single meshing operation. It only records direct information from this meshing operation and does not include the grid's historical position in global space. Local codes can use binary encoding, which consists of two parts:

[0118] Axis: A code used to identify the division direction, recording the selected axis for the current layer division. For example, the X-axis is represented by 00, the Y-axis by 01, and the Z-axis by 10.

[0119] Sub-block bit: A code used to identify the position of a sub-grid. When two sub-grids are bisected along the target axis, it's necessary to distinguish whether it's "left" or "right" (or top / bottom, front / back). For example, left, top, and front can be represented by 1, and right, bottom, and back by 0.

[0120] Next, the axis position and the sub-block position are concatenated to form the local code of the sub-grid. For example, the local code of the upper half of the sub-grid after dividing along the Y-axis is 011. That is, the local code is a three-bit binary code.

[0121] Finally, the local codes of the subgrid and the local codes of all parent grid nodes related to the subgrid partitioning process are concatenated hierarchically to obtain the global code of the subgrid. The global code is an absolute and unique identifier for a grid within the entire hierarchical spatial data structure. It is formed by concatenating the local codes of all partitioning steps performed from the top-level root grid to the current grid in hierarchical order. For example, the local code of the upper half of the subgrid (first layer) after partitioning along the Y-axis is 011, and the local code of the right half of the subgrid (second layer) after further partitioning along the X-axis is 001. Finally, the global code of the second-layer subgrid is 011 001. Each grid's global code uniquely defines its position and depth (level) in the entire space. A longer code length indicates that the grid has been partitioned more times, its level is deeper, and its volume is smaller.

[0122] For example, please refer to Figure 3 , Figure 3 This is a schematic diagram of the global encoding structure of the grid in an embodiment of this application. For example... Figure 3 As shown, the bit order of the global encoded bit string is as follows: the coarse level (near the root) is in the high bit, and the fine level (near the leaf) is in the low bit. That is, the earlier the partition occurs, the closer its local code (3 bits) is to the high bit; the later the partition occurs, the closer its local code (3 bits) is to the low bit.

[0123] In this way, the local code of a sub-grid is adjacent to the local code of its parent grid, thus generating bit order and hierarchy depth. That is, for the global code of a grid, by truncating its last bit segment (i.e., the last 3 bits), the global code of the parent grid above that grid can be obtained. For example, the global code of the parent grid above grid 011 001 100 is 011 001. Simultaneously, by flipping the last bit (i.e., the last 1 bit), the global code of the adjacent sibling grid of that grid can be obtained. Sibling grids refer to two sub-grids generated in a single partitioning process. For example, the global code of the sibling grid of grid 011 001 100 is 011 001 101.

[0124] This application embodiment organizes and stores the spatial elements (POIs, building surfaces, oblique photogrammetry / 3D models / BIM, etc.) contained in each grid cell according to the unique grid code of that cell, constructing a unified index structure based on the code. Through this index system, multi-source heterogeneous data of different sources and types can be uniformly mapped to a standardized 3D grid framework, realizing simplified expression and efficient management of complex scene data.

[0125] To better implement the above methods, this application also provides a spatial grid adaptive partitioning device, which can be integrated into an electronic device, such as a terminal or server. The terminal can be a tablet computer, a smart Bluetooth device, a laptop computer, or a personal computer; the server can be a single server or a server cluster composed of multiple servers.

[0126] For example, in this embodiment, the method of this application embodiment will be described in detail by taking the spatial grid adaptive partitioning device specifically integrated into the server as an example.

[0127] For example, such as Figure 4 As shown, the spatial mesh adaptive partitioning device may include a vertex density acquisition module 410, a multivariate environmental factor prediction module 420, a mesh to be partitioned determination module 430, and a mesh partitioning module 440, as follows:

[0128] Vertex density acquisition module 410 is used to acquire multi-source heterogeneous 3D data of the target region and preprocess the multi-source heterogeneous 3D data to obtain the vertex density of each vertex in the target region.

[0129] The multivariate environmental factor prediction module 420 is used to predict the multivariate environmental factors corresponding to each vertex based on the vertex coordinates of each vertex and a pre-trained continuous expression model of environmental factors.

[0130] The grid determination module 430 is used to determine whether a grid is a grid to be divided based on the vertex density of each vertex and multiple environmental factors for any grid within the target area.

[0131] The mesh generation module 440 is used to further divide the mesh to be divided based on the longest axis bisection strategy.

[0132] Optionally, the mesh determination module 430 includes:

[0133] The importance calculation submodule is used to determine the importance of any grid within the target area based on the vertex density of each vertex and multiple environmental factors.

[0134] The importance determination submodule is used to determine a grid as a grid to be divided if the importance of the grid is greater than or equal to the division threshold.

[0135] Optionally, the importance calculation submodule includes:

[0136] The first computing unit is used to determine the geometric complexity of any grid within the target area based on preset geometric weights and the sum of vertex densities of all vertices within the grid.

[0137] The second calculation unit is used to determine the environmental complexity of the grid based on the multivariate environmental factors of each vertex and the environmental weights corresponding to each multivariate environmental factor.

[0138] The comprehensive computing unit is used to determine the importance of the mesh based on geometric complexity and environmental complexity.

[0139] Optionally, the mesh generation module 440 is specifically used for:

[0140] Divide the mesh into two sub-mesh along the target axis, which refers to the direction of the longest axis among the three axes of length, width, and height of the mesh;

[0141] The device also includes:

[0142] The local coding construction module is used to determine the local coding of a sub-mesh based on its axis position and sub-block position. The axis position refers to the coding representation corresponding to the target axis, and the sub-block position refers to the coding representation corresponding to the partition position of the sub-mesh on the target axis. The partition position can be one of left-right partition, top-bottom partition, or front-back partition.

[0143] The global encoding construction module is used to concatenate the local encoding of the sub-mesh and the local encoding of the parent mesh node to which the sub-mesh belongs in a hierarchical manner to obtain the global encoding of the sub-mesh. The parent mesh node refers to the mesh related to the sub-mesh division process.

[0144] Optionally, the continuous expression model of environmental factors includes at least a location encoding module and a multilayer perceptron regression module; the multivariate environmental factor prediction module 420 includes:

[0145] The input submodule is used to input the vertex coordinates of each vertex into the position encoding module;

[0146] The high-dimensional transformation submodule is used in the position encoding module to map the vertex coordinates of each vertex to a high-dimensional space through a random Gaussian matrix, and then to obtain a high-dimensional feature vector through sine and cosine function transformations.

[0147] The regression prediction submodule is used to input high-dimensional feature vectors into the multilayer perceptron regression module, perform nonlinear transformations through multiple hidden layers, and then obtain the multivariate environmental factors corresponding to the vertex coordinates of each vertex through the output layer regression.

[0148] The output submodule is used to output the multivariate environment factors corresponding to the vertex coordinates of each vertex.

[0149] Optionally, the vertex density acquisition module 410 includes:

[0150] The coordinate unification submodule is used to project multi-source heterogeneous 3D data onto the same coordinate system to obtain the initial data of the target area.

[0151] The data cleaning submodule is used to clean the initial data to obtain valid data.

[0152] The density calculation submodule is used to determine the vertex density of each vertex within the target area based on valid data.

[0153] Optionally, the valid data includes point feature data, polygon model data, and 3D model data; the density calculation submodule includes:

[0154] The first type of density calculation unit is used for: determining the volume of the axis-aligned outer bounding box of the surface model data and the 3D model data; counting the number of original vertices contained in the surface model data and the 3D model data; determining the original vertex density based on the volume of the axis-aligned outer bounding box and the number of original vertices; taking the corner points of the axis-aligned outer bounding box as representative vertices of the surface model data and the 3D model data, and determining the vertex density of the representative vertices based on the original vertex density;

[0155] The second type of density calculation unit is used to: determine the vertex density of point feature data to be one for point feature data;

[0156] The vertex density statistics unit is used to: obtain the vertex density of each vertex in the target area based on the vertex density of representative vertices of all face model data and 3D model data, and the vertex density of all point feature data within the target area.

[0157] In practice, each of the above units can be implemented as an independent entity or can be arbitrarily combined to be implemented as the same or several entities. For the specific implementation of each of the above units, please refer to the previous method embodiments, which will not be repeated here.

[0158] In the embodiments of this application, the terms "module" or "unit" refer to a computer program or part of a computer program that has a predetermined function and works with other related parts to achieve a predetermined goal, and can be implemented wholly or partially using software, hardware (such as processing circuitry or memory), or a combination thereof. Similarly, a processor (or multiple processors or memory) can be used to implement one or more modules or units. Furthermore, each module or unit can be part of an overall module or unit that includes the functionality of that module or unit.

[0159] This application also provides an electronic device, which can be a terminal, a server, or other similar device. The terminal can be a tablet computer, a smart Bluetooth device, a laptop computer, a personal computer, etc.; the server can be a single server or a server cluster composed of multiple servers, etc.

[0160] In some embodiments, the spatial grid adaptive partitioning device can also be integrated into multiple electronic devices. For example, the spatial grid adaptive partitioning device can be integrated into multiple servers, and the spatial grid adaptive partitioning method of this application can be implemented by multiple servers.

[0161] In this embodiment, as Figure 5 As shown, it illustrates a structural schematic diagram of the electronic device involved in the embodiments of this application, specifically:

[0162] The electronic device may include components such as a processor 510 with one or more processing cores, a memory 520 with one or more computer-readable storage media, a power supply 530, an input module 540, and a communication module 550. Those skilled in the art will understand that... Figure 5 The electronic device structure shown does not constitute a limitation on the electronic device and may include more or fewer components than shown, or combine certain components, or have different component arrangements. Wherein:

[0163] The processor 510 is the control center of the electronic device. It connects various parts of the electronic device via various interfaces and lines, and performs various functions and processes data by running or executing software programs and / or modules stored in the memory 520, and by calling data stored in the memory 520, thereby performing overall detection of the electronic device. In some embodiments, the processor 510 may include one or more processing cores; in some embodiments, the processor 510 may integrate an application processor and a modem processor, wherein the application processor mainly handles the operating system, user interface, and applications, and the modem processor mainly handles wireless communication. It is understood that the modem processor may also not be integrated into the processor 510.

[0164] The memory 520 can be used to store software programs and modules. The processor 510 executes various functional applications and data processing by running the software programs and modules stored in the memory 520. The memory 520 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, application programs required for at least one function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the electronic device, etc. In addition, the memory 520 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device. Accordingly, the memory 520 may also include a memory controller to provide the processor 510 with access to the memory 520.

[0165] The electronic device also includes a power supply 530 that supplies power to the various components. In some embodiments, the power supply 530 can be logically connected to the processor 510 through a power management system, thereby enabling functions such as managing charging, discharging, and power consumption through the power management system. The power supply 530 may also include one or more DC or AC power supplies, recharging systems, power fault detection circuits, power converters or inverters, power status indicators, and other arbitrary components.

[0166] The electronic device may also include an input module 540, which can be used to receive input digital or character information and generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function control.

[0167] The electronic device may also include a communication module 550. In some embodiments, the communication module 550 may include a wireless module, through which the electronic device can perform short-range wireless transmission, thereby providing users with wireless broadband internet access. For example, the communication module 550 can be used to help users send and receive emails, browse web pages, and access streaming media.

[0168] Although not shown, the electronic device may also include a display unit, etc., which will not be described in detail here. Specifically, in this embodiment, the processor 510 in the electronic device loads the executable files corresponding to the processes of one or more applications into the memory 520 according to the following instructions, and the processor 510 runs the applications stored in the memory 520, thereby realizing the various functions in the above-mentioned spatial grid adaptive partitioning method.

[0169] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be performed by instructions, or by instructions controlling related hardware. These instructions can be stored in a computer-readable storage medium and loaded and executed by a processor.

[0170] Therefore, embodiments of this application provide a computer-readable storage medium storing a plurality of instructions that can be loaded by a processor to execute steps in any of the spatial grid adaptive partitioning methods provided in embodiments of this application.

[0171] The storage medium may include: read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0172] According to one aspect of this application, a computer program product or computer program is provided, comprising computer instructions stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the electronic device to perform the spatial grid adaptive partitioning method provided in the above embodiments.

[0173] Since the instructions stored in the storage medium can execute the steps of any of the spatial grid adaptive partitioning methods provided in the embodiments of this application, the beneficial effects that any of the spatial grid adaptive partitioning methods provided in the embodiments of this application can achieve can be realized. For details, please refer to the previous embodiments, which will not be repeated here.

[0174] The above provides a detailed description of a spatial grid adaptive partitioning method, apparatus, device, and medium provided in the embodiments of this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A spatial grid adaptive partitioning method, characterized in that, The method includes: Acquire multi-source heterogeneous 3D data of the target region, and preprocess the multi-source heterogeneous 3D data to obtain the vertex density of each vertex in the target region; Based on the vertex coordinates of each vertex, the multivariate environmental factors corresponding to each vertex are predicted based on the pre-trained continuous expression model of environmental factors. The multivariate environmental factors refer to a variety of physical or environmental parameters that affect low-altitude flight applications within the target area, reflecting the three-dimensional spatial environmental conditions of the target area. The multivariate environmental factors include wind speed, temperature, light intensity, magnetic field strength, and air pollution index. For any grid within the target region, the geometric complexity of the grid is determined based on a preset geometric weight and the sum of the vertex densities of all vertices within the grid. The environmental complexity of the grid is determined based on the multivariate environmental factors of each vertex in the grid and the environmental weights corresponding to each multivariate environmental factor. The importance of the mesh is determined based on the geometric complexity and the environmental complexity. If the importance of the grid is greater than or equal to the partitioning threshold, then the grid is determined to be a grid to be partitioned; The grid to be divided is divided into two sub-grids along the target axis, where the target axis refers to the direction of the longest axis among the length, width, and height axes of the grid to be divided. For any of the sub-mesh, a step is performed to determine the geometric complexity of the mesh based on a preset geometric weight and the sum of the vertex densities of each vertex within the mesh, until the importance of the sub-mesh is less than a set subdivision threshold, at which point the subdivision stops.

2. The spatial grid adaptive partitioning method as described in claim 1, characterized in that, The method further includes: Based on the axis position and sub-block position of the sub-mesh, the local code of the sub-mesh is determined. The axis position refers to the code representation corresponding to the target axis, and the sub-block position refers to the code representation corresponding to the partition position of the sub-mesh on the target axis. The partition position can be one of left-right partition, top-bottom partition, or front-back partition. The local code of the sub-mesh and the local code of the parent mesh node to which the sub-mesh belongs are concatenated hierarchically to obtain the global code of the sub-mesh. The parent mesh node refers to the mesh related to the sub-mesh division process.

3. The spatial grid adaptive partitioning method as described in claim 1, characterized in that, The continuous expression model of environmental factors includes at least a location encoding module and a multilayer perceptron regression module; Based on the vertex coordinates of each vertex, the multivariate environmental factors corresponding to each vertex are predicted using a pre-trained continuous representation model of environmental factors, including: Input the vertex coordinates of each vertex into the position encoding module; In the position encoding module, the vertex coordinates of each vertex are mapped to a high-dimensional space through a random Gaussian matrix, and then transformed by sine and cosine functions to obtain a high-dimensional feature vector; The high-dimensional feature vector is input into the multilayer perceptron regression module, and then undergoes nonlinear transformation through multiple hidden layers. Finally, the multivariate environmental factors corresponding to the vertex coordinates of each vertex are obtained through regression through the output layer. Output the multivariate environmental factors corresponding to the vertex coordinates of each vertex.

4. A spatial grid adaptive partitioning method as described in any one of claims 1-3, characterized in that, The multi-source heterogeneous 3D data is preprocessed to obtain the vertex density of each vertex within the target region, including: The multi-source heterogeneous 3D data is projected onto the same coordinate system to obtain the initial data of the target region; The initial data is cleaned to obtain valid data; Based on the valid data, the vertex density of each vertex within the target region is determined.

5. The spatial grid adaptive partitioning method as described in claim 4, characterized in that, The valid data includes point feature data, surface model data, and 3D model data; Based on the valid data, the vertex density of each vertex within the target region is determined, including: For the surface model data and the 3D model data, determine the volume of the axis-aligned outer bounding box of the surface model data and the 3D model data; The number of original vertices contained in the surface model data and the 3D model data is counted. The original vertex density is determined based on the volume of the axis-aligned outer bounding box and the original number of vertices; The corner points of the outer bounding box aligned with the axis are used as representative vertices of the face model data and the 3D model data, and the vertex density of the representative vertices is determined based on the original vertex density. For the point feature data, the vertex density of the point feature data is determined to be one; Based on the vertex density of the representative vertices of all the surface model data and the 3D model data within the target area, and the vertex density of all the point feature data, the vertex density of each vertex within the target area is obtained.

6. A spatial grid adaptive partitioning device, characterized in that, The device includes: The vertex density acquisition module is used to acquire multi-source heterogeneous 3D data of the target region and preprocess the multi-source heterogeneous 3D data to obtain the vertex density of each vertex in the target region. The multivariate environmental factor prediction module is used to predict the multivariate environmental factors corresponding to each vertex based on the vertex coordinates of each vertex and a pre-trained continuous expression model of environmental factors. The multivariate environmental factors refer to a variety of physical or environmental parameters that affect low-altitude flight applications within the target area, reflecting the three-dimensional environmental conditions of the target area. The multivariate environmental factors include wind speed, temperature, light intensity, magnetic field strength, and air pollution index. The first computing unit is used to determine the geometric complexity of any grid within the target area based on preset geometric weights and the sum of vertex densities of all vertices within the grid. The second calculation unit is used to determine the environmental complexity of the grid based on the multivariate environmental factors of each vertex and the environmental weights corresponding to each multivariate environmental factor. A comprehensive computing unit is used to determine the importance of the mesh based on geometric complexity and environmental complexity; The importance determination submodule is used to determine a grid as a grid to be divided if the importance of the grid is greater than or equal to the division threshold. The mesh generation module is used to divide the mesh into two sub-mesh along the target axis, which refers to the direction of the longest axis among the three axes of length, width, and height of the mesh. For any sub-mesh, the module performs a step of determining the geometric complexity of the mesh based on preset geometric weights and the sum of vertex densities of all vertices in the mesh, until the importance of the sub-mesh is less than a set division threshold, at which point the subdivision stops.

7. An electronic device, characterized in that, It includes a processor and a memory, the memory storing multiple instructions; the processor loads instructions from the memory to execute the steps in a spatial grid adaptive partitioning method as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a plurality of instructions adapted for loading by a processor to perform the steps in a spatial grid adaptive partitioning method as described in any one of claims 1 to 5.

Images