Warehouse three-dimensional model modularization construction method and system based on digital twinning

Abstract

The invention relates to the technical field of three-dimensional modeling, in particular to a warehouse three-dimensional model modularization construction method and system based on digital twinning, and the method comprises the steps: dividing a three-dimensional space into a plurality of grid units according to the distribution density of material point location data; setting an assembly priority for each three-dimensional component according to the support dependency relationship; initializing the wave function states of the grid units, calculating the Shannon entropy of each grid unit, and performing weighted correction on the Shannon entropy by using the assembly priority to obtain a weighted entropy value; the grid unit with the minimum weighted entropy value is selected for state collapse so as to determine the grid unit as a specific three-dimensional component; and updating the feasible state set of the adjacent grid units according to an adjacency constraint rule until the states of all the grid units are determined, and generating a warehouse three-dimensional model. According to the technical scheme, the warehouse three-dimensional model which accords with the physical law and is high in precision can be rapidly generated.

Description

Technical Field

[0001] This application relates to the field of 3D modeling technology, and in particular to a modular construction method and system for 3D warehouse models based on digital twins. Background Technology

[0002] With the development of smart grids and digital twin technologies, the management and control of electrical materials in warehouses is becoming increasingly intelligent. As a key link in achieving visualized management and control of materials, constructing high-fidelity 3D models of warehouses is of great significance for improving warehouse management efficiency, optimizing space utilization, and ensuring operational safety. Therefore, how to automatically construct 3D models of large-scale warehouse scenarios while ensuring model accuracy has become an urgent problem to be solved.

[0003] The existing process of building 3D models often requires collecting image data of the warehouse or using LiDAR to scan the warehouse to obtain point cloud data, thereby reconstructing the 3D model of the warehouse.

[0004] However, both laser scanning and image data acquisition rely on expensive acquisition equipment, and the massive point cloud or image data acquired requires complex processing, making it difficult to meet the needs of rapid model updates under frequent changes in warehouse materials, and thus unable to quickly and accurately construct 3D warehouse models. Summary of the Invention

[0005] To address the technical problem of the inability to quickly and accurately construct 3D warehouse models, this application provides a modular construction method and system for 3D warehouse models based on digital twins, which can quickly generate 3D warehouse models that conform to physical laws and have high accuracy.

[0006] In a first aspect, this application provides a method for modularly constructing a 3D warehouse model based on digital twins. The method includes: acquiring material location data of the warehouse; dividing the 3D space into multiple grid cells according to the distribution density of the material location data, wherein the size of the grid cell is negatively correlated with the distribution density; establishing a component model library containing multiple 3D components; defining adjacency constraint rules between each 3D component; and setting assembly priority for each 3D component based on support dependencies; initializing the wavefunction state of the grid cell according to the wavefunction collapse algorithm; calculating the Shannon entropy of each grid cell; and using the assembly priority to weight and correct the Shannon entropy to obtain a weighted entropy value; selecting the grid cell with the smallest weighted entropy value for state collapse; determining the grid cell as a specific 3D component based on the material location data; updating the feasible state set of adjacent grid cells according to the adjacency constraint rules; repeating the state collapse and update steps until the state of all grid cells is determined; and instantiating each 3D component to generate a 3D warehouse model.

[0007] By generating mesh cells whose size is negatively correlated with density based on the distribution density of material location data, adaptive discretization in three-dimensional space is achieved, significantly reducing the total number of meshes while ensuring assembly accuracy. At the same time, by defining adjacency constraint rules and assembly priorities, and using assembly priority to weighted correct Shannon entropy, the physical support logic corresponding to the assembly priority is followed during wave function collapse, fundamentally eliminating topological errors such as component suspension and ensuring the physical authenticity of the generated model.

[0008] Preferably, dividing the three-dimensional space into multiple grid units based on the distribution density of material location data includes: dividing the three-dimensional space into multiple macro regions, counting the number of material location data in each macro region to construct a material density field; normalizing the material density field to obtain the distribution density of each location in the three-dimensional space; calculating the corresponding grid step size based on the distribution density of each location, and dividing each macro region according to the grid step size to generate non-uniform grid units.

[0009] The density of material distribution at various locations in three-dimensional space was quantified, providing accurate data support for the subsequent generation of non-uniform meshes. Then, the mesh step size was calculated based on the distribution density, realizing a smooth transition of mesh size with material density, ensuring that details can be captured in dense material areas, while effectively reducing data redundancy in sparse areas.

[0010] Preferably, the method for calculating the grid step size includes: setting a basic grid step size and a density sensitivity coefficient; using the basic grid step size as the numerator, the sum of the products of 1 and the distribution density and the density sensitivity coefficient as the denominator, the ratio of the two as the first step size, and the maximum value between the first step size and the preset minimum step size as the grid step size.

[0011] Preferably, the step of setting assembly priority for each three-dimensional component based on support dependencies includes: the three-dimensional components include basic components, skeleton components, structural components, and material components; the assembly priority of the basic components is 0, including the ground and walls; the assembly priority of the skeleton components is 1, including shelf uprights and stacker crane tracks; the assembly priority of the structural components is 2, including shelf beams and shelves; and the assembly priority of the material components is 3, including pallets and goods.

[0012] Preferably, the step of using assembly priority to weight the Shannon entropy to obtain a weighted entropy value includes: calculating a priority decay coefficient, wherein the priority decay coefficient is negatively correlated with the assembly priority of the three-dimensional component with the highest probability in the candidate component set of the mesh cell; and multiplying the Shannon entropy of the mesh cell by the priority decay coefficient to obtain a weighted entropy value.

[0013] By using assembly priority to correct Shannon entropy, the weighted entropy value of basic components with lower assembly priority is significantly reduced. Under the minimum entropy priority selection mechanism of wave function collapse algorithm, the state of basic components such as ground and columns is ensured to be processed and determined first, thereby simulating the real physical construction process and ensuring the structural stability of the generated model.

[0014] Preferably, initializing the wavefunction state of the grid cell according to the wavefunction collapse algorithm includes: determining whether there is material location data in the grid cell; in response to the existence of material location data, setting the initial existence probability of the corresponding material component in the grid cell to a first probability value, wherein the first probability value is greater than the sum of the probability values ​​of the other feasible components; in response to the absence of material location data, maintaining the grid cell in a fully superimposed state, wherein the probability of each three-dimensional component other than the material component is the same in the fully superimposed state.

[0015] Preferably, during the repeated execution of the state collapse and update steps, the construction method further includes a dynamic subdivision step: calculating the spatial change rate of Shannon entropy between the current grid cell and its adjacent grid cells to obtain the conflict gradient; in response to the conflict gradient satisfying a preset subdivision condition, splitting the current grid cell into multiple sub-grid cells; and re-executing the state collapse and update steps within the scope of the sub-grid cells.

[0016] Dynamic subdivision of mesh cells is introduced during state collapse and update. The spatial change rate of Shannon entropy is used to identify conflict regions with complex topological structures. Mesh splitting and re-collapse are performed only on these regions, avoiding the huge computational overhead caused by global subdivision. At the same time, it can automatically capture the geometric edges and complex interfaces of components, significantly improving the detail of key parts of the model.

[0017] Preferably, the method for calculating the conflict gradient includes: obtaining the maximum Shannon entropy of all grid cells in the initial iteration, normalizing the Shannon entropy of the current grid cell using the maximum Shannon entropy; calculating the difference between the normalized entropy values ​​of the current grid cell and its neighboring grid cells in each direction of the three-dimensional space, calculating the ratio of the difference to the grid cell size, and taking the average of all ratios as the conflict gradient.

[0018] By utilizing the concept of image gradient, the degree of spatial change in state determinism is quantified, enabling precise identification of the boundary region between deterministic and uncertain states, thus providing an objective and accurate basis for dynamic grid subdivision.

[0019] Preferably, the preset subdivision conditions include: the conflict gradient of the current grid cell is greater than a preset gradient threshold, the normalized entropy value of the current grid cell is greater than a preset uncertainty threshold, and the number of candidate components in the current grid cell is greater than or equal to 2.

[0020] In a second aspect, this application also provides a modular construction system for a 3D warehouse model based on digital twins, including a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the modular construction method for a 3D warehouse model based on digital twins according to the first aspect of this application.

[0021] The technical solution of this application has the following beneficial technical effects: By adaptively dividing non-uniform grid cells based on the distribution density of material location data, small-scale grids are used in densely populated areas to ensure model accuracy, while large-scale grids are used in open areas to reduce computational load, thus balancing modeling accuracy and computational efficiency. Furthermore, by defining adjacency constraint rules and assembly priorities between components, and using assembly priorities to weight and correct the Shannon entropy in the wavefunction collapse algorithm, the algorithm is guided to prioritize the determination of basic support components such as the ground and columns, ensuring that the generated warehouse 3D model conforms to physical gravity logic and avoiding errors such as component suspension. Finally, by introducing a dynamic subdivision mechanism based on the spatial change rate of Shannon entropy, grid subdivision is performed only in conflict areas with complex topological structures, achieving high-fidelity capture of key interface details. Ultimately, relying on a pre-established component model library and material location data, a warehouse 3D digital twin model that conforms to physical laws and has high accuracy is automatically generated. Attached Figure Description

[0022] Figure 1 This is a flowchart of a modular construction method for a 3D warehouse model based on digital twins, according to an embodiment of this application.

[0023] Figure 2 This is a schematic diagram of a shelf column in the component model library according to an embodiment of this application.

[0024] Figure 3 This is a schematic diagram of a tray in the component model library according to an embodiment of this application.

[0025] Figure 4 This is a schematic diagram of a three-dimensional model of a warehouse according to an embodiment of this application.

[0026] Figure 5 This is a structural block diagram of a modular construction system for a warehouse 3D model based on digital twins, according to an embodiment of this application. Detailed Implementation

[0027] According to a first aspect of this application, this application provides a component-based construction method for a 3D warehouse model based on digital twins. Figure 1 This is a flowchart illustrating a component-based method for constructing a 3D warehouse model using digital twins, according to an embodiment of this application. Figure 1As shown, the modular construction method for the 3D warehouse model based on digital twins includes steps S101 to S104, which are described in detail below.

[0028] S101, acquire material location data of the warehouse, and divide the three-dimensional space into multiple grid units according to the distribution density of the material location data. The size of the grid unit is negatively correlated with the distribution density.

[0029] In one embodiment, material location data refers to the set of coordinates recording the storage locations of materials and equipment in the power material warehouse, which typically originates from ledger records in the material management system or CAD design drawings. Distribution density is used to characterize the density of materials existing in a local area within three-dimensional space.

[0030] To resolve the trade-off between accuracy and computational efficiency brought about by a uniform mesh in large-scale warehouse scenarios, this step adopts an adaptive discretization strategy. In areas with dense materials, such as shelving areas, smaller mesh cells are used to ensure assembly accuracy; in open areas, such as aisles, larger mesh cells are used to reduce computational load.

[0031] Specifically, the step of dividing the three-dimensional space into multiple grid units based on the distribution density of material location data includes: dividing the three-dimensional space into multiple macro regions, counting the number of material location data in each macro region to construct a material density field; normalizing the material density field to obtain the distribution density of each location in the three-dimensional space; calculating the corresponding grid step size based on the distribution density of each location, and dividing each macro region based on the grid step size to generate non-uniform grid units.

[0032] The macroscopic region is a block larger than the final grid cell scale. For example, a warehouse can be divided into blocks of 10 meters × 10 meters × 10 meters. By counting the number of points within each block and normalizing the data, a distribution density with values ​​between 0 and 1 is obtained.

[0033] To accurately establish the negative correlation between grid size and density, the method for calculating the grid step size includes: setting a basic grid step size and a density sensitivity coefficient; using the basic grid step size as the numerator, the sum of the products of 1 and the distribution density and the density sensitivity coefficient as the denominator, the ratio of the two as the first step size, and the maximum value between the first step size and the preset minimum step size as the grid step size.

[0034] The grid step size satisfies the following relationship: ; in, Macroscopic region in three-dimensional space The grid step size at that point represents the macroscopic region. Internal grid size; The base grid step size, exemplified by a value of 2 meters, represents the maximum grid size for an open area; This is the density sensitivity coefficient, used to control the mesh's sensitivity to density; for example, it is set to 5. For macro regions Distribution density at location; The minimum grid step size can be 0.5 meters.

[0035] Understandably, when the distribution density at a location is high, the denominator increases significantly, resulting in a smaller calculated grid step size, for example, reduced to about 1 meter, thereby achieving high-precision fine meshing; conversely, when the density is close to 0, the grid step size remains the basic step size.

[0036] In this way, adaptive mesh generation in three-dimensional space is achieved, which significantly reduces the total number of meshes while ensuring the assembly accuracy of core equipment, thus solving the problem of large computational load in large-scale warehouse scenarios.

[0037] S102, establish a component model library containing multiple three-dimensional components, define the adjacency constraint rules between each three-dimensional component, and set the assembly priority for each three-dimensional component according to the support dependency relationship.

[0038] In one embodiment, the component model library contains basic components for building a 3D model of a warehouse, and each basic component is a 3D component. These basic components are basic units that can be processed by the wave function collapse algorithm, and are used to realize the component-based construction of the 3D model of the warehouse.

[0039] The adjacency constraint rule refers to the interface type defined for the six faces (top, bottom, left, right, front, and back) of each three-dimensional component. Only components with matching interface types can be adjacent in space. For example, the upper surface interface of a shelf beam can only be connected to the lower surface interface of a pallet. This constraint is used to constrain the assembly relationship of each three-dimensional component.

[0040] To prevent the generation of models that do not conform to the logic of physical gravity, such as generating models with basic components suspended in mid-air, an assembly priority is introduced. Specifically, setting the assembly priority for each 3D component based on support dependencies includes: the 3D components include basic components, skeleton components, structural components, and material components; the assembly priority of the basic components is 0, including the ground and walls; the assembly priority of the skeleton components is 1, including shelf columns and stacker crane tracks; the assembly priority of the structural components is 2, including shelf beams and shelves; and the assembly priority of the material components is 3, including pallets and goods.

[0041] Please see Figure 2 This is a schematic diagram of a shelf column in the component model library according to an embodiment of this application. It belongs to a type of skeleton component. Please refer to [link / reference]. Figure 3This is a schematic diagram of a pallet in the component model library according to the embodiments of this application. It belongs to a type of material component. The assembly priority of the shelf column is 2, while the assembly priority of the pallet is 3.

[0042] Understandably, the smaller the assembly priority value, the more fundamental the 3D component is in the physical structure and the higher it should be prioritized. For example, the ground must be in place before the columns can be installed, then the beams, and finally the pallets.

[0043] In this way, by digitally describing the connection logic and priority labels between 3D components, adjacency constraint rules and assembly priorities that must be followed are provided to prevent the generation of erroneous models that do not conform to the logic of physical gravity.

[0044] S103: Initialize the wave function state of the grid cells according to the wave function collapse algorithm, calculate the Shannon entropy of each grid cell, and use the assembly priority to perform weighted correction on the Shannon entropy to obtain the weighted entropy value.

[0045] In one embodiment, wave function state initialization refers to setting each mesh cell to a superposition state of all possible components. A superposition state is a state in which a mesh cell can be a variety of three-dimensional components.

[0046] To fully utilize the original business data, the wave function state of the grid cell is initialized according to the wave function collapse algorithm, including: determining whether there is material location data in the grid cell; in response to the existence of material location data, setting the initial existence probability of the corresponding material component in the grid cell to a first probability value, the first probability value being greater than the sum of the probability values ​​of the other feasible components; in response to the absence of material location data, maintaining the grid cell in a fully superimposed state, in which the probability of each three-dimensional component other than the material component is the same.

[0047] For example, the first probability value can be set to 1; in this way, the algorithm can be guided to generate materials first at the recorded locations, and in the unrecorded locations, they are in a superposition state, waiting for subsequent collapse calculations.

[0048] In standard wavefunction collapse algorithms, the mesh with the minimum Shannon entropy is typically selected for collapse. Shannon entropy reflects the uncertainty of the mesh state. To ensure the generation of a warehouse 3D model that conforms to physical gravity logic, the Shannon entropy is weighted and corrected using assembly priority to obtain a weighted entropy value. This includes: calculating a priority decay coefficient, which is negatively correlated with the assembly priority of the 3D component with the highest probability in the candidate component set of the mesh cell; and multiplying the Shannon entropy of the mesh cell by the priority decay coefficient to obtain the weighted entropy value.

[0049] Grid cells weighted entropy value The calculation formula is as follows: ; in, For grid cells Correct the previous Shannon entropy; For grid cells The assembly priority of the 3D component with the highest probability in the current candidate component set; The attenuation sensitivity is defined as a value ranging from 0 to 1, and in this embodiment, the attenuation sensitivity is defined as 1. The assembly priority is the maximum value, which is set to 3 in this embodiment.

[0050] The priority attenuation coefficient ranges from 0 to 1. If the mesh element... Assembly priority of the 3D component with the highest probability in the current candidate component set The smaller the value, the higher the priority for assembly; in this case, the corresponding priority attenuation coefficient... The smaller the size, the smaller the mesh cells. The lower the final weighted entropy value, the better.

[0051] In this way, the wave function collapse algorithm selects the grid with the smallest weighted entropy value for collapse. At this time, the grid cells of the basic support structure will be processed and determined first, that is, the basic components such as the ground and walls with a priority of 0 will be assembled, which fundamentally eliminates the phenomenon of components being suspended.

[0052] S104: Select the grid cell with the smallest weighted entropy value for state collapse, and determine the grid cell as a specific three-dimensional component based on the material location data; update the feasible state set of adjacent grid cells according to the adjacency constraint rules, and repeat the state collapse and update steps until the state of all grid cells is determined, and then instantiate each three-dimensional component to generate the warehouse three-dimensional model.

[0053] In one embodiment, in the wave function collapse algorithm, state collapse refers to determining a grid cell in a superposition state as a single component; the feasible state set is the set of candidate components for the grid cell, including three-dimensional components that can be placed, and each three-dimensional component corresponds to a probability value; updating the feasible state set of neighboring grid cells means that once a grid cell determines its corresponding three-dimensional component, candidate components in its neighboring grids that do not conform to the adjacency rules are immediately removed.

[0054] To address the issue that the current mesh cannot accurately represent complex interface details, the construction method further includes a dynamic subdivision step during the repeated execution of state collapse and update steps: calculating the spatial change rate of Shannon entropy between the current mesh cell and its adjacent mesh cells to obtain the conflict gradient; in response to the conflict gradient satisfying a preset subdivision condition, splitting the current mesh cell into multiple sub-mesh cells; and re-executing the state collapse and update steps within the scope of the sub-mesh cells.

[0055] Specifically, the method for calculating the conflict gradient includes: obtaining the maximum Shannon entropy of all grid cells in the initial iteration, normalizing the Shannon entropy of the current grid cell using the maximum Shannon entropy; calculating the difference between the normalized entropy values ​​of the current grid cell and its neighboring grid cells in each direction of the three-dimensional space, calculating the ratio of the difference to the grid cell size, and taking the average of all ratios as the conflict gradient.

[0056] Using the current grid cell as the grid cell For example, grid cells Conflict gradient The satisfying relation is: ; In the formula, For grid cells A set of adjacent grid cells in all directions. Represents any adjacent grid cell, The number of adjacent grid cells. Adjacent grid cells The normalized entropy value, For grid cells The normalized entropy value, and For grid cells and adjacent grid cells The grid size.

[0057] The conflict gradient, borrowing the concept of image gradient, is used to identify spatial boundaries where state determinism changes drastically. If the entropy values ​​of neighboring grid cells around the current grid cell are similar, the conflict gradient of the current grid cell is close to 0, indicating that the current grid cell has a gentle state and does not require subdivision. If one side is a definite pillar and the other side is completely unknown air, the conflict gradient of the current grid cell will be very high, indicating that the current grid cell has complex interface connections or topological conflicts and requires subdivision.

[0058] It should be noted that the preset subdivision conditions include: the conflict gradient of the current grid cell is greater than a preset gradient threshold, the normalized entropy value of the current grid cell is greater than a preset uncertainty threshold, and the number of candidate components in the current grid cell is greater than or equal to 2. For example, the gradient threshold can be set to 1.5, and the uncertainty threshold can be set to 0.3. This ensures that subdivision is only performed in truly conflicting regions, i.e., grid cells with uncertain states and many solutions, avoiding unnecessary computational overhead. Specifically, the current grid can be split into... A smaller subgrid is created to further subdivide the current grid.

[0059] The state collapse and update steps are continuously executed until the state of each mesh cell is determined, i.e., the 3D components within each mesh cell are identified. At this point, the 3D components and their orientation information after each mesh collapse are read. The corresponding 3D components are then called from the component model library, instantiated at the corresponding coordinate positions, and rendered and optimized to obtain the complete 3D model of the warehouse. Please refer to [link to relevant documentation]. Figure 4 This is a schematic diagram of a three-dimensional warehouse model according to an embodiment of this application.

[0060] In this way, by subdividing the mesh, the model can automatically capture the geometric edges and complex interfaces of the 3D components. While ensuring the realism of key parts, it avoids the huge computational overhead caused by global subdivision, and finally generates a realistic and accurate 3D model of the warehouse.

[0061] According to a second aspect of this application, this application also provides a modular construction system for a 3D warehouse model based on digital twins. Figure 5 This is a structural block diagram of a modular construction system for a 3D warehouse model based on digital twins, according to an embodiment of this application. Figure 5 As shown, the system 50 includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement a modular construction method for a warehouse 3D model based on digital twins according to the first aspect of this application. The system also includes other components well-known to those skilled in the art, such as a communication bus and a communication interface. Their configurations and functions are known in the art and will not be described further here.

[0062] It should be noted that, for those skilled in the art, several modifications and improvements can be made without departing from the concept of this application, and these all fall within the scope of protection of this application.

Claims

1. A component-based construction method for a 3D warehouse model based on digital twins, characterized in that, The construction method includes: acquiring material location data of the warehouse, dividing the three-dimensional space into multiple grid units according to the distribution density of the material location data, wherein the size of the grid unit is negatively correlated with the distribution density; Establish a component model library containing multiple 3D components, define the adjacency constraint rules between each 3D component, and set the assembly priority for each 3D component based on the support dependency relationship; The wave function state of the grid cells is initialized according to the wave function collapse algorithm, the Shannon entropy of each grid cell is calculated, and the Shannon entropy is weighted and corrected using assembly priority to obtain the weighted entropy value. The grid cell with the smallest weighted entropy value is selected for state collapse. Based on the material location data, this grid cell is determined as a specific three-dimensional component. The feasible state set of adjacent grid cells is updated according to the adjacency constraint rules. The state collapse and update steps are repeated until the state of all grid cells is determined. Then, each three-dimensional component is instantiated to generate the three-dimensional model of the warehouse.

2. The modular construction method for a warehouse 3D model based on digital twins according to claim 1, characterized in that, The method of dividing the three-dimensional space into multiple grid units based on the distribution density of material location data includes: The three-dimensional space is divided into multiple macro regions, and the number of material point data in each macro region is counted to construct a material density field. The material density field is normalized to obtain the distribution density of each location in the three-dimensional space. The corresponding grid step size is calculated based on the distribution density of each location, and each macro region is divided according to the grid step size to generate non-uniform grid cells.

3. The modular construction method for a warehouse 3D model based on digital twins according to claim 2, characterized in that, The method for calculating the grid step size includes: Set the basic grid step size and density sensitivity coefficient; take the basic grid step size as the numerator, take the sum of 1 and the product of the distribution density and the density sensitivity coefficient as the denominator, take the ratio of the two as the first step size, and take the maximum value between the first step size and the preset minimum step size as the grid step size.

4. The modular construction method for a warehouse 3D model based on digital twins according to claim 1, characterized in that, The method of setting assembly priorities for each 3D component based on support dependencies includes: the 3D components include basic components, skeleton components, structural components, and material components; the assembly priority of the basic components is 0, including the ground and walls; the assembly priority of the skeleton components is 1, including shelf uprights and stacker crane tracks; the assembly priority of the structural components is 2, including shelf beams and shelves; and the assembly priority of the material components is 3, including pallets and goods.

5. The modular construction method for a warehouse 3D model based on digital twins according to claim 1, characterized in that, The step of using assembly priority to weight the Shannon entropy to obtain a weighted entropy value includes: Calculate the priority decay coefficient, which is negatively correlated with the assembly priority of the three-dimensional component with the highest probability in the candidate component set of the mesh cell; The weighted entropy value is obtained by multiplying the Shannon entropy of the grid cell by the priority decay coefficient.

6. The modular construction method for a warehouse 3D model based on digital twins according to claim 1, characterized in that, The wavefunction state of the mesh element is initialized according to the wavefunction collapse algorithm, including: Determine whether material location data exists within the grid cell; In response to the existence of material location data, the initial existence probability of the corresponding material component in the grid cell is set to a first probability value, which is greater than the sum of the probability values ​​of the other feasible components. In response to the absence of material location data, the grid cell is kept in a fully superimposed state, in which all three-dimensional components except material components have the same probability.

7. The modular construction method for a warehouse 3D model based on digital twins according to claim 1, characterized in that, The construction method further includes a dynamic subdivision step during the repeated execution of the state collapse and update steps: Calculate the spatial rate of change of Shannon entropy between the current grid cell and its neighboring grid cells to obtain the conflict gradient; in response to the conflict gradient satisfying the preset subdivision condition, split the current grid cell into multiple sub-grid cells; and re-execute the state collapse and update steps within the sub-grid cells.

8. The modular construction method for a warehouse 3D model based on digital twins according to claim 7, characterized in that, The method for calculating the conflict gradient includes: Obtain the maximum Shannon entropy of all grid cells in the first iteration, and normalize the Shannon entropy of the current grid cell using the maximum Shannon entropy; Calculate the difference between the normalized entropy values ​​of the current mesh cell and its neighboring mesh cells in each direction of the three-dimensional space, calculate the ratio of the difference to the mesh cell size, and take the mean of all ratios as the conflict gradient.

9. The component-based construction method for a warehouse 3D model based on digital twins according to claim 7, characterized in that, The preset subdivision conditions include: the conflict gradient of the current grid cell is greater than a preset gradient threshold, the normalized entropy value of the current grid cell is greater than a preset uncertainty threshold, and the number of candidate components in the current grid cell is greater than or equal to 2.

10. A modular construction system for warehouse 3D models based on digital twins, characterized in that, It includes a processor and a memory, the memory storing computer program instructions, which, when executed by the processor, implement the modular construction method for a warehouse 3D model based on digital twins according to any one of claims 1 to 9.

Images