A digital twin modeling method, device, equipment and storage medium

Abstract

The application provides a kind of digital twin modeling method, device, equipment and storage medium, it is related to digital twin modeling technical field, the method comprises: obtaining initial point cloud data;Obtain preprocessed point cloud data;According to the preprocessed point cloud data, obtain first point cloud feature information;Carry out object classification and segmentation processing, obtain second point cloud feature information;Generate interactive twin model;Obtain real-time state information, and update interactive twin model according to real-time state information.Processing.Through the generation of point cloud map, the convolution feature extraction processing based on point cloud data is realized, the global and local features of point cloud can be efficiently extracted, the efficiency and accuracy of modeling are improved, and high modeling accuracy is conducive to ensuring the accuracy of interactive twin model to accurately reflect the actual scene, while high modeling efficiency is conducive to realizing the synchronous change of interactive twin model and actual scene, improving the real-time of digital twin monitoring and interaction, and can meet the machine room monitoring and management demand.

Description

Technical Field

[0001] This invention relates to the field of digital twin modeling technology, and in particular to a digital twin modeling method, apparatus, device, and storage medium. Background Technology

[0002] As a crucial component of network infrastructure, data centers are responsible for the operation and maintenance of critical equipment. With the increasing scale and complexity of data centers, traditional two-dimensional blueprint design and manual management methods are showing limitations, failing to meet the demands for efficient design, refined management, and real-time monitoring. By constructing a three-dimensional model of the data center, various information elements are mapped into a virtual space, facilitating data center monitoring and management.

[0003] In existing technologies, 3D modeling techniques mainly rely on methods such as geometric modeling, image reconstruction, and voxel modeling. However, these methods have shortcomings when dealing with complex computer room scenes: geometric modeling relies on manually constructing geometric structures, which is suitable for regular objects, but it is difficult to guarantee the accuracy and integrity of the model when facing the complex environment of a computer room; image reconstruction methods rely on generating 3D models from images taken from multiple angles, but the reconstruction accuracy is unstable due to factors such as lighting and shooting angle, making it difficult to apply to complex computer room scenes; although voxel modeling methods can handle complex geometric shapes, the computation and storage costs increase significantly with the increase of scene complexity, resulting in low processing efficiency.

[0004] Therefore, existing 3D modeling of data center scenarios suffers from insufficient model accuracy and low modeling efficiency, failing to meet the requirements for high-precision monitoring and real-time interaction of data centers. Summary of the Invention

[0005] This invention provides a digital twin modeling method, apparatus, device, and storage medium to solve the problems of insufficient model accuracy and low modeling efficiency in the existing 3D modeling of computer room scenarios, which cannot meet the requirements for high-precision monitoring and real-time interaction of computer rooms.

[0006] This invention provides a digital twin modeling method, comprising:

[0007] Obtain initial point cloud data;

[0008] The initial point cloud data is preprocessed to obtain preprocessed point cloud data;

[0009] Based on the preprocessed point cloud data, a dynamic point cloud map is generated and convolutional feature extraction is performed on the dynamic point cloud map to obtain the first point cloud feature information.

[0010] Based on the first point cloud feature information, object classification and segmentation are performed to obtain the second point cloud feature information;

[0011] Based on the second point cloud feature information, an interactive twin model is generated;

[0012] Obtain real-time status information and update the interactive twin model based on the real-time status information.

[0013] According to a digital twin modeling method provided by the present invention, the step of obtaining first point cloud feature information by generating a dynamic point cloud map based on the preprocessed point cloud data and performing convolutional feature extraction processing on the dynamic point cloud map includes:

[0014] Based on the preprocessed point cloud data, determine the initial processed point cloud information;

[0015] Based on the processed point cloud information, a first preset number of nearest neighbor points are determined for each point to construct the dynamic point cloud map;

[0016] Based on the dynamic point cloud map, calculate the point feature information of each point;

[0017] The point feature information is aggregated to obtain point cloud update feature information;

[0018] Based on the point cloud update feature information, the processed point cloud information is updated to obtain new point cloud update feature information until the number of times the point cloud update feature information is obtained reaches a preset number. Then, the point cloud update feature information is subjected to global pooling processing to obtain the first point cloud feature information.

[0019] According to a digital twin modeling method provided by the present invention, the step of generating an interactive twin model based on the second point cloud feature information includes:

[0020] Based on the second point cloud feature information, generate triangular mesh model information;

[0021] The triangular mesh model information is optimized to obtain mesh optimization model information;

[0022] The interactive twin model is generated by rendering based on the mesh optimization model information.

[0023] According to a digital twin modeling method provided by the present invention, the step of generating triangular mesh model information based on the second point cloud feature information includes:

[0024] Based on the second point cloud feature information, a second preset number of nearest neighbor points are determined for each point to obtain point neighborhood information;

[0025] Based on the neighborhood information of the points, calculate the local plane and normal vector corresponding to each point to generate a triangular mesh surface;

[0026] The triangular mesh model is generated based on each of the triangular mesh surfaces.

[0027] According to a digital twin modeling method provided by the present invention, the step of optimizing the triangular mesh model information to obtain mesh optimization model information includes:

[0028] The triangular mesh model information is simplified to obtain the first mesh model information;

[0029] The first mesh model information is smoothed to obtain the second mesh model information;

[0030] The second mesh model information is subjected to detail enhancement processing to obtain the mesh optimization model information.

[0031] According to a digital twin modeling method provided by the present invention, after generating triangular mesh model information based on the second point cloud feature information, the method further includes:

[0032] Obtain texture information;

[0033] Based on the texture information, the triangular mesh model information is subjected to texture mapping processing.

[0034] According to a digital twin modeling method provided by the present invention, the step of preprocessing the initial point cloud data to obtain preprocessed point cloud data includes:

[0035] The initial point cloud data is globally registered based on a preset registration algorithm to obtain the first point cloud data.

[0036] The first point cloud data is denoised to obtain the second point cloud data;

[0037] The second point cloud data is downsampled to obtain the preprocessed point cloud data.

[0038] The present invention also provides a digital twin modeling apparatus, comprising:

[0039] The data acquisition module is used to acquire initial point cloud data;

[0040] The data preprocessing module is used to preprocess the initial point cloud data to obtain preprocessed point cloud data;

[0041] The feature extraction module is used to obtain first point cloud feature information by generating a dynamic point cloud map based on the preprocessed point cloud data and performing convolutional feature extraction processing on the dynamic point cloud map.

[0042] The 3D model construction module is used to perform object classification and segmentation based on the first point cloud feature information, obtain the second point cloud feature information, and generate an interactive twin model based on the second point cloud feature information.

[0043] The model optimization module is used to optimize the model.

[0044] A real-time rendering interaction module is used to generate display information and perform interactive operations based on the interaction twin model.

[0045] The model update module is used to acquire real-time status information and update the interactive twin model based on the real-time status information.

[0046] The data acquisition module, the data preprocessing module, the 3D model construction module, the model optimization module, the real-time rendering interaction module, and the model update module work together to implement the above-mentioned digital twin modeling method.

[0047] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement a digital twin modeling method as described above.

[0048] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a digital twin modeling method as described above.

[0049] This invention provides a digital twin modeling method, apparatus, device, and storage medium, which has at least the following beneficial effects: First, it acquires initial point cloud data, which accurately reflects the geometric shape and details of objects, providing a data foundation for precise modeling. Second, it preprocesses the initial point cloud data to remove redundant and abnormal data, reducing data processing volume and improving modeling efficiency. Third, it performs convolutional feature extraction based on the preprocessed point cloud data by generating a dynamic point cloud map to obtain first point cloud feature information, achieving point cloud-based feature extraction. Fourth, it performs object classification and segmentation based on the first point cloud feature information to identify and segment the point cloud features of different objects, obtaining second point cloud feature information. This facilitates subsequent object-based modeling and avoids object aggregation in the model. Fifth, it generates an interactive twin model based on the second point cloud feature information. The interactive twin model is updated according to real-time state information, reflecting changes in the real scene in real time, realizing the function of a digital twin. Through the interactive twin model, it is possible to monitor and interact with the actual scene based on a 3D model. Therefore, by generating point cloud maps, convolutional feature extraction based on point cloud data can be achieved, which can efficiently extract global and local features of point clouds, improve the efficiency and accuracy of modeling. High modeling accuracy is conducive to ensuring the accuracy of the interactive twin model to accurately reflect the actual scene. At the same time, high modeling efficiency is conducive to realizing the synchronous changes between the interactive twin model and the actual scene, improving the real-time monitoring and interaction of digital twins, and meeting the needs of data center monitoring and management. Attached Figure Description

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

[0051] Figure 1 This is a flowchart illustrating a digital twin modeling method provided by the present invention.

[0052] Figure 2 This is a schematic diagram of the processing procedure of a digital twin modeling device provided by the present invention.

[0053] Figure 3 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0054] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0055] The following is combined with Figure 1 A digital twin modeling method according to the present invention includes:

[0056] S100: Acquire initial point cloud data;

[0057] S200: Preprocess the initial point cloud data to obtain preprocessed point cloud data;

[0058] S300: Based on the preprocessed point cloud data, generate a dynamic point cloud map and perform convolutional feature extraction processing on the dynamic point cloud map to obtain first point cloud feature information;

[0059] S400: Based on the first point cloud feature information, perform object classification and segmentation processing to obtain the second point cloud feature information;

[0060] S500: Generate an interactive twin model based on the second point cloud feature information;

[0061] S600: Obtain real-time status information and update the interactive twin model based on the real-time status information.

[0062] Initial point cloud data is acquired, accurately reflecting the geometry and details of objects, providing a data foundation for precise modeling. Preprocessing of the initial point cloud data removes redundant and outlier data, reducing processing volume and improving modeling efficiency. Based on the preprocessed point cloud data, convolutional feature extraction is performed using a dynamic point cloud map to obtain first-order point cloud feature information, achieving point cloud-based feature extraction. Based on the first-order point cloud feature information, object classification and segmentation are performed to identify and segment the point cloud features of different objects, obtaining second-order point cloud feature information. This facilitates subsequent object-based modeling, avoiding object aggregation in the model. Based on the second-order point cloud feature information, an interactive twin model is generated. This interactive twin model is updated based on real-time state information, reflecting changes in the real-world scene and realizing the function of a digital twin. Through the interactive twin model, the effect of monitoring and interacting with the actual scene based on a 3D model can be achieved.

[0063] Therefore, by generating point cloud maps, convolutional feature extraction based on point cloud data can be achieved, which can efficiently extract global and local features of point clouds, improve the efficiency and accuracy of modeling. High modeling accuracy is conducive to ensuring the accuracy of the interactive twin model to accurately reflect the actual scene. At the same time, high modeling efficiency is conducive to realizing the synchronous changes between the interactive twin model and the actual scene, improving the real-time monitoring and interaction of digital twins, and meeting the needs of data center monitoring and management.

[0064] It is important to emphasize that point cloud data cannot be directly processed for convolutional feature extraction. This invention overcomes the limitation of point cloud data in this process by using points as nodes in a graph and determining the edges between nodes based on the positional relationships between points, thereby generating a dynamic point cloud graph. Convolutional feature extraction is then performed using this dynamic point cloud graph. Leveraging the high precision of point cloud data in reflecting the geometry and details of real-world objects, the final interactive twin model achieves higher accuracy. Furthermore, convolutional feature extraction based on the dynamic point cloud graph efficiently extracts both global and local features, significantly improving modeling efficiency, especially when handling irregular and sparse point cloud data.

[0065] Interactive twin models map various information from the physical data center into a virtual space through 3D modeling, enabling real-time linkage and interaction between the physical and virtual spaces. Furthermore, the interactive twin model is updated synchronously using real-time status information to achieve the effect of a digital twin, thereby improving the management efficiency and monitoring accuracy of the data center and facilitating operation and maintenance.

[0066] In some embodiments of the present invention, the initial point cloud data may be generated by laser scanning, such as by scanning the computer room with a three-dimensional lidar. The initial point cloud data may be in formats such as LAS, PLY, and PCD.

[0067] It is understandable that real-time status information refers to the current status information of the actual scene. The real-time status information and the initial point cloud data correspond to the same actual scene, such as the same data center. This allows the interactive twin model to be updated based on the real-time status information, reflecting the synchronous status of the actual scene and achieving the effect of real-time monitoring. This facilitates the monitoring of the operating status of various devices in the data center. Furthermore, by interacting with the interactive twin model, information can be obtained more intuitively and efficiently, improving management efficiency. When an abnormal fault is detected based on the interactive twin model, the location of the fault and related operating status information can be quickly determined, facilitating maintenance personnel to quickly locate the fault and perform maintenance operations.

[0068] In some embodiments of the present invention, real-time status information may be generated by sensors and monitoring devices in the actual scene, such as environmental sensors, temperature sensors, humidity sensors, power operation monitoring devices, etc. in a computer room.

[0069] In some embodiments of the digital twin modeling method of the present invention, step S200 includes:

[0070] The initial point cloud data is globally registered based on a preset registration algorithm to obtain the first point cloud data.

[0071] The first point cloud data is denoised to obtain the second point cloud data;

[0072] The second point cloud data is downsampled to obtain the preprocessed point cloud data.

[0073] The initial point cloud data includes point cloud data of objects at different locations. A preset registration algorithm is used to perform global registration on the initial point cloud data to unify the data into the same coordinate system, forming the first point cloud data. This helps ensure spatial consistency and facilitates subsequent processing. The initial point cloud data is generated by scanning the objects. During the scanning process, due to environmental interference, the initial point cloud data includes noise points, outliers, and isolated points. Denoising processing removes these outliers, outliers, and noise points from the first point cloud data, forming the second point cloud data. This allows the second point cloud data to more accurately reflect the actual objects and reduces the amount of subsequent data processing, thus improving processing efficiency. By downsampling the second point cloud data, preprocessed point cloud data is obtained. This reduces the density of the point cloud data without compromising accuracy, further reducing the data volume and improving processing efficiency, thus ensuring real-time performance.

[0074] In some embodiments of the present invention, the preset registration algorithm can be an implementation of an algorithm capable of global registration, such as the ICP (Iterative Closest Point) algorithm, to achieve global registration processing of the initial point cloud data.

[0075] In some embodiments of the present invention, algorithms such as statistical filtering and radius filtering can be used to denoise the first point cloud data.

[0076] In some embodiments of the present invention, downsampling processing of the second point cloud data can be achieved using algorithms such as accelerated grid downsampling.

[0077] In some embodiments of the digital twin modeling method of the present invention, step S300 includes:

[0078] Based on the preprocessed point cloud data, determine the initial processed point cloud information;

[0079] Based on the processed point cloud information, a first preset number of nearest neighbor points are determined for each point to construct the dynamic point cloud map;

[0080] Based on the dynamic point cloud map, calculate the point feature information of each point;

[0081] The point feature information is aggregated to obtain point cloud update feature information;

[0082] Based on the point cloud update feature information, the processed point cloud information is updated to obtain new point cloud update feature information until the number of times the point cloud update feature information is obtained reaches a preset number. Then, the point cloud update feature information is subjected to global pooling processing to obtain the first point cloud feature information.

[0083] Based on preprocessed point cloud data, initial point cloud information for processing is determined. According to the coordinates of each point in the processed point cloud information, a first preset number of nearest neighbor points are determined. The coordinates of these points are used as node features of the graph, and the distances between points and their nearest neighbors are used as edge features, thus forming a dynamic point cloud graph that reflects local feature information. Based on the dynamic point cloud graph, the point feature information corresponding to each point is calculated, and then the point feature information is aggregated to obtain updated point cloud feature information, which can be understood as updating the point cloud data. Based on the updated point cloud feature information, the processed point cloud information is updated, i.e., new processed point cloud information is determined. The dynamic point cloud graph is constructed again, and point feature information is calculated again to obtain new updated point cloud feature information. This process is repeated to achieve convolutional feature extraction. When the convolution iteration reaches a preset number, i.e., the number of times the updated point cloud feature information is obtained reaches a preset number, global pooling is performed on the features of each point in the finally obtained updated point cloud feature information to obtain the first point cloud feature information. This first feature information not only retains the local feature information of the points but also reflects global feature information, facilitating subsequent classification and segmentation processing.

[0084] In this way, by repeatedly constructing a dynamic point cloud map and calculating features based on the processed point cloud information, convolutional feature extraction based on the dynamic point cloud map is achieved. This integrates the local neighborhood point information of a point, captures features such as local geometric structure and topological relationships between points, and performs a preset number of rounds of processing, which can successively extract higher-level and more abstract features. In this process, the dynamic point cloud map is dynamically updated as the feature information of the point cloud changes, so as to capture new local spatial features, which helps the first point cloud feature information to more accurately reflect the features of the object.

[0085] In some embodiments of the present invention, the point cloud update feature information can be calculated using the following expression:

[0086]

[0087] in, This refers to the point feature information of the i-th point when acquiring the point cloud update feature information for the l-th time. Let i be the set of nearest neighbors of the i-th point. ( ) is a nonlinear function for feature combination (which can be a multilayer perceptron MLP), and Aggregate is an aggregation process (such as summation, finding the maximum value, etc.).

[0088] In some embodiments of the present invention, multiple EdgeConv layers can be stacked to achieve the generation of dynamic point cloud maps and the convolutional feature extraction processing based on the dynamic point cloud maps. The number of EdgeConv layers is equal to the preset number of times, that is, equal to the number of times the point cloud update feature information is obtained. The preprocessed point cloud data is used as the input data of the first EdgeConv layer, the output data of the previous EdgeConv layer is used as the input data of the next EdgeConv layer, and the output data of the last EdgeConv layer is the first point cloud feature information.

[0089] In some embodiments of the present invention, the DGCNN processing model can be used to implement a process of obtaining first point cloud feature information by generating a dynamic point cloud map based on preprocessed point cloud data and performing convolutional feature extraction processing based on the dynamic point cloud map. Multiple EdgeConv layers can be stacked in the DGCNN processing model.

[0090] In some embodiments of the present invention, constructing a dynamic point cloud graph can be achieved by finding K (a first preset number) nearest neighbors for each point based on Euclidean distance or other metrics, thereby constructing a local neighborhood for each point. This process treats the point cloud as a graph structure, where points are nodes in the graph, and the edges connecting points to their neighbors in the local neighborhood constitute the edges of the graph. Since this process is based on the dynamic determination between points, each time the point cloud information is updated according to the point cloud update feature information, new adjacency relationships are constructed, thereby dynamically forming different graphs, i.e., new dynamic point cloud graphs.

[0091] Since the initial point cloud data is obtained through scanning, and the point cloud data between different objects is continuous, object classification and segmentation processing is performed on the first point cloud feature information to identify different objects and separate the point cloud features of different objects for subsequent modeling processing.

[0092] In some embodiments of the digital twin modeling method of the present invention, step S500 includes:

[0093] Based on the second point cloud feature information, generate triangular mesh model information;

[0094] The triangular mesh model information is optimized to obtain mesh optimization model information;

[0095] The interactive twin model is generated by rendering based on the mesh optimization model information.

[0096] Based on the second point cloud feature information after object classification and segmentation, surface reconstruction is performed on the point cloud features of different objects to construct and integrate object models, generating triangular mesh model information. By optimizing the triangular mesh model information to appropriately reduce the number of polygons, redundancy and complexity are reduced while maintaining model accuracy as much as possible. Obtaining optimized mesh model information helps avoid an excessively large final model and ensures processing efficiency. Rendering is then performed based on the optimized mesh model information to obtain an interactive twin model, facilitating display and interactive operations.

[0097] In some embodiments of the digital twin modeling method of the present invention, the step of generating triangular mesh model information based on the second point cloud feature information includes:

[0098] Based on the second point cloud feature information, a second preset number of nearest neighbor points are determined for each point to obtain point neighborhood information;

[0099] Based on the neighborhood information of the points, calculate the local plane and normal vector corresponding to each point to generate a triangular mesh surface;

[0100] The triangular mesh model is generated based on each of the triangular mesh surfaces.

[0101] Based on the second point cloud feature information, and using a second preset number of nearest neighbor points around each point, neighborhood information is obtained. This reflects the local feature information around each point. Then, based on the point's neighborhood information, and using the position coordinates and features of each point, local planes and normal vectors are calculated to generate continuous triangular mesh surfaces, enabling object modeling. The triangular mesh surfaces of each object are integrated to form a triangular mesh model, thus modeling the actual scene. In this way, starting from local features, triangular mesh surfaces are constructed, and continuous triangular mesh surfaces model objects, ultimately forming a triangular mesh model. This approach helps ensure the accuracy of local details, and the form of the triangular mesh surface accurately reflects the modeling details, giving the triangular mesh model high precision and meeting modeling requirements.

[0102] In some embodiments of the present invention, generating a triangular mesh surface based on the point neighborhood information can be achieved using a surface reconstruction algorithm, such as the Poisson surface reconstruction algorithm.

[0103] In some embodiments of the digital twin modeling method of the present invention, the step of optimizing the triangular mesh model information to obtain mesh optimization model information includes:

[0104] The triangular mesh model information is simplified to obtain the first mesh model information;

[0105] The first mesh model information is smoothed to obtain the second mesh model information;

[0106] The second mesh model information is subjected to detail enhancement processing to obtain the mesh optimization model information.

[0107] By simplifying the triangular mesh model to reduce the number of redundant polygons and obtaining the first mesh model information, the complexity of the mesh is reduced while preserving model details. This helps reduce the amount and complexity of data in subsequent model processing, improving processing efficiency. Smoothing the first mesh model information and adjusting vertex positions to make the mesh surface smoother, the second mesh model information is obtained. This helps eliminate irregular planes on the model surface caused by interference, improving model accuracy. Detail enhancement processing is then performed on the second mesh model information to enhance the details of important objects, obtaining an optimized mesh model that highlights key details while maintaining model simplicity. Therefore, through mesh simplification, smoothing, and detail enhancement, the model is optimized, improving the quality of the final interactive twin model.

[0108] In some embodiments of the present invention, when performing detail enhancement processing, classification information from the object classification and segmentation process can be combined, i.e., the type of the object can be identified, so as to perform detail enhancement processing on important and key objects.

[0109] In some embodiments of the present invention, mesh simplification can be performed using algorithms such as Quadric Edge Collapse Decimation, and smoothing can be performed using algorithms such as Laplacian smoothing.

[0110] In some embodiments of the digital twin modeling method of the present invention, after generating triangular mesh model information based on the second point cloud feature information, the method further includes:

[0111] Obtain texture information;

[0112] Based on the texture information, the triangular mesh model information is subjected to texture mapping processing.

[0113] When acquiring initial point cloud data, you can acquire colored point clouds or corresponding image data, and then obtain the corresponding texture information. Mapping the texture onto the triangular mesh model helps improve the quality of the model and the visual realism of the final interactive twin model.

[0114] The present invention provides a digital twin modeling device, which can be referred to in correspondence with the digital twin modeling device described above.

[0115] refer to Figure 2 The present invention also provides a digital twin modeling apparatus, comprising:

[0116] The data acquisition module is used to acquire initial point cloud data;

[0117] The data preprocessing module is used to preprocess the initial point cloud data to obtain preprocessed point cloud data;

[0118] The feature extraction module is used to obtain first point cloud feature information by generating a dynamic point cloud map based on the preprocessed point cloud data and performing convolutional feature extraction processing on the dynamic point cloud map.

[0119] The 3D model construction module is used to perform object classification and segmentation based on the first point cloud feature information, obtain the second point cloud feature information, and generate an interactive twin model based on the second point cloud feature information.

[0120] The model optimization module is used to optimize the model.

[0121] A real-time rendering interaction module is used to generate display information and perform interactive operations based on the interaction twin model.

[0122] The model update module is used to acquire real-time status information and update the interactive twin model based on the real-time status information.

[0123] The data acquisition module, the data preprocessing module, the 3D model construction module, the model optimization module, the real-time rendering interaction module, and the model update module work together to implement the above-mentioned digital twin modeling method.

[0124] The data acquisition module obtains initial point cloud data, which can realistically reflect the geometry of the object and accurately reflect the object's details, providing a data foundation for accurate modeling.

[0125] The data preprocessing module preprocesses the initial point cloud data to remove redundant and abnormal data, reducing the amount of data processing and improving modeling efficiency.

[0126] The feature extraction module performs convolutional feature extraction processing based on the preprocessed point cloud data by generating a dynamic point cloud map to obtain the first point cloud feature information, thereby realizing feature extraction processing based on point cloud.

[0127] The 3D model building module performs object classification and segmentation based on the first point cloud feature information to identify and segment the point cloud features of different objects, thereby obtaining the second point cloud feature information. This facilitates subsequent object-based modeling and avoids object aggregation in the model. Based on the second point cloud feature information, an interactive twin model is generated.

[0128] The model optimization module performs model optimization during the generation of the interactive twin model.

[0129] The real-time rendering interaction module performs rendering processing based on the interactive twin model, generates corresponding model display information, and allows interactive operations on the model display information.

[0130] The model update module updates the interactive twin model based on real-time status information, which can reflect changes in the real scene to the interactive twin model in real time, realizing the function of digital twin. Through the interactive twin model, the effect of monitoring and interacting with the actual scene based on the 3D model can be achieved.

[0131] Therefore, by generating point cloud maps, convolutional feature extraction based on point cloud data can be achieved, which can efficiently extract global and local features of point clouds, improve the efficiency and accuracy of modeling. High modeling accuracy is conducive to ensuring the accuracy of the interactive twin model to accurately reflect the actual scene. At the same time, high modeling efficiency is conducive to realizing the synchronous changes between the interactive twin model and the actual scene, improving the real-time monitoring and interaction of digital twins, and meeting the needs of data center monitoring and management.

[0132] In some embodiments of the present invention, the real-time rendering interaction module can perform rendering processing based on the interactive twin model to realize virtual reality (VR) display, augmented reality (AR) display and other functions, allowing users to perform interactive operations in a more three-dimensional and realistic virtual environment.

[0133] In some embodiments of the present invention, the model update module can acquire sensor detection data and automatically update the parameters and associated operating status information in the interactive twin model.

[0134] In some embodiments of the digital twin modeling device of the present invention, it further includes: a data storage and management module, which is used to store point cloud data, model data, feature attribute information and historical data, and manage user access permissions and operation permissions to ensure data security and confidentiality.

[0135] In some embodiments of the present invention, the feature extraction module may include a DGCNN processing model, which has multiple stacked EdgeConv layers. The DGCNN processing model is obtained through a pre-training process. During model training, data augmentation techniques, such as random rotation, scaling, and flipping, are used to improve the generalization ability of the DGCNN processing model. Simultaneously, optimization algorithms, such as Adam and SGD, combined with regularization techniques, such as Dropout, Batch, and Normalization, can be used to prevent overfitting and optimize the model. Then, the performance of the trained model is verified by evaluating the accuracy and precision of the model's output results, using methods such as cross-validation, to determine whether the model meets the usage requirements.

[0136] Figure 3 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 3 As shown, the electronic device may include a processor 810, a communications interface 820, a memory 830, and a communication bus 840. The processor 810, communications interface 820, and memory 830 communicate with each other via the communication bus 840. The processor 810 can call logical instructions stored in the memory 830 to execute the aforementioned digital twin modeling method.

[0137] Furthermore, the logical instructions in the aforementioned memory 830 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0138] On the other hand, the present invention also provides a computer program product, the computer program product including a computer program that can be stored on a non-transitory computer-readable storage medium, and when the computer program is executed by a processor, the computer is able to execute a digital twin modeling method provided by the above methods.

[0139] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform a digital twin modeling method provided by the methods described above.

[0140] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0141] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0142] In the description of this invention, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0143] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0144] All actions involving the acquisition of signals, information, or data in this application are carried out in accordance with the relevant data protection laws and policies of the country where the application is located, and with the authorization of the owner of the relevant device.

[0145] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A digital twin modeling method, characterized in that, include: Obtain initial point cloud data; The initial point cloud data is preprocessed to obtain preprocessed point cloud data; Based on the preprocessed point cloud data, a dynamic point cloud map is generated and convolutional feature extraction is performed on the dynamic point cloud map to obtain the first point cloud feature information. Based on the first point cloud feature information, object classification and segmentation are performed to obtain the second point cloud feature information; Based on the second point cloud feature information, an interactive twin model is generated; Obtain real-time status information and update the interactive twin model based on the real-time status information; The step of obtaining first point cloud feature information based on the preprocessed point cloud data, generating a dynamic point cloud map, and performing convolutional feature extraction processing on the dynamic point cloud map includes: Based on the preprocessed point cloud data, determine the initial processed point cloud information; Based on the processed point cloud information, a first preset number of nearest neighbor points are determined for each point to construct the dynamic point cloud map; Based on the dynamic point cloud map, calculate the point feature information of each point; The point feature information is aggregated to obtain point cloud update feature information; Based on the point cloud update feature information, update the processed point cloud information to obtain new point cloud update feature information until the number of times the point cloud update feature information is obtained reaches a preset number, and perform global pooling processing on the point cloud update feature information to obtain the first point cloud feature information. Among them, the first point cloud feature information represents the local feature information and global feature information of the point; The step of generating an interactive twin model based on the second point cloud feature information includes: Based on the second point cloud feature information, generate triangular mesh model information; The triangular mesh model information is optimized to obtain mesh optimization model information; The interactive twin model is generated by rendering based on the mesh optimization model information. The optimization process of the triangular mesh model information to obtain mesh optimization model information includes: The triangular mesh model information is simplified to obtain the first mesh model information; The first mesh model information is smoothed to obtain the second mesh model information; By combining the classification information from object classification and segmentation, the items identified as important and key in the second mesh model information are subjected to detail enhancement processing to obtain the mesh optimization model information.

2. The digital twin modeling method according to claim 1, characterized in that, The step of generating triangular mesh model information based on the second point cloud feature information includes: Based on the second point cloud feature information, a second preset number of nearest neighbor points are determined for each point to obtain point neighborhood information; Based on the neighborhood information of the points, calculate the local plane and normal vector corresponding to each point to generate a triangular mesh surface; The triangular mesh model is generated based on each of the triangular mesh surfaces.

3. The digital twin modeling method according to claim 2, characterized in that, After generating the triangular mesh model information based on the second point cloud feature information, the method further includes: Obtain texture information; Based on the texture information, the triangular mesh model information is subjected to texture mapping processing.

4. The digital twin modeling method according to claim 1, characterized in that, The step of preprocessing the initial point cloud data to obtain preprocessed point cloud data includes: The initial point cloud data is globally registered based on a preset registration algorithm to obtain the first point cloud data. The first point cloud data is denoised to obtain the second point cloud data; The second point cloud data is downsampled to obtain the preprocessed point cloud data.

5. A digital twin modeling device, characterized in that, include: The data acquisition module is used to acquire initial point cloud data; The data preprocessing module is used to preprocess the initial point cloud data to obtain preprocessed point cloud data; The feature extraction module is used to obtain first point cloud feature information by generating a dynamic point cloud map based on the preprocessed point cloud data and performing convolutional feature extraction processing on the dynamic point cloud map. The 3D model construction module is used to perform object classification and segmentation based on the first point cloud feature information, obtain the second point cloud feature information, and generate an interactive twin model based on the second point cloud feature information. The model optimization module is used to optimize the model. A real-time rendering interaction module is used to generate display information and perform interactive operations based on the interaction twin model. The model update module is used to acquire real-time status information and update the interactive twin model based on the real-time status information. The data acquisition module, the data preprocessing module, the 3D model construction module, the model optimization module, the real-time rendering interaction module, and the model update module cooperate to implement a digital twin modeling method as described in any one of claims 1 to 4.

6. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements a digital twin modeling method as described in any one of claims 1 to 4.

7. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements a digital twin modeling method as described in any one of claims 1 to 4.

Images