Three-dimensional point cloud modeling method and system for digital twinning of power system and medium
By combining lidar scanning with preset reference points in the power system with deep learning, the problems of low accuracy and poor adaptability in 3D reconstruction in existing technologies have been solved, achieving high-precision 3D point cloud reconstruction and lightweight processing.
Patent Information
- Application Number
- CN202510829760.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-20
- Publication Date
- 2025-11-18
AI Technical Summary
Existing 3D reconstruction methods for power systems have low reconstruction accuracy, cannot reproduce detailed structures, have poor adaptability, and are insufficient in handling occlusion. Furthermore, deep learning algorithms lack optimal solution criteria when reconstructing complex 3D structures, resulting in missing details and blurred edges.
By pre-setting reference points at different vertical heights on the target object, multi-angle point cloud data is obtained using LiDAR scanning. Coordinate transformation and optimization are performed, noise is removed by combining nearest neighbor search and grid method, and feature extraction and segmentation are carried out using deep learning algorithms to generate a high-precision 3D point cloud model.
It achieves high-precision 3D point cloud reconstruction, enriches 3D structural features, solves the problems of missing local details and blurred edges of complex structures, and constructs a lightweight 3D model.
Smart Images
Figure CN120976406A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of digital twinning, in particular to a three-dimensional point cloud modeling method and system for digital twinning of a power system and a medium. BACKGROUND
[0002] In recent years, with the development of unmanned aerial vehicles and three-dimensional reconstruction technology, using unmanned aerial vehicles to realize aerial inspection and three-dimensional modeling of power assets has become a new technical direction. The model is an important part of digital twinning and an important prerequisite for realizing the function of digital twinning. How to construct the digital twinning model of the power data communication network is of great significance.
[0003] The current common three-dimensional reconstruction method of the power system is to use an unmanned aerial vehicle to carry an RGB camera, plan a comprehensive shooting of the target area, obtain two-dimensional image data, generate three-dimensional point clouds using a structured light motion algorithm, and further construct a three-dimensional model of the line facility. This method has low reconstruction accuracy and can only generate a simplified line overview, which cannot present the detailed structure. It has poor adaptability and is insufficient in handling occlusion, which is prone to have visual angle dead zones.
[0004] Another method is based on laser radar scanning and computer three-dimensional point cloud processing technology, which can accurately reconstruct the three-dimensional structure and spatial orientation of the object to be measured, and measure the geometric parameters more accurately, so it has gradually been applied in the field of digital twinning of the power system.
[0005] Multi-image stereo matching (MVS) is a key technology in the field of computer vision for recovering the three-dimensional structure of a scene from multiple images taken from different perspectives. It reconstructs the three-dimensional model of the scene by analyzing multiple images taken from different perspectives. At present, the development of multi-image stereo matching in point cloud processing mainly relies on deep learning methods, which can be divided into the following four types: projection-based methods, discretization-based methods, hybrid-based methods, and point-based methods. Projection-based methods project three-dimensional point clouds onto two-dimensional planes, converting three-dimensional problems into two-dimensional image processing tasks, and using existing convolutional neural networks for processing. However, due to the compression of three-dimensional point clouds into two-dimensional space, some spatial information may be lost when dealing with complex geometric structures. Discretization-based methods map point cloud data into a three-dimensional voxel grid, forming a structure similar to a three-dimensional image, and then use a three-dimensional convolutional neural network (3D CNN) for processing. However, in high-resolution point clouds, the discretization process may result in excessive computational load and storage requirements, making it difficult to balance resolution and computational efficiency, and the discretization process may also cause loss of local geometric information. Hybrid-based methods combine the characteristics of projection and discretization, allowing the use of both two-dimensional and three-dimensional feature information to improve the model's perception and generalization ability, but the network structure is complex and requires high computational resources.
[0006] However, the general three-dimensional point cloud reconstruction technology based on laser radar scanning, when fusing, usually lacks local detail features in a photo at an angle for a relatively complex three-dimensional structure; and the deep learning algorithm lacks a judgment standard for an optimal solution when calculating, resulting in an adverse result of missing details, blurring edges of a complex three-dimensional structure or overall integration. SUMMARY
[0007] The present application aims to provide a three-dimensional point cloud modeling method and system for power system digital twinning to solve the above technical problems.
[0008] The present application provides a three-dimensional point cloud modeling method for power system digital twinning, comprising the steps of: S1, initial point cloud data acquisition: preset at least M basic reference points P on the target object which are not at the same vertical height, M≥3, denoted as reference points ; acquire initial point cloud data of the target object at at least three angles through laser radar scanning; wherein each angle of the laser radar scanning can display at least N of the basic reference points, N≥3, denoted as reference points ; S2, optimization of initial point cloud data: S21, take the initial point cloud data at one angle as the to-be-optimized point cloud data, and select at least one initial point cloud data from the initial point cloud data at other angles as the transformed point cloud data; S22, take the space where the to-be-optimized point cloud data is located as the reference space, construct a basic coordinate system, and convert the points in the transformed point cloud data to the basic coordinate system according to the known position information of the reference points in the transformed point cloud data in the coordinate system of the transformed point cloud data, and combine the points in the transformed point cloud data with the points of the to-be-optimized point cloud data to form new point cloud data; S23, perform noise reduction and redundancy removal optimization on the new point cloud data to obtain optimized point cloud data; S24, repeat the steps S21-S23 for the initial point cloud data at each angle to obtain the optimized point cloud data of the initial point cloud data at each angle; wherein the data combination of the to-be-optimized point cloud data and the transformed point cloud data in the step S21 is not repeated; S3, fusion modeling: perform three-dimensional point cloud reconstruction according to the optimized point cloud data at the at least three angles of the target object to obtain a three-dimensional structure model of the target object.
[0009] Preferably, the reference points are points not on the same straight line.
[0010] Preferably, in the step S23, the outlier points in the new point cloud data are determined by a nearest neighbor search algorithm, including: dividing the data space into three dimensions by using a kd-tree, sequentially extracting each sample point in the new point cloud data, and according to the sample point and a set radius R, counting all data points in the data set with a distance less than R from the sample point, and when the number of points in the neighborhood of the sample point is less than a set threshold X, the sample point is determined as an outlier; and the outlier points are deleted to obtain optimized point cloud data after noise reduction. Preferably, the points in the new point cloud data are removed by a grid method. S231, a three-dimensional grid with a preset pixel size is created; S232, the three-dimensional grid is randomly placed in the new point cloud data, when the number of points in the grid is greater than a preset number, the center point of the grid is replaced with the points in the grid as new coordinate points, and when the number of points in the grid is less than or equal to the preset number, the points in the grid remain unchanged; S233, after a preset number of iterations is reached, the optimized point cloud data after removing redundancy is obtained.
[0011] Preferably, the number of transformed point cloud data is 1.
[0012] Preferably, the step S3 includes: S31, the known position information of the basic reference point P is taken as an initial solution, and the optimized point cloud data of the at least three angles is registered by an iterative closest point algorithm to obtain registered point cloud data; S32, feature extraction and feature segmentation are performed on the registered point cloud data by a deep learning algorithm; S33, according to the target feature parameters extracted in the step S32, a building model is reconstructed.
[0013] Preferably, the step S32 includes: a Point CNN point cloud segmentation network is trained by a three-dimensional image of a target object of a known power system, the target object main body and internal object are segmented from the registered point cloud data by using the trained segmentation network, the surrounding environment and the internal object are cropped and removed, and a pure registered point cloud data point cloud model is obtained; a training loss function is: ; wherein, is a global geometric semantic consistency loss, is a local geometric semantic consistency loss, and a and b are corresponding weights; , is a geometric weighting coefficient, N is the number of points, represents the divergence of the i-th sample, is the target semantic distribution of the i-th sample, is the predicted semantic distribution of the i-th sample. , is the neighborhood of point i, and respectively represent the feature vectors of points i and j, represents the square of the Euclidean distance between the feature vectors of points i and j; carrying out point cloud slicing processing and extracting target feature parameters.
[0014] Preferably, the step S33 comprises: generating a BIM rough model according to the extracted target feature parameters, refining the BIM model, and finally establishing a textured target BIM model for the model mapping.
[0015] The application also provides a three-dimensional point cloud modeling system for digital twinning of a power system, comprising a memory, a processor, a computer program stored in the memory and executable on the processor, and the processor executes the computer program to realize the steps of the three-dimensional point cloud modeling method for digital twinning of a power system according to any one of the above.
[0016] The application also provides a computer readable storage medium storing a computer program, and the computer program is executed by a processor to realize the steps of the three-dimensional point cloud modeling method for digital twinning of a power system according to any one of the above.
[0017] In the present application, the following steps are adopted: S1, initial point cloud data acquisition: preset at least M basic reference points P on the target object which are not at the same vertical height, M≥3, denoted as reference point ; obtain initial point cloud data of at least three angles of the target object through laser radar scanning; wherein each angle of the laser radar scanning can display at least N of the basic reference points, N≥3, denoted as reference point ; S2, optimization of initial point cloud data: S21, take the initial point cloud data of one angle as the to-be-optimized point cloud data, and select at least one initial point cloud data from the initial point cloud data of other angles as the transformed point cloud data; S22, take the space where the to-be-optimized point cloud data is located as the reference space, construct a basic coordinate system, and according to the reference points S21, obtaining the initial point cloud data of each angle of the target object; S22, converting the initial point cloud data of each angle to the base coordinate system according to the known position information of the transformation point cloud data in the coordinate system of the transformation point cloud data, converting the points in the transformation point cloud data to the base coordinate system, and combining the points of the initial point cloud data of each angle with the points of the to-be-optimized point cloud data to form new point cloud data; S23, performing noise reduction and redundancy removal optimization on the new point cloud data to obtain optimized point cloud data; S24, repeating the steps of steps S21-S23 on the initial point cloud data of each angle to obtain the optimized point cloud data of the initial point cloud data of each angle; wherein the data combination of the to-be-optimized point cloud data and the transformation point cloud data in step S21 is not repeated; S3, fusion modeling: performing three-dimensional point cloud reconstruction according to the optimized point cloud data of the at least three angles of the target object to obtain a three-dimensional structure model of the target object.
[0018] By artificially adding the reference point, the coordinate conversion of the initial point cloud data under different angles can be realized to realize the pre-fusion matching of the point cloud data under multiple different angles.
[0019] Meanwhile, the reference point can be used as a feature coordinate reference value participating in registration to solve the problem that the iterative closest point algorithm (ICP algorithm) is prone to fall into a local optimal solution in the registration process.
[0020] When the point cloud data under multiple different angles is fused for the first time, only the fusion of points is performed, and no registration and feature extraction are performed; only noise reduction and redundancy removal processing are performed on the point cloud data; the high-efficiency neighbor searching algorithm is used to count the number of neighbor points to distinguish and remove the outlier noise of the point cloud; the grid method is used to effectively filter out the redundant data points, the three-dimensional morphological features of the target object are maintained, a three-dimensional model with low data amount is constructed, and the lightweight processing of three-dimensional modeling is realized. BRIEF DESCRIPTION OF DRAWINGS
[0021] The accompanying drawings, which are part of this application, serve to further understand the application, and the illustrative embodiments of the application and the description thereof are used to explain the application, but do not constitute an improper limitation on the application. Obviously, the accompanying drawings in the following description are only some embodiments, and other drawings can be obtained from these drawings without creative labor for those skilled in the art. In the drawings: Figure 1 It is a flowchart of the three-dimensional point cloud modeling method for power system digital twinning in an embodiment of the application.
[0022] Figure 2 It is a hardware structure schematic diagram of a system running the three-dimensional point cloud modeling method for power system digital twinning in an embodiment of the application.
[0023] The implementation, functional characteristics and advantages of the application will be further described with reference to the embodiments and the accompanying drawings. Detailed Implementation
[0024] The technical problems solved by the embodiments of the present invention, the technical solutions adopted, and the technical effects achieved will be clearly and completely described below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other equivalent or obvious variations of embodiments obtained by those skilled in the art without creative effort fall within the protection scope of the present invention. The embodiments of the present invention can be embodied in various different ways as defined and covered by the claims.
[0025] It should be noted that many specific details are given in the following description for ease of understanding. However, it is obvious that the present invention may be implemented without these specific details.
[0026] It should be noted that, in the absence of explicit limitations or conflicts, the various embodiments and their technical features in this invention can be combined with each other to form a technical solution.
[0027] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
[0028] In the following description, the use of suffixes such as "module," "part," or "unit" to denote elements is solely for the purpose of illustrative purposes and has no specific meaning in itself. Therefore, "module," "part," or "unit" may be used interchangeably.
[0029] The invention will now be described in further detail with reference to the accompanying drawings.
[0030] Please combine Figure 1 and Figure 2 , Figure 1 A method for three-dimensional point cloud modeling of a power system digital twin, as described in one embodiment of the present invention, includes the following steps: S1, Initial point cloud data acquisition: Pre-determine at least M basic reference points P on the target object that are not at the same vertical height, where M≥3, and denoted as reference points. The initial point cloud data of the target object is acquired from at least three angles using lidar scanning; wherein, at least N of the basic reference points, where N≥3, are displayed at each angle scanned by the lidar, and are denoted as reference points. ; S2, Optimization of initial point cloud data: S21, take the initial point cloud data from one angle as the point cloud data to be optimized, and select at least one initial point cloud data from the initial point cloud data from other angles as the transformed point cloud data. S22, constructing a basic coordinate system with the space where the point cloud data to be optimized as a reference space, converting points in the transformed point cloud data to the basic coordinate system according to known position information of reference points in the transformed point cloud data, and combining the points in the transformed point cloud data with points in the point cloud data to be optimized to form new point cloud data; S23, performing noise reduction and redundancy elimination optimization on the new point cloud data to obtain optimized point cloud data; S23, performing noise reduction and redundancy elimination optimization on the new point cloud data to obtain optimized point cloud data; S24, repeating the steps S21-S23 for the initial point cloud data of each angle to obtain optimized point cloud data of the initial point cloud data of each angle; wherein the data combination of the point cloud data to be optimized and the transformed point cloud data in the step S21 is not repeated; S3, fusion modeling: performing three-dimensional point cloud reconstruction according to the optimized point cloud data of the target object at the at least three angles to obtain a three-dimensional structure model of the target object.
[0031] Preferably, the reference points are points not on the same straight line.
[0032] Preferably, in the step S23, outlier points in the new point cloud data are determined by a nearest neighbor search algorithm, including: dividing the data space into three dimensions by kd-tree, sequentially extracting each sample point in the new point cloud data, and counting all data points with a distance less than R from the sample point according to the sample point and a set radius R, when the number of points in the neighborhood of the sample point is less than a set threshold X, the sample point is determined as an outlier; deleting the outlier points to obtain the optimized point cloud data after noise reduction.
[0033] Preferably, the points in the new point cloud data are removed by a grid method: S231, creating a three-dimensional grid with a preset pixel size; S232, randomly placing the three-dimensional grid in the new point cloud data, when the number of points in the grid is greater than a preset number, replacing the center of gravity of the grid with the points in the grid as new coordinate points, and when the number of points in the grid is less than or equal to the preset number, keeping the points in the grid unchanged; S233, after reaching a preset number of iterations, obtaining the optimized point cloud data after removing redundancy.
[0034] Preferably, the number of transformed point cloud data is 1.
[0035] Preferably, the step S3 includes: S31, registering the optimized point cloud data of the at least three angles by an iterative closest point algorithm with the known position information of the basic reference point P as an initial solution to obtain registered point cloud data; S32, performing feature extraction and feature segmentation on the registered point cloud data by a deep learning algorithm; S33, reconstructing a building model according to the target feature parameters extracted in step S32.
[0036] Preferably, the step S32 comprises: training the Point CNN point cloud segmentation network through a three-dimensional image of the target object of the known power system, and segmenting the target object and internal objects from the registered point cloud data by using the trained segmentation network, and cutting and removing the surrounding environment and internal objects to obtain a pure registered point cloud data point cloud model; preferably, the training loss function is: ; wherein, is a global geometric semantic consistency loss, is a local geometric semantic consistency loss, and a and b are corresponding weights; , is a geometric weighting coefficient, N is the number of points, represents the divergence of the i-th sample, is the target semantic distribution of the i-th sample, is the predicted semantic distribution of the i-th sample; , is the neighborhood of point i, and respectively represent the feature vectors of points i and j, represents the square of the Euclidean distance between the feature vectors of points i and j; performing point cloud slicing processing and extracting target feature parameters; specifically, first calculate the minimum cubic hexahedron space of the target point cloud, and adjust the point cloud posture of the obtained target point cloud model so that the long side of the minimum cubic hexahedron of the target point cloud is parallel to the X axis, then set the slice thickness along the X axis, Y axis and Z axis direction respectively to slice the target point cloud, and then extract the related parameters of the target structure in the slice point cloud as the target feature parameters.
[0037] Preferably, the step S33 comprises: generating a BIM rough model according to the extracted target feature parameters, refining the BIM model, and finally establishing a textured target BIM model for the model map. Specifically, according to the extracted target feature parameters, a BIM rough model is generated according to the IFC standardized modeling format, the BIM model is refined by referring to the plan and elevation section view, and finally a textured target BIM model is established for the model map.
[0038] By artificially adding reference points, coordinate conversion of initial point cloud data under different angles can be realized to achieve pre-fusion matching of point cloud data under multiple different angles.
[0039] Meanwhile, the reference points can be used as feature coordinate reference values for participating in registration to solve the problem that the iterative closest point algorithm (ICP algorithm) is prone to local optimal solution in the registration process.
[0040] In the first fusion of point cloud data under multiple different angles, only point fusion is performed, and registration and feature extraction are not performed; only noise reduction and redundancy removal processing are performed on the point cloud data; the high-efficiency neighbor searching algorithm is used to count the number of neighbor points to identify and remove the outlier noise of the point cloud; the grid method is used to effectively filter out redundant data points, maintain the three-dimensional morphological features of the target object, construct a three-dimensional model with low data volume, and realize lightweight processing of three-dimensional modeling.
[0041] The application further provides a three-dimensional point cloud modeling system for digital twinning of a power system, which is established on a computer system and specifically comprises a memory 61, a processor 62, and a computer program 63 stored in the memory 61 and executable on the processor 62, and the processor 62 implements the steps of the three-dimensional point cloud modeling method for digital twinning of a power system according to any one of the above embodiments when executing the computer program 63.
[0042] The application further provides a computer readable storage medium storing a computer program, and the computer program implements the steps of the three-dimensional point cloud modeling method for digital twinning of a power system according to any one of the above embodiments when executed by a processor.
[0043] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above functional units and modules is exemplified, and in actual application, the above functions can be completed by different functional units and modules according to needs, that is, the internal structure of the device is divided into different functional units or modules to complete all or part of the functions described above. Each functional unit and module in the embodiment can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit, and the integrated unit can be realized in the form of hardware or software. In addition, the specific names of each functional unit and module are only for mutual distinction, and do not limit the protection scope of the application. The specific working process of the units and modules in the system can refer to the corresponding process in the foregoing method embodiments, and will not be repeated here.
[0044] In the above embodiments, the description of each embodiment has its own focus, and the parts not described or recorded in a certain embodiment can be referred to the relevant description of other embodiments.
[0045] Those skilled in the art can understand that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. A person skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.
[0046] In the embodiments provided by the present application, it should be understood that the disclosed apparatus / terminal device and method can be implemented in other ways. For example, the apparatus / terminal device embodiments described above are merely schematic, for example, the division of the modules or units is merely a logical function division, and there can be another division manner in actual implementation, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some interface, device or unit, and can be electrical, mechanical or in other forms.
[0047] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, that is, they can be located in one place, or can be distributed on a plurality of network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the present embodiment.
[0048] In addition, each functional unit in each embodiment of the present application can be integrated into a processing unit, or each unit can exist physically independently, or two or more units can be integrated into one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.
[0049] The integrated module / unit, if realized in the form of a software function unit and sold or used as an independent product, can be stored in a computer-readable storage medium. Based on such understanding, all or part of the processes in the above-mentioned embodiment methods can also be completed by a computer program instructing related hardware, and the computer program can be stored in a computer-readable storage medium. The computer program can implement the steps of each method embodiment when executed by a processor. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or some intermediate forms. The computer-readable medium can include any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc. It should be noted that the computer-readable medium can include or exclude contents according to the requirements of legislation and patent practice in the jurisdiction, for example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium does not include electrical carrier signals and telecommunication signals.
[0050] Therefore, from any viewpoint, the embodiments should be regarded as exemplary and non-limiting, and the scope of the present application is defined by the appended claims rather than the above description, and therefore all changes falling within the meaning and scope of the equivalent elements of the application file are intended to be included in the present application.
[0051] In the description of the present specification, the description referring to the terms "an embodiment", "another embodiment", "other embodiments", or "a first embodiment to an Xth embodiment" means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the present specification, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, method steps or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.
[0052] The above-mentioned embodiment numbers of the present application are only for description, not representing the advantages or disadvantages of the embodiments.
[0053] The above-described embodiments are only used to illustrate the technical solutions of the present application, and are not intended to limit the present application; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.
Claims
1. A method for three-dimensional point cloud modeling of a digital twin of a power system, characterized in that, Including the following steps: S1, Initial point cloud data acquisition: Pre-determine at least M basic reference points P on the target object that are not at the same vertical height, where M≥3, and denoted as reference points. Initial point cloud data of the target object is acquired from at least three angles using lidar scanning; wherein, at least N of the basic reference points, where N≥3, are displayed at each angle scanned by the lidar, denoted as reference points. ; S2, Optimization of initial point cloud data: S21, take the initial point cloud data from one angle as the point cloud data to be optimized, and select at least one initial point cloud data from the initial point cloud data from other angles as the transformed point cloud data. S22, using the space where the point cloud data to be optimized is located as the reference space, construct a basic coordinate system, and based on the reference points in the transformed point cloud data... Using the known position information in the coordinate system of the transformed point cloud data, the points in the transformed point cloud data are transformed to the base coordinate system, and together with the points in the point cloud data to be optimized, they form new point cloud data. S23, perform noise reduction and redundancy removal optimization on the new point cloud data to obtain optimized point cloud data; S24, repeat steps S21-S23 for the initial point cloud data at each angle to obtain optimized point cloud data for the initial point cloud data at each angle; wherein, the data combination of the point cloud data to be optimized and the transformed point cloud data in step S21 is not repeated; S3, Fusion Modeling: Based on the optimized point cloud data of the target object from at least three angles, perform three-dimensional point cloud reconstruction to obtain a three-dimensional structural model of the target object.
2. The three-dimensional point cloud modeling method for power system digital twins according to claim 1, characterized in that, The reference point Points that are not on the same straight line.
3. The three-dimensional point cloud modeling method for power system digital twins according to claim 1, characterized in that, In step S23, outliers in the new point cloud data are determined using a nearest neighbor search algorithm, including: dividing the data space into three dimensions using a kd-tree, extracting each sample point in the new point cloud data sequentially, and based on the sample point and a set radius R, collecting all data points in the data set whose distance to the sample point is less than R. When the number of points in the neighborhood of a sample point is less than a set threshold X, it is determined to be an outlier; deleting the outlier results in denoised optimized point cloud data.
4. The three-dimensional point cloud modeling method for power system digital twins according to claim 3, characterized in that, The points in the new point cloud data are redundantly removed using a grid method: S231, Create a 3D mesh with a preset pixel size; S232, the three-dimensional mesh is randomly placed in the new point cloud data. When the number of points in the mesh is greater than a preset number, the centroid of the mesh is used to replace the points in the mesh as new coordinate points. When the number of points in the mesh is less than or equal to the preset number, the points in the mesh remain unchanged. S233, after reaching the preset number of iterations, obtain the optimized point cloud data after removing redundancy.
5. The three-dimensional point cloud modeling method for power system digital twins according to claim 1, characterized in that, The number of transformed point cloud data is 1.
6. The three-dimensional point cloud modeling method for power system digital twins according to claim 1, characterized in that, Step S3 includes: S31, using the known position information of the basic reference point P as the initial solution, the optimized point cloud data of the at least three angles are registered using the iterative nearest point algorithm to obtain registered point cloud data; S32 uses deep learning algorithms to extract and segment features from registered point cloud data. S33. Based on the target object feature parameters extracted in step S32, the building model is reconstructed.
7. The method for three-dimensional point cloud modeling of power system digital twins according to claim 6, characterized in that, Step S32 includes: The PointCNN point cloud segmentation network is trained using 3D images of target objects in a known power system. The trained segmentation network is then used to segment the main body and internal objects of the target object from the registered point cloud data. Surrounding environments and internal objects are then cropped to remove them, resulting in a clean registered point cloud model. The training loss function is: ;in, For global geometric semantic consistency loss, For local geometric semantic consistency loss, a and b are the corresponding weights; , These are geometric weighting coefficients, where N is the number of points. This represents the divergence of the i-th sample. It is the target semantic distribution of the i-th sample. It is the predicted semantic distribution of the i-th sample; , It is the neighborhood of point i. and Let i and j represent the eigenvectors of points i and j, respectively. The square of the Euclidean distance between the eigenvectors of points i and j is represented. Perform point cloud slicing and extract feature parameters of the target object.
8. The three-dimensional point cloud modeling method for power system digital twins according to claim 6, characterized in that, Step S33 includes: generating a rough BIM model based on the extracted target object feature parameters, refining the BIM model, and finally creating a textured BIM model of the target object for the model texture mapping.
9. A three-dimensional point cloud modeling system for digital twins of power systems, characterized in that, The method includes 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 the steps of the three-dimensional point cloud modeling method for a digital twin of a power system as described in any one of claims 1-8.
10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the three-dimensional point cloud modeling method for power system digital twins as described in any one of claims 1 to 8.