Substation equipment multi-view point cloud data fusion three-dimensional modeling method and system

Abstract

The invention provides a substation equipment multi-view point cloud data fusion three-dimensional modeling method and system. The method comprises the steps of obtaining original point cloud data under multiple views and synchronously updating an equipment surface feature topological relation; setting a registration reference surface matched with the geometric structure of the equipment according to the spatial distribution characteristics of the point cloud; carrying out rigid alignment by adopting a point cloud registration algorithm based on a feature weight; and calling a curvature constraint-based three-dimensional reconstruction algorithm through the registration error parameters to generate a high-precision grid model. Wherein the registration reference plane setting needs to judge the point cloud coverage integrity and the overlapping area condition, and when the condition is not met, the collection view angle is adjusted or the point cloud data is complemented; the registration condition is judged through a curvature consistency index; the three-dimensional reconstruction process comprises voxel down-sampling, curved surface function construction and grid optimization, and the final model needs to be subjected to electrical safety verification. According to the method, high-precision three-dimensional modeling of the substation equipment can be realized, and the geometric feature reduction degree and the electrical safety verification reliability are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of three-dimensional modeling, and more specifically, to a method and system for fusion of multi-view point cloud data of substation equipment. Background Art

[0002] In the field of 3D modeling of substation equipment, the demand for accurate modeling of substation equipment is increasing with the development of power systems. Traditional modeling methods rely primarily on the acquisition of point cloud data from a single perspective. While this method can capture some geometric information about the equipment, due to viewing angle limitations, it often fails to fully capture all surface features of the equipment, resulting in insufficient model integrity and accuracy. Furthermore, single-perspective point cloud data is prone to data loss and inaccurate topological relationships when processing complex structures, impacting subsequent equipment maintenance and safety assessments.

[0003] Although existing point cloud data acquisition technology is constantly improving, challenges still exist in multi-view data fusion. Multi-view point cloud data fusion technology aims to collect point cloud data from different angles and effectively integrate this data to generate a more complete and accurate three-dimensional model. However, current technology still needs to be improved in terms of the registration accuracy of point cloud data, feature extraction, and the efficiency and quality of model reconstruction. For example, when processing substation equipment with complex geometric features, traditional point cloud registration algorithms often have difficulty accurately aligning point cloud data from different perspectives, resulting in large registration errors. At the same time, when processing point cloud data, existing technologies are not timely and accurate enough in updating the topological relationship of equipment surface features, which affects the final quality of the model.

[0004] In the process of implementing the embodiments of the present invention, there are at least the following problems or defects in the existing technology: First, when fusing multi-view point cloud data, the existing technology has difficulty in effectively handling the problems of data overlap and missing from different perspectives, resulting in insufficient integrity and accuracy of the model. Secondly, when faced with complex geometric structures, the traditional point cloud registration algorithm has low registration accuracy and cannot meet the needs of high-precision modeling. In addition, during the point cloud data processing process, the existing technology does not update the topological relationship of the surface features of the equipment in a timely and accurate manner, which affects the final quality of the model. Finally, after generating the three-dimensional model, the existing technology lacks the link to perform electrical safety verification on the model, and cannot ensure the safety and reliability of the model in actual application. Summary of the Invention

[0005] The present invention provides a method and system for fusion of multi-view point cloud data into three-dimensional modeling of substation equipment.

[0006] In a first aspect of the present invention, a method for 3D modeling of substation equipment by fusion of multi-view point cloud data is provided, comprising:

[0007] Obtain original point cloud datasets of substation equipment from multiple perspectives, and synchronously update the topological relationships of equipment surface features through point cloud acquisition equipment;

[0008] Based on the spatial distribution characteristics of the original point cloud data set, setting multiple sets of point cloud registration reference planes so that the multiple sets of point cloud registration reference planes match the geometric structure of the substation equipment;

[0009] Invoking a feature weight-based point cloud registration algorithm to rigidly align the multiple sets of point cloud data; and

[0010] A registration error parameter of each group of the multiple groups of point cloud data is obtained, and a three-dimensional reconstruction algorithm based on curvature constraints is called to generate a high-precision three-dimensional mesh model of the substation equipment.

[0011] Furthermore, setting a plurality of sets of point cloud registration reference planes based on the spatial distribution characteristics of the original point cloud dataset includes:

[0012] Determine whether the point cloud data under the current viewing angle covers the complete surface of the device and whether there are overlapping areas between adjacent point clouds, and update the topological relationship of the device surface features. When there is complete coverage and overlapping areas, directly select the current point cloud data point set and set the alignment reference plane;

[0013] When there is incomplete coverage but overlapping areas, adjust the point cloud acquisition perspective and rescan the missing areas to update the point cloud dataset;

[0014] When there is complete coverage but no overlapping area, the point cloud completion method based on feature similarity is called to extract supplementary point clouds from adjacent perspectives and add them to the current point cloud dataset. Then, it is determined whether the point cloud dataset meets the registration conditions. When the conditions are met, the current point cloud data set is selected and the registration reference plane is set. Otherwise, the scanning path planning method based on curvature difference is called to obtain the optimal acquisition angle θ to adjust the posture of the point cloud acquisition device. After updating the point cloud dataset, it returns to the step of calling the point cloud registration algorithm based on feature weights.

[0015] Furthermore, determining whether the point cloud dataset meets the registration conditions includes:

[0016] Calculate the curvature consistency index C of the overlapping area of ​​the point cloud. If C≥C min , then it is determined that the registration conditions are met and a multi-view point cloud registration sequence is generated;

[0017] Otherwise, a point cloud completion instruction is generated and the method returns to the step of calling the point cloud completion method based on feature similarity, where C is the curvature consistency index, C min is the preset curvature threshold.

[0018] Furthermore, before obtaining the original point cloud dataset of the substation equipment at multiple perspectives, the following steps are also included:

[0019] Perform 3D spatial modeling of the substation equipment structure to generate the equipment local coordinate system using the following parameterized equations:

[0020]

[0021] Among them, R is the device reference radius, α is the surround scanning angle, Δ x , Δ y , Δ z is the equipment installation position offset, k is the height scaling factor, and β is the vertical scanning inclination angle; the three-dimensional spatial model includes the insulator string area, the busbar connection area, the transformer casing surface and the circuit breaker assembly, wherein the insulator string area is discretized by a cylindrical coordinate system, the busbar connection area is fitted by a B-spline curve, the transformer casing surface is parameterized by non-uniform rational B-splines (NURBS), and the circuit breaker assembly is decomposed into multiple sub-point cloud blocks by rigid components.

[0022] Furthermore, the point cloud completion method based on feature similarity includes:

[0023] Set the point cloud curvature difference threshold δ to obtain the missing area point cloud set P under the current perspective missing ;

[0024] Extract the adjacent view point cloud set P adjacent Zhong and P missing The subset of point clouds with curvature matching is used to filter candidate point clouds using the following similarity metric formula:

[0025]

[0026] Among them, S is the similarity measure, p i 、p j is a point in the point cloud, k is the curvature of the point cloud; project the matching point cloud to the current view coordinate system and fill it to P missing , generating a complete registration reference surface.

[0027] Furthermore, the scanning path planning method based on curvature difference includes:

[0028] Calculate the maximum curvature gradient of the current missing area and the adjacent point cloud The deflection angle θ of the acquisition device is determined by the following formula:

[0029]

[0030] in, is the maximum curvature gradient, are the partial derivatives of curvature in the x and y directions respectively; adjust the acquisition equipment to scan in the θ direction until the curvature consistency index C of the coverage area reaches C min .

[0031] Furthermore, the point cloud registration algorithm based on feature weights includes:

[0032] Define the point cloud feature weight matrix W, initialize the registration transformation matrix T, the iteration error threshold ε, and the maximum number of iterations N max ;

[0033] For the kth iteration, calculate the transformed point cloud set Q k =T k P, calculate the current iteration error E(T k ):

[0034]

[0035] Among them, w i is the feature weight of the i-th point, is the point cloud coordinate after transformation, is the target point cloud coordinate, T is the registration transformation matrix, ε is the iterative error threshold, N max is the maximum number of iterations;

[0036] If E(T k )<ε or k≥N max , output the optimal transformation matrix T opt ; Otherwise, solve T by singular value decomposition k+1 , iterative execution until convergence, where ε is the preset iterative error threshold, N max is the maximum number of allowed iterations.

[0037] Furthermore, the curvature-constrained 3D reconstruction algorithm includes:

[0038] The registered point cloud is downsampled to a voxel grid and the surface is smoothed using the moving least squares (MLS) method;

[0039] Construct the implicit surface function F(x,y,z)=0 and solve the surface gradient field through Poisson's equation:

[0040]

[0041] in, is the Laplace operator, F is the implicit surface function, It is the normal vector field of the point cloud; the isosurface is extracted to generate the initial mesh, and the mesh topology is optimized using the edge collapse algorithm to retain the geometric features at the curvature mutation point.

[0042] Furthermore, the optimized grid model needs to undergo electrical safety verification, including:

[0043] Calculate the minimum air gap d between the device surface and adjacent conductors min , if d min <d safe (safety threshold), the violation area is marked in the model and warning information is output; otherwise, the final 3D model and BOM list are generated, where d min The minimum air gap between the equipment surface and the adjacent conductor, d safe The preset safety threshold.

[0044] In a second aspect of the present invention, a system for fusion of multi-view point cloud data of substation equipment is provided, comprising:

[0045] The point cloud acquisition module is used to obtain the original point cloud data set of substation equipment from multiple perspectives and synchronously update the topological relationship of the equipment surface features through the point cloud acquisition device;

[0046] A registration reference plane generation module is used to set multiple sets of point cloud registration reference planes based on the spatial distribution characteristics of the original point cloud data set, so that the multiple sets of point cloud registration reference planes match the geometric structure of the substation equipment;

[0047] a point cloud registration module, configured to call a feature weight-based point cloud registration algorithm to rigidly align the plurality of point cloud data sets; and

[0048] The verification output module is used to obtain the registration error parameters of each group of the multiple groups of point cloud data, and call the three-dimensional reconstruction algorithm based on curvature constraints to generate a high-precision three-dimensional mesh model of the substation equipment.

[0049] The above embodiments of the present invention have at least the following beneficial effects:

[0050] 1. Through multi-view point cloud acquisition and intelligent registration reference surface dynamic generation technology, the scanning blind area problem caused by the complex structure of substation equipment is effectively solved, and non-destructive and complete modeling of key areas such as insulator strings and busbar connection parts is achieved. The geometric integrity and topological accuracy of the 3D model are improved, providing a reliable digital foundation for subsequent equipment status analysis.

[0051] 2. The innovative method of combining a feature-weighted adaptive registration algorithm with a dynamic verification of curvature consistency overcomes the matching deviation problem of traditional registration technology when processing equipment surface features. In particular, it achieves sub-millimeter registration accuracy on complex structures such as transformer housing surfaces and circuit breaker components, thereby improving the geometric fidelity of the 3D reconstructed model.

[0052] 3. By integrating the curvature-constrained reconstruction algorithm with intelligent safety gap detection technology, the digital expression of the electrical safety characteristics of the equipment surface is realized simultaneously during the 3D modeling process, solving the problems of low efficiency and subjective errors in traditional manual detection. It provides an intelligent analysis method for the insulation performance evaluation and operation status monitoring of substation equipment, and improves the reliability and safety of power equipment operation and maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] The above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily apparent by reading the following detailed description with reference to the accompanying drawings, in which several embodiments of the present invention are shown by way of example and not limitation, in which:

[0054] Figure 1 A schematic diagram of the process of a method for fusion of multi-view point cloud data for 3D modeling of substation equipment provided by one embodiment of the present invention;

[0055] Figure 2 A schematic diagram of the structure of a multi-view point cloud data fusion 3D modeling system for substation equipment provided by one embodiment of the present invention;

[0056] Figure 3 The figure schematically shows the structure of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0057] The principles and spirit of the present invention will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are provided solely to enable those skilled in the art to better understand and implement the present invention, and are not intended to limit the scope of the present invention in any way. Rather, these embodiments are provided to make the present invention more thorough and complete, and to fully convey the scope of the present invention to those skilled in the art.

[0058] Those skilled in the art will appreciate that the embodiments of the present invention may be implemented as a system, apparatus, device, method, or computer program product. Therefore, the present invention may be implemented in the following forms: entirely in hardware, entirely in software (including firmware, resident software, microcode, etc.), or in a combination of hardware and software.

[0059] It should be noted that any number of elements in the drawings is for illustration only and not for limitation, and any naming is only for distinction and does not have any limiting meaning.

[0060] Reference below Figure 1 , Figure 1 This is a flow chart of a method for fusion 3D modeling of multi-view point cloud data for substation equipment provided by an embodiment of the present invention. Figure 1As shown, a 3D modeling method for substation equipment fusion of multi-view point cloud data includes:

[0061] S1. Obtain the original point cloud dataset of the substation equipment from multiple perspectives, and synchronously update the topological relationship of the equipment surface features through the point cloud acquisition device;

[0062] S2. Based on the spatial distribution characteristics of the original point cloud data set, setting multiple sets of point cloud registration reference planes so that the multiple sets of point cloud registration reference planes match the geometric structure of the substation equipment;

[0063] S3, calling a feature weight-based point cloud registration algorithm to rigidly align the multiple sets of point cloud data; and

[0064] S4. Obtaining a registration error parameter of each of the multiple groups of point cloud data, and calling a curvature-constrained three-dimensional reconstruction algorithm to generate a high-precision three-dimensional mesh model of the substation equipment.

[0065] It should be noted that the present invention proposes a 3D modeling method for substation equipment using multi-view point cloud data fusion. Its core lies in generating a high-precision 3D mesh model through the acquisition, registration, and fusion of multi-view point cloud data. This process first requires acquiring a raw point cloud dataset of the substation equipment from multiple viewpoints. Point cloud data here refers to a collection of points on the equipment surface acquired using a 3D scanning device, with each point containing spatial coordinate information. Simultaneously, the point cloud acquisition device synchronously updates the topological relationships of the equipment surface features, namely, the geometric features and connectivity of the equipment surface, to ensure real-time and accurate data. Next, based on the spatial distribution characteristics of the raw point cloud dataset, multiple sets of point cloud registration datum planes are set to match the geometric structure of the substation equipment. Spatial distribution characteristics here refer to the spatial distribution of the point cloud data, including its density and distribution range. By invoking a feature-weighted point cloud registration algorithm, the multiple sets of point cloud data are rigidly aligned, adjusting their position and orientation to align them spatially. Finally, the registration error parameters of each set of point cloud data are obtained, and a curvature-constrained 3D reconstruction algorithm is called to generate a high-precision 3D mesh model of the substation equipment. The curvature constraint here means that the curvature information of the point cloud data is considered during the model reconstruction process to ensure the geometric features and details of the model.

[0066] Specifically, the multi-view point cloud data in the present invention refers to a set of point cloud data obtained by scanning the substation equipment from different angles. The point cloud data of each view contains the surface information of the equipment at that view. When setting the point cloud registration reference plane, it is necessary to consider the geometric structure of the equipment, such as the shape, size and surface features of the equipment. For the point cloud registration algorithm based on feature weights, feature weights refer to the weight values ​​assigned to each point according to the geometric features of the point cloud, such as curvature and density. These weight values ​​are used to improve the accuracy of the registration during the point cloud registration process. When obtaining the registration error parameters, these parameters include the distance error and angle error between the point clouds, which are used to evaluate the quality of the registration. The three-dimensional reconstruction algorithm based on curvature constraints uses the curvature information of the point cloud to constrain the surface shape of the model during the model reconstruction process to ensure that the geometric features and details of the model are accurately reconstructed. In actual applications, the settings of these algorithms and parameters need to be adjusted according to the specific substation equipment and scanning environment to achieve the best modeling effect.

[0067] Preferably, when acquiring the raw point cloud dataset for substation equipment, a high-precision 3D laser scanner can be used to scan the data to ensure data accuracy and integrity. When setting the reference plane for point cloud registration, the position and orientation of the reference plane can be determined based on the device's geometry, such as feature points like edges and corners. When invoking a feature-weighted point cloud registration algorithm, the curvature and density of each point can be calculated, and then weighted accordingly. For example, points with greater curvature, such as edges or protrusions, can be assigned higher weights to ensure accurate alignment of these important features during the registration process. In a curvature-constrained 3D reconstruction algorithm, the curvature of the point cloud is calculated to construct an implicit surface function. The surface gradient field is then solved using the Poisson equation. Finally, isosurfaces are extracted to generate an initial mesh. The mesh topology is then optimized using an edge-collapse algorithm to preserve geometric features at locations with sudden changes in curvature. In practice, these steps can be implemented using professional 3D modeling software. Input parameters include point cloud data and device geometry parameters, and the output is a high-precision 3D mesh model.

[0068] In some embodiments, setting multiple sets of point cloud registration reference planes based on the spatial distribution characteristics of the original point cloud dataset includes:

[0069] Determine whether the point cloud data under the current viewing angle covers the complete surface of the device and whether there are overlapping areas between adjacent point clouds, and update the topological relationship of the device surface features. When there is complete coverage and overlapping areas, directly select the current point cloud data point set and set the alignment reference plane;

[0070] When there is incomplete coverage but overlapping areas, adjust the point cloud acquisition perspective and rescan the missing areas to update the point cloud dataset;

[0071] When there is complete coverage but no overlapping area, the point cloud completion method based on feature similarity is called to extract supplementary point clouds from adjacent perspectives and add them to the current point cloud dataset. Then, it is determined whether the point cloud dataset meets the registration conditions. When the conditions are met, the current point cloud data set is selected and the registration reference plane is set. Otherwise, the scanning path planning method based on curvature difference is called to obtain the optimal acquisition angle θ to adjust the posture of the point cloud acquisition device. After updating the point cloud dataset, it returns to the step of calling the point cloud registration algorithm based on feature weights.

[0072] It should be noted that when performing three-dimensional modeling of multi-view point cloud data fusion of substation equipment, the present invention pays special attention to the spatial distribution characteristics of the original point cloud data set, and sets multiple sets of point cloud registration reference planes accordingly. The purpose of this process is to ensure that the point cloud data can accurately reflect the geometric structure of the equipment, thereby improving the accuracy of modeling. Specifically, it is necessary to determine whether the point cloud data under the current perspective can completely cover the surface of the equipment, and whether there are overlapping areas between adjacent point clouds. The overlapping area here refers to the part where point cloud data from different perspectives overlap each other in space, which is crucial for subsequent point cloud registration. At the same time, it is also necessary to synchronously update the topological relationship of the equipment surface features, that is, the geometric features and connection relationships of the equipment surface, to ensure the accuracy and integrity of the data. If the point cloud data from the current perspective cannot completely cover the surface of the equipment, or there is no overlapping area, corresponding measures need to be taken, such as adjusting the point cloud acquisition perspective or adopting point cloud completion methods, to ensure the quality and availability of the point cloud data.

[0073] Specifically, the spatial distribution characteristics in the present invention refer to the distribution of point cloud data in space, including the density and distribution range of the point cloud. When judging whether the point cloud data covers the complete surface of the device, it can be evaluated by analyzing the spatial distribution range of the point cloud data and the actual size of the device. For example, if a part of the device is not covered by the point cloud data, then it can be considered that the point cloud data does not completely cover the surface of the device. For the overlapping area of ​​adjacent point clouds, it can be determined by calculating the spatial intersection between point cloud data from different perspectives. If there is an overlapping area, the current point cloud data point set can be directly selected and the alignment reference plane can be set. If there is no overlapping area, it is necessary to call the point cloud completion method based on feature similarity to extract supplementary point clouds from adjacent perspectives and add them to the current point cloud data set. The feature similarity here refers to the degree of similarity between the geometric features in the point cloud data, such as curvature, density, etc. By calculating the similarity of these features, point cloud data similar to the missing area of ​​the current perspective can be extracted from the point cloud data of adjacent perspectives to fill the missing part. When determining whether a point cloud dataset satisfies registration, it is necessary to calculate the curvature consistency index C of the overlapping area of ​​the point cloud. This is a parameter that measures the degree of curvature similarity of the point cloud data in the overlapping area. If C is greater than or equal to the preset curvature threshold Cmin , then it can be considered that the point cloud data meets the registration conditions.

[0074] Preferably, when setting the point cloud registration reference plane, the original point cloud data can be preprocessed first, such as removing noise points, filling holes, etc., to improve the data quality. When judging whether the point cloud data covers the complete surface of the device, three-dimensional space modeling software can be used to assist in the analysis. By comparing the point cloud data with the three-dimensional model of the device, the coverage can be intuitively judged. When calculating the curvature consistency index C, algorithms such as the moving least squares method MLS can be used to calculate the curvature of the point cloud, and then the value of C can be determined through statistical analysis. If C is less than C min , indicating that the curvature consistency of the point cloud data in the overlapping area is insufficient, and it is necessary to further adjust the point cloud acquisition perspective or adopt a point cloud completion method. When adjusting the point cloud acquisition perspective, the optimal acquisition angle θ can be calculated based on the geometric structure of the device and the distribution of the point cloud data. Then, the posture of the point cloud acquisition device can be adjusted, the missing area can be rescanned, and the point cloud dataset can be updated. During the point cloud completion process, a curvature difference threshold δ can be set to filter the point cloud subset that matches the curvature of the missing area. Through these specific operation steps and parameter settings, the quality of the point cloud data can be effectively improved, providing a reliable data foundation for subsequent 3D modeling.

[0075] In some embodiments, determining whether the point cloud dataset satisfies the registration condition includes:

[0076] Calculate the curvature consistency index C of the overlapping area of ​​the point cloud. If C≥C min , then it is determined that the registration conditions are met and a multi-view point cloud registration sequence is generated;

[0077] Otherwise, a point cloud completion instruction is generated and the method returns to the step of calling the point cloud completion method based on feature similarity, where C is the curvature consistency index, C min is the preset curvature threshold.

[0078] It should be noted that the present invention uses the curvature consistency index C as a key parameter when judging whether the point cloud data set meets the registration conditions. This index is used to evaluate whether the geometric features of the point cloud data in the overlapping area are similar enough to determine whether point cloud registration can be performed. Specifically, when the calculated curvature consistency index C is greater than or equal to the preset curvature threshold C min When , it indicates that the geometric features of the point cloud data in the overlapping area have high consistency, meet the registration conditions, and can generate a multi-view point cloud registration sequence. On the contrary, if C is less than C min, it is necessary to trigger the point cloud completion command and return to the point cloud completion method based on feature similarity to further improve the point cloud data. This method effectively ensures the quality of the point cloud data and the accuracy of the registration, providing a solid foundation for subsequent 3D modeling.

[0079] Specifically, the curvature consistency index C is a quantitative parameter that measures the degree of similarity of geometric features of point cloud data in the overlapping area. When calculating C, it is necessary to calculate the curvature of the point cloud data in the overlapping area. Curvature is a geometric property that describes the degree of curvature of the point cloud surface. The preset curvature threshold C min It is a benchmark value used to determine whether the point cloud data meets the registration conditions. It is usually set according to the geometric complexity of the equipment and the modeling accuracy requirements. For example, for substation equipment with complex geometric structures, C min It needs to be set high to ensure high registration accuracy. In practice, the process of calculating C involves calculating the curvature of each point in the overlapping area of ​​the point cloud data, and then statistically analyzing the consistency of these curvature values. If the degree of consistency is high, that is, the curvature values ​​of most points are similar, the C value is large; conversely, if the degree of consistency is low, the C value is small. In this way, it is possible to accurately determine whether the point cloud data meets the registration requirements.

[0080] Preferably, when calculating the curvature consistency index C, algorithms such as the moving least squares method MLS can be used to calculate the curvature of the point cloud. The specific steps include: first, performing local fitting on the point cloud data in the overlapping area to obtain the curvature estimate of each point; then, calculating the standard deviation or variance of these curvature values ​​to evaluate the consistency of the curvature. If the standard deviation or variance is small, it means that the curvature values ​​are more concentrated and the consistency is high, and the C value is large at this time; conversely, if the standard deviation or variance is large, it means that the curvature values ​​are dispersed and the consistency is low, and the C value is small at this time. In actual applications, C can be adjusted according to the specific geometric characteristics of the equipment and the modeling accuracy requirements. min For example, for modeling tasks with high accuracy requirements, C min Set it high to ensure high registration accuracy. Furthermore, when triggering the point cloud completion command, you can select an appropriate completion method based on the geometric characteristics of the missing area, such as interpolation-based or model-based completion, to improve completion accuracy and efficiency. These specific steps and parameter settings can effectively improve the quality of point cloud data, providing a reliable data foundation for subsequent 3D modeling.

[0081] In some embodiments, before obtaining the original point cloud dataset of the substation equipment at multiple perspectives, the method further includes:

[0082] Perform 3D spatial modeling of the substation equipment structure to generate the equipment local coordinate system using the following parameterized equations:

[0083]

[0084] Among them, R is the device reference radius, α is the surround scanning angle, Δ x , Δ y , Δ z is the equipment installation position offset, k is the height scaling factor, and β is the vertical scanning inclination angle; the three-dimensional spatial model includes the insulator string area, the busbar connection area, the transformer casing surface and the circuit breaker assembly, wherein the insulator string area is discretized by a cylindrical coordinate system, the busbar connection area is fitted by a B-spline curve, the transformer casing surface is parameterized by non-uniform rational B-splines (NURBS), and the circuit breaker assembly is decomposed into multiple sub-point cloud blocks by rigid components.

[0085] It should be noted that in the present invention, in order to more accurately obtain the original point cloud data set of the substation equipment, the substation equipment structure is first modeled in three-dimensional space. This process generates the local coordinate system of the equipment through parameterized equations, thereby providing an accurate reference framework for subsequent point cloud data acquisition and processing. The parameterized equation is a mathematical model used to describe the position and shape of the equipment in three-dimensional space. In this way, it can be ensured that the point cloud data matches the actual geometric structure of the equipment, improving the accuracy and efficiency of modeling. The three-dimensional space model includes various key parts of the equipment, such as the insulator string area, the busbar connection area, the transformer casing surface and the circuit breaker assembly, etc. These parts are modeled using different mathematical methods to adapt to their complex geometric shapes.

[0086] Specifically, each parameter in the parameterized equation has a clear physical meaning. For example, R represents the device base radius, which is a basic parameter of the device geometry; α is the surround scanning angle, which is used to describe the rotation angle of the scanning device in the horizontal direction; Δ x , Δ y , Δ z is the device installation position offset, these parameters are used to adjust the position of the device in three-dimensional space; k is the height scaling factor, used to adjust the size of the device in the vertical direction; β is the vertical scanning tilt angle, used to describe the tilt angle of the scanning device in the vertical direction. These parameters together define the position and shape of the device in three-dimensional space. In the three-dimensional space model, the insulator string area is discretized using a cylindrical coordinate system. This method is suitable for describing components with axisymmetric structures. The busbar connection area is fitted using B-spline curves. B-spline curves are a flexible curve fitting method suitable for describing complex curve shapes. The transformer housing surface is parameterized using non-uniform rational B-spline NURBS. NURBS is a powerful surface modeling method that can accurately represent complex free-form surfaces. The circuit breaker assembly is decomposed into multiple sub-point cloud blocks through rigid components. This method is suitable for describing complex structures composed of multiple rigid components.

[0087] Preferably, when constructing a three-dimensional spatial model, professional three-dimensional modeling software such as AutoCAD, SolidWorks or Rhinoceros can be used. These software provide a wealth of modeling tools and functions, which can easily input the above parameters to generate an accurate three-dimensional model. For example, for the insulator string area, a series of discrete points can be defined in a cylindrical coordinate system and its surface can be generated using an interpolation method. For the busbar connection area, a set of control points can be input and a curve can be generated using a B-spline curve fitting tool. For the transformer housing surface, a control point grid of the NURBS surface can be defined and the weight factor can be adjusted to obtain the desired surface shape. After the three-dimensional model is generated, it can be exported to a standard three-dimensional model file format, such as STL or OBJ, for use in subsequent point cloud data processing. In actual applications, the settings of these parameters need to be adjusted according to the specific substation equipment and scanning environment to ensure the accuracy and applicability of the model. Through these specific operation steps and parameter settings, the quality of the point cloud data can be effectively improved, providing a reliable data foundation for subsequent three-dimensional modeling.

[0088] In some embodiments, the point cloud completion method based on feature similarity includes:

[0089] Set the point cloud curvature difference threshold δ to obtain the missing area point cloud set P under the current perspective missing ;

[0090] Extract the adjacent view point cloud set P adjacent Zhong and P missing The subset of point clouds with curvature matching is used to filter candidate point clouds using the following similarity metric formula:

[0091]

[0092] Among them, S is the similarity measure, p i 、p j is a point in the point cloud, κ is the curvature of the point cloud; project the matching point cloud to the current view coordinate system and fill it to P missing , generating a complete registration reference surface.

[0093] It should be noted that, in the present invention, the point cloud completion method based on feature similarity is to solve the problem of missing point cloud data under certain viewing angles. This method analyzes the missing area point cloud set P under the current viewing angle. missing , and the point cloud set P from adjacent viewpoints adjacent Extract the subset of point clouds that match the curvature to fill the missing area. Curvature matching here means finding point cloud data with similar geometric features to the missing area by calculating the curvature of the point cloud. Similarity measurement formula S(pi ,p j ) is used to quantify the similarity between two point cloud points, where the larger the S value, the higher the similarity between the two points. In this way, the missing parts in the point cloud data can be effectively filled, improving the completeness and accuracy of the point cloud data.

[0094] Specifically, the point cloud curvature difference threshold δ is a parameter used to screen candidate point clouds, which defines the maximum allowable difference between point cloud curvatures. missing It refers to the point cloud area that cannot be scanned in the current viewing angle, and the adjacent viewing angle point cloud set P adjacent It refers to the point cloud data obtained from other perspectives, which contains the information of the missing area. The similarity measurement formula S(p i ,p j ) in the i and p j Represent two points in the point cloud, and κ represents the curvature of the point cloud, that is, the degree of curvature of the point cloud surface. By calculating the difference in curvature between the two points and substituting it into the similarity metric formula, a similarity metric S between 0 and 1 can be obtained. When the S value is large, it indicates that the curvature of the two points is similar, and they can be considered to match in terms of geometric features. In this way, a subset of point clouds that match the curvature of the missing region can be filtered from the point cloud set of adjacent viewpoints and projected into the coordinate system of the current viewpoint to fill the missing region.

[0095] Preferably, when implementing the point cloud completion method based on feature similarity, the original point cloud data can be preprocessed first, such as removing noise points and outliers, to improve data quality. When calculating the curvature of the point cloud, algorithms such as the moving least squares method MLS can be used. These algorithms can calculate the curvature of each point based on the local geometric structure of the point cloud. When screening candidate point clouds, an appropriate curvature difference threshold δ can be set according to the needs of the actual application. For example, in modeling tasks with high precision requirements, δ can be set to a smaller value to ensure that the screened point cloud has highly similar geometric features to the missing area. When projecting the matching point cloud to the current perspective coordinate system, the coordinate transformation and alignment of the point cloud need to be considered to ensure the consistency of the filled point cloud data in space. Through these specific operation steps and parameter settings, the accuracy and efficiency of point cloud completion can be effectively improved, providing a more complete and accurate data foundation for subsequent three-dimensional modeling.

[0096] In some embodiments, the curvature difference-based scanning path planning method includes:

[0097] Calculate the maximum curvature gradient of the current missing area and the adjacent point cloud The deflection angle θ of the acquisition device is determined by the following formula:

[0098]

[0099] in, is the maximum curvature gradient, are the partial derivatives of curvature in the x and y directions respectively; adjust the acquisition equipment to scan in the θ direction until the curvature consistency index C of the coverage area reaches C min .

[0100] It should be noted that the curvature difference-based scanning path planning method in this invention aims to optimize the scanning path of the point cloud acquisition device to ensure coverage of all critical areas of the device during the scanning process, particularly those with significant curvature variation. This method calculates the maximum curvature gradient between the currently missing area and the adjacent point cloud to determine the acquisition device's deflection angle θ, thereby adjusting the acquisition device's posture to scan the missing area. This process ensures the integrity and accuracy of the point cloud data, providing a high-quality data foundation for subsequent 3D modeling. In this way, the efficiency and quality of point cloud acquisition can be effectively improved, reducing data loss and duplicate scanning.

[0101] Specifically, the maximum curvature gradient It is a parameter that measures the degree of change in the curvature of the point cloud. It represents the maximum rate of change of the curvature in space. The deflection angle θ is calculated based on the maximum curvature gradient and is used to guide the scanning direction of the acquisition device. When calculating the deflection angle θ, it is necessary to consider the change of curvature in different directions, that is, the partial derivative of the curvature and These partial derivatives reflect the changing trend of curvature in the x and y directions. By calculating the ratio of these partial derivatives, the deflection angle θ can be obtained, thereby determining the optimal scanning direction of the acquisition device. In practical applications, the acquisition device needs to adjust its posture according to the calculated deflection angle θ to ensure that the missing area can be scanned. When the curvature consistency index C of the coverage area reaches the preset threshold C min , it means that the scan has covered all key areas and you can stop the supplementary scan.

[0102] Preferably, when implementing the scanning path planning method based on curvature difference, the original point cloud data can be preprocessed first to remove noise points and outliers to improve data quality. When the deflection angle θ is determined, a suitable threshold C can be set according to the actual application requirements. min For example, in modeling tasks that require high accuracy, C minSet it high to ensure scanning coverage of all critical areas. When adjusting the acquisition device's posture, an automated control system can be used to automatically adjust the device's orientation based on the calculated deflection angle θ. These specific steps and parameter settings can effectively improve the efficiency and quality of point cloud acquisition, providing a more complete and accurate data foundation for subsequent 3D modeling.

[0103] In some embodiments, the feature weight-based point cloud registration algorithm includes:

[0104] Define the point cloud feature weight matrix W, initialize the registration transformation matrix T, the iteration error threshold ε, and the maximum number of iterations N max ;

[0105] For the kth iteration, calculate the transformed point cloud set Q k =T k P, calculate the current iteration error E(T k ):

[0106]

[0107] Among them, w i is the feature weight of the i-th point, is the point cloud coordinate after transformation, is the target point cloud coordinate, T is the registration transformation matrix, ε is the iterative error threshold, N max is the maximum number of iterations;

[0108] If E(T k )<ε or k≥N max , output the optimal transformation matrix T opt ; Otherwise, solve T by singular value decomposition k+1 , iterative execution until convergence, where ε is the preset iterative error threshold, N max is the maximum number of allowed iterations.

[0109] It should be noted that the feature weight-based point cloud registration algorithm in the present invention is a method for improving the registration accuracy of point cloud data. The algorithm defines the point cloud feature weight matrix W, initializes the registration transformation matrix T, the iterative error threshold ε and the maximum number of iterations N max , to gradually optimize the alignment process of point cloud data. In each iteration, the transformed point cloud set Q is calculated k and the target point cloud set P target The weighted error function E(T) between them is used to update the registration transformation matrix T according to this error. This process continues until the error E(T) is less than the iteration error threshold ε or the maximum number of iterations N is reached. maxIn this way, the algorithm can effectively align point cloud data from different perspectives to generate a high-precision 3D model.

[0110] Specifically, the point cloud feature weight matrix W is a matrix used to store the weights of each point feature. These weights are determined by the curvature and density of the point cloud. The registration transformation matrix T is a matrix used to describe the spatial transformation of point cloud data, including operations such as translation and rotation. The iteration error threshold ε is a parameter used to control the iteration accuracy of the algorithm. When the error is less than the threshold, the registration is considered to be sufficiently accurate. The maximum number of iterations N max It is a parameter used to prevent the algorithm from falling into an infinite loop and specifies the maximum number of iterations of the algorithm. In each iteration, the transformed point cloud set Q is calculated. k and the target point cloud set P target The weighted error function E(T) between can quantify the error of the current registration. i represents the feature weight of the i-th point, Represents the coordinates of the point cloud after transformation, Represents the target point cloud coordinates. Solving T by singular value decomposition SVD can obtain the optimal registration transformation matrix T opt , thereby achieving accurate alignment of point cloud data.

[0111] Preferably, when implementing a feature-weighted point cloud registration algorithm, the raw point cloud data can first be preprocessed, such as removing noise points and outliers, to improve data quality. When defining the point cloud feature weight matrix W, the weight of each point can be calculated based on the curvature and density of the point cloud. For example, points with greater curvature, such as edges or protrusions of a device, can be assigned higher weights to ensure that these important features are accurately aligned during the registration process. When initializing the registration transformation matrix T, the identity matrix can be used as the initial value, indicating that no transformation has occurred to the point cloud data in the initial state. When setting the iterative error threshold ε, an appropriate value can be selected based on the requirements of the actual application. For example, in modeling tasks with high accuracy requirements, ε can be set to a smaller value. In each iteration, the registration transformation matrix can be gradually optimized by calculating the weighted error function E(T) and solving T using SVD. Through these specific operation steps and parameter settings, the accuracy and efficiency of point cloud registration can be effectively improved, providing a more accurate data foundation for subsequent 3D modeling.

[0112] In some embodiments, the curvature-constrained 3D reconstruction algorithm includes:

[0113] The registered point cloud is downsampled to a voxel grid and the surface is smoothed using the moving least squares (MLS) method;

[0114] Construct the implicit surface function F(x,y,z)=0 and solve the surface gradient field through Poisson's equation:

[0115]

[0116] in, is the Laplace operator, F is the implicit surface function, It is the normal vector field of the point cloud; the isosurface is extracted to generate the initial mesh, and the mesh topology is optimized using the edge collapse algorithm to retain the geometric features at the curvature mutation point.

[0117] It should be noted that the curvature-constrained 3D reconstruction algorithm in this invention is a method for generating a high-precision 3D mesh model from registered point cloud data. This algorithm reduces the data volume by downsampling the voxel grid, smoothes the point cloud surface using the moving least squares method (MLS), constructs an implicit surface function, and solves the surface gradient field using the Poisson equation. Finally, it extracts isosurfaces to generate an initial mesh, and optimizes the mesh topology using an edge-collapse algorithm. This process not only preserves the geometric features at locations with sudden changes in curvature but also improves the model's accuracy and quality, ensuring that the 3D model accurately reflects the actual geometry of the substation equipment.

[0118] Specifically, voxel grid downsampling is a data simplification technology that divides the point cloud data into a three-dimensional grid, retaining only one point in each grid cell, thereby reducing the amount of data and improving processing efficiency. Moving least squares MLS is a surface smoothing technology that smoothes the point cloud surface through local polynomial fitting to reduce noise and irregularities. The implicit surface function is a mathematical model used to describe the surface in three-dimensional space. By solving the surface gradient field by the Poisson equation, an accurate representation of the surface can be obtained. The Poisson equation is a partial differential equation used to solve the surface gradient field, where the Laplace operator represents the change in curvature of the surface and the point cloud normal vector field represents the normal direction of the point cloud. By solving the Poisson equation, the implicit surface function can be obtained, and then the isosurface can be extracted to generate the initial mesh. The edge collapse algorithm is a mesh optimization technology that optimizes the mesh topology by reducing the number of edges in the mesh while retaining important geometric features, such as details at the curvature mutation point.

[0119] Preferably, when implementing a three-dimensional reconstruction algorithm based on curvature constraints, the registered point cloud data can be preprocessed first, such as removing noise points and outliers, to improve data quality. When downsampling the voxel grid, the appropriate voxel size can be set according to the needs of the actual application. For example, in modeling tasks with high accuracy requirements, the voxel size can be set smaller. When using the moving least squares method MLS to smooth the surface, the appropriate polynomial order and local fitting range can be selected to achieve the best smoothing effect. When constructing an implicit surface function, the surface gradient field can be solved by the Poisson equation. The specific steps include calculating the normal vector field of the point cloud and then using the Poisson equation to solve the implicit surface function. When extracting the isosurface to generate the initial mesh, an isosurface extraction algorithm, such as the Marching Cubes algorithm, can be used to extract the mesh model from the implicit surface function. Finally, the mesh topology is optimized by the edge collapse algorithm, retaining the geometric features at the curvature mutation point to ensure the accuracy and quality of the model. Through these specific operation steps and parameter settings, the accuracy and efficiency of 3D reconstruction can be effectively improved, providing high-quality 3D models for subsequent equipment maintenance and safety assessment.

[0120] In some embodiments, the optimized grid model needs to undergo electrical safety verification, including:

[0121] Calculate the minimum air gap d between the device surface and adjacent conductors min , if d min <d safe (safety threshold), the violation area is marked in the model and warning information is output; otherwise, the final 3D model and BOM list are generated, where d min The minimum air gap between the equipment surface and the adjacent conductor, d safe The preset safety threshold.

[0122] It should be noted that after generating a high-precision 3D mesh model of the substation equipment, the present invention further performs an electrical safety check. This process is mainly to ensure that the generated 3D model complies with electrical safety standards in actual applications, especially whether the minimum air gap between the equipment surface and adjacent conductors meets safety requirements. By calculating the minimum air gap d between the equipment surface and adjacent conductors, the minimum air gap d between the equipment surface and adjacent conductors is calculated. min and the preset safety threshold d safe By comparing, we can determine whether the model meets the safety standards. min Less than d safe , indicating that there are safety hazards in the model, it is necessary to mark the illegal areas in the model and output warning information; otherwise, generate the final 3D model and BOM list to ensure the safety and reliability of the model.

[0123] Specifically, the minimum air gap d minRefers to the shortest air distance between the surface of the equipment and the adjacent conductor. This is a key electrical safety parameter used to prevent short circuits or discharges between electrical equipment. The preset safety threshold d_safe is a minimum allowable air gap value set according to electrical safety standards to ensure the safety of the equipment during operation. min When d is , the three-dimensional space analysis method can be used to calculate the shortest distance between the device surface and the adjacent conductor in the model. min Less than d safe , indicating that the air gap between devices is insufficient to meet safety requirements and that appropriate measures need to be taken, such as adjusting the equipment layout or adding insulation materials. The BOM is a bill of materials for equipment, containing all components and material information for equipment maintenance and management.

[0124] Preferably, when implementing electrical safety verification, professional 3D modeling and analysis software, such as SolidWorks or AutoCAD, can be used. These software provide powerful 3D spatial analysis functions, which can easily calculate the minimum air gap d between the equipment surface and adjacent conductors. min In the calculation of d min When you enter the 3D model of the device and the position information of the adjacent conductors, the software will automatically calculate the shortest air distance. min Less than d safe , the violation areas can be marked in the model, such as by highlighting or adding annotations, so that users can intuitively identify potential safety hazards. At the same time, warning information is output to remind users to take appropriate measures. min Greater than or equal to d safe , indicating that the model meets safety standards, the final 3D model and BOM can be generated. BOM generation can be achieved through the software's reporting function. Simply input the equipment's component and material information, and the software will automatically generate a detailed list. These specific steps and parameter settings effectively ensure the safety and reliability of the generated 3D model in practical applications, safeguarding the safe operation of substation equipment.

[0125] The above embodiments of the present invention have the following beneficial effects:

[0126] 1. Through multi-view point cloud acquisition and intelligent registration reference surface dynamic generation technology, the scanning blind area problem caused by the complex structure of substation equipment is effectively solved, and non-destructive and complete modeling of key areas such as insulator strings and busbar connection parts is achieved. The geometric integrity and topological accuracy of the 3D model are improved, providing a reliable digital foundation for subsequent equipment status analysis.

[0127] 2. The innovative method of combining a feature-weighted adaptive registration algorithm with a dynamic verification of curvature consistency overcomes the matching deviation problem of traditional registration technology when processing equipment surface features. In particular, it achieves sub-millimeter registration accuracy on complex structures such as transformer housing surfaces and circuit breaker components, thereby improving the geometric fidelity of the 3D reconstructed model.

[0128] 3. By integrating the curvature-constrained reconstruction algorithm with intelligent safety gap detection technology, the digital expression of the electrical safety characteristics of the equipment surface is realized simultaneously during the 3D modeling process, solving the problems of low efficiency and subjective errors in traditional manual detection. It provides an intelligent analysis method for the insulation performance evaluation and operation status monitoring of substation equipment, and improves the reliability and safety of power equipment operation and maintenance.

[0129] like Figure 2 As shown, some embodiments provide a multi-view point cloud data fusion 3D modeling system for substation equipment, the system comprising:

[0130] Point cloud acquisition module 201 is used to obtain the original point cloud data set of the substation equipment under multiple viewing angles, and synchronously update the topological relationship of the equipment surface features through the point cloud acquisition device;

[0131] A registration reference plane generating module 202 is configured to set a plurality of point cloud registration reference planes based on the spatial distribution characteristics of the original point cloud data set, so that the plurality of point cloud registration reference planes match the geometric structure of the substation equipment;

[0132] a point cloud registration module 203, configured to call a feature weight-based point cloud registration algorithm to perform rigid alignment on the plurality of sets of point cloud data; and

[0133] The verification output module 204 is used to obtain the registration error parameters of each group of the multiple groups of point cloud data, and call the three-dimensional reconstruction algorithm based on curvature constraints to generate a high-precision three-dimensional mesh model of the substation equipment.

[0134] It is understandable that the modules and references recorded in the substation equipment multi-view point cloud data fusion 3D modeling system Figure 1 The steps in the multi-view point cloud data fusion 3D modeling method for substation equipment correspond to each other. Therefore, the operations, features, and beneficial effects described above for the multi-view point cloud data fusion 3D modeling method for substation equipment are also applicable to the multi-view point cloud data fusion 3D modeling system for substation equipment and the modules included therein, and will not be repeated here.

[0135] Reference below Figure 3, which shows a schematic structural diagram of an electronic device 300 suitable for implementing some embodiments of the present invention. The electronic devices in some embodiments of the present invention may include, but are not limited to, mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 3 The terminal device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0136] like Figure 3 As shown, the electronic device 300 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 301, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 302 or a program loaded from a storage device 308 into a random access memory (RAM) 303. Various programs and data required for the operation of the electronic device 300 are also stored in the RAM 303. The processing device 301, the ROM 302, and the RAM 303 are connected to each other via a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0137] Typically, the following devices may be connected to the I / O interface 305: an input device 306 including, for example, a touch screen, a touchpad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 307 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 308 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 309. The communication device 309 may allow the electronic device 300 to communicate with other devices wirelessly or by wire to exchange data. Figure 3 The electronic device 300 is shown with various devices, but it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed instead. Figure 3 Each block shown in the figure may represent one device, or may represent multiple devices as needed.

[0138] Furthermore, the storage medium of the embodiment of the present application stores program instructions that can implement all the above methods, wherein the program instructions can be stored in the above storage medium in the form of a software product, including a number of instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) or a processor to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, or a terminal device such as a computer, a server, a mobile phone, or a tablet.

[0139] The above descriptions are merely some preferred embodiments of the present invention and an illustration of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present invention is not limited to the technical solutions formed by a specific combination of the above-mentioned technical features, but should also encompass other technical solutions formed by any combination of the above-mentioned technical features or their equivalents without departing from the above-mentioned inventive concept. For example, a technical solution formed by mutually replacing the above-mentioned features with (but not limited to) technical features having similar functions disclosed in the embodiments of the present invention.

Claims

1. A 3D modeling method for substation equipment based on multi-view point cloud data fusion, characterized in that: include: Obtain original point cloud datasets of substation equipment from multiple perspectives, and synchronously update the topological relationships of equipment surface features through point cloud acquisition equipment; Based on the spatial distribution characteristics of the original point cloud data set, setting multiple sets of point cloud registration reference planes so that the multiple sets of point cloud registration reference planes match the geometric structure of the substation equipment; Invoking a feature weight-based point cloud registration algorithm to rigidly align the multiple sets of point cloud data; as well as A registration error parameter of each group of the multiple groups of point cloud data is obtained, and a three-dimensional reconstruction algorithm based on curvature constraints is called to generate a high-precision three-dimensional mesh model of the substation equipment.

2. The method for 3D modeling of substation equipment using multi-view point cloud data fusion according to claim 1, characterized in that: Setting multiple sets of point cloud registration reference planes based on the spatial distribution characteristics of the original point cloud dataset includes: Determine whether the point cloud data under the current viewing angle covers the complete surface of the device and whether there are overlapping areas between adjacent point clouds, and update the topological relationship of the device surface features. When there is complete coverage and overlapping areas, directly select the current point cloud data point set and set the alignment reference plane; When there is incomplete coverage but overlapping areas, adjust the point cloud acquisition perspective and rescan the missing areas to update the point cloud dataset; When there is complete coverage but no overlapping area, the point cloud completion method based on feature similarity is called to extract supplementary point clouds from adjacent perspectives and add them to the current point cloud dataset. Then, it is determined whether the point cloud dataset meets the registration conditions. When the conditions are met, the current point cloud data set is selected and the registration reference plane is set. Otherwise, the scanning path planning method based on curvature difference is called to obtain the optimal acquisition angle θ to adjust the posture of the point cloud acquisition device. After updating the point cloud dataset, it returns to the step of calling the point cloud registration algorithm based on feature weights.

3. The method for 3D modeling of substation equipment using multi-view point cloud data fusion according to claim 2, characterized in that: Determining whether a point cloud dataset meets the registration conditions includes: Calculate the curvature consistency index C of the overlapping area of ​​the point cloud. If C≥C min , then it is determined that the registration conditions are met and a multi-view point cloud registration sequence is generated; Otherwise, a point cloud completion instruction is generated and the method returns to the step of calling the point cloud completion method based on feature similarity, where C is the curvature consistency index, C min is the preset curvature threshold.

4. The method for 3D modeling of substation equipment using multi-view point cloud data fusion according to claim 2, characterized in that: Before obtaining the original point cloud dataset of substation equipment from multiple perspectives, the following steps are also included: Perform 3D spatial modeling of the substation equipment structure to generate the equipment local coordinate system using the following parameterized equations: Among them, R is the device reference radius, α is the surround scanning angle, Δ x , Δ y , Δ z is the equipment installation position offset, k is the height scaling factor, and β is the vertical scanning inclination angle; the three-dimensional space model includes the insulator string area, the busbar connection area, the transformer casing surface and the circuit breaker assembly, wherein the insulator string area is discretized by the cylindrical coordinate system, the busbar connection area is fitted by the B-spline curve, the transformer casing surface is parameterized by the non-uniform rational B-spline, and the circuit breaker assembly is decomposed into multiple sub-point cloud blocks by the rigid components.

5. The method for 3D modeling of substation equipment using multi-view point cloud data fusion according to claim 3, characterized in that: The point cloud completion method based on feature similarity includes: Set the point cloud curvature difference threshold δ to obtain the missing area point cloud set P under the current perspective missing ; Extract the adjacent view point cloud set P adjacent Zhong and P missing The subset of point clouds with curvature matching is used to filter candidate point clouds using the following similarity metric formula: Among them, S is the similarity measure, p i 、p j is a point in the point cloud, k is the curvature of the point cloud; project the matching point cloud to the current view coordinate system and fill it to P missing , generating a complete registration reference surface.

6. The method for 3D modeling of substation equipment using multi-view point cloud data fusion according to claim 5, characterized in that: The scanning path planning method based on curvature difference includes: Calculate the maximum curvature gradient of the current missing area and the adjacent point cloud The deflection angle θ of the acquisition device is determined by the following formula: in, is the maximum curvature gradient, are the partial derivatives of curvature in the x and y directions respectively; adjust the acquisition equipment to scan in the θ direction until the curvature consistency index C of the coverage area reaches C min .

7. The method for 3D modeling of substation equipment using multi-view point cloud data fusion according to claim 3, characterized in that: The point cloud registration algorithm based on feature weights includes: Define the point cloud feature weight matrix W, initialize the registration transformation matrix T, the iteration error threshold ε, and the maximum number of iterations N max ; For the kth iteration, calculate the transformed point cloud set Q k =T k P, calculate the current iteration error E(T k ): Among them, w i is the feature weight of the i-th point, is the point cloud coordinate after transformation, is the target point cloud coordinate, T is the registration transformation matrix, ε is the iterative error threshold, N max is the maximum number of iterations; If E(T k )<ε or k≥N max , output the optimal transformation matrix T opt ; Otherwise, solve T by singular value decomposition k+1 , iterative execution until convergence, where ε is the preset iterative error threshold, N max is the maximum number of allowed iterations.

8. The method for 3D modeling of substation equipment using multi-view point cloud data fusion according to claim 7, characterized in that: The three-dimensional reconstruction algorithm based on curvature constraint includes: The registered point cloud is downsampled to a voxel grid and the surface is smoothed using the moving least squares method; Construct the implicit surface function F(x,y,z)=0 and solve the surface gradient field through Poisson's equation: in, is the Laplace operator, F is the implicit surface function, It is the normal vector field of the point cloud; the isosurface is extracted to generate the initial mesh, and the mesh topology is optimized using the edge collapse algorithm to retain the geometric features at the curvature mutation point.

9. The method for 3D modeling of substation equipment using multi-view point cloud data fusion according to claim 8, characterized in that: The optimized grid model needs to undergo electrical safety verification, including: Calculate the minimum air gap d between the device surface and adjacent conductors min , if d min <d safe , then mark the illegal area in the model and output warning information; otherwise, generate the final 3D model and BOM list, where d min The minimum air gap between the equipment surface and the adjacent conductor, d safe The preset safety threshold.

10. A 3D modeling system for substation equipment using multi-view point cloud data fusion, characterized in that: include: The point cloud acquisition module is used to obtain the original point cloud data set of substation equipment from multiple perspectives and synchronously update the topological relationship of the equipment surface features through the point cloud acquisition device; A registration reference plane generation module is used to set multiple sets of point cloud registration reference planes based on the spatial distribution characteristics of the original point cloud data set, so that the multiple sets of point cloud registration reference planes match the geometric structure of the substation equipment; A point cloud registration module, configured to call a feature weight-based point cloud registration algorithm to perform rigid alignment on the plurality of sets of point cloud data; as well as The verification output module is used to obtain the registration error parameters of each group of the multiple groups of point cloud data, and call the three-dimensional reconstruction algorithm based on curvature constraints to generate a high-precision three-dimensional mesh model of the substation equipment.

Images