Electric power facility three-dimensional model lightweight method based on curved surface reconstruction

Abstract

The invention relates to the technical field of digital modeling, in particular to an electric power facility three-dimensional model lightweight method based on curved surface reconstruction. The method comprises the following steps: collecting electric power facility point cloud data, electric power facility image data and electric power facility monitoring tilt attitude data, and carrying out point cloud-image space matching and tilt registration optimization processing to generate point cloud-image space matching data; based on the registration point cloud-image space data, carrying out sampling dimension reduction processing on the electric power facility point cloud to generate electric power facility point cloud dimension reduction data; according to the electric power facility point cloud dimension reduction data, three-dimensional grid model reconstruction and electric power facility curved surface lightweight model construction processing are carried out, and an electric power facility curved surface lightweight model is generated; and executing operation and maintenance simulation operation of the electric power facility curved surface lightweight model through the electric power facility curved surface lightweight model. According to the invention, lightweight processing of the three-dimensional model reflecting each structure operation and maintenance place of the electric power facility is realized.

Description

Technical Field

[0001] This invention relates to the field of digital modeling technology, and in particular to a lightweight method for three-dimensional models of power facilities based on surface reconstruction. Background Technology

[0002] With the continuous development and intelligent transformation of the power industry, the scale of power facilities is constantly expanding and their structure is becoming increasingly complex. To achieve efficient management, precise operation and maintenance, and fault prediction of power facilities, 3D modeling technology, with its advantage of intuitively presenting the spatial structure and operating status of facilities, has become an important means of digital construction in the power sector. By constructing 3D models of power facilities, operation and maintenance personnel can more clearly grasp information such as equipment layout and connection relationships, providing strong support for scientifically formulating maintenance plans and quickly locating fault points. However, existing lightweight methods for 3D power facility models based on surface reconstruction still have significant limitations. Due to the wide variety of power facilities and the significant differences in structure and function between different types, lightweight 3D power facility models generally lack flexibility and adaptability, making it difficult to perform personalized modeling for the characteristics of various facilities and failing to meet the diverse application needs of the power industry. Summary of the Invention

[0003] Based on this, the present invention provides a lightweight method for three-dimensional models of power facilities based on surface reconstruction to solve at least one of the above-mentioned technical problems.

[0004] To achieve the above objectives, a lightweight method for 3D models of power facilities based on surface reconstruction is proposed, comprising the following steps:

[0005] Step S1: Acquire power facility structure data; design and process dynamically adjustable UAV scanning parameters based on power facility structure data to generate dynamically adjustable UAV scanning parameter information; control the UAV equipment to collect power facility point cloud data, power facility image data, and power facility monitoring tilt attitude data through dynamically adjusted UAV scanning parameter information.

[0006] Step S2: Perform point cloud-image spatial matching processing based on power facility point cloud data and power facility image data to generate point cloud-image spatial matching data; use power facility monitoring tilt attitude data to optimize the tilt attitude accuracy registration of point cloud-image spatial matching data to generate registered point cloud-image spatial data.

[0007] Step S3: Calculate the curvature of the local spatial point cloud based on the registered point cloud-image spatial data to generate local spatial point cloud curvature data; perform sampling and dimensionality reduction processing on the power facility point cloud of the local spatial point cloud curvature data to generate power facility point cloud dimensionality reduction data; perform 3D mesh model reconstruction processing based on the power facility point cloud dimensionality reduction data to generate a power facility 3D mesh model.

[0008] Step S4: Divide the 3D mesh model of the power facility into a multi-level mesh structure to generate multi-level mesh structure data of the power facility; based on the multi-level mesh structure data of the power facility, perform surface lightweight model construction processing of the power facility to generate a surface lightweight model of the power facility.

[0009] Step S5: Use the power facility operation and maintenance monitoring cloud platform to perform operation and maintenance monitoring of power facilities to obtain power facility operation and maintenance monitoring data; based on the power facility operation and maintenance monitoring data, transfer the power facility surface lightweight model to execute the operation and maintenance simulation operation of the power facility surface lightweight model.

[0010] Furthermore, step S1 includes the following steps:

[0011] Step S11: Obtain power facility structure data;

[0012] Step S12: Identify key structural data of power facilities through power facility structure data; and perform structural characteristic analysis of power facilities based on power facility structure data to generate power facility structural characteristic data, wherein the power facility structural characteristic data includes key structural data of power facilities, outline characteristic data of power facilities, and geometric characteristic data of power facilities.

[0013] Step S13: Design and process the UAV scanning parameters for dynamic adjustment based on the structural characteristic data of power facilities, and generate information on dynamically adjusted UAV scanning parameters;

[0014] Step S14: Perform dynamic scanning operation on the UAV equipment based on the dynamically adjusted UAV scanning parameter information, wherein the UAV equipment includes a high-precision laser point cloud device, a high-precision image monitoring device, and a tilt sensor;

[0015] Step S15: Use the drone equipment performing the drone dynamic scanning operation to collect point cloud data of power facilities, image data of power facilities, and tilt attitude data of power facilities monitoring respectively.

[0016] Furthermore, step S2 includes the following steps:

[0017] Step S21: Extract time-series synchronized feature nodes from the power facility point cloud data and the power facility image data;

[0018] Step S22: Use fuzzy logic algorithm to perform fuzzy logic transformation on the time-series synchronized feature nodes of power facility point cloud data and power facility image data to generate fuzzy logic data with feature node precision deviation.

[0019] Step S23: Calculate the point cloud-image transformation matrix based on the fuzzy logic data of feature node precision deviation to generate the point cloud-image transformation matrix;

[0020] Step S24: Perform point cloud-image spatial matching processing on the power facility point cloud data and power facility image data through the point cloud-image transformation matrix to generate point cloud-image spatial matching data; use the power facility monitoring tilt attitude data to optimize the tilt attitude accuracy registration of the point cloud-image spatial matching data to generate registered point cloud-image spatial data.

[0021] Furthermore, step S3 includes the following steps:

[0022] Step S31: Perform intelligent identification processing on the registered point cloud-image spatial data to generate monitoring power facility structure type data;

[0023] Step S32: Perform adaptive local spatial partitioning processing on the registered point cloud-image spatial data to generate adaptive local spatial data of power facilities;

[0024] Step S33: Calculate the curvature of the local spatial point cloud based on the adaptive local spatial data of the power facilities to generate local spatial point cloud curvature data;

[0025] Step S34: Set the sampling rate of power facility structure based on the monitored power facility structure type data;

[0026] Step S35: Perform sampling and dimensionality reduction processing on the local spatial point cloud curvature data using the power facility structure sampling rate to generate power facility point cloud dimensionality reduction data;

[0027] Step S36: Perform 3D mesh model reconstruction processing based on the dimensionality reduction data of power facility point cloud to generate a 3D mesh model of power facility.

[0028] Furthermore, step S32 includes the following steps:

[0029] Step S321: Set the spatial node entropy attribute of power facility structure type based on the monitored power facility structure type data;

[0030] Step S322: Analyze the spatial node density entropy value of the registered point cloud-image spatial data by using the spatial node entropy value attribute of the power facility structure type, and generate spatial node density entropy value data;

[0031] Step S323: Perform adaptive local spatial partitioning of power facilities on the registered point cloud-image spatial data using spatial node density entropy data to generate adaptive local spatial data of power facilities.

[0032] Furthermore, step S4 includes the following steps:

[0033] Step S41: Based on the monitored power facility structure type data, perform multi-level grid structure division of the power facility three-dimensional grid model to generate multi-level grid structure data of the power facility, which includes primary main grid structure data, secondary connecting grid structure data and tertiary detailed grid structure data.

[0034] Step S42: Reconstruct the surface of the primary main mesh structure based on the primary main mesh structure data to generate primary main surface structure data;

[0035] Step S43: Extract spatial node features of the secondary connected mesh structure based on the secondary connected mesh structure data to obtain secondary connected mesh structure feature data;

[0036] Step S44: Perform tertiary detail mesh structure attribute parameter mapping processing based on the tertiary detail mesh structure data to generate tertiary detail mesh structure mapping data;

[0037] Step S45: The lightweight surface model of the power facility is constructed by using the primary main surface structure data, the secondary connected mesh structure feature data, and the tertiary detailed mesh structure mapping data.

[0038] Furthermore, step S42 includes the following steps:

[0039] Step S421: Extract the point cloud coordinates of the primary main grid structure from the primary main grid structure data to obtain the point cloud coordinate data of the primary main grid structure;

[0040] Step S422: Initialize the main mesh structure surface using the preset main structure surface order and first-level main mesh structure point cloud coordinate data, and generate the main mesh structure surface parameters;

[0041] Step S423: Optimize the surface parameters of the main mesh structure to obtain the optimized surface parameters of the main mesh structure;

[0042] Step S424: Optimize the surface parameters of the main mesh structure to reconstruct the surface of the primary main mesh structure and generate primary main surface structure data.

[0043] Furthermore, step S423 includes the following steps:

[0044] Based on the point cloud coordinate data of the primary main mesh structure, the geometric distance error and normal vector error of the main mesh structure are calculated respectively, and the geometric distance error data and normal vector error data are obtained respectively.

[0045] The bidirectional error function of the main mesh structure is defined by using geometric distance error data and normal vector error data to obtain the bidirectional error function of the main mesh structure;

[0046] The surface parameters of the main mesh structure are optimized by using a two-way error function of the main mesh structure to obtain the optimized surface parameters of the main mesh structure.

[0047] Furthermore, step S43 includes the following steps:

[0048] The coordinates of the nodes of the secondary connected mesh structure data are extracted to obtain the coordinate data of the nodes of the secondary connected mesh. The spatial difference of the nodes of the secondary connected mesh is calculated to generate the spatial difference data of the nodes of the secondary connected mesh. Based on the spatial difference data of the nodes of the secondary connected mesh, the spatial node features of the secondary connected mesh structure data are extracted to obtain the feature data of the secondary connected mesh structure.

[0049] Furthermore, step S44 includes the following steps:

[0050] The coordinates of the nodes in the tertiary detailed mesh structure data are extracted to obtain the node coordinate data. The attribute parameters of the tertiary detailed mesh structure are analyzed based on the data to generate the tertiary detailed mesh structure parameters. The mapping values ​​of the tertiary detailed mesh structure parameters to the node coordinate data are then processed to generate the mapping data.

[0051] The beneficial effects of this application are as follows: By deeply analyzing the structural data of power facilities, this invention accurately identifies key structural data and comprehensively analyzes structural characteristics, generating structural characteristic data containing key structures, contours, and geometric features, providing a reliable basis for the design of UAV scanning parameters. Based on this, dynamically adjusting the UAV scanning parameters allows for the reasonable setting of parameters such as scanning height, speed, and angle according to the structural characteristics of different power facilities. This ensures that the high-precision laser point cloud equipment, high-precision image monitoring equipment, and tilt sensor carried by the UAV can efficiently collect comprehensive and accurate power facility point cloud data, image data, and monitor tilt attitude data, avoiding data redundancy or omission of key information due to unreasonable parameters, thus building a solid data foundation for subsequent model construction. The invention extracts temporal synchronization feature nodes from the power facility point cloud data and image data, establishing a key connection for matching the two types of data. Fuzzy logic algorithms are used to convert the precision deviation of feature nodes, effectively handling the uncertainties and errors in laser point cloud data and images. Based on this, the transformation matrix of the point cloud and image is calculated, and spatial matching of the point cloud and image is completed, enabling precise spatial correspondence between the point cloud and image data. By combining tilt attitude monitoring data of power facilities with tilt attitude accuracy registration optimization, errors caused by equipment attitude changes are further eliminated, significantly improving the matching accuracy and consistency of point cloud and image spatial data. This generates high-quality registered point cloud-image spatial data, providing accurate and reliable data support for subsequent point cloud processing and 3D model reconstruction, ensuring the accuracy and reliability of model construction. Intelligent analysis of the registered point cloud-image spatial data enables precise identification of power facility structure types, providing a targeted basis for subsequent processing. Based on the structure type data, spatial node entropy attributes are set, and after spatial node density entropy analysis, adaptive local spatial partitioning is achieved, allowing point cloud curvature calculation to focus on key areas, closely matching the actual structural characteristics of power facilities. By setting the sampling rate based on the structure type data, local spatial point cloud curvature data is sampled and dimensionality reduced, effectively reducing the amount of point cloud data while retaining key information, thus lowering data processing complexity and storage costs. Finally, a 3D mesh model is reconstructed based on the dimensionality-reduced data. The generated model ensures both accuracy and efficiency, laying a solid foundation for subsequent model optimization and application. Based on the structural type data of power facilities, the 3D mesh model is divided into multi-level structures, subdividing the model into a primary main mesh, a secondary connecting mesh, and a tertiary detailed mesh structure. This hierarchical processing method can accurately represent the different structural levels of power facilities. For each level of mesh structure, surface reconstruction, feature extraction, and attribute parameter mapping are performed. By accurately extracting point cloud coordinates, initializing and optimizing surface parameters, and combining error calculation and bidirectional error function optimization, high accuracy of the main mesh structure surface reconstruction is ensured. Node coordinate extraction and spatial difference calculation are used to accurately obtain the features of the connecting mesh structure. Key attributes of the detailed mesh structure are preserved through attribute parameter analysis and mapping.Finally, a lightweight surface model is constructed by integrating data from all levels. This significantly reduces the model's data volume while maintaining its integrity and detail, improving its performance in storage, transmission, and visualization applications. Real-time monitoring of power facilities using a power facility operation and maintenance monitoring cloud platform efficiently collects various operation and maintenance monitoring data, covering multiple dimensions such as equipment operating status and environmental parameters. This data is then transmitted to the lightweight surface model of the power facility to perform operation and maintenance simulations. Leveraging the efficient data processing capabilities of the lightweight model, the operating status of power facilities under different conditions can be quickly simulated, intuitively presenting potential faults and equipment performance trends. This combination not only achieves deep integration of operation and maintenance data with the 3D model but also provides power facility operation and maintenance personnel with a visual and interactive decision support tool. This helps in the timely formulation of scientific and reasonable operation and maintenance strategies, improving the intelligence level and efficiency of power facility operation and maintenance management, and ensuring the stable and reliable operation of the power system.

[0052] Therefore, this invention achieves accurate data collection on different types of power facilities by acquiring structural data of power facilities and analyzing their structural characteristics, and dynamically adjusting the scanning parameters of UAVs. In the data processing stage, fuzzy logic algorithms and adaptive local spatial partitioning techniques are used to match the point clouds and images of monitored power facilities, perform point cloud dimensionality reduction, and reconstruct 3D mesh models based on the characteristics of various power facility data. Based on the monitored power facility structural type data, a multi-level mesh structure is partitioned, and the main, connecting, and detailed mesh structures are processed layer by layer to construct a lightweight curved surface model that fits the characteristics of different facilities. Combined with an operation and maintenance monitoring cloud platform, data and model linkage is achieved to conduct operation and maintenance simulation operations. For power facilities with different structures and functions, personalized processing is achieved throughout the entire process from data acquisition to model construction and application, significantly improving the flexibility and adaptability of lightweight 3D power facility models. This meets the diverse application needs of the power industry and effectively solves the problem that existing technologies cannot perform personalized modeling for the characteristics of various facilities. Attached Figure Description

[0053] Figure 1 This is a schematic diagram of the steps of a method for lightweighting a three-dimensional model of a power facility based on surface reconstruction according to the present invention.

[0054] Figure 2 for Figure 1 A detailed flowchart illustrating the implementation steps of step S2.

[0055] Figure 3 for Figure 1 A detailed flowchart illustrating the implementation steps of step S3.

[0056] Figure 4 for Figure 1 A detailed flowchart illustrating the implementation steps of step S4.

[0057] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0058] The technical method of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0059] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. Functional entities may be implemented in software, or in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods. The term "and / or" as used herein includes any and all combinations of one or more of the associated items listed.

[0060] To achieve the above objectives, please refer to Figures 1 to 4 This invention provides a lightweight method for three-dimensional models of power facilities based on surface reconstruction. In the embodiments of this invention, please refer to... Figure 1 The diagram shown is a flowchart illustrating the steps of a method for lightweighting a 3D power facility model based on surface reconstruction according to the present invention. The method includes the following steps:

[0061] Step S1: Acquire power facility structure data; design and process dynamically adjustable UAV scanning parameters based on power facility structure data to generate dynamically adjustable UAV scanning parameter information; control the UAV equipment to collect power facility point cloud data, power facility image data, and power facility monitoring tilt attitude data through dynamically adjusted UAV scanning parameter information.

[0062] In this embodiment of the invention, power facility structural data is retrieved from the substation design drawing archive. This data is stored in the form of CAD drawings and includes detailed design parameters for equipment such as transformers, circuit breakers, disconnectors, and transmission lines. Examples include the transformer's external dimensions (5 meters long, 3 meters wide, and 4 meters high), the connection relationships of each component, and the height of the transmission line towers (20 meters) and the crossarm spacing (1.5 meters). Simultaneously, the power facility structural data is supplemented and improved by combining the actual installation location coordinates (latitude, longitude, and altitude) of the equipment recorded in the power facility operation and maintenance management system. Based on the power facility structural data, dynamically adjusted UAV scanning parameters are designed and processed. For transformers that are relatively large and structurally simple, the UAV scanning height is set to 8 meters (12 meters) above the top of the transformer, the flight speed is 4 meters per second, and the scanning angle is set to 360 degrees horizontally and 0-90 degrees vertically to ensure comprehensive coverage of the transformer's top, sides, and bottom. To ensure scanning accuracy, the scanning frequency of the high-precision laser point cloud device was set to 1500 Hz, achieving a point cloud density of 150 points / square meter. The shooting resolution of the high-precision image monitoring device was set to 6000×4000 pixels, with a shooting interval of 0.6 seconds. For power transmission line towers, considering their slender and vertically distributed characteristics, the drone was set to perform a circling scan at a distance of 3 meters from the tower, with the flight speed reduced to 2 meters per second. The scanning angle was adjusted to 360 degrees horizontally and 15-85 degrees vertically, and the emission power of the laser point cloud device was increased to 90W to enhance the scanning effect on small parts. These parameter settings generated dynamically adjusted drone scanning parameter information. This dynamically adjusted drone scanning parameter information was then input into a DJI Matrice 350RTK drone equipped with a high-precision laser point cloud device (such as Riegl VUX-SYS), a high-precision image monitoring device (such as Sony ILCE-7RM4A), and a tilt sensor (such as XSens MTi-G-710). The drone scans the substation's power facilities according to a pre-set route and parameters. During the scan, a laser point cloud device emits a laser beam at a frequency of 1500 Hz to collect point cloud data of the power facility's surface in real time; an image monitoring device captures high-definition images at 0.6-second intervals; and a tilt sensor records the drone's tilt attitude data in real time at a frequency of 100 Hz, including roll, pitch, and yaw angles. After the scan is completed, the drone transmits the collected power facility point cloud data (PCD format), image data (JPEG format), and monitored tilt attitude data (CSV format) to the ground data processing center via a 5G data transmission module.

[0063] Step S2: Perform point cloud-image spatial matching processing based on power facility point cloud data and power facility image data to generate point cloud-image spatial matching data; use power facility monitoring tilt attitude data to optimize the tilt attitude accuracy registration of point cloud-image spatial matching data to generate registered point cloud-image spatial data.

[0064] In this embodiment of the invention, point cloud data and image data of transformer power facilities collected from a substation are used as examples to extract time-synchronous feature nodes from the point cloud data and image data. Each point in the point cloud data records a collection timestamp (accurate to milliseconds), and the EXIF ​​information of the image data also includes the shooting time (accurate to milliseconds). Data segments with timestamp errors within ±80 milliseconds are selected. In these data segments, locations with obvious geometric features, such as the edge endpoints of the transformer's heat sink and the lifting lugs on the top of the oil tank, are used as feature nodes. The three-dimensional coordinates (X, Y, Z) of these feature nodes are extracted from the point cloud data, and the pixel coordinates (u, v) of the corresponding feature nodes are identified in the image data using an edge detection algorithm. A fuzzy logic algorithm is used to perform fuzzy logic transformation on the time-synchronous feature nodes to account for feature node accuracy deviations. Taking the transformer heat sink endpoint feature node as an example, the actual spatial position calculated from the three-dimensional coordinates of this feature node in the point cloud data is compared with the spatial position after pixel coordinate transformation in the image data. Position deviation thresholds are set: deviations within 0-6 cm are defined as "low deviation"; deviations within 6-18 cm are defined as "medium deviation"; and deviations greater than 18 cm are defined as "high deviation". Simultaneously, time synchronization error thresholds are set: errors within ±80 milliseconds are "high synchronization", errors within 80-200 milliseconds are "medium synchronization", and errors greater than 200 milliseconds are "low synchronization". A fuzzy logic rule table is established based on the degree of deviation and synchronization. For example, when the deviation is "low deviation" and the synchronization is "high synchronization", the output fuzzy logic value is "excellent"; when the deviation is "high deviation" and the synchronization is "low synchronization", the output is "poor". This method performs fuzzy logic transformation on the feature node precision deviation, generating fuzzy logic data for the feature node precision deviation. Then, the point cloud-image transformation matrix is ​​calculated based on the feature node precision deviation fuzzy logic data. Taking multiple transformer feature nodes as an example, for each feature node, the three-dimensional coordinates (X1, Y1, Z1) in the point cloud data and the transformed three-dimensional coordinates (X2, Y2, Z2) in the image data are used as input. Employing the least squares principle, the rotation matrix R (containing rotation angle parameters around the X, Y, and Z axes) and translation vector T (translation distance parameters in the X, Y, and Z directions) are continuously adjusted to minimize the sum of squared coordinate errors of all feature nodes in the point cloud space and image space. Specifically, the point cloud coordinates are transformed using the matrix operation [X2,Y2,Z2,1]=[X1,Y1,Z1,1]×[R|T]. After multiple iterations, when the sum of squared errors converges to below 0.02 square centimeters, the final rotation matrix R and translation vector T are determined, thus generating the point cloud-image transformation matrix. This matrix is ​​then used to perform point cloud-image space matching processing on the point cloud data and image data of transformer power facilities.Matrix multiplication is performed on each point cloud data point (X, Y, Z) to obtain the transformed coordinates (X', Y', Z'). These transformed point cloud coordinates are then projected onto the image plane to establish a correspondence between the point cloud data and the image data, generating point cloud-image spatial matching data. Next, tilt attitude data from power facility monitoring is introduced, such as roll angle θ = 4 degrees, pitch angle φ = -2 degrees, and yaw angle ψ = 3 degrees recorded when a UAV scans a transformer. This attitude angle data is converted into rotation matrices, which are then used to rotate and translate the generated point cloud-image spatial matching data. Specifically, the point cloud coordinates in the matching data are rotated using a rotation matrix, and then translated based on the actual distance (8 meters) between the UAV and the transformer, further aligning the point cloud data and image data in space to ultimately generate registered point cloud-image spatial data.

[0065] Step S3: Calculate the curvature of the local spatial point cloud based on the registered point cloud-image spatial data to generate local spatial point cloud curvature data; perform sampling and dimensionality reduction processing on the power facility point cloud of the local spatial point cloud curvature data to generate power facility point cloud dimensionality reduction data; perform 3D mesh model reconstruction processing based on the power facility point cloud dimensionality reduction data to generate a power facility 3D mesh model.

[0066] In this embodiment of the invention, local spatial point cloud curvature calculation is performed on the registered point cloud-image spatial data. The transformer point cloud data is divided into multiple local regions according to spatial location. The size of each region is set as a spherical range with a radius of 0.5 meters centered at a certain point. For the point cloud data in each local region, one point is selected, and its geometric relationship with surrounding neighboring points (e.g., the 20 closest points to the selected point) is calculated. The curvature of the point is determined by calculating the normal vector of the local surface formed by these neighboring points and the average distance from the point to the neighboring points. If the average distance from the point to the neighboring points is small and the normal vector of the neighboring points changes significantly, it indicates that the curvature of the point is large. The above calculation is performed on all points in each local region to generate local spatial point cloud curvature data containing the curvature value of each point. The local spatial point cloud curvature data is then subjected to sampling and dimensionality reduction processing of power facility point clouds. The sampling rate is set to 20%, and an adaptive curvature sampling method is used. For areas with high curvature, such as the edges of transformer heat sinks and connection points, the sampling density is appropriately increased to preserve the detailed features of these areas; for areas with low curvature, such as the surface of the transformer's main casing, the sampling density is reduced. In practice, the point cloud data is sorted from highest to lowest curvature value. The top 10% of points with the highest curvature values ​​are selected first, and then one point is selected every four points from the remaining points, achieving a 20% sampling rate. Redundant points are removed, generating dimensionality-reduced point cloud data for the power facilities. A 3D mesh model is then reconstructed based on this dimensionality-reduced point cloud data. The Delaunay triangulation algorithm is used to select three adjacent points from the dimensionality-reduced point cloud data to form an initial triangle. Then, other points are added gradually, and based on the positional relationship between the points and existing triangles, it is determined whether a new triangle can be formed, and the new triangle cannot intersect with existing triangles. This process is repeated until all points have been processed, ultimately constructing a 3D mesh model of the transformer. Other power facilities, such as circuit breakers and switchgear, are processed in the same way to generate a 3D mesh model of the entire substation's power facilities.

[0067] Step S4: Divide the 3D mesh model of the power facility into a multi-level mesh structure to generate multi-level mesh structure data of the power facility; based on the multi-level mesh structure data of the power facility, perform surface lightweight model construction processing of the power facility to generate a surface lightweight model of the power facility.

[0068] In this embodiment of the invention, the three-dimensional mesh model of the power facility is divided into a multi-level mesh structure based on the monitored structural type data of the power facility. For transformers, the main body, such as the tank and core, which constitute the main shape and load-bearing function, is divided into a first-level main mesh structure; the pipes and supporting structures connecting the tank and radiator are divided into a second-level connecting mesh structure; and the small components such as the heat sinks and valves on the transformer are divided into a third-level detailed mesh structure. For circuit breakers, the main components such as the contacts and arc-extinguishing chamber are classified as a first-level main mesh structure; the connecting rods between the contacts and the operating mechanism are classified as a second-level connecting mesh structure; and the plating on the contact surface and the fine-tuning screws of the operating mechanism are classified as a third-level detailed mesh structure. Through this division method, the entire three-dimensional mesh model of the power facility is systematically divided into multi-level mesh structure data of the power facility, including first-level main mesh structure data, second-level connecting mesh structure data, and third-level detailed mesh structure data. Based on the multi-level mesh structure data of the power facility, a surface lightweight model of the power facility is constructed. For the first-level main mesh structure, the B-spline surface fitting method is used for surface reconstruction. Taking the transformer main tank as an example, uniformly distributed points are selected from the primary main mesh structure data as control points. B-spline surfaces are generated based on these control points. By adjusting the order and parameters of the surface, it is made to fit the point cloud data as closely as possible, generating the primary main surface structure data. For the secondary connected mesh structure, its spatial node features, such as the position and angle of connection points, are extracted. Connecting components are added to the primary main surface structure data using these features, clearly showing the connection relationships between components. For the tertiary detailed mesh structure, its attribute parameters (such as the material of the heat sink and the opening pressure of the valve) are mapped to the corresponding nodes. This detailed information is retained in the model, but its geometric representation is simplified; for example, the heat sink is replaced with a simplified planar model. Finally, the primary main surface structure data, the secondary connected mesh structure feature data, and the tertiary detailed mesh structure mapping data are integrated to generate a lightweight model of the power facility surface. This model retains the key structures and details of the power facility while reducing the amount of data and improving the model's processing efficiency.

[0069] Step S5: Use the power facility operation and maintenance monitoring cloud platform to perform operation and maintenance monitoring of power facilities to obtain power facility operation and maintenance monitoring data; based on the power facility operation and maintenance monitoring data, transfer the power facility surface lightweight model to execute the operation and maintenance simulation operation of the power facility surface lightweight model.

[0070] In this embodiment of the invention, a power facility operation and maintenance monitoring cloud platform is used to monitor and process the power facilities of the substation. This cloud platform is connected to various sensor networks within the substation to collect real-time operational data of the power facilities. For example, temperature sensors, oil level sensors, and winding deformation sensors are installed on the transformer. The temperature sensor collects transformer oil temperature data every 10 minutes, the oil level sensor monitors the oil level in the tank in real time, and the winding deformation sensor collects winding deformation data every hour. For circuit breakers, contact temperature sensors and opening / closing time monitoring sensors are installed. The contact temperature sensor collects contact temperature every 5 minutes, and the opening / closing time monitoring sensor records the time and action details of each opening / closing operation. The cloud platform summarizes and analyzes this data to generate power facility operation and maintenance monitoring data. For example, for transformer oil temperature data, the cloud platform judges it based on a preset temperature threshold (e.g., the normal operating temperature range is 40-80℃). If the oil temperature exceeds 80℃, it is marked as abnormal, and the time and temperature value of the abnormality are recorded. For circuit breaker opening and closing time data, the cloud platform analyzes whether the time of each operation is within the specified range (e.g., the opening time should be between 30-60 milliseconds). If it exceeds the range, an early warning is issued. Power facility operation and maintenance monitoring data is transmitted to a lightweight curved surface model of the power facility to perform operation and maintenance simulation operations. Taking abnormal transformer oil temperature as an example, the abnormal oil temperature data is input into the lightweight curved surface model to simulate the impact of rising oil temperature on various components of the transformer. Through heat conduction simulation algorithms, the temperature distribution changes of components such as the transformer core and windings after the oil temperature rise are calculated, predicting the location and extent of possible thermal faults. Simultaneously, based on the simulation results, areas where faults may occur are displayed in the model, highlighted in red, providing operation and maintenance personnel with intuitive fault warning information. For cases of abnormal circuit breaker opening and closing times, the model simulates the impact of abnormal opening and closing times on contact wear, arc extinguishing effect, etc., predicting possible fault types (e.g., contact erosion, arc extinguishing failure, etc.) and providing corresponding maintenance suggestions, such as adjusting the preload of the opening and closing springs and checking the wear condition of the contact surfaces. In this way, operation and maintenance monitoring data are used to simulate operations in a lightweight curved surface model, which can help identify potential faults in power facilities in advance and provide a scientific basis for operation and maintenance decisions.

[0071] Furthermore, step S1 includes the following steps:

[0072] Step S11: Obtain power facility structure data;

[0073] Step S12: Identify key structural data of power facilities through power facility structure data; and perform structural characteristic analysis of power facilities based on power facility structure data to generate power facility structural characteristic data, wherein the power facility structural characteristic data includes key structural data of power facilities, outline characteristic data of power facilities, and geometric characteristic data of power facilities.

[0074] Step S13: Design and process the UAV scanning parameters for dynamic adjustment based on the structural characteristic data of power facilities, and generate information on dynamically adjusted UAV scanning parameters;

[0075] Step S14: Perform dynamic scanning operation on the UAV equipment based on the dynamically adjusted UAV scanning parameter information, wherein the UAV equipment includes a high-precision laser point cloud device, a high-precision image monitoring device, and a tilt sensor;

[0076] Step S15: Use the drone equipment performing the drone dynamic scanning operation to collect point cloud data of power facilities, image data of power facilities, and tilt attitude data of power facilities monitoring respectively.

[0077] In this embodiment of the invention, taking high-voltage transmission line power facilities as an example, structural data of power facilities such as transmission line towers, insulator strings, and conductors are retrieved from the power GIS (Geographic Information System) database. The tower data stored in the GIS database includes structured information such as base coordinates (latitude and longitude, altitude), tower type (e.g., cat-head type straight tower ZMC1-33), tower height (33 meters), and crossarm dimensions (8 meters long, 0.5 meters wide). Insulator string data includes the number of insulator discs (28 discs), diameter of a single insulator disc (255 mm), and thickness (146 mm). Conductor data covers parameters such as conductor type (LGJ-400 / 50 steel-cored aluminum stranded wire), diameter (27.63 mm), and safe working tension. Simultaneously, by reading the three-dimensional CAD design drawings of the power facilities, detailed parameters of their three-dimensional geometric structure are obtained, ensuring the integrity and accuracy of the power facility structural data and providing a precise data foundation for subsequent processing. Taking transmission line towers as an example, the tower structural data obtained in step S11 is processed. Key structural data of the transmission line towers are extracted using data parsing algorithms, identifying the tower head, main tower body, and crossarms as critical structures. Taking the tower head as an example, point cloud data analysis software (such as CloudCompare) is used to identify the tower head structure by recognizing unique geometric features (such as triangular structure distribution) in the tower point cloud data. When calculating the tower's structural contour characteristics, a contour extraction algorithm is used to identify boundary points in the tower point cloud data, obtaining the tower's contour curve coordinates at a sampling interval of 0.01 meters to form contour characteristic data. When calculating the structural geometric characteristics, a three-dimensional geometric calculation method is used to measure parameters such as the length, angle, and volume of each component of the tower. For example, the length error of the main tower body is controlled within ±0.02 meters, and the angle error is controlled within ±0.5 degrees, generating power facility structural characteristic data that includes key structures, contour characteristics, and geometric characteristics. Dynamically adjustable UAV scanning parameters are designed for the structural characteristic data of transmission line towers. For the 33-meter-high cat-head shaped linear tower, based on its structural characteristics, the UAV scanning height was set to 38 meters (5 meters above the top of the tower to avoid collision risk), and the flight speed was set to 4 meters per second (to ensure the continuity and overlap of the scan data). The scanning route adopted a spiral circling method, with a horizontal scanning angle of 360 degrees and a vertical scanning angle range of 15 degrees to 75 degrees (to ensure that the top, middle, and bottom structures of the tower can be completely scanned). For the insulator string, due to its delicate structure, the UAV was set to perform local scanning at a distance of 2 meters from the insulator string, with the flight speed reduced to 1.5 meters per second, using a zigzag route, a horizontal scanning angle of 180 degrees, and the vertical scanning angle automatically adjusted according to the insulator string tilt angle (by reading the insulator string tilt parameters obtained in step S12).Simultaneously, the point cloud density of the high-precision laser point cloud device was set to 200 points / square meter, and the scanning frequency to 2000 Hz; the shooting resolution of the high-precision image monitoring device was set to 6000×4000 pixels, and the shooting interval was 0.5 seconds, generating dynamically adjustable UAV scanning parameter information containing the above parameters. A UAV was used as the scanning platform, equipped with the high-precision laser point cloud device, the high-precision image monitoring device, and a tilt sensor. The dynamically adjusted UAV scanning parameter information was input into the UAV flight control system. Before takeoff, the UAV obtained centimeter-level positioning accuracy (horizontal accuracy ±1cm, vertical accuracy ±1.5cm) through the RTK real-time dynamic positioning system. The UAV scanned power facilities such as transmission line towers and insulator strings according to the set flight path and parameters. During the scanning process, the laser point cloud device emits a laser beam at a frequency of 2000 Hz to scan the surface of the power facilities. The return time and reflection intensity of each laser pulse are recorded to form point cloud data. The image monitoring equipment captures high-definition images at 0.5-second intervals. The tilt sensor collects data such as the UAV's three-axis acceleration, angular velocity, and attitude angle in real time at a frequency of 100 Hz, ensuring that the collected power facility data is comprehensive and accurate. The collected power facility point cloud data (stored in PCD format, containing the three-dimensional coordinates X, Y, Z of each point and the reflection intensity value), image data (JPEG format, containing RGB color information), and monitored tilt attitude data (stored in CSV format, containing timestamps, three-axis acceleration values, three-axis angular velocity values, roll angle, pitch angle, and yaw angle) are transmitted to the ground data processing server via a 5G data transmission module. The server uses a RAID 5 disk array for data storage to ensure data security. During data transmission, the TCP / IP protocol is used, the data transmission buffer size is set to 128MB, and the transmission rate is kept stable at over 500Mbps to ensure that the data is transmitted to the ground quickly and completely, providing raw data support for subsequent point cloud-image spatial matching processing, model building, and other operations.

[0078] Furthermore, as an embodiment of the present invention, reference is made to... Figure 2 As shown, Figure 1 A detailed flowchart illustrating the implementation steps of step S2 is provided in this embodiment. Step S2 includes:

[0079] Step S21: Extract time-series synchronized feature nodes from the power facility point cloud data and the power facility image data;

[0080] In this embodiment of the invention, point cloud data and image data of power facilities on transmission line towers are used as examples. The point cloud data is stored in PCD format, containing the three-dimensional coordinates and reflection intensity of each point, with timestamps accurate to milliseconds. The image data is in JPEG format, with the shooting time embedded in the EXIF ​​information also accurate to milliseconds. By parsing the timestamp information in the data files, data segments with the same timestamp or an error within ±50 milliseconds are selected. In these data segments, locations with obvious geometric features, such as the top of the tower head and the end of the crossarm, are used as feature nodes. The three-dimensional coordinates of these feature nodes are extracted from the point cloud data. In the image data, an edge detection method is used to determine the pixel coordinates of the corresponding feature nodes by identifying the turning points of the tower outline, thereby extracting the time-synchronized feature nodes of the power facility point cloud data and image data.

[0081] Step S22: Use fuzzy logic algorithm to perform fuzzy logic transformation on the time-series synchronized feature nodes of power facility point cloud data and power facility image data to generate fuzzy logic data with feature node precision deviation.

[0082] In this embodiment of the invention, taking the feature node at the top of a tower as an example, the actual position calculated from the three-dimensional coordinates of the feature node in the point cloud data is compared with its spatial position after pixel coordinate conversion in the image data. Due to equipment accuracy and measurement errors, there is a positional deviation between the two. A positional deviation threshold is set: a deviation within 0-5 cm is defined as "low deviation"; within 5-15 cm is defined as "medium deviation"; and greater than 15 cm is defined as "high deviation". Simultaneously, a time synchronization error threshold is set: an error within ±50 milliseconds is defined as "high synchronization", 50-150 milliseconds as "medium synchronization", and greater than 150 milliseconds as "low synchronization". A fuzzy logic rule table is established based on the degree of deviation and synchronization. For example, when the deviation is "low deviation" and the synchronization is "high synchronization", the output fuzzy logic value is "excellent"; when the deviation is "high deviation" and the synchronization is "low synchronization", the output is "poor". This method performs fuzzy logic conversion on the feature node precision deviation to generate fuzzy logic data for the feature node precision deviation.

[0083] Step S23: Calculate the point cloud-image transformation matrix based on the fuzzy logic data of feature node precision deviation to generate the point cloud-image transformation matrix;

[0084] In this embodiment of the invention, based on the fuzzy logic data of feature node precision deviation, taking multiple feature nodes at the tower head and crossarm as examples, for each feature node, the three-dimensional coordinates (X1, Y1, Z1) in the point cloud data and the transformed three-dimensional coordinates (X2, Y2, Z2) in the image data are used as input. Utilizing the principle of least squares, by continuously adjusting the rotation matrix R (containing rotation angle parameters around the X, Y, and Z axes) and the translation vector T (translation distance parameters in the X, Y, and Z directions), the sum of squared coordinate errors of all feature nodes in the point cloud space and image space is minimized. Specifically, the point cloud coordinates are transformed using the matrix operation [X2, Y2, Z2, 1] = [X1, Y1, Z1, 1] × [R|T] (where [R|T] is the transformation matrix containing the rotation matrix and translation vector). After multiple iterative calculations, when the sum of squared errors converges to below a set threshold (e.g., 0.01 square centimeters), the final rotation matrix R and translation vector T are determined, thereby generating the point cloud-image transformation matrix.

[0085] Step S24: Perform point cloud-image spatial matching processing on the power facility point cloud data and power facility image data through the point cloud-image transformation matrix to generate point cloud-image spatial matching data; use the power facility monitoring tilt attitude data to optimize the tilt attitude accuracy registration of the point cloud-image spatial matching data to generate registered point cloud-image spatial data.

[0086] In this embodiment of the invention, taking the data processing of power facilities in a substation as an example, after obtaining the point cloud-image transformation matrix in step S23, it is applied to the point cloud data and image data of the power facilities of the substation transformer. The point cloud data contains the three-dimensional coordinates (X, Y, Z) and reflection intensity information of the surface points of each component of the transformer, while the image data records the RGB color information of the transformer's appearance. For each point cloud data point, the original three-dimensional coordinates (X, Y, Z) are multiplied by the point cloud-image transformation matrix through matrix multiplication to obtain the transformed three-dimensional coordinates (X', Y', Z'). The transformed point cloud coordinates (X', Y', Z') are converted into pixel coordinates (u, v) on the image plane according to the projection principle of the image. Specifically, the transformation process is as follows: using the principle of similar triangles, combined with the camera's intrinsic parameter matrix (including parameters such as focal length and optical center coordinates, assuming that the camera's intrinsic parameter matrix has known fixed parameters, such as focal length f = 50mm, optical center coordinates (cx = 320, cy = 240)), the three-dimensional coordinates are projected onto the two-dimensional image plane. For example, for a point cloud data point (X', Y', Z'), its pixel coordinates (u, v) on the image are calculated using the formulas u = f*X' / Z' + cx and v = f*Y' / Z' + cy. This establishes the correspondence between point cloud data and image data, generating point cloud-image spatial matching data. Next, optimization is performed using tilt attitude data monitored by power facilities. When the UAV scans a transformer, its onboard tilt sensor records a roll angle θ = 5 degrees, a pitch angle φ = -3 degrees, and a yaw angle ψ = 2 degrees. These angle data are converted into rotation matrices. The calculation of the rotation matrix follows the Euler angle rotation principle: first, the yaw angle ψ is rotated around the Z-axis, then the pitch angle φ is rotated around the Y-axis, and finally the roll angle θ is rotated around the X-axis, resulting in the corresponding rotation matrix R. For each point cloud coordinate (X', Y', Z') in the point cloud-image spatial matching data, rotation is performed using matrix multiplication (X”, Y”, Z”) = R*(X', Y', Z'). Simultaneously, based on the actual distance d = 8 meters between the UAV and the transformer (assuming this distance was determined during scanning parameter settings), the rotated coordinates are translated, with the translation vector T = [0, 0, d]. The final adjusted coordinates are obtained using (X”', Y”', Z”') = (X”, Y”, Z”) + T. This process is repeated for all point cloud coordinates to achieve more precise spatial alignment between the point cloud data and the image data, ultimately generating registered point cloud-image spatial data.

[0087] Furthermore, as an embodiment of the present invention, reference is made to... Figure 3 As shown, Figure 1 A detailed flowchart illustrating the implementation steps of step S3 is provided in this embodiment. Step S3 includes:

[0088] Step S31: Perform intelligent identification processing on the registered point cloud-image spatial data to generate monitoring power facility structure type data;

[0089] In this embodiment of the invention, the registered point cloud-image spatial data includes data on facilities such as wind turbine towers, nacelles, blades, and transmission lines. By analyzing the spatial distribution and geometric features of the point cloud data, towers that are cylindrical with a height much greater than their diameter (e.g., 80 meters high and 4 meters in diameter) are identified as tower structure types based on their regular cylindrical geometric shape and size proportions. Nacelles with a specific top shape, containing components such as generators and gearboxes, and with dimensions of approximately 10 meters long, 4 meters wide, and 3 meters high, are identified as nacelle structure types based on the unique shape and size of their combined components. Blades are identified as blade structure types based on their slender, twisted shape (up to 50 meters long and 2 meters wide at maximum). Transmission lines are identified based on the slender and continuously distributed point cloud features. These identification results are compiled into monitoring power facility structure type data, clearly recording the structure type name, corresponding point cloud data range, and key geometric parameters of each facility.

[0090] Step S32: Perform adaptive local spatial partitioning processing on the registered point cloud-image spatial data to generate adaptive local spatial data of power facilities;

[0091] In this embodiment of the invention, based on the monitored power facility structure type data, a spatial node entropy attribute is set for the tower structure type. Since the tower surface is relatively smooth and the spatial node distribution is relatively uniform, the node density weight in the spatial node entropy attribute is set to 0.3, and the node distribution uniformity weight is set to 0.7. Spatial node density entropy data is obtained by calculating the spatial distance between each node and its 10 adjacent nodes in the point cloud data. If the average distance between nodes in a certain area is 0.2 meters and the standard deviation of the distance is less than 0.05 meters, then the node distribution in that area is considered uniform. Based on the spatial node density entropy data, the tower point cloud data is divided into local spaces of 10 meters in height, resulting in 8 adaptive local spaces. For the nacelle structure, due to the complexity of its internal components, the node density weight is set to 0.7, and the node distribution uniformity weight is set to 0.3. According to the function and spatial location of the components, the nacelle point cloud data is divided into 5 adaptive local spaces, such as the generator compartment and gearbox compartment, generating adaptive local space data for the power facility.

[0092] Step S33: Calculate the curvature of the local spatial point cloud based on the adaptive local spatial data of the power facilities to generate local spatial point cloud curvature data;

[0093] In this embodiment of the invention, taking the adaptive local spatial data of the power facilities of a wind farm tower as an example, within each local space, for each point in the point cloud data, 20 neighboring points within a radius of 0.5 meters are selected. The curvature of the local spatial point cloud is determined by calculating the normal vector of the plane formed by the point and its neighboring points, and the average distance from the point to its neighboring points. Specifically, the normal vector of the plane formed by three adjacent points is calculated using the vector cross product. If the angle between two adjacent normal vectors changes more significantly, and the average distance from the point to its neighboring points is smaller, it indicates a greater curvature at that point. For example, at the flange connection at the top of the tower, due to structural changes, the curvature of the point cloud is significantly higher than that of the tower body. The above calculation is performed on all points within each local space to generate local spatial point cloud curvature data containing the curvature value of each point, accurately recording the position of each point in space and the corresponding curvature magnitude.

[0094] Step S34: Set the sampling rate of power facility structure based on the monitored power facility structure type data;

[0095] In this embodiment of the invention, based on the monitoring data of power facility structure types, a lower structural sampling rate of 10% is set for the wind turbine tower of a wind farm, as its structure is relatively regular and mainly used for support; for the nacelle interior, due to the presence of numerous precision components, a higher structural sampling rate of 30% is set to ensure model accuracy; and for the blades, considering the importance of their surface shape to aerodynamic performance, a sampling rate of 20% is set. The sampling rate is set according to the functional importance and geometric complexity of each structural type. The more critical the function and the more complex the geometry, the higher the sampling rate, to ensure that key information is retained during the lightweighting process, generating power facility structure sampling rate data corresponding to each structural type.

[0096] Step S35: Perform sampling and dimensionality reduction processing on the local spatial point cloud curvature data using the power facility structure sampling rate to generate power facility point cloud dimensionality reduction data;

[0097] In this embodiment of the invention, based on the local spatial point cloud curvature data of wind farm power facilities, for tower structures, points with higher curvature values ​​and uniformly distributed points are selected from the local spatial point cloud curvature data at a 10% structural sampling rate. Specifically, the tower point cloud data is first sorted by curvature value from largest to smallest, and the top 5% of points with the highest curvature values ​​are selected. Then, one point is selected every nine points from the remaining points to achieve a 10% sampling rate. Redundant points are removed to generate dimensionality-reduced point cloud data for the tower's power facilities. For nacelle structures, at a 30% sampling rate, points reflecting the outlines of internal components and connections, as well as points with significant curvature changes, are prioritized. Points in empty internal areas are removed to complete the sampling and dimensionality reduction of the nacelle point cloud data. For blade structures, at a 20% sampling rate, points in key areas of blade surface curvature changes and points uniformly distributed along the blade's length and width are retained to generate dimensionality-reduced point cloud data for the blade's power facilities, effectively reducing the amount of point cloud data while preserving key model features.

[0098] Step S36: Perform 3D mesh model reconstruction processing based on the dimensionality reduction data of power facility point cloud to generate a 3D mesh model of power facility.

[0099] In this embodiment of the invention, the power point cloud dimensionality reduction data of power facilities is used as input. For the point cloud dimensionality reduction data of the tower, a triangular meshing method is adopted to connect the points in the point cloud data into triangular patches according to their spatial position. Starting from the bottom, three adjacent points are connected into triangles in sequence. The validity of the mesh is ensured by judging whether the newly generated triangle intersects with the existing triangles. During the connection process, based on the curvature information of the points, the triangular patches are densified for areas with large curvature, such as the tower flange connection, so that the mesh fits the actual shape better. For the point cloud dimensionality reduction data of the nacelle and blades, the triangular meshing method is also adopted. Based on the geometric features and point cloud distribution of each component, a three-dimensional mesh model is constructed. Finally, the three-dimensional mesh models of each component such as the tower, nacelle, and blades are integrated to generate a complete three-dimensional mesh model of the wind farm power facility, accurately presenting the geometric structure and spatial morphology of the power facility.

[0100] Furthermore, step S32 includes the following steps:

[0101] Step S321: Set the spatial node entropy attribute of power facility structure type based on the monitored power facility structure type data;

[0102] Step S322: Analyze the spatial node density entropy value of the registered point cloud-image spatial data by using the spatial node entropy value attribute of the power facility structure type, and generate spatial node density entropy value data;

[0103] Step S323: Perform adaptive local spatial partitioning of power facilities on the registered point cloud-image spatial data using spatial node density entropy data to generate adaptive local spatial data of power facilities.

[0104] In this embodiment of the invention, after acquiring data on the structural types of monitored power facilities, spatial node entropy attributes are set for different structural types. For transformer structures, due to their smooth outer surface and complex internal structure, a node density weight of 0.4 and a node distribution uniformity weight of 0.6 are set. For transmission line tower structures, considering their obvious vertical distribution characteristics and large differences in the size of each component, a node density weight of 0.6 and a node distribution uniformity weight of 0.4 are set. For insulator string structures, due to their slender shape and high structural repeatability, a node density weight of 0.3 and a node distribution uniformity weight of 0.7 are set. Simultaneously, a node density calculation radius of 0.3 meters and a uniformity calculation reference distance of 1 meter are defined. These parameters are combined into spatial node entropy attributes for power facility structural types for subsequent analysis. Based on the set spatial node entropy attributes, the substation registration point cloud-image spatial data are analyzed. Taking transformer point cloud data as an example, during the node density calculation, the number of neighboring points is counted within a spherical area with a radius of 0.3 meters, centered on each point. If a point has 15 neighboring points, its node density is 15 / (4 / 3*π*0.3*0.3*0.3). When calculating the node distribution uniformity, 10 neighboring points within 1 meter of the current point are selected, and the standard deviation of the distance between these points is calculated. If the standard deviation is 0.1 meters, the distribution uniformity index for that point is calculated according to a preset formula. The node density and distribution uniformity are weighted and summed with weights of 0.4 and 0.6, respectively, to obtain the spatial node density entropy value for each point. After completing the above calculations for all transformer point cloud data, spatial node density entropy value data covering the entropy value of each point is generated. The same processing is performed on the point cloud data of transmission line towers and insulator strings. Based on the spatial node density entropy value data, the substation registration point cloud-image spatial data is divided. For transformer point cloud data, points with similar entropy values ​​and adjacent spatial locations are grouped into a local space. For example, points with entropy values ​​between 0.6 and 0.8 and a spatial distance of no more than 0.5 meters are grouped into the local space of the transformer casing; points with entropy values ​​between 0.2 and 0.4 are grouped into the local space of the transformer core and windings. For transmission line tower point cloud data, based on entropy value changes and distribution along the height, the ground foundation (with high and stable entropy values) is grouped into a local space, the tower body is grouped into different local spaces every 5 meters in height with entropy value fluctuations within ±0.1, and the crossarm is grouped into a separate local space. For insulator string point cloud data, based on the standard of every 10 insulators and an entropy value difference of less than 0.2, it is divided into multiple elongated local spaces, ultimately generating adaptive local spatial data of power facilities including transformers, transmission line towers, and insulator strings.

[0105] Furthermore, as an embodiment of the present invention, reference is made to... Figure 4 As shown, Figure 1A detailed flowchart illustrating the implementation steps of step S4 is provided in this embodiment. Step S4 includes:

[0106] Step S41: Based on the monitored power facility structure type data, perform multi-level grid structure division of the power facility three-dimensional grid model to generate multi-level grid structure data of the power facility, which includes primary main grid structure data, secondary connecting grid structure data and tertiary detailed grid structure data.

[0107] In this embodiment of the invention, based on monitoring data on the structural types of power facilities, the completed three-dimensional mesh model of substation power facilities is divided. The main components constituting the external shape and primary function, such as the transformer casing and oil tank, as well as the cabinet frame of large high-voltage switchgear, are classified as the primary main mesh structure. The busbars connecting the transformer and the high-voltage switchgear, cable trays, and supporting steel frames between various devices are classified as the secondary connection mesh structure. Smaller components, such as heat sinks on the transformer and indicator lights and buttons on the high-voltage switchgear, are classified as the tertiary detailed mesh structure. By clearly defining the level of each component, multi-level structural data of the power facility mesh, including primary main mesh structure data, secondary connection mesh structure data, and tertiary detailed mesh structure data, is generated, achieving systematic layering of the three-dimensional mesh model of the substation power facilities.

[0108] Step S42: Reconstruct the surface of the primary main mesh structure based on the primary main mesh structure data to generate primary main surface structure data;

[0109] In this embodiment of the invention, the primary main grid structure data of a substation transformer is used as an example for surface reconstruction. The point cloud coordinate data of the primary main grid structure of the transformer shell is extracted, and the X, Y, and Z coordinate values ​​of each point on the shell surface in three-dimensional space are accurately obtained, with the error controlled within ±0.01 meters. Based on a preset primary structure surface order of second order, these point cloud coordinate data are used to initialize the primary grid structure surface, initially constructing the surface shape of the transformer shell. Subsequently, by calculating the vertical distance from each point cloud coordinate to the initialized surface, and using the minimum total distance as the optimization objective, the positions of the surface control points are manually adjusted to gradually make the surface conform to the point cloud data. After multiple adjustments, when the average distance from all point cloud coordinates to the surface is less than 0.005 meters, the optimized surface parameters of the primary grid structure are determined. Finally, based on these parameters, the surface reconstruction of the primary main grid structure is completed, generating primary main surface structure data that accurately reflects the shape of the transformer shell.

[0110] Step S43: Extract spatial node features of the secondary connected mesh structure based on the secondary connected mesh structure data to obtain secondary connected mesh structure feature data;

[0111] In this embodiment of the invention, spatial node features are extracted from the secondary connection grid structure data of the busbars connecting transformers and high-voltage switchgear in a substation. First, the node coordinate data of the secondary connection grid structure of the busbars are extracted to obtain the three-dimensional coordinates of each connection point and bend point of the busbars. Then, the spatial differences between adjacent nodes in the X, Y, and Z directions are calculated. For example, for a horizontally oriented busbar segment, the distance difference between adjacent nodes in the Y direction (horizontal direction) and the Z direction (vertical direction) is calculated. If the distance difference between adjacent nodes in the Z direction of a certain busbar segment suddenly increases, it indicates that a bend exists at that location. Based on these spatial difference data, key features such as the connection position and bend angle of the busbars, as well as the node distribution pattern, are identified, thereby obtaining secondary connection grid structure feature data containing information such as the connection method and spatial orientation of the busbars, clearly presenting the structural characteristics of the busbars.

[0112] Step S44: Perform tertiary detail mesh structure attribute parameter mapping processing based on the tertiary detail mesh structure data to generate tertiary detail mesh structure mapping data;

[0113] In this embodiment of the invention, indicator lights on a substation high-voltage switchgear are used as three-level detailed mesh structure data for attribute parameter mapping processing. First, the node coordinate data of the indicator light's three-level detailed mesh structure is extracted, accurately recording the three-dimensional coordinates of each point on the indicator light's outer shell outline. Then, the indicator light's attribute parameters are analyzed, including its color (red indicates a fault), rated voltage (220V), and power (5W). These attribute parameters are mapped to the corresponding node coordinate data, adding attribute fields such as color, voltage, and power to the node data. For example, when describing the coordinate data of a node on the indicator light's outer shell, the indicator light is simultaneously labeled as red, rated voltage 220V, and power 5W, generating three-level detailed mesh structure mapping data containing node coordinates and detailed attribute parameters, thus closely linking the indicator light's geometric information with its functional attributes.

[0114] Step S45: The lightweight surface model of the power facility is constructed by using the primary main surface structure data, the secondary connected mesh structure feature data, and the tertiary detailed mesh structure mapping data.

[0115] In this embodiment of the invention, in a substation scenario, a lightweight surface model is constructed by integrating primary main surface structure data, secondary connecting mesh structure feature data, and tertiary detailed mesh structure mapping data. First, based on the primary main surface structure data of the transformer casing and high-voltage switchgear cabinet frame, the main model framework of the substation's main equipment is built. Next, according to the secondary connecting mesh structure feature data such as busbars, connecting components are accurately added to the main model. Following the connection positions and directions recorded in the feature data, components such as busbars and cable trays are precisely connected to the corresponding equipment. Finally, the tertiary detailed mesh structure mapping data, such as high-voltage switchgear indicator lights, is integrated into the model. Detailed components such as indicator lights are added at appropriate locations and assigned attributes such as color and voltage. By progressively integrating data from each level, a lightweight surface model of power facilities is generated that retains the key structures and details of the substation's power facilities while significantly reducing the amount of data.

[0116] Furthermore, step S42 includes the following steps:

[0117] Step S421: Extract the point cloud coordinates of the primary main grid structure from the primary main grid structure data to obtain the point cloud coordinate data of the primary main grid structure;

[0118] Step S422: Initialize the main mesh structure surface using the preset main structure surface order and first-level main mesh structure point cloud coordinate data, and generate the main mesh structure surface parameters;

[0119] Step S423: Optimize the surface parameters of the main mesh structure to obtain the optimized surface parameters of the main mesh structure;

[0120] Step S424: Optimize the surface parameters of the main mesh structure to reconstruct the surface of the primary main mesh structure and generate primary main surface structure data.

[0121] In this embodiment of the invention, the primary main mesh structure data of a power facility, namely a transformer, is processed. The primary main mesh structure covers the main parts of the transformer, such as the outer shell and the oil tank. Point cloud coordinates are extracted from the mesh structure of these parts through precise data analysis. Specifically, during data acquisition, the position of each point in three-dimensional space is recorded point by point with millimeter precision. For each point on the transformer outer shell, its X, Y, and Z coordinate values ​​are recorded. For example, at a corner point of the outer shell, its coordinates are precisely recorded as (12345, 67890, 23456), with an error controlled within ±1 millimeter. After comprehensive extraction of the entire primary main mesh structure, point cloud coordinate data containing all point cloud coordinates is obtained, providing a precise data foundation for subsequent surface initialization. Based on the preset order of the main structure surface (set to third order), the obtained transformer primary main mesh structure point cloud coordinate data is used for main mesh structure surface initialization. The third-order surface can better fit the complex shape of the transformer. Taking the circular oil tank of a transformer as an example, a series of points uniformly distributed on the surface of the tank are selected from the point cloud coordinate data. Based on the coordinate positions of these points and combined with the mathematical model of a third-order surface, the initial parameters of the surface are determined. Specifically, the curvature, center position, and other parameters of the surface are determined by calculating the spatial relationship between these points. For example, the radius and center position of the tank surface are determined by calculating the distance and angle from multiple points to a certain center point. After a series of calculations and adjustments, the surface parameters of the main mesh structure are generated. These parameters initially describe the surface morphology of the transformer's primary main mesh structure. The surface parameters of the main mesh structure are then optimized. Taking the outer shell surface of the transformer as an example, the generated surface parameters are applied to the actual point cloud coordinate data, and the deviation between each point cloud coordinate and the surface is calculated. The deviation is calculated by measuring the vertical distance from the point cloud coordinate to the surface. For example, for a point on the outer shell, the vertical distance from it to the current surface is measured. If the distance is large, it indicates that the fit between the point and the surface is poor. The surface parameters were manually adjusted with the goal of minimizing the sum of the squares of the deviations between all point cloud coordinates and the surface. By continuously fine-tuning parameters such as curvature and position, the deviations were gradually reduced. After multiple iterations, when the average deviation between all point cloud coordinates and the surface was less than 0.5 mm, the optimization goal was considered achieved, and the optimized surface parameters for the main mesh structure were obtained. At this point, the surface could more accurately fit the actual shape of the transformer's primary main mesh structure. Based on the optimized surface parameters, the primary main mesh structure of the substation transformer was reconstructed. Using the optimized surface parameters as a basis, the surfaces of the transformer's outer shell, oil tank, and other primary main components were reconstructed in three-dimensional space. For the outer shell surface, the shape of the shell was accurately drawn based on the optimized curvature and position parameters, ensuring a high degree of conformity with the actual transformer shell. During the reconstruction process, millimeters were used as the unit of precision to ensure that every detail of the surface was accurately represented.For example, every edge and corner on the outer shell is precisely drawn according to the optimized parameters. In this way, the surface reconstruction of the primary main mesh structure is completed, generating primary main surface structure data that accurately reflects the shape of the primary main body of the transformer, providing accurate main structure information for the subsequent construction of a surface lightweight model of power facilities.

[0122] Furthermore, step S423 includes the following steps:

[0123] Based on the point cloud coordinate data of the primary main mesh structure, the geometric distance error and normal vector error of the main mesh structure are calculated respectively, and the geometric distance error data and normal vector error data are obtained respectively.

[0124] The bidirectional error function of the main mesh structure is defined by using geometric distance error data and normal vector error data to obtain the bidirectional error function of the main mesh structure;

[0125] The surface parameters of the main mesh structure are optimized by using a two-way error function of the main mesh structure to obtain the optimized surface parameters of the main mesh structure.

[0126] In this embodiment of the invention, error calculation is performed using the point cloud coordinate data of the primary main mesh structure of a substation transformer casing as an example. For each point cloud coordinate, its corresponding projection point on the surface of the main mesh structure is determined. The geometric distance error is obtained by measuring the straight-line distance between the point cloud coordinate and the projection point. For example, if the straight-line distance between a point cloud coordinate and the projection point in three-dimensional space is 3 mm, then the geometric distance error of that point is recorded as 3 mm. Simultaneously, the angle between the normal vector at the point cloud coordinate and the surface normal vector at the corresponding projection point is calculated, and the angle value is used as the normal vector error. For example, if the angle between the two normal vectors is 5 degrees, then the normal vector error of that point is 5 degrees. By traversing all point cloud coordinates, complete geometric distance error data and normal vector error data are obtained. These data accurately reflect the deviations in position and direction between the point cloud and the surface. Based on the obtained geometric distance error data and normal vector error data, a bidirectional error function is defined for the main mesh structure of the substation transformer casing. The geometric distance error is weighted at 0.6, and the normal vector error at 0.4. The geometric distance error of each point is multiplied by 0.6, and the normal vector error by 0.4. The two results are then summed to obtain the comprehensive error value for that point. The comprehensive error values ​​of all points are summed to construct the bidirectional error function for the main mesh structure. For example, for 100 points on the transformer casing, the comprehensive error value for each point is calculated, and these 100 comprehensive error values ​​are summed to obtain the current value of the bidirectional error function. This function comprehensively considers the positional and directional errors between the point cloud and the surface, providing a quantitative basis for subsequent parameter optimization. Using the defined bidirectional error function for the main mesh structure, the surface parameters of the main mesh structure of the substation transformer casing are optimized. The curvature and position parameters of the surface are used as optimization objects. These parameters are manually adjusted, and after each adjustment, the geometric distance error and normal vector error of all point cloud coordinates are recalculated to obtain new bidirectional error function values. For example, the curvature parameter is first increased by 0.1, the error is recalculated, and a new bidirectional error function value is obtained, which is then compared with the value before adjustment. If the new value is less than the value before adjustment, it indicates that the adjustment direction is correct, and the adjustment continues in this direction; if the new value is greater than the value before adjustment, the adjustment direction is changed. By continuously repeating the process of adjusting parameters, calculating errors, and comparing function values, the surface parameters are continuously optimized until the bidirectional error function value reaches its minimum. The surface parameters determined at this point are the optimized surface parameters of the main mesh structure, enabling the main mesh structure surface to more accurately conform to the actual shape of the transformer shell.

[0127] Furthermore, step S43 includes the following steps:

[0128] The coordinates of the nodes of the secondary connected mesh structure data are extracted to obtain the coordinate data of the nodes of the secondary connected mesh. The spatial difference of the nodes of the secondary connected mesh is calculated to generate the spatial difference data of the nodes of the secondary connected mesh. Based on the spatial difference data of the nodes of the secondary connected mesh, the spatial node features of the secondary connected mesh structure data are extracted to obtain the feature data of the secondary connected mesh structure.

[0129] In this embodiment of the invention, the busbar connecting the transformer and the switchgear is used as the secondary connection grid structure data. The busbar grid structure consists of numerous nodes, and the coordinates of these nodes are extracted. Using a high-precision three-dimensional measuring device, with millimeters as the unit of precision, the spatial position of each node on the busbar is accurately measured. For example, at a bend in the busbar, the coordinates of a node are measured as (1234.5, 678.9, 234.5). Following the order from one end to the other and from top to bottom, all nodes on the busbar are measured sequentially, and the X, Y, and Z coordinate values ​​of each node are recorded to form complete secondary connection grid node coordinate data, providing accurate basic data for subsequent spatial difference calculations. Based on the extracted secondary connection grid node coordinate data, the spatial difference of the connection grid nodes on the busbar is calculated. Two adjacent nodes are selected, and their coordinate differences in the X, Y, and Z directions are calculated. For example, for adjacent nodes A (1234.5, 678.9, 234.5) and B (1235.2, 679.1, 234.8) on a busbar, the difference in the X direction is calculated to be 1235.2 - 1234.5 = 0.7 mm, the difference in the Y direction is 679.1 - 678.9 = 0.2 mm, and the difference in the Z direction is 234.8 mm.

[0130] -234.5 = 0.3 mm. This calculation is performed on all adjacent nodes on the busbar, recording the differences in the three directions for each pair of adjacent nodes to generate spatial difference data for the connecting grid nodes. This data reflects the spatial positional changes between nodes on the busbar. Based on the spatial difference data of the connecting grid nodes, spatial node features are extracted from the secondary connecting grid structure data of the busbar. If the spatial difference between adjacent nodes in a certain segment of the busbar is large and relatively stable in the X direction, while the differences in the Y and Z directions are small, it indicates that this segment of the busbar mainly extends in the X direction. If the differences between adjacent nodes in the X, Y, and Z directions all change significantly, it indicates that there may be a bend or branch at that location. By analyzing these difference data, spatial node features such as the connection position, bending angle, and extension direction of the busbar are identified. For example, when it is found that the difference between several consecutive adjacent nodes gradually decreases in the X direction and gradually increases in the Y direction, it can be determined that this is a bending area of ​​the busbar, and the angle and position information of this bending area are measured and recorded. These identified features were organized into secondary connected grid structure feature data, which accurately described the spatial structural characteristics of the busbar.

[0131] Furthermore, step S44 includes the following steps:

[0132] The coordinates of the nodes in the tertiary detailed mesh structure data are extracted to obtain the node coordinate data. The attribute parameters of the tertiary detailed mesh structure are analyzed based on the data to generate the tertiary detailed mesh structure parameters. The mapping values ​​of the tertiary detailed mesh structure parameters to the node coordinate data are then processed to generate the mapping data.

[0133] In this embodiment of the invention, the buttons on the high-voltage switchgear of a substation are used as the data for processing in a three-level detailed mesh structure. The three-level detailed mesh structure of the button consists of multiple triangular facets, and the vertex of each facet is a node. When extracting the coordinates of these nodes, the position of each node in three-dimensional space is precisely measured in millimeters. The X, Y, and Z coordinate values ​​of each node on the button surface are obtained using a three-dimensional measuring device. For example, the coordinates of a node on the circular surface of the top of the button are (50, 30, 20), with the error controlled within ±0.5 mm. The coordinate values ​​of all nodes are recorded sequentially to form complete three-level detailed mesh structure node coordinate data, which accurately reflects the position information of the button's geometric shape in three-dimensional space. Attribute parameter analysis is performed on the specific three-level detailed mesh structure data of the buttons on the high-voltage switchgear of the substation. The attribute information of the buttons is obtained from the equipment design documents and technical data, including the button's color (green indicates normal operation), type (self-reset button), rated operating voltage (24V), and operating current (1A). At the same time, combined with the functional requirements of the buttons in actual operation, their operating frequency, response time, and other characteristic parameters are analyzed. These attribute information are organized and categorized, clarifying the specific values ​​and meanings of each attribute. This generates a three-level detailed mesh structure parameter that includes button color, type, electrical parameters, etc., comprehensively describing the button's function and performance attributes. The three-level detailed mesh structure parameter is then mapped to the extracted node coordinate data. Taking a button on a substation high-voltage switchgear as an example, attribute fields are added to each node in the node coordinate data. For nodes representing the button surface, a color attribute field is added to its coordinate data and assigned the value "green," a type attribute field is added and assigned the value "self-reset button," and a rated operating voltage attribute field is added and assigned the value "24V," etc. In this way, the button's attribute parameters are mapped one-to-one with the corresponding node coordinates, ensuring that each node not only contains spatial location information but also is associated with corresponding functional attribute information. This ultimately generates complete three-level detailed mesh structure mapping data, achieving the fusion of geometric and attribute information, and providing detailed data support for the subsequent construction of accurate lightweight models of power facility surfaces.

[0134] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0135] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A lightweight method for three-dimensional models of power facilities based on surface reconstruction, characterized in that, Includes the following steps: Step S1: Acquire power facility structure data; design and process dynamically adjustable UAV scanning parameters based on power facility structure data to generate dynamically adjustable UAV scanning parameter information; By dynamically adjusting the drone scanning parameters, the drone equipment is controlled to collect point cloud data, image data, and tilt attitude data of power facilities. Step S2: Perform point cloud-image spatial matching processing based on power facility point cloud data and power facility image data to generate point cloud-image spatial matching data; use power facility monitoring tilt attitude data to optimize the tilt attitude accuracy registration of point cloud-image spatial matching data to generate registered point cloud-image spatial data. Step S3: Calculate the curvature of the local spatial point cloud based on the registered point cloud-image spatial data to generate local spatial point cloud curvature data; perform sampling and dimensionality reduction processing on the power facility point cloud of the local spatial point cloud curvature data to generate power facility point cloud dimensionality reduction data; perform 3D mesh model reconstruction processing based on the power facility point cloud dimensionality reduction data to generate a power facility 3D mesh model. Step S4: Divide the 3D mesh model of the power facility into a multi-level mesh structure to generate multi-level mesh structure data of the power facility; based on the multi-level mesh structure data of the power facility, perform surface lightweight model construction processing of the power facility to generate a surface lightweight model of the power facility. Step S5: Use the power facility operation and maintenance monitoring cloud platform to perform operation and maintenance monitoring of power facilities to obtain power facility operation and maintenance monitoring data; based on the power facility operation and maintenance monitoring data, transfer the power facility surface lightweight model to execute the operation and maintenance simulation operation of the power facility surface lightweight model.

2. The lightweight method for three-dimensional models of power facilities based on surface reconstruction according to claim 1, characterized in that, Step S1 includes the following steps: Step S11: Obtain power facility structure data; Step S12: Identify key structural data of power facilities through power facility structure data; and perform structural characteristic analysis of power facilities based on power facility structure data to generate power facility structural characteristic data, wherein the power facility structural characteristic data includes key structural data of power facilities, outline characteristic data of power facilities, and geometric characteristic data of power facilities. Step S13: Design and process the UAV scanning parameters for dynamic adjustment based on the structural characteristic data of power facilities, and generate information on dynamically adjusted UAV scanning parameters; Step S14: Perform dynamic scanning operation on the UAV equipment based on the dynamically adjusted UAV scanning parameter information, wherein the UAV equipment includes a high-precision laser point cloud device, a high-precision image monitoring device, and a tilt sensor; Step S15: Use the drone equipment performing the drone dynamic scanning operation to collect point cloud data of power facilities, image data of power facilities, and tilt attitude data of power facilities monitoring respectively.

3. The lightweight method for three-dimensional models of power facilities based on surface reconstruction according to claim 1, characterized in that, Step S2 includes the following steps: Step S21: Extract time-series synchronized feature nodes from the power facility point cloud data and the power facility image data; Step S22: Use fuzzy logic algorithm to perform fuzzy logic transformation on the time-series synchronized feature nodes of power facility point cloud data and power facility image data to generate fuzzy logic data with feature node precision deviation. Step S23: Calculate the point cloud-image transformation matrix based on the fuzzy logic data of feature node precision deviation to generate the point cloud-image transformation matrix; Step S24: Perform point cloud-image spatial matching processing on the power facility point cloud data and power facility image data through the point cloud-image transformation matrix to generate point cloud-image spatial matching data; use the power facility monitoring tilt attitude data to optimize the tilt attitude accuracy registration of the point cloud-image spatial matching data to generate registered point cloud-image spatial data.

4. The lightweight method for three-dimensional models of power facilities based on surface reconstruction according to claim 1, characterized in that, Step S3 includes the following steps: Step S31: Perform intelligent identification processing on the registered point cloud-image spatial data to generate monitoring power facility structure type data; Step S32: Perform adaptive local spatial partitioning processing on the registered point cloud-image spatial data to generate adaptive local spatial data of power facilities; Step S33: Calculate the curvature of the local spatial point cloud based on the adaptive local spatial data of the power facilities to generate local spatial point cloud curvature data; Step S34: Set the sampling rate of power facility structure based on the monitored power facility structure type data; Step S35: Perform sampling and dimensionality reduction processing on the local spatial point cloud curvature data using the power facility structure sampling rate to generate power facility point cloud dimensionality reduction data; Step S36: Perform 3D mesh model reconstruction processing based on the dimensionality reduction data of power facility point cloud to generate a 3D mesh model of power facility.

5. The lightweight method for three-dimensional models of power facilities based on surface reconstruction according to claim 4, characterized in that, Step S32 includes the following steps: Step S321: Set the spatial node entropy attribute of power facility structure type based on the monitored power facility structure type data; Step S322: Analyze the spatial node density entropy value of the registered point cloud-image spatial data by using the spatial node entropy value attribute of the power facility structure type, and generate spatial node density entropy value data; Step S323: Perform adaptive local spatial partitioning of power facilities on the registered point cloud-image spatial data using spatial node density entropy data to generate adaptive local spatial data of power facilities.

6. The lightweight method for three-dimensional models of power facilities based on surface reconstruction according to claim 4, characterized in that, Step S4 includes the following steps: Step S41: Based on the monitored power facility structure type data, perform multi-level grid structure division of the power facility three-dimensional grid model to generate multi-level grid structure data of the power facility, which includes primary main grid structure data, secondary connecting grid structure data and tertiary detailed grid structure data. Step S42: Reconstruct the surface of the primary main mesh structure based on the primary main mesh structure data to generate primary main surface structure data; Step S43: Extract spatial node features of the secondary connected mesh structure based on the secondary connected mesh structure data to obtain secondary connected mesh structure feature data; Step S44: Perform tertiary detail mesh structure attribute parameter mapping processing based on the tertiary detail mesh structure data to generate tertiary detail mesh structure mapping data; Step S45: The lightweight surface model of the power facility is constructed by using the primary main surface structure data, the secondary connected mesh structure feature data, and the tertiary detailed mesh structure mapping data.

7. The lightweight method for three-dimensional models of power facilities based on surface reconstruction according to claim 6, characterized in that, Step S42 includes the following steps: Step S421: Extract the point cloud coordinates of the primary main grid structure from the primary main grid structure data to obtain the point cloud coordinate data of the primary main grid structure; Step S422: Initialize the main mesh structure surface using the preset main structure surface order and first-level main mesh structure point cloud coordinate data, and generate the main mesh structure surface parameters; Step S423: Optimize the surface parameters of the main mesh structure to obtain the optimized surface parameters of the main mesh structure; Step S424: Optimize the surface parameters of the main mesh structure to reconstruct the surface of the primary main mesh structure and generate primary main surface structure data.

8. The lightweight method for three-dimensional models of power facilities based on surface reconstruction according to claim 7, characterized in that, Step S423 includes the following steps: Based on the point cloud coordinate data of the primary main mesh structure, the geometric distance error and normal vector error of the main mesh structure are calculated respectively, and the geometric distance error data and normal vector error data are obtained respectively. The bidirectional error function of the main mesh structure is defined by using geometric distance error data and normal vector error data to obtain the bidirectional error function of the main mesh structure; The surface parameters of the main mesh structure are optimized by using a two-way error function of the main mesh structure to obtain the optimized surface parameters of the main mesh structure.

9. The lightweight method for three-dimensional models of power facilities based on surface reconstruction according to claim 6, characterized in that, Step S43 includes the following steps: The coordinates of the nodes of the secondary connected mesh structure data are extracted to obtain the coordinate data of the nodes of the secondary connected mesh. The spatial difference of the nodes of the secondary connected mesh is calculated to generate the spatial difference data of the nodes of the secondary connected mesh. Based on the spatial difference data of the nodes of the secondary connected mesh, the spatial node features of the secondary connected mesh structure data are extracted to obtain the feature data of the secondary connected mesh structure.

10. The lightweight method for three-dimensional models of power facilities based on surface reconstruction according to claim 6, characterized in that, Step S44 includes the following steps: The coordinates of the nodes in the tertiary detailed mesh structure data are extracted to obtain the node coordinate data. The attribute parameters of the tertiary detailed mesh structure are analyzed based on the data to generate the tertiary detailed mesh structure parameters. The mapping values ​​of the tertiary detailed mesh structure parameters to the node coordinate data are then processed to generate the mapping data.

Images