Rapid rural energy facility modeling method based on three-dimensional laser point cloud

Abstract

The invention relates to the technical field of three-dimensional modeling, in particular to a rural energy facility rapid modeling method based on three-dimensional laser point cloud. The method comprises the following steps: acquiring point cloud data of energy facilities in a rural area by using three-dimensional laser scanning equipment to obtain point cloud data of the rural energy facilities; performing target identification according to the rural energy facility point cloud data to obtain rural energy facility target data; constructing a three-dimensional model according to the rural energy facility target data to obtain a rural energy facility three-dimensional model; and performing refined modeling according to the rural energy facility three-dimensional model to obtain a rural energy facility refined model. According to the method, the energy facility model with high precision, rich details and complete physical attributes is generated by combining geometric refinement, topological optimization and physical attribute simulation, so that the modeling efficiency is remarkably improved, and technical support is provided for various application scenes such as operation and maintenance management, fault detection and simulation analysis of facilities.

Description

Technical Field

[0001] The invention relates to the technical field of three-dimensional modeling, and in particular to a method for rapid modeling of rural energy facilities based on three-dimensional laser point clouds. Background Art

[0002] With the rapid development of the new energy industry, energy infrastructure in rural areas (such as solar power generation equipment, wind power generation equipment, transformers and distribution systems, etc.) is becoming increasingly complex and large-scale. The operation and maintenance of rural energy facilities involves a lot of inspection, monitoring and maintenance work. Due to the wide distribution of facilities and complex environmental conditions, it is very difficult to rely on traditional methods for regular inspection and maintenance, and it is difficult to achieve real-time monitoring and fault prediction, resulting in high maintenance costs and low efficiency. Summary of the invention

[0003] In order to solve the above technical problems, the present invention proposes a method for rapid modeling of rural energy facilities based on three-dimensional laser point cloud to solve at least one of the above technical problems.

[0004] The present application provides a method for rapid modeling of rural energy facilities based on three-dimensional laser point cloud, the method comprising:

[0005] S1. Collect point cloud data of energy facilities in rural areas using 3D laser scanning equipment to obtain point cloud data of rural energy facilities;

[0006] S2. Perform target recognition based on the rural energy facility point cloud data to obtain rural energy facility target data;

[0007] S3, constructing a three-dimensional model according to the target data of rural energy facilities to obtain a three-dimensional model of rural energy facilities;

[0008] S4. Perform detailed modeling based on the three-dimensional model of rural energy facilities to obtain a detailed model of rural energy facilities.

[0009] The three-dimensional laser scanning equipment in the present invention can quickly and comprehensively capture the spatial data of rural energy facilities, has the advantages of high precision and high resolution, and ensures that the geometric information of the facilities is complete. Compared with traditional measurement methods, laser scanning technology can more efficiently process large areas, especially in rural environments, reduce manual measurement errors, and improve the speed and accuracy of data acquisition. The use of computer vision and artificial intelligence technology to automatically identify point cloud data effectively reduces the workload of manual marking and classification, saving a lot of time. Using the identified facility data, a three-dimensional model of rural energy facilities can be quickly generated to show the structure and layout of the facilities, providing visual support for operation and maintenance management. The three-dimensional model is highly operable and can help users intuitively understand the geometric form and position relationship of the facilities, providing a reliable basis for subsequent optimization design, monitoring and maintenance decisions. The present invention can greatly improve the modeling speed, avoid the complexity of traditional manual modeling, and reduce the time cost of model generation. The refined modeling step further refines the preliminary three-dimensional model to ensure that the structural details, textures and materials of the facilities are more accurately restored, thereby improving the accuracy and authenticity of the model. This step can enhance the complexity and detail of the 3D model, making the model closer to the real facility. It is suitable for scenarios that require high-precision modeling, such as operation and maintenance planning, equipment inspection, and maintenance plan formulation.

[0010] Optionally, S1 includes:

[0011] S11. Use laser scanning equipment to determine the scanning range and perform multi-dimensional spatial positioning to ensure that all energy facilities in the rural area are covered and obtain scanning space coordinate data;

[0012] S12, using a three-dimensional laser scanning device to collect high-density point cloud data by scanning spatial coordinate data to obtain an initial point cloud data set;

[0013] S13, optimizing the scanning perspective of the initial point cloud data set to obtain perspective point cloud data;

[0014] S14, performing multi-source data fusion on the view point cloud data to obtain fused point cloud data;

[0015] S15. Time-synchronize the fused point cloud data to obtain rural energy facility point cloud data.

[0016] In the present invention, the working range of the scanning equipment can be accurately determined by multi-dimensional spatial positioning technology, ensuring that every detail of the facility is within the scanning range, avoiding omissions, and reducing the scanning difficulty caused by the complex terrain and irregular distribution of facilities in rural areas. By predetermining the scanning range, the system can plan the best scanning path, reduce unnecessary repeated scanning, improve efficiency and save equipment energy consumption. The three-dimensional laser scanning equipment can collect high-density point cloud data, generate high-resolution facility geometry information, and ensure that every detail of the facility (such as pipelines, cables, equipment shape, etc.) is accurately recorded. Through viewing angle optimization, the scanning blind area caused by the position or angle limitation of the equipment is reduced to ensure that all parts of the facility can be completely scanned. By fusing the data of multiple scanning devices or different scanning angles, the complete geometric data of the facility can be obtained, avoiding the problem of local data loss caused by a single perspective or device. If the energy facility is in a dynamic state (for example, the rotation of wind power generation equipment), time synchronization can ensure the real-time and accuracy of the point cloud data, and provide accurate basic data for dynamic modeling or operation and maintenance.

[0017] Optionally, S2 includes:

[0018] S21, performing preliminary filtering according to the rural energy facility point cloud data to obtain point cloud filtering data;

[0019] S22, performing density estimation and segmentation according to the point cloud filtering data to obtain point cloud segmentation data;

[0020] S23, extracting morphological features according to the point cloud segmentation data to obtain facility point cloud morphological feature data;

[0021] S24, performing energy signal analysis based on the rural energy facility point cloud morphological feature data to obtain facility point cloud energy feature data;

[0022] S25, performing multi-target clustering based on the rural energy facility point cloud energy feature data to obtain facility clustering data;

[0023] S26, performing deep target detection according to the facility clustering data to obtain facility deep detection classification data;

[0024] S27, extracting edge features according to the facility depth detection classification data to obtain facility edge feature data;

[0025] S28. Semantically annotate the rural energy facility point cloud data based on the facility edge feature data and the facility depth detection classification data to obtain the rural energy facility target data.

[0026] In the present invention, high-quality point cloud data is retained through the filtering step, making the data cleaner and providing a better basis for segmentation and feature extraction. By estimating and segmenting according to the point cloud density, the various parts of the energy facility can be effectively divided to ensure that each facility or facility component is represented by independent data. The density estimation method is suitable for facilities with complex shapes or different sizes, and can effectively segment facility parts of different shapes and sizes to improve segmentation accuracy. Through morphological feature extraction, the key geometric structures of the facility, such as cylinders (pipelines), planes (solar panels) or complex surfaces (wind power equipment), can be identified. By combining the energy signals of the facility (such as the light intensity of photovoltaic panels, the rotational energy of wind power equipment, etc.), the morphology and function of the facility can be more accurately distinguished, effectively improving the accuracy of classification and recognition. Multi-target clustering can combine morphological features and energy features, and perform clustering processing according to the different functions or features of the facility. Through clustering, the system can efficiently classify similar facility types. Deep target detection can combine morphological features, energy features and clustering information to more accurately identify specific facility types, such as transformers, wind turbines or solar panels. Edge feature extraction can capture the edge details of the facility model and ensure the geometric accuracy of the boundary part, especially when identifying facilities with clear boundaries (such as photovoltaic panels), which improves the refinement of the model. By combining deep detection classification data and edge features, high-precision semantic annotation can be performed, so that different types of facilities have clear labels and categories in the point cloud data, supporting further analysis and modeling.

[0027] Optionally, S3 includes:

[0028] S31, reconstructing the network according to the rural energy facility target data to obtain a preliminary facility network model;

[0029] S32, performing surface refinement on the preliminary facility network model to obtain a facility surface refinement model;

[0030] S33, performing texture mapping on the facility surface refinement model to obtain facility mapping data;

[0031] S34, performing geometric correction according to the facility mapping data to obtain a facility geometric reference model;

[0032] S35, embedding structural features according to the facility geometric reference model to obtain a facility embedded structural feature model;

[0033] S36. Physical attributes are assigned according to the facility embedded structural feature model to obtain a three-dimensional model of rural energy facilities.

[0034] In the present invention, through the network reconstruction method, a preliminary geometric network model can be efficiently generated from the target data of the facility, which greatly reduces the manual intervention of modeling and provides a way to quickly generate the facility framework. Through the surface refinement operation, a more refined facility surface model can be generated to capture more details, especially the complex features of the facility surface (such as seams, edges, curved surfaces, etc.). Through texture mapping, the model surface can obtain a realistic texture effect, improve the visual quality of the facility three-dimensional model, and make the model closer to the appearance of the actual facility during the display and simulation process. Texture mapping can give appropriate surface effects to different materials (such as metal, plastic, glass, etc.), so that the model is more detailed when reflecting the material properties of the facility. Through geometric correction, it can be ensured that the model geometry is more consistent with the actual facility, correct the geometric deviation caused by data noise or measurement error, and obtain a more accurate facility benchmark geometry model. Through structural feature embedding, the model not only expresses the geometric shape, but also incorporates the internal structure information of the facility, such as support frame, pipeline, connection node, etc., so that the model is not only suitable for appearance display, but also for functional analysis and structural performance simulation. By assigning physical properties (such as material density, electrical conductivity, thermal conductivity, etc.) to the three-dimensional model of the facility, the model becomes not only a tool for geometric expression, but can also be used for applications such as physical simulation and energy analysis.

[0035] Optionally, the geometric correction comprises:

[0036] Performing surface fitting geometry correction according to the facility mapping data to obtain a first facility geometry reference model;

[0037] Perform shape constraint geometry correction according to the facility mapping data to obtain a second facility geometry reference model;

[0038] Performing spatial variation geometry correction according to the facility mapping data to obtain a third facility geometry reference model;

[0039] Acquire rural energy facility image data;

[0040] Extract rural energy facility features and estimate rural energy facility parameters based on rural energy facility image data to obtain rural energy facility image feature data and rural energy facility parameter data;

[0041] The first facility geometric reference model, the second facility geometric reference model and the third facility geometric reference model are fused according to the rural energy facility image feature data and the rural energy facility parameter data to obtain a facility geometric reference model.

[0042] The geometric correction method of surface fitting in the present invention can correct the uneven area of ​​the facility model surface to ensure that the surface is smooth and conforms to the geometric characteristics of the actual facility. Surface fitting can adapt to complex surface shapes and generate more realistic facility geometric models. Shape-constrained geometric correction ensures that the facility model is consistent with the expected geometric shape by applying geometric constraints (such as cylinders, planes, squares, etc.). It is very effective in standardized components of energy facilities (such as pipes, support rods, etc.), and can significantly improve the geometric accuracy of the model. Spatial variation geometric correction can be corrected according to the dynamic changes of the facility in the actual environment (such as wind force, load, etc.), reflecting the geometric changes of the facility under different use conditions. By acquiring the image data of the facility, the reference dimension of the model correction is increased, so that the model can combine three-dimensional point cloud data with image information to improve the comprehensiveness of geometric correction. Image features and parameter estimation can add more geometric parameters of the facility to the model, such as length, width, height, shape ratio, etc., so that the model not only reflects the appearance, but also has available physical structure information. The fusion model retains the surface features, standard shape features and spatial variation features of the facility, so that it can simultaneously reflect the state of the facility under static and dynamic conditions, and has higher practicality.

[0043] Optionally, the surface fitting geometry correction includes:

[0044] Perform multi-scale surface fitting initialization according to the facility mapping data to obtain multi-scale surface fitting data;

[0045] Perform local curvature optimization according to multi-scale surface fitting data to obtain multi-scale surface optimization data;

[0046] Perform B-spline surface fitting according to the multi-scale surface optimization data to obtain B-spline surface data;

[0047] Perform a topological consistency check on the B-spline surface data to obtain the surface topological consistency data;

[0048] Perform global error minimization based on the surface topology consistency data to obtain surface error minimization data;

[0049] Matching and mapping are performed based on the B-spline surface data and the surface error minimization data to obtain the first facility geometric reference model.

[0050] The multi-scale surface fitting initialization in the present invention can quickly generate the basic geometric model of the facility surface, provide the surface fitting results at different resolutions, and effectively capture the overall structure and geometric form of the facility. Through local curvature optimization, high curvature areas (such as bends, corners, etc.) can be refined during the surface fitting process to capture more local geometric details, which can ensure the balance between details and large-scale structures of the model. B-spline surface fitting can flexibly express complex surface forms by approximating the surface using B-spline functions, especially the processing effect on bends and complex surfaces is significant, and a smoother and more accurate surface model is generated. Topological consistency check can ensure that the surface model maintains coherence and integrity during the generation process, avoids the generation of broken, discontinuous or self-intersecting surfaces, and ensures the consistency of the model topological structure. Global error minimization can correct local errors in the surface fitting process, minimize the gap between the surface model and the actual geometric form of the facility, thereby improving the overall accuracy of the model. By matching and mapping B-spline surface data and error minimization data, a highly accurate facility geometric reference model can be generated. The model not only retains the smoothness of the surface fitting, but also ensures its consistency with the actual geometric shape of the facility by minimizing the error.

[0051] Optionally, the local curvature optimization comprises:

[0052] The curvature is calculated according to the multi-scale surface fitting data to obtain the multi-scale surface curvature data;

[0053] Performing curvature partitioning on the multi-scale surface fitting data according to the multi-scale surface curvature data to obtain curvature partitioning data;

[0054] Extracting curvature partition features according to the curvature partition data to obtain curvature partition feature data;

[0055] Constructing a curvature map according to the curvature partition characteristic data and the curvature partition data to obtain curvature map data;

[0056] Refining the curvature partition data in a neighboring area according to the curvature image data to obtain curvature partition refined data;

[0057] Perform curvature smoothing control according to curvature partitioning and refinement data to obtain curvature smoothing data;

[0058] The boundary curvature is optimized according to the curvature smoothing data to obtain multi-scale surface optimization data.

[0059] In the present invention, by calculating the curvature of different parts of the surface, the geometric features of each area in the facility model can be accurately captured. The calculation of the curvature provides an important basis for geometric features, ensuring that the model's details in the local area are more accurate. Through curvature partitioning, the model can be divided into different areas, so that the areas of high curvature and low curvature can be processed separately, effectively reducing the amount of calculation. More detailed optimization is performed on the high curvature area, while the low curvature area is simplified, which can balance the processing accuracy and efficiency. By extracting the curvature features in the partition, the key geometric features of the model can be identified, such as sharp corners, inflection points or areas with drastic surface changes of high curvature. These feature data can help improve the fineness of the model and ensure that the key structural parts are optimized. The curvature map provides a visual representation of the curvature distribution in the model, which is convenient for analyzing and understanding the geometric changes in different areas of the model. Through the curvature map, it is possible to intuitively identify which areas need to be further optimized and which areas can be simplified. By refining the adjacent areas of the high curvature areas in the curvature map, the local details of the model can be further enhanced to ensure that the geometry of these areas is more in line with the structural characteristics of the actual facilities. Curvature smoothing control can eliminate sharp transitions and abrupt geometric changes in the model, making the model smoother and more natural overall, especially in large curved areas, where smoothing can eliminate unnecessary noise or errors. Boundary curvature optimization can accurately process the edges and contours of the model to ensure that the geometry of these areas is consistent with the actual facility, especially in edge areas such as sharp corners and contours. The optimized model can more accurately reflect the appearance of the facility.

[0060] Optionally, the shape constraint geometry correction comprises:

[0061] Perform preliminary shape classification according to the facility mapping data to obtain shape classification data;

[0062] Perform minimum envelope shape detection according to shape classification data to obtain minimum envelope data;

[0063] Perform shape constraint matching according to the minimum envelope data to obtain shape constraint matching data;

[0064] Perform local refinement and global shape correction according to shape constraint matching data to obtain shape refinement geometry data and shape correction data respectively;

[0065] Perform edge constraint optimization according to shape refinement geometry data to obtain edge constraint data;

[0066] Data fusion is performed based on the edge constraint data and the shape correction data to obtain a second facility geometric reference model.

[0067] The preliminary classification in the present invention helps to quickly distinguish the key geometric forms in the facility, making the model processing more accurate, effectively extracting the geometric characteristics of different parts, and ensuring that subsequent operations are more targeted. Minimum envelope shape detection can effectively determine the overall shape outline of the facility, so that the model can be simplified and optimized while maintaining the main geometric features. The minimum envelope reduces redundant details and helps to better perform shape correction. Shape constraint matching uses predefined shape templates (such as cylinders, rectangles, planes, etc.) for geometric matching, so that the key structural parts of the model are more consistent with the geometric shape of the actual facility and correct geometric deviations. Local refinement enables the model to obtain higher geometric accuracy in key areas (such as corners, edges, and connection points), ensuring that the geometric features of the details are not overly simplified, especially the key points of the structure can be more accurately reflected. Through edge constraint optimization, the contour and edge parts of the model can more accurately reflect the true shape of the facility, especially geometric features such as edges and edges can be more clearly expressed to avoid edge blur or distortion. Through the data fusion step, the shape correction data is combined with the edge optimization results to ensure that the facility model can retain the fine edge features and reflect the correction results of the overall structure, making the model more accurate and complete.

[0068] Optionally, the space-varying geometry correction comprises:

[0069] Capture spatial changes based on facility mapping data to obtain spatial change field data;

[0070] Perform deformation analysis and modeling based on the spatial variation field data to obtain deformation model data;

[0071] Perform local deformation optimization according to the deformation model data to obtain local deformation optimization data;

[0072] Perform spatial consistency constraints based on local deformation optimization data to obtain spatial consistency data;

[0073] Physical simulation deformation correction is performed based on the spatial consistency data to obtain the third facility geometric reference model.

[0074] In the present invention, by capturing spatial changes, the deformation and spatial displacement of facilities under different environmental conditions, such as wind load, temperature change, mechanical stress, etc., can be recorded, so that the model is not only a static geometric representation, but also can reflect the changes of facilities at different times and states. By analyzing the spatial variation field data, the deformation characteristics of the facilities under different conditions can be accurately calculated, and the generated deformation model can reflect the real physical morphological changes of the facilities, so that the model is not limited to the geometric appearance, but also includes dynamic behavior. Through local deformation optimization, the key areas of the facilities (such as joints, load-bearing components, etc.) can be refined to ensure that the geometric models of these areas can truly reflect the morphological changes of the facilities under different load conditions and improve the accuracy of the local models. Spatial consistency constraints can ensure that local deformation will not affect the coherence and consistency of the overall model, and prevent excessive local deformation from causing overall distortion of the model or loss of physical consistency. Through physical simulation deformation correction, the deformation part in the model can be further optimized to ensure that it conforms to physical laws, such as elastic deformation and plastic deformation of materials, to ensure that the geometric model is not only accurate at the geometric level, but also reflects the behavior of the actual facilities in terms of physical properties.

[0075] Optionally, S4 includes:

[0076] S41, generating a fine network based on the three-dimensional model of rural energy facilities to obtain a fine grid model of the facilities;

[0077] S42, performing topology optimization according to the facility refined grid model to obtain a facility topology optimization model;

[0078] S43, calculating material reflectivity according to the facility topology optimization model to obtain a facility reflectivity model;

[0079] S44, performing collision detection and correction according to the facility reflectivity model to obtain a facility collision correction model;

[0080] S45, performing detail texture enhancement according to the facility collision correction model to obtain a facility detail texture enhancement model;

[0081] S46. Material aging simulation is performed based on the facility detail texture enhancement model to obtain a refined model of rural energy facilities.

[0082] In the present invention, through fine network generation, a high-resolution mesh model can be generated to capture the fine geometric features of the facility, such as pipes, brackets, blades of wind power equipment, etc., so that the model has higher accuracy in the display and simulation process. Topology optimization can reduce the redundant parts of the model, remove unnecessary geometric details, and optimize the coherence and stability of the structure. The optimized model has higher structural rationality, reduces the calculation burden, and improves the simulation efficiency. By calculating the reflectivity of the material, the model can more realistically express the appearance of the facility, especially the visual effects of high-reflectivity materials such as metal and glass, which are closer to the actual performance under simulated lighting conditions. Through collision detection, possible geometric penetration or overlap problems in the facility model can be identified to ensure the physical consistency of the model. Collision correction can avoid unreasonable physical phenomena such as overlap or intersection of structures during simulation and visualization. Detail texture enhancement adds real-world surface details to the model, such as scratches on metal surfaces, rust on aging equipment, etc. These details enhance the realism of the model and make it more attractive in applications such as visualization and virtual reality. Through material aging simulation, the aging of the facility in long-term use can be reflected, such as surface rust, weathering, material degradation, etc. Aging simulation not only increases the realism of the model, but can also be used for simulation of facility maintenance and life assessment.

[0083] The purpose of the present invention is that the three-dimensional laser scanning equipment can efficiently collect large-scale, high-density point cloud data of rural energy facilities, and capture the precise geometry, surface details and spatial position of the facilities. Compared with traditional measurement methods, this method can still ensure high-precision acquisition under irregular terrain and complex facility layout conditions. By combining machine learning and computer vision technology, different energy facilities (such as wind turbines, solar panels, transformers, etc.) are automatically identified from point cloud data. Automatic identification greatly reduces manual intervention and improves the efficiency of data processing of large-scale energy facilities. During the construction process, topology optimization is used to ensure the geometric and structural consistency of the model, avoid the accumulation of redundant data, and ensure the integrity and operability of the model. The optimized three-dimensional model achieves a good balance between high precision and low redundancy, laying the foundation for refined modeling. Refined modeling further refines and optimizes the geometric details of the facility (such as corners, curved surfaces, connection points, etc.) through steps such as multi-scale geometric correction and local curvature optimization. The high-resolution geometric detail representation makes the model not only realistic in appearance, but also reflects the complex structure inside the facility. In the process of refined modeling, physical properties are simulated in combination with material properties (such as reflectivity, material characteristics, and aging status), so that the model is not only a geometric representation, but also has the optical, thermal, and mechanical properties of the material. The high-precision three-dimensional model and facility characteristic analysis generated by this method can provide support for future facility planning, especially for the site selection, design, and structural optimization of new facilities. Provide a strong data foundation. BRIEF DESCRIPTION OF THE DRAWINGS

[0084] Other features, objects and advantages of the present application will become more apparent by reading the detailed description of non-limiting implementations made with reference to the following drawings:

[0085] Figure 1 A flowchart showing a method for rapid modeling of rural energy facilities based on three-dimensional laser point cloud according to an embodiment of the present invention is provided;

[0086] Figure 2 A flowchart showing a method for collecting point cloud data of rural energy facilities according to an embodiment of the present invention is provided;

[0087] Figure 3 A flowchart showing a method for identifying a rural energy facility target according to an embodiment of the present invention is provided;

[0088] Figure 4 A flowchart showing a method for constructing a three-dimensional model of rural energy facilities according to an embodiment of the present invention is provided;

[0089] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0090] The technical method of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0091] In addition, the drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The functional entities can be implemented in software form, or in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0092] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0093] The facility area is scanned using a 3D laser scanning device to obtain facility point cloud data. The area covers 5,000 square meters, the point cloud data acquisition accuracy is 5mm, and approximately 400,000 points are generated per square meter, so the total data volume is approximately 2 billion points. The scanning area is a solar power station with an area of ​​5,000 square meters. The facilities include 300 solar panels, 20 transformers, and 2 control centers. The laser scanning equipment performs full coverage scanning at 0.1 second intervals, and the total acquisition time is 20 minutes. It is approximately 50GB, with extremely high data density, which can accurately describe the morphology and spatial relationship of the facility.

[0094] Extract facility geometry from point cloud data, and use machine learning and computer vision algorithms to automatically identify facility targets in point clouds. After removing redundant points (such as weeds, animals, and other noise), use density segmentation and feature extraction algorithms to identify energy facilities. By analyzing the characteristics of planar structures, the solar panels are distinguished and 300 solar panels and their locations are identified. Morphological analysis is performed on non-planar areas in the point cloud to identify complex geometric bodies such as transformers and control centers. After point cloud recognition, the 300 solar panels in the facility are identified as rectangular planes with a size of 1.6 meters by 1 meter, and the position and tilt angle are accurately identified. Each solar panel generates about 80,000 point data, 20 transformers generate 500,000 point data, and 2 control centers generate 800,000 point data. It is about 200MB, containing the recognition results and corresponding geometric data of each facility.

[0095] A 3D geometric model is constructed based on the identified facility geometry data, and the 3D model of the facility is automatically generated. The point cloud data of each facility is surface fitted to generate a preliminary 3D model. Topology optimization is used to ensure that the geometric connections of each part are correct. Each solar panel is corrected to a standard rectangle through a shape matching algorithm to ensure the shape consistency of the model. The 300 solar panels are geometrically reconstructed, and the generated model has the specific position, angle, size and other information of each solar panel, and is corrected according to the errors of the actual facility. The size of the initially generated 3D geometric model is 10GB, of which each solar panel is about 100KB, the transformer model is 300KB, and the control center is 500KB.

[0096] The generated 3D geometric model is refined to improve the geometric accuracy and surface details of the model and add physical properties. Surface details are processed for each model, such as refining the edges of the solar panel, adding brackets, bolts and other detailed structures, to ensure that the model not only reflects the overall shape, but also reflects the specific details. Reflectivity calculation and aging simulation are performed in combination with the physical properties of the material such as reflectivity, thermal conductivity, and durability. For solar panels, the surface reflectivity is set to 85%, which gradually decreases with the years of use. The surface reflectivity of the solar panel is simulated to drop to 75% after 10 years of use, and the wear and slight damage on the surface are simulated. The refined solar panel model: contains information such as brackets, connectors, reflectivity, etc. of each panel. The data volume increases to 200KB / block, and the total model size is about 30GB. Through the material aging algorithm, the aging process of the facility in the next 10 years is predicted to provide data support for the operation and maintenance of the facility.

[0097] After data is collected by handheld / unmanned aerial vehicle (UAV) using a 3D laser point cloud device, preprocessing is first performed, including noise filtering, point cloud thinning and other operations. Some noise in the point cloud data comes from wind or reflection. The neighborhood average filter is used to reduce the noise impact and ensure the clear geometric outline of the facility. The solar panels are extracted from the complex environment through the point cloud segmentation algorithm. During the recognition process, the size of the facility is first estimated based on the point cloud density, and the facility is classified according to its geometric features (such as planarity and aspect ratio). In this way, different types of facilities are identified and classified.

[0098] The identified facility point cloud data enters the 3D reconstruction process, and the geometric model of each facility is generated using a surface fitting algorithm. For complex facilities (such as transformers and control centers), B-spline surface fitting is used to enhance surface details through local refinement.

[0099] During the topology optimization process, the geometric connections within the facility are identified to ensure that the topology of the model is consistent with the actual facility design. For example, the connection between the solar panel bracket and the ground, and the control center and the transformer are optimized to ensure the integrity of the geometric structure.

[0100] In the data processing stage, reflectivity calculation models are used to simulate the material properties of facilities, such as the surface reflectivity of solar panels. For facilities with a long service life, the wear and tear of their surfaces is simulated through material aging models to predict future maintenance needs. The material aging model is constructed based on expert knowledge or databases for data modeling, including mathematical expressions and corresponding logical processing. Collision detection ensures that there is no unreasonable interlacing or overlapping between facility models, and corrects areas with potential collisions.

[0101] See also Figures 1 to 4The present application provides a method for rapid modeling of rural energy facilities based on three-dimensional laser point cloud, the method comprising:

[0102] S1. Collect point cloud data of energy facilities in rural areas using 3D laser scanning equipment to obtain point cloud data of rural energy facilities;

[0103] Specifically, a three-dimensional laser scanning device (such as a fixed laser scanner or a mobile scanner) is used to perform multi-angle and multi-directional scanning of rural energy facilities. During the scanning process, the device emits a laser beam and obtains the spatial coordinates of each point by measuring the difference in reflection time to form point cloud data. To ensure coverage and accuracy, multiple scanning positions need to be merged, and the point cloud data of each scanning position are merged into a whole using registration technology to obtain complete point cloud data of rural energy facilities. In the process of collecting point cloud data, it is necessary to filter the interference data in the environment, such as removing irrelevant noise points such as vegetation and buildings.

[0104] S2. Perform target recognition based on the rural energy facility point cloud data to obtain rural energy facility target data;

[0105] Specifically, the point cloud data is pre-processed, including removing isolated points, sparse points, and smoothing, to enhance the usability of the data. Geometric feature extraction methods are used to identify the characteristics of rural energy facilities, such as by calculating the curvature, normal vector and other geometric information of the local area of ​​the point cloud. In the feature recognition process of energy facilities, different types of facility structures can be identified by region, such as transformers, transmission towers, energy storage equipment, etc. Through the analysis of spatial relationships, the correlation structure and distance between various energy facilities are identified, and the target data of rural energy facilities is obtained. The data at this time contains the location information and geometric features of each facility.

[0106] S3, constructing a three-dimensional model according to the target data of rural energy facilities to obtain a three-dimensional model of rural energy facilities;

[0107] Specifically, according to the target data identified in S2, key geometric features are extracted and the facilities are modeled. The corresponding three-dimensional geometric structure is generated using the point coordinates in the point cloud, such as by generating meshes, surfaces, etc., to gradually build a rough three-dimensional model of the energy facility. In the process of model building, the integrity of the structure needs to be considered, and the model should be optimized by filling in the details, such as generating the connection parts of the pipeline, the support structure of the equipment, etc. The constructed three-dimensional model should retain the geometric information in the original point cloud data to ensure that the size and shape of the model conform to the physical characteristics of the actual energy facility.

[0108] S4. Perform detailed modeling based on the three-dimensional model of rural energy facilities to obtain a detailed model of rural energy facilities.

[0109] Specifically, based on the rough three-dimensional model, refinement is performed to increase the details of the model. This includes enhancing the surface texture, material, color, etc. of the energy facilities to improve the visual effect and authenticity of the model. In the refined modeling, the structural part of the facility is further refined, such as accurate geometric modeling of complex parts such as cables and brackets of the transformer. The dynamic properties of the energy facility are supplemented, such as adding moving parts of the facility or simulating the time-varying characteristics of the facility according to actual usage. Finally, the overall refined model is verified and optimized to ensure that the accuracy of the model meets the requirements of practical applications, including error analysis and adjustment of the model to minimize the gap between the refined model and the real facility.

[0110] Optionally, S1 includes:

[0111] S11. Use laser scanning equipment to determine the scanning range and perform multi-dimensional spatial positioning to ensure that all energy facilities in the rural area are covered and obtain scanning space coordinate data;

[0112] Specifically, before scanning, the geographical location and coverage of rural energy facilities are first determined through geographic information systems (GIS) or drone images to ensure that the scanned area is large enough to cover all energy facilities. Use GNSS (Global Navigation Satellite System) or IMU (Inertial Measurement Unit) equipment for accurate multi-dimensional spatial positioning to determine the absolute position and attitude of the laser scanning equipment to ensure the positioning accuracy of the scanning equipment in the entire area. Set up multiple scanning equipment location points around energy facilities in rural areas to ensure that all facilities are within the scannable range. The area is a 200m × 200m substation and transmission line facility area. The location of each scanning device is determined by the GNSS system, and the three-dimensional spatial coordinates are recorded. Assume that the first scanning device is located at (84.12345E, 38.98765W, 75 meters), that is, the longitude, latitude and altitude coordinates.

[0113] S12, using a three-dimensional laser scanning device to collect high-density point cloud data by scanning spatial coordinate data to obtain an initial point cloud data set;

[0114] Specifically, the laser scanning equipment scans at multiple predetermined locations simultaneously, and obtains the surface information of the energy facility by emitting a high-frequency laser beam and measuring the time difference of the reflected light. In order to obtain higher density point cloud data, a smaller angle step value and a higher resolution can be set during scanning to ensure that the details of the surface of each facility are accurately captured. The device begins to scan point by point, collecting millions of points per second to form an initial point cloud data set. In this 200 m × 200 m area, the amount of data collected is 2,000,000 points, with a point spacing of approximately 0.5 cm. The point cloud data contains three-dimensional coordinates (x, y, z) and reflection intensity information, forming an initial point cloud data set containing (x, y, z, I).

[0115] S13, optimizing the scanning perspective of the initial point cloud data set to obtain perspective point cloud data;

[0116] Specifically, the point cloud data of different perspectives are optimized to ensure that the data coverage at different perspectives is uniform and there are no missing areas. The point cloud data of different scanning positions can be aligned by matching feature points. According to the angle distribution of the scanning equipment, the noise points caused by equipment occlusion or blind spots of the viewing angle are removed. By analyzing the multi-perspective data of the scanning equipment, it is identified that some parts of the energy facilities have blind spots, especially the top of the taller transmission tower or the back area of ​​the equipment. By increasing the scanning angle, the scanning data is supplemented. The number of points of the optimized perspective point cloud data is 1,800,000 points, eliminating duplicate and useless scanning information.

[0117] S14, performing multi-source data fusion on the viewpoint point cloud data to obtain fused point cloud data;

[0118] Specifically, multi-source data fusion refers to aligning and splicing point cloud data at different locations and different scanning perspectives to form a unified 3D point cloud data set. Data fusion is performed using feature points or reference points, and all scanned data is fused into a whole through coordinate transformation. During the fusion process, it is necessary to process the resolution and coordinate system differences of scanned data from different devices or different locations and unify them into the same reference coordinate system. By aligning the key feature points in each perspective (such as the corners of the transformer, the top of the transmission tower, etc.), the data from different devices and perspectives are fused into one. After the fusion process, the number of point cloud data points obtained is 1,700,000, and duplicate and overlapping areas have been removed.

[0119] S15. Time-synchronize the fused point cloud data to obtain rural energy facility point cloud data.

[0120] Specifically, time synchronization is to ensure that the data collected by multiple scanning devices are on the same time axis. Especially when the scanning area is large and multiple devices are used, time synchronization can avoid position differences in the data collected by different devices at different time points. By recording the timestamp during the data collection process, the data of all scanning devices are synchronized using a unified time base. Two scanning devices scanned the same area at different time periods, with a time difference of about 5 seconds. Through timestamp correction, the two sets of data are aligned to the same time base. A complete point cloud dataset after time synchronization processing is obtained, which contains 1,650,000 valid points of the entire rural energy facility.

[0121] Optionally, S2 includes:

[0122] S21, performing preliminary filtering according to the rural energy facility point cloud data to obtain point cloud filtering data;

[0123] Specifically, the original point cloud data is initially filtered using parameters such as distance and reflection intensity to remove noise and irrelevant points. For example, data points of ground vegetation, buildings or other non-energy facilities are removed. By setting a distance threshold, irrelevant points far away from the device are removed, especially noise points, isolated points and boundary points during the scanning process. The collected point cloud data is 1,750,000 points. After preliminary filtering, points with low reflection intensity and far away from the target are removed, and 1,500,000 valid point clouds remain.

[0124] S22, performing density estimation and segmentation according to the point cloud filtering data to obtain point cloud segmentation data;

[0125] Specifically, the point density of each point in its neighborhood is calculated to identify density differences in different areas. Density-based estimates can help identify boundaries between different structures or objects. The point cloud is divided into different regions based on density information. Lower density areas may represent the edges or noise points of the device, while high-density areas usually correspond to the core structure of the device. For each point, the number of points in its surrounding neighborhood (assuming a radius of 5 cm) is calculated to determine its local density. It is identified that the high-density area contains approximately 1,200,000 points and the low-density area contains 300,000 points. The high-density area is segmented from the low-density area according to density to obtain 1,200,000 points of point cloud segmentation data.

[0126] S23, extracting morphological features according to the point cloud segmentation data to obtain facility point cloud morphological feature data;

[0127] Specifically, the geometric features of each segmented area are analyzed, including curvature, normal vector, surface shape, etc. Through these morphological features, different components of the facility can be identified, such as the casing of the transformer, the support structure of the transmission tower, etc. According to the geometric information of the point cloud, the main morphological features of the equipment are extracted, such as planes, surfaces, corners and other geometric features. Through the neighborhood of each point in the point cloud, its local curvature is calculated to identify the flat parts and complex structures on the surface of the equipment. The morphological feature data of the equipment is extracted, including curved surface areas (such as equipment casings) and angular areas (such as tower support structures). The point cloud morphological feature data at this time contains approximately 1,000,000 points.

[0128] S24, performing energy signal analysis based on the rural energy facility point cloud morphological feature data to obtain facility point cloud energy feature data;

[0129] Specifically, the point cloud data is regarded as a signal, and the different parts of the facility are identified by analyzing its energy characteristics. The energy signal is calculated based on information such as the reflection intensity and position change of the point to identify the different components of the facility. The point cloud is analyzed in the frequency domain to calculate the frequency characteristics of each part and extract important energy change areas. The energy characteristics are calculated using the reflection intensity and point cloud density, and the energy changes corresponding to different areas of the point cloud are analyzed. For example, the metal surface of the equipment may show a stronger energy signal, while the energy signal of the supporting structure is relatively weak. After the energy signal analysis, high-energy areas (such as metal parts) and low-energy areas (such as brackets, cables, etc.) of the equipment are extracted. The data set contains 900,000 points.

[0130] S25, performing multi-target clustering based on the rural energy facility point cloud energy feature data to obtain facility clustering data;

[0131] Specifically, cluster analysis is performed using the spatial position, energy characteristics, and morphological characteristics of the point cloud to identify different equipment or equipment parts. The point cloud is divided into multiple different groups through a clustering algorithm, each group representing a target or equipment part. According to the clustering results, facilities such as transmission towers, transformers, and cables are separated. By calculating the similarity of the point cloud data, facility components such as transformers, transmission towers, and cables are identified and clustered into different targets. Assume that three main cluster groups are obtained after clustering, and each group contains point data of different facilities. The obtained cluster data includes three main facilities, the transformer group contains 300,000 points, the transmission tower group contains 400,000 points, and the cable group contains 200,000 points.

[0132] Specifically, it is necessary to obtain the energy characteristics of the point cloud data. The energy characteristic data is characteristic information calculated based on the reflection intensity of laser scanning, point cloud density, etc., which can reflect the reflection ability and density differences of different materials or structures in laser scanning. For example, the metal transformer shell may have a higher reflectivity, while the rubber insulation material has a lower reflectivity, and the different parts of the facility are distinguished by features. For example, the point cloud energy characteristic data contains the reflection intensity value of each point. The reflectivity of the transformer shell is above 0.7, while the structural reflectivity of the transmission tower is around 0.5. By analyzing the energy characteristics in the point cloud data, the most relevant features for facility segmentation are selected, such as reflection intensity, point cloud density, and spatial position. The features will be used as input to the clustering algorithm to distinguish the various parts of the facility. In the point cloud data of the transformer, significant differences in reflection intensity and density can be found. For example, the metal transformer has a higher reflection intensity, while the connected bracket part has a higher density. Algorithms include K-means clustering, DBSCAN (density-based spatial clustering), etc. In point cloud data processing, DBSCAN is more applicable because it can handle irregular shapes and filter noise points. K-means: Suitable for use when the facility category is known in advance, and the number of facility categories can be directly specified. DBSCAN: Suitable for irregular facility shapes and does not require the number of categories to be specified in advance. Using the DBSCAN algorithm to cluster based on reflection intensity and density, clusters of different facility components such as transformers and transmission towers can be identified. Input the energy feature data into the selected clustering algorithm for multi-target clustering. The clustering process will segment different parts of the facility based on the density and energy features of the point cloud data, ensuring that the point cloud groups of each category represent different parts of the facility. For example, the clustering results divide transformers and transmission towers into two different categories. The transmission tower part includes the base and support poles, while the transformer part is clustered into a relatively independent set of point cloud data. During the clustering process, there will be some isolated noise points or non-facility-related data. These irrelevant points can be filtered out by threshold setting or further processing. Assign labels to each group in the clustering results to ensure that each part of the facility is labeled as a specific category (such as transformers, transmission towers, etc.). For the clustered results, isolated points or points not related to the facilities will be filtered out, and finally labels will be assigned to the clustered facility parts, such as "transformer housing", "transmission tower base", "transmission tower support", etc. After clustering is completed, the clustering results are output as facility clustering data. Each cluster group represents a component of the facility, and these data can be used for target detection, semantic annotation, and 3D modeling. The clustering data structure usually includes category labels and point cloud information contained in each category (such as spatial coordinates, energy characteristics, etc.).

[0133] S26, performing deep target detection according to the facility clustering data to obtain facility deep detection classification data;

[0134] Specifically, through in-depth analysis of cluster data, the specific parts of each facility are further identified and classified. Combined with the size, shape and location of the facility, the specific functional components of the equipment, such as the transformer housing, radiator, bracket, etc., are distinguished. The targets in each cluster group are classified and marked. By analyzing the geometric features of transformers and transmission towers, the components of each facility are identified. The classifications detected for transformers include: housing (200,000 points), radiator (50,000 points) and bracket (50,000 points). The classification data obtained through deep detection covers the specific components of transformers, transmission towers and cables.

[0135] Specifically, the input facility clustering data includes clustered point clouds or mesh data, representing the various components of the facility, such as cables, brackets, transformers, etc. The clustering data provides a preliminary grouping for target detection, and target detection further performs fine-grained analysis on each grouping. Key geometric features, such as the size, shape, and spatial position of each part of the facility, are extracted from the clustering data, and the features will be used in the pre-processing step of target detection. For example, transformers, transmission towers, and cables have been divided into different groups in the clustering data. By extracting the geometric features of each clustering group (such as the size and shape of the transformer), it is prepared for deep target detection. Deep target detection usually uses target detection algorithms based on deep neural networks, such as YOLO (You Only Look Once) and Faster R-CNN. The clustering data of the facility is converted into an input format recognizable by the deep learning network for target detection training and inference. Use existing pre-trained models, such as models trained on datasets such as urban facilities and industrial equipment, as initial weights. Pre-trained models can accelerate the training process of target detection, especially in the case of similar facilities (such as substation equipment detection). The YOLO network is selected as the base model to detect different components of transformers and transmission towers, such as cables, insulators and other key components. The clustered data is fed into the deep learning model to train the target detection network. The network will perform forward propagation based on the input data and gradually learn the different components of the facility. Based on the detection results, the model predicts the different components of the facility and assigns them to the corresponding categories (such as transformers, high-voltage cables, support structures, etc.). After training, new data is fed into the model, and the network automatically detects the various components of the facility and outputs the category and location of each detection target. The inference process uses the final weights of the network to classify the new data. For example, by feeding the facility clustering data into the YOLO model, the model predicts different components such as the transformer housing, cables, and brackets, and assigns a classification label to each component (such as "transformer housing", "high-voltage cable", etc.). The detection results usually contain multiple overlapping prediction boxes or regions. In order to improve the accuracy, post-processing steps are required, such as the NMS algorithm to remove redundant detection boxes and optimize the detection results. Ensure that the output classification results are both accurate and non-repetitive. By adjusting the threshold of the model or optimizing the parameters, the accuracy of target detection can be further improved, and missed detections and false detections can be reduced. For example, in deep target detection, the casing and bracket of the transformer may have multiple overlapping prediction boxes. After NMS processing, the optimal detection box is retained, and the duplicate detection results are removed, outputting the most accurate facility target detection results. According to the detection results, each component of the facility is assigned a corresponding category label and spatially annotated. Deep target detection can not only identify the category of the facility, but also locate the spatial position of each target. The detection results of each facility are converted into deep detection classification data, including the target name, spatial coordinates, and recognition confidence of each category.

[0136] S27, extracting edge features according to the facility depth detection classification data to obtain facility edge feature data;

[0137] Specifically, edge detection methods are used to identify the edge contours and key connecting parts of the facility. Edge detection is mainly based on the curvature change or normal vector difference in the point cloud to extract the geometric boundaries of the equipment. The edge features of the contours, joints and other key areas of transformers and transmission towers are extracted. Key edge points are extracted in the transformer housing and bracket areas. The edge feature dataset contains 150,000 points, representing the main geometric contours of the equipment. As in mesh-based edge extraction, in mesh data, edge features are usually defined by discontinuous geometric boundaries (such as acute angles between two faces) or the boundaries of the mesh. Edge lines can be extracted by finding the angle difference between adjacent faces or discontinuous areas in the vertex neighborhood. If the angle difference between adjacent faces exceeds a certain threshold (such as 30 degrees or 45 degrees), these areas can be considered edge features. Free edges in the mesh that are not connected to other patches can also be used as edge features. Alternatively, using point cloud-based edge extraction, if the input data is a point cloud, the geometric relationship between neighborhood points is used to detect edges. By analyzing the density and direction changes of the point cloud, the edge points of the facility can be identified. For example, at the edge of a transformer housing, the point cloud can change direction or density significantly.

[0138] S28. Semantically annotate the rural energy facility point cloud data based on the facility edge feature data and the facility depth detection classification data to obtain the rural energy facility target data.

[0139] Specifically, the deep detection classification data and edge feature data are combined to semantically annotate the specific parts of each facility. For example, the transformer casing, radiator, support structure and other functional components are annotated. Automated labeling is performed through rules or learning-based models to assign corresponding semantic labels to each facility component. Transformers, transmission towers and cables are semantically annotated to mark the specific functional components of each facility. The annotated target dataset contains 1,200,000 points, completing the semantic annotation of the facilities.

[0140] Optionally, S3 includes:

[0141] S31, reconstructing the network according to the rural energy facility target data to obtain a preliminary facility network model;

[0142] Specifically, a preliminary 3D model is formed by converting point cloud data into a mesh structure. The process of network reconstruction is mainly based on triangulation, which converts point cloud data into a network model composed of triangular facets. The point cloud data is connected into a polygonal mesh, and the overall framework and shape of the facility are formed by point-by-point connection. Delaunay triangulation or other neighborhood-based reconstruction methods are applied. The point cloud data contains 1,200,000 points. After triangulation meshing, a preliminary facility network model consisting of approximately 600,000 triangular facets is generated, which preliminarily reflects the overall structure and appearance of the facility.

[0143] S32, performing surface refinement on the preliminary facility network model to obtain a facility surface refinement model;

[0144] Specifically, based on the preliminary mesh model, the surface is further refined. By increasing the mesh density or adjusting the mesh structure, the smoothness and accuracy of the model are improved, making the details of the facility clearer. By means of local refinement and mesh optimization, irregular surface areas such as sharp angles or rough surfaces are corrected to generate a smoother surface. For example, for the complex structure of the transmission tower, the number of triangular facets is locally increased, increasing the initial 600,000 facets to 1,000,000 facets, which improves the accuracy of surface details, especially optimizing curved areas and edges.

[0145] S33, performing texture mapping on the facility surface refinement model to obtain facility mapping data;

[0146] Specifically, the real texture is mapped onto the refined 3D model surface through the collected image data or the existing texture library. In the texture mapping process, it is necessary to ensure the accurate matching of the texture with the geometric surface to avoid texture distortion or distortion. The 3D surface is unfolded into a 2D UV coordinate system so that the image texture can be correctly attached to the surface of the facility. UV coordinates are generated for the outer shell area of ​​the transmission tower and transformer to map the real metal surface texture onto the equipment. By comparing with the original image, it is ensured that the texture is not obviously stretched or distorted. After the texture mapping is completed, the corresponding mapping map data is generated to make the surface of the facility look more realistic.

[0147] S34, performing geometric correction according to the facility mapping data to obtain a facility geometric reference model;

[0148] Specifically, local geometric corrections are performed by analyzing the geometric errors that occur after texture mapping. In particular, geometric shape corrections are performed on edge areas or uneven parts covered by the texture to ensure consistency between the model and the texture. Geometric correction technology is used to repair possible misalignments and misalignments on the surface, making the entire geometric form more consistent with the physical facility. It was found that the heat sink part of the transformer had a slight texture misalignment during the mapping process, and fine-tuning was performed on this area to make the geometric baseline model more accurate. The corrected baseline model is consistent with the actual appearance of the facility.

[0149] S35, embedding structural features according to the facility geometric reference model to obtain a facility embedded structural feature model;

[0150] Specifically, key structural features of the facility are embedded on the basis of the geometric reference model. For example, the supporting structure of the transmission tower and the cable interface of the transformer are embedded, which reflect the functionality and structural integrity of the facility. For different facility components, more detailed geometry is added, such as secondary structures such as bolts and brackets, to increase the realism of the model. The structural brackets and supporting cables of the transmission tower are feature embedded to ensure that their structure is consistent with the actual facility. Adding more details, such as the joints of the brackets and the cable connectors of the transformer, increases the functionality of the model.

[0151] S36. Physical attributes are assigned according to the facility embedded structural feature model to obtain a three-dimensional model of rural energy facilities.

[0152] Specifically, physical properties such as material, weight, density, conductivity, etc. are assigned to the model to ensure that the model is not only realistic in appearance, but also has physical simulation capabilities. These properties will be used for tasks such as engineering analysis and simulation. According to the actual material of the facility, the corresponding physical properties are selected. For example, the shell of the transformer is made of metal material and the corresponding conductivity and density properties are assigned. The metal part of the transmission tower is assigned the conductivity property, and the metal density property is assigned to the transformer shell, so that the model is not only visually realistic, but also can be used for physical simulation. The generated 3D model includes both geometric details and physical properties.

[0153] Optionally, the geometric correction comprises:

[0154] Performing surface fitting geometry correction according to the facility mapping data to obtain a first facility geometry reference model;

[0155] Specifically, for areas with more complex surfaces on the facility, surface fitting technology is used to correct the geometry of the 3D model. By fitting the point cloud data of the surface, a smoother and continuous surface is reconstructed to avoid irregularities and errors on the model surface. For rounded surfaces in facilities such as transformers and transmission towers, more precise geometric surfaces are fitted to ensure that the mapping data and geometric shapes match well. For example, for the cylindrical shell of the transformer, point cloud data is used for surface fitting, which eliminates the irregular triangular mesh on the surface and makes the surface of the shell smoother. The generated first facility geometric reference model has an accurate surface morphology.

[0156] Perform shape constraint geometry correction according to the facility mapping data to obtain a second facility geometry reference model;

[0157] Specifically, shape constraint corrections are performed on components with clear geometric shapes. For example, the supporting structures of transmission towers are usually straight lines or shapes with clear geometric shapes. Shape constraint technology is used to ensure that these structures in the model strictly follow geometric specifications. Use known geometric rules to constrain specific shapes in the model, such as the linearity and verticality of support rods, to ensure that the model meets actual mechanical and engineering constraints. In the support part of the transmission tower, shape constraints are used to ensure that the connection points between the supports are strictly straight structures to avoid unnecessary deformation in the model. The generated second facility geometric reference model has accurate geometry.

[0158] Performing spatial variation geometry correction according to the facility mapping data to obtain a third facility geometry reference model;

[0159] Specifically, according to the spatial layout of the facility, spatial adjustments are made to the geometric errors caused by perspective, light changes, etc. In particular, in the height of different parts of the equipment or in complex multi-layer structure areas, the geometry of the model is adjusted to adapt to the actual spatial changes. Spatial transformations are applied in the model to ensure that the spatial relationship between the various parts of the equipment is correct, especially the relative position between the transmission tower and the cable is accurate. For example, between the transmission tower and the connected high-voltage cable, the geometric position of the cable is corrected through spatial adjustment so that the height and connection point position of the cable are more in line with the real scene, and the third facility geometric reference model is obtained.

[0160] Acquire rural energy facility image data;

[0161] Specifically, multi-view image data of rural energy facilities can be obtained through drones or ground-based camera equipment. High-resolution images of transformers, transmission towers, and cables can be taken from different angles using drones or ground-based camera equipment to capture the appearance information of all key facilities.

[0162] Extract rural energy facility features and estimate rural energy facility parameters based on rural energy facility image data to obtain rural energy facility image feature data and rural energy facility parameter data;

[0163] Specifically, the geometric and texture features of rural energy facilities, such as edges, contours, surface textures, etc., are extracted from the image, and the features provide a reference for the correction of the three-dimensional model. Based on the image data, the key geometric parameters of the facility are estimated, such as the height of the transformer, the inclination angle of the transmission tower, the length and curvature of the connecting cable, etc. The edge features of the transformer shell are extracted from the image, and it is identified that the inclination angle of the transmission tower is about 5 degrees and the length of the cable is 50 meters. The obtained image feature data includes the geometric contours of the equipment, and the parameter data includes specific values ​​such as height and angle.

[0164] The first facility geometric reference model, the second facility geometric reference model and the third facility geometric reference model are fused according to the rural energy facility image feature data and the rural energy facility parameter data to obtain a facility geometric reference model.

[0165] Specifically, the geometric model obtained by surface fitting, shape constraint, and spatial variation correction is fused with image features and parameter data to ensure that the geometric shape of the model is consistent with the actual facility. The geometric parameters from the image are compared and corrected with the various correction data of the model, and the model details are adjusted to make it more consistent with the actual situation. By comparing the image data with the geometric correction data in the model, the surface shape of the transformer housing is adjusted to ensure that the tilt angle is consistent with the actual facility status. The fused facility geometric reference model can accurately reflect the spatial position, shape, and functional characteristics of the equipment.

[0166] Optionally, the surface fitting geometry correction includes:

[0167] Perform multi-scale surface fitting initialization according to the facility mapping data to obtain multi-scale surface fitting data;

[0168] Specifically, the curved areas of the facility are fitted at different scales. First, a larger scale grid is used to approximate the overall shape of the facility, and then the scale is gradually reduced to increase the details of the model to improve the accuracy of the surface fitting. This multi-scale processing method can effectively reduce local errors and capture the overall outline of the facility. The model is fitted step by step, from a rough overall fit to a more refined local fit. The transformer housing is initially fitted with a larger triangular grid to generate preliminary multi-scale surface fitting data. The coarse grid is used to capture the overall curvature of the housing and initially form the global shape of the model.

[0169] Perform local curvature optimization according to multi-scale surface fitting data to obtain multi-scale surface optimization data;

[0170] Specifically, on the basis of multi-scale fitting, the local curvature of the surface is further analyzed, and the areas with higher or more complex curvatures are optimized to ensure more accurate fitting in the curved parts. By adjusting the local curvature of each point, the surface can be smoothed and local errors can be reduced. According to the change in curvature, the distribution of local points and grids is readjusted to make the fitting more consistent with the actual geometric form. For example, at the top of the transformer casing, the local curvature is higher. After local optimization, the grid density in this area is increased, making the surface smoother and reducing errors. The optimized multi-scale surface data contains more local details.

[0171] Perform B-spline surface fitting according to the multi-scale surface optimization data to obtain B-spline surface data;

[0172] Specifically, B-spline curves are used to construct the smooth surface of the facility. B-spline is a smooth curve based on control points. By interpolating or approximating point cloud data, a surface model with good continuity is constructed, which is particularly suitable for complex surface fitting. Appropriate control points are selected to perform spline fitting on the surface to make the model smoother and more accurate in geometry. The optimized surface data is converted into a B-spline surface and fitted to the circular shell part of the transformer. During the fitting process, by selecting control points, the generated B-spline surface data contains the overall smooth characteristics of the surface.

[0173] Perform a topological consistency check on the B-spline surface data to obtain the surface topological consistency data;

[0174] Specifically, check the topology of the surface to ensure that the B-spline surface has no self-intersections, cracks, or discontinuous areas in topology. Analyze the connectivity and structure of the surface mesh to ensure that the topology of the surface is consistent with the actual structure of the facility. If inconsistent topology or cracks or overlapping areas are found, perform local repairs. Check the shell surface of the transformer to ensure that all parts of the surface are connected and there are no breaks. After the topology consistency check, generate topology consistency data to ensure the coherence and correct topology of the surface.

[0175] Perform global error minimization based on the surface topology consistency data to obtain surface error minimization data;

[0176] Specifically, based on the topological consistency data, a global error minimization process is performed (using an iterative optimization algorithm, such as the Least Squares Method, Laplacian Smoothing, etc.) to optimize the fitting effect of the entire surface. The error minimization algorithm is used to reduce the global error between the fitted surface and the original point cloud data, making the fitted surface more accurate. The distribution of local and global errors is analyzed to ensure that the error of the surface is minimized globally. Within the global range of the surface, the vertex position of each mesh is adjusted so that the error between the surface and the original data is controlled at the millimeter level. The generated surface error minimization data reduces the error to an acceptable range, ensuring the accuracy of the surface fitting.

[0177] Matching and mapping are performed based on the B-spline surface data and the surface error minimization data to obtain the first facility geometric reference model.

[0178] Specifically, the B-spline surface data is matched with the error minimization data, and the two are merged through a mapping operation to form the final geometric reference model. It is ensured that the fitted surface is not only smooth and accurate, but also meets the actual engineering requirements. The fitted B-spline surface and the result of global error minimization are fused to make the geometric reference model have higher accuracy. By comparing the B-spline surface with the error minimization data, the two are combined to ensure that the geometric model of the transformer housing is both smooth and accurate. The generated first facility geometric reference model contains all the surface information after detail optimization.

[0179] Optionally, the local curvature optimization comprises:

[0180] The curvature is calculated according to the multi-scale surface fitting data to obtain the multi-scale surface curvature data;

[0181] Specifically, the curvature of the local area of ​​each point is calculated, mainly the Gaussian curvature and the average curvature of each point on the point cloud or surface. Gaussian curvature describes the local curvature of the surface, while the average curvature reflects the smoothness of the curvature of the surface in different directions. The curvature is calculated in the local area of ​​each point through the preset neighborhood radius to ensure that the surface details are accurately described. The vertices of each triangular facet are selected on the transformer casing, and the Gaussian curvature and the average curvature of each point are calculated based on the point cloud data of its neighborhood. The result generates multi-scale surface curvature data, and the area with a larger curvature value indicates that the surface curvature in this area is obvious.

[0182] Performing curvature partitioning on the multi-scale surface fitting data according to the multi-scale surface curvature data to obtain curvature partitioning data;

[0183] Specifically, the surface is partitioned according to the curvature data, and the surface is divided into several regions according to different curvature values. High curvature regions are usually located where the surface is more curved (such as sharp corners or edges), while low curvature regions are located on relatively smooth surfaces. A curvature threshold is set to divide the curvature values ​​into different intervals for classifying different surface regions. If the curvature of certain curved parts on the transformer casing is higher than the set threshold, these areas are divided into high curvature areas, and the smoother parts of the casing are divided into low curvature areas. The generated curvature partitioning data includes the division results of high and low curvature areas.

[0184] Extracting curvature partition features according to the curvature partition data to obtain curvature partition feature data;

[0185] Specifically, feature extraction is performed on the curvature partition data to extract local geometric features of each partition, such as bending direction, curvature change rate, etc. The features of high curvature areas may include sharp corners, protrusions or depressions, while the features of low curvature areas are relatively smooth. The local features within each partition are analyzed, especially in high curvature areas, to extract key features of surface changes. In the high curvature area of ​​the transformer casing, the sharp angle features of the protrusions and the local bending features with large curvature changes are extracted to generate curvature partition feature data.

[0186] Constructing a curvature map according to the curvature partition characteristic data and the curvature partition data to obtain curvature map data;

[0187] Specifically, a color or grayscale image is used to represent the curvature values ​​of different regions, with high curvature regions represented by dark colors and low curvature regions represented by light colors, so as to analyze the curvature distribution more clearly. In the curvature image, the curvature distribution of the transformer housing is displayed in a color gradient, with high curvature regions presented as dark colors and low curvature regions as light colors.

[0188] Specifically, the surface of the facility can be regarded as composed of discrete point clouds or grids, and each point (or grid vertex) is used as a "node" in the graph. The lines ("edges") between nodes represent the geometric relationship between adjacent points, especially the change in curvature. The graph model is established by analyzing the curvature difference between adjacent nodes. The node represents each point (or triangular mesh vertex) in the curvature partition. Each node carries the corresponding curvature information, and the node attributes include its position (coordinates) and curvature value. The edge represents the spatial relationship between adjacent nodes, especially in areas with large curvature changes. The weight of the edge can be determined according to the curvature difference between the two nodes. The weight of the edge can indicate that the greater the curvature difference, the higher the weight of the edge, and the weight of the edge in areas with similar curvature is lower. Each grid point is extracted from the curvature partition data as a node. Each node saves its spatial position, curvature value, and information about neighboring nodes. For example, the surface of a transformer housing contains 5,000 vertices, each vertex is a node, and its curvature value and three-dimensional spatial coordinates are saved. The generated node graph data includes 5,000 nodes. Connect adjacent nodes as edges, and assign weights based on the curvature differences between adjacent vertices. Edges with large curvature differences can be assigned higher weights, edges with small curvature differences can be assigned lower weights, and even edges in smooth areas can be removed. For example, the curvature difference between adjacent vertices A and B is 0.5, while the curvature difference between vertices B and C is 0.1, so a larger weight (such as 0.5) is assigned between AB, and a smaller weight (such as 0.1) is assigned between BC. The generated edge information contains the start and end points of each edge and its weight. By constructing a complete graph of all nodes and their adjacent edges. This graph can intuitively reflect the local curvature changes and global relationships of the facility surface. For example, areas with large curvature changes will form dense nodes and high-weight edges, while smooth areas are relatively sparse and have low edge weights. For example, areas with significant curvature changes on the transformer casing (such as edges and connection points) will appear as high-density node-edge structures in the graph, while flat areas of the casing will form low-density node-edge graph structures. The generated node-edge graph data can represent the overall geometric curvature characteristics of the facility.

[0189] Refining the curvature partition data in a neighboring area according to the curvature image data to obtain curvature partition refined data;

[0190] Specifically, according to the high curvature areas in the curvature image, these areas are refined, and the mesh density or surface nodes are increased to improve the accuracy of the surface in the high curvature areas. The refinement is mainly concentrated in the high curvature areas and the adjacent medium curvature areas to ensure smooth transitions. More mesh details are added in the areas with sudden changes in curvature to capture subtle changes in the surface. The high curvature areas of the transformer casing (such as connection points and edge parts) are locally encrypted, and the refined surface is more refined, reducing surface errors. The generated curvature partition refinement data contains higher density mesh information.

[0191] Perform curvature smoothing control according to curvature partitioning and refinement data to obtain curvature smoothing data;

[0192] Specifically, a smooth transition is made between the refined high curvature area and its adjacent area to ensure the continuity of the surface. Through curvature smoothing control, the surface mutation is reduced, especially when the high curvature area transitions to the low curvature area, and a natural smooth transition is maintained. Smooth control ensures that the transition curve between the high curvature and low curvature areas is smooth without obvious angles or breaks. By adjusting the curvature transition of the shell, the transition between the raised area of ​​the transformer and the surrounding low curvature area becomes more natural and smooth, reducing the curvature mutation, and the generated curvature smoothing data ensures the overall smoothness of the surface.

[0193] The boundary curvature is optimized according to the curvature smoothing data to obtain multi-scale surface optimization data.

[0194] Specifically, the curvature of the surface boundary area is optimized to ensure that the curvature of the boundary part is smoothly connected with the curvature of the internal area. The optimization of the boundary mainly focuses on the continuity and curvature consistency of the surface edge to reduce the sudden change or discontinuity of the edge curvature. The boundary of the surface is optimized to keep it consistent with the smoothness of the overall surface, ensuring that the curvature optimization of the edge meets the actual requirements of the facility. The curvature of the edge area of ​​the transformer casing is optimized to keep the edge curvature consistent with other parts, avoiding sudden changes in the curvature of the edge area, and the generated multi-scale surface optimization data has a high degree of smoothness and curvature consistency.

[0195] Optionally, the shape constraint geometry correction comprises:

[0196] Perform preliminary shape classification according to the facility mapping data to obtain shape classification data;

[0197] Specifically, based on the facility mapping data, different parts of the facility are preliminarily classified according to shape features. Generally, shapes can be divided into standard geometric shapes (such as cylinders, rectangles, planes) and complex shapes (such as curved surfaces or irregular structures). By analyzing the contours and boundary features of each part of the facility, the geometric shapes are classified to determine whether they are basic geometric bodies (such as straight lines, arcs) or complex shapes. On the transformer casing, the supporting structure is preliminarily classified as a straight line or rectangular part, while the main body of the casing is circular. The generated shape classification data contains the geometric shape category information of each component.

[0198] Perform minimum envelope shape detection according to shape classification data to obtain minimum envelope data;

[0199] Specifically, based on the shape classification, the minimum envelope detection is performed on each classified area to find the minimum outer envelope geometry of each shape. The minimum envelope is used to simplify the shape description and provide a reference for subsequent shape constraints. Use the most suitable geometric shape (such as rectangle, cylinder) for outer envelope fitting to ensure that the shape is constrained within the envelope. For example, the transformer support structure is fitted with a rectangular outer envelope to obtain the minimum envelope data of the support part; the outer shell is fitted with a cylindrical outer envelope. The generated minimum envelope data contains the minimum geometric envelope of each shape.

[0200] Perform shape constraint matching according to the minimum envelope data to obtain shape constraint matching data;

[0201] Specifically, the detected minimum envelope data is constrained to match the actual shape of the facility to ensure that each shape conforms to the geometric characteristics within its envelope. For example, the straight part should strictly maintain straightness and the cylindrical part should conform to the cylindrical shape. Through geometric alignment and shape constraints, each shape is ensured to be adjusted within its envelope to maintain the overall shape regularity. The support structure of the transformer is shape-constrained so that the straight lines of the support part conform to its rectangular envelope and the curved surface of the shell conforms to its cylindrical envelope. The generated shape constraint matching data ensures that the main shape of the facility conforms to the geometric constraints.

[0202] Perform local refinement and global shape correction according to shape constraint matching data to obtain shape refinement geometry data and shape correction data respectively;

[0203] Specifically, local refinement is performed in areas of high curvature or complex geometric structures to increase the mesh density and improve the geometric accuracy of these areas. Refinement is mainly targeted at complex shapes, such as sharp corners, connections or curved areas. The geometry of the entire facility is globally corrected to ensure the continuity and consistency of the overall shape, especially to maintain geometric regularity in large-scale planes, straight lines or circular areas. Local refinement is performed at the connection of the transformer casing to increase the mesh density and make the curved surface of the casing more refined; global corrections are performed on the overall straight line part of the supporting structure to ensure its overall geometric regularity. The generated shape refinement geometry data and shape correction data describe the details after local and global corrections, respectively.

[0204] Perform edge constraint optimization according to shape refinement geometry data to obtain edge constraint data;

[0205] Specifically, the edge of the facility is optimized with constraints to ensure that the geometry of the edge meets the actual engineering requirements. The purpose of the optimization is to ensure that the shape of the edge remains smooth or linear to avoid distortion or fracture. According to the overall shape requirements of the facility, the geometry of the edge is corrected, especially to ensure the smooth connection between the edge and the overall geometry. Constrained optimization is performed on the edge area of ​​the transformer casing to ensure that the edge curve is consistent with the main body of the casing and there is no obvious deformation or fracture. The generated edge constraint data ensures the geometric accuracy of the edge.

[0206] Data fusion is performed based on the edge constraint data and the shape correction data to obtain a second facility geometric reference model.

[0207] Specifically, the edge constraint data is fused with the shape correction data to form a complete geometric model. During the fusion process, the transition between the edge part and the overall geometric shape is ensured to be natural, and the accuracy of the local details is maintained. During the fusion process, the relationship between the local details and the global shape is adjusted to ensure that the model maintains continuity and consistency globally. The edge data of the transformer shell is fused with the overall surface shape to ensure that the shape of the shell edge matches the global geometric form of the shell. The resulting second facility geometric reference model contains both local fine shapes and overall regularity and consistency.

[0208] Optionally, the space-varying geometry correction comprises:

[0209] Capture spatial changes based on facility mapping data to obtain spatial change field data;

[0210] Specifically, the changes in the shape of the facility in different spatial dimensions are analyzed through the facility mapping data, capturing the deformation of the facility caused by external forces, environment or manufacturing process. By capturing the changes on the surface or inside of the facility, spatial variation field data is generated. Combined with laser scanning and image data, the changes in facilities in different spatial positions are captured to form a multi-dimensional variation field. On a transmission tower, due to the slight displacement caused by the external environment (such as wind load), by analyzing the changes in the surface of the facility from multiple perspectives, the offset of the top of the transmission tower is captured as 5 mm. The generated spatial variation field data includes these subtle spatial displacement information.

[0211] Perform deformation analysis and modeling based on the spatial variation field data to obtain deformation model data;

[0212] Specifically, the local or overall deformation of the facility is analyzed based on the spatial variation field data. The local deformation characteristics of the facility under the action of external forces or the environment, such as bending, stretching or compression, are analyzed. Through this analysis, a deformation model is formed to describe the specific deformation state of the facility. Through modeling tools, the captured deformation information is converted into a geometric model, and the deformation model can represent the local or global deformation of the facility. For example, at the top of the transmission tower, due to wind load, the top structure has a slight tilt. Based on the captured offset information, a deformation model is constructed to describe the deformation characteristics and offset of the area. The generated deformation model data reflects the degree of deformation of different parts.

[0213] Perform local deformation optimization according to the deformation model data to obtain local deformation optimization data;

[0214] Specifically, based on the deformation model, the local deformation area is optimized, focusing on the adjustment and correction of the area with larger deformation. The goal of optimization is to ensure that the deformation area of ​​the facility is within a physically reasonable range while maintaining geometric accuracy. The deformation area is locally encrypted or reconstructed to make the deformation smoother and more natural in space, reducing local distortion or irregular deformation caused by external forces. In the inclined area at the top of the transmission tower, the local structure is optimized to reduce local irregular deformation and make the deformation transition smoother. The optimized local deformation optimization data ensures the rationality of the deformation area while maintaining the overall geometric consistency of the facility.

[0215] Perform spatial consistency constraints based on local deformation optimization data to obtain spatial consistency data;

[0216] Specifically, after local optimization, the shape of the entire facility is constrained for spatial consistency. Ensure that the local deformation area is spatially consistent with the surrounding geometric structure to avoid unreasonable spatial deformation of the overall structure due to local adjustments. According to the geometric requirements of the facility, adjust the local and global connection parts to ensure that the deformation area is well connected with the spatial form of the overall facility and ensure the overall spatial consistency of the facility. In the top area of ​​the transmission tower, after local deformation optimization, the connection between this area and other parts is checked and corrected for consistency to ensure that the inclination and deformation of the top will not affect the structural integrity of other parts of the transmission tower. The generated spatial consistency data ensures the consistency of the facility shape on a global scale.

[0217] Physical simulation deformation correction is performed based on the spatial consistency data to obtain the third facility geometric reference model.

[0218] Specifically, physical simulation tools are used to correct the deformation of the facility, and the deformed area is further corrected based on the physical properties under the action of external forces (such as the rigidity, elasticity, stress, etc. of the material) to ensure that the deformation is within the physically acceptable range. According to the material properties of the facility, the geometric structure of the deformed area is adjusted to conform to the laws of physics, such as ensuring that the structure maintains uniform stress distribution after deformation. Through physical simulation tools, the deformation of the transmission tower under wind load is analyzed to ensure that the top tilt will not affect the safety of the structure, and it is corrected. The adjusted structure is more stable, and the generated third facility geometric reference model includes the deformation state after physical correction.

[0219] Optionally, S4 includes:

[0220] S41, generating a fine network based on the three-dimensional model of rural energy facilities to obtain a fine grid model of the facilities;

[0221] Specifically, based on the preliminary 3D model, a high-precision mesh structure is generated. The geometry of the model is refined by increasing the mesh density, especially in complex or critical areas, including refining the original triangular mesh or using interpolation techniques to generate finer meshes. For the curved surfaces and edge parts of the facility, local encryption technology is used to generate higher-resolution meshes to ensure the accuracy of the model in these areas. For example, a higher-density triangular mesh is generated on the surface of the transformer casing. The initial mesh contains 20,000 facets. After refinement, the generated facility refined mesh model contains 80,000 facets, which more accurately describes the curved surface and edge details of the casing.

[0222] S42, performing topology optimization according to the facility refined grid model to obtain a facility topology optimization model;

[0223] Specifically, the topology is optimized on the basis of the fine grid, by reducing unnecessary faces or optimizing the grid structure to ensure that the geometry of the model is more in line with practical applications while reducing computational complexity. The optimization process includes simplifying unimportant parts and adjusting the grid structure to improve efficiency while maintaining accuracy. Redundant faces are simplified, especially in flat areas or areas with little geometric change, while high accuracy is maintained for complex areas. For example, topological simplification of the flat area of ​​the transformer casing reduces 30% of redundant faces, while the edges and complex structures of the casing maintain high accuracy. The model after topology optimization has a more reasonable grid structure, which saves resources and maintains accuracy.

[0224] S43, calculating material reflectivity according to the facility topology optimization model to obtain a facility reflectivity model;

[0225] Specifically, based on the material properties of the facility, the optical reflectivity of each part of the surface is calculated. By analyzing the characteristics of the materials (such as metal, plastic, coating, etc.), the light reflection behavior of different areas is calculated to ensure that the model performs realistically under lighting. According to the reflection characteristics of different materials, the corresponding reflectivity model is generated, especially between smooth surfaces and rough surfaces, the difference in reflectivity is large. For the metal shell area of ​​the transformer, its high reflectivity value is calculated, while for its coating and rubber bracket area, the reflectivity value is lower. The generated facility reflectivity model accurately describes the reflection characteristics of different material areas.

[0226] S44, performing collision detection and correction according to the facility reflectivity model to obtain a facility collision correction model;

[0227] Specifically, through the collision detection algorithm in the virtual environment, it is detected whether there are collisions or overlaps between the parts of the facility model, especially whether the gaps between different parts are sufficient. During the detection process, it is checked whether the model conforms to the geometric constraints of the physical world. For the detected collision or overlap problems, local geometric corrections are performed to adjust the distance and position between the parts to avoid interference in actual operation. At the connection of the transmission tower, geometric overlaps between the support structure and the cable were found. These positions were corrected through collision detection, and the spacing between the support and the cable was adjusted. The generated facility collision correction model eliminates the geometric interference problem.

[0228] S45, performing detail texture enhancement according to the facility collision correction model to obtain a facility detail texture enhancement model;

[0229] Specifically, the visual details of the facility are enhanced by refining its surface texture. For example, the surface of the facility is made more realistic by adding surface microstructures, cracks, wear marks, etc. Use high-resolution texture maps or normal maps to enhance the visual effect. Perform high-resolution texture mapping on the refined model surface to ensure that the model can show realistic material details under changes in lighting and viewing angles. For example, subtle wear textures and metal surface details are added to the transformer shell, and high-resolution normal maps are applied to the bracket part. The generated facility detail texture enhancement model is more visually realistic.

[0230] S46. Material aging simulation is performed based on the facility detail texture enhancement model to obtain a refined model of rural energy facilities.

[0231] Specifically, by simulating the material aging process of facilities under long-term use or environmental influences, adding details such as rust, fading, and surface wear, the model is made more consistent with the actual state after long-term use. The aging simulation is performed according to the aging laws of different materials to generate realistic aging effects. According to time and environmental factors (such as climate, pollution, etc.), the surface texture and material of the facilities are dynamically adjusted to generate a model that conforms to the service life. The metal part of the transmission tower is simulated for corrosion. Over time, rust appears on the metal surface; the coating of the transformer casing begins to fade. The generated refined model of rural energy facilities includes the material changes after aging simulation, showing the true state of the facilities after long-term use.

[0232] Therefore, from any point of view, the embodiments should be regarded as illustrative and non-restrictive, and the scope of the present invention is limited by the attached application documents rather than the above description, and it is intended that all changes falling within the meaning and scope of equivalent elements of the application documents are included in the present invention.

[0233] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be 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 present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A method for rapid modeling of rural energy facilities based on three-dimensional laser point cloud, characterized in that: The method comprises: S1. Collect point cloud data of energy facilities in rural areas using 3D laser scanning equipment to obtain point cloud data of rural energy facilities; S2. Perform target recognition based on the rural energy facility point cloud data to obtain rural energy facility target data; S3, constructing a three-dimensional model according to the target data of rural energy facilities to obtain a three-dimensional model of rural energy facilities; S4. Perform detailed modeling based on the three-dimensional model of rural energy facilities to obtain a detailed model of rural energy facilities.

2. The method according to claim 1, characterized in that S1 includes: Use laser scanning equipment to determine the scanning range and perform multi-dimensional spatial positioning to ensure that all energy facilities in rural areas are covered and obtain scanning space coordinate data; Use a 3D laser scanning device to collect high-density point cloud data by scanning spatial coordinate data to obtain an initial point cloud data set; Optimize the scanning perspective of the initial point cloud data set to obtain perspective point cloud data; Perform multi-source data fusion on the viewpoint point cloud data to obtain fused point cloud data; The fused point cloud data is time synchronized to obtain the point cloud data of rural energy facilities.

3. The method according to claim 1, characterized in that S2 include: Perform preliminary filtering based on the rural energy facility point cloud data to obtain point cloud filtering data; Density estimation and segmentation are performed based on the point cloud filtering data to obtain point cloud segmentation data; Perform morphological feature extraction based on point cloud segmentation data to obtain facility point cloud morphological feature data; Energy signal analysis is performed based on the morphological characteristic data of rural energy facility point clouds to obtain energy characteristic data of facility point clouds; Multi-objective clustering is performed based on the energy characteristic data of rural energy facility point clouds to obtain facility clustering data; Perform deep target detection based on facility clustering data to obtain facility deep detection classification data; Extract edge features based on facility depth detection classification data to obtain facility edge feature data; The rural energy facility point cloud data is semantically annotated based on the facility edge feature data and facility depth detection classification data to obtain the rural energy facility target data.

4. The method according to claim 1, characterized in that S3 include: Reconstruct the network based on the rural energy facility target data to obtain a preliminary facility network model; Performing surface refinement on the preliminary facility network model to obtain a facility surface refinement model; Perform texture mapping on the facility surface refinement model to obtain facility mapping data; Perform geometric correction according to the facility mapping data to obtain a facility geometric reference model; Embedding structural features according to the facility geometric benchmark model to obtain a facility embedded structural feature model; Physical attributes are assigned according to the facility embedded structural feature model to obtain a three-dimensional model of rural energy facilities.

5. The method according to claim 4, characterized in that The geometry corrections include: Performing surface fitting geometry correction according to the facility mapping data to obtain a first facility geometry reference model; Perform shape constraint geometry correction according to the facility mapping data to obtain a second facility geometry reference model; Performing spatial variation geometry correction according to the facility mapping data to obtain a third facility geometry reference model; Acquire rural energy facility image data; Extract rural energy facility features and estimate rural energy facility parameters based on rural energy facility image data to obtain rural energy facility image feature data and rural energy facility parameter data; The first facility geometric reference model, the second facility geometric reference model and the third facility geometric reference model are fused according to the rural energy facility image feature data and the rural energy facility parameter data to obtain a facility geometric reference model.

6. The method according to claim 5, characterized in that The surface fitting geometry correction includes: Perform multi-scale surface fitting initialization according to the facility mapping data to obtain multi-scale surface fitting data; Perform local curvature optimization according to multi-scale surface fitting data to obtain multi-scale surface optimization data; Perform B-spline surface fitting according to the multi-scale surface optimization data to obtain B-spline surface data; Perform a topological consistency check on the B-spline surface data to obtain the surface topological consistency data; Perform global error minimization based on the surface topology consistency data to obtain surface error minimization data; Matching and mapping are performed based on the B-spline surface data and the surface error minimization data to obtain the first facility geometric reference model.

7. The method according to claim 6, characterized in that The local curvature optimization includes: The curvature is calculated according to the multi-scale surface fitting data to obtain the multi-scale surface curvature data; Performing curvature partitioning on the multi-scale surface fitting data according to the multi-scale surface curvature data to obtain curvature partitioning data; Extracting curvature partition features according to the curvature partition data to obtain curvature partition feature data; Constructing a curvature map according to the curvature partition characteristic data and the curvature partition data to obtain curvature map data; Refining the curvature partition data in a neighboring area according to the curvature image data to obtain curvature partition refined data; Perform curvature smoothing control according to curvature partitioning and refinement data to obtain curvature smoothing data; The boundary curvature is optimized according to the curvature smoothing data to obtain multi-scale surface optimization data.

8. The method according to claim 5, characterized in that The shape constraint geometry correction includes: Perform preliminary shape classification according to the facility mapping data to obtain shape classification data; Perform minimum envelope shape detection according to shape classification data to obtain minimum envelope data; Perform shape constraint matching according to the minimum envelope data to obtain shape constraint matching data; Perform local refinement and global shape correction according to shape constraint matching data to obtain shape refinement geometry data and shape correction data respectively; Perform edge constraint optimization according to shape refinement geometry data to obtain edge constraint data; Data fusion is performed based on the edge constraint data and the shape correction data to obtain a second facility geometric reference model.

9. The method according to claim 5, characterized in that The spatially varying geometry corrections include: Capture spatial changes based on facility mapping data to obtain spatial change field data; Perform deformation analysis and modeling based on the spatial variation field data to obtain deformation model data; Perform local deformation optimization according to the deformation model data to obtain local deformation optimization data; Perform spatial consistency constraints based on local deformation optimization data to obtain spatial consistency data; Physical simulation deformation correction is performed based on the spatial consistency data to obtain the third facility geometric reference model.

10. The method according to claim 1, characterized in that S4 include: Generate a fine network based on the three-dimensional model of rural energy facilities to obtain a fine grid model of the facilities; Perform topology optimization based on the refined mesh model of the facility to obtain a facility topology optimization model; Calculate the material reflectivity according to the facility topology optimization model to obtain a facility reflectivity model; Perform collision detection and correction according to the facility reflectivity model to obtain a facility collision correction model; Perform detail texture enhancement according to the facility collision correction model to obtain a facility detail texture enhancement model; Material aging simulation is performed based on the facility detail texture enhancement model to obtain a refined model of rural energy facilities.

Images