A method and system for processing wall panels for three-dimensional steel structures in substations.

Abstract

This invention discloses a processing method and system for three-dimensional steel structure wall panels in substations, applicable to substations with three-dimensional steel structure wall panels assembled within their topological structure. The method includes: denoising and simplifying initial image data of the three-dimensional steel structure wall panels to be processed; generating a three-dimensional spatial depth map corresponding to the wall panels; analyzing each pixel in the three-dimensional spatial depth map to obtain a marked three-dimensional spatial depth map; performing pixel completion processing on the initial image data to obtain second image data; inputting the second image data and engineering attribute data into a splitting strategy model for analysis and processing to obtain a target splitting strategy; and generating a processing strategy for the three-dimensional steel structure wall panels to be processed based on the target splitting strategy. The processing method for three-dimensional steel structure wall panels in substations provided by this invention significantly improves the processing efficiency of these wall panels.

Description

Technical Field

[0001] This invention relates to the field of wall panel technology for three-dimensional steel structures in substations, and in particular to a processing method and system for wall panels used in three-dimensional steel structures in substations. Background Technology

[0002] The processing of prefabricated steel structure wall panels for substations is a key step in ensuring efficient assembly and shortening the construction cycle, directly affecting the overall project's progress and construction quality.

[0003] In existing technologies, the processing methods for wall panels used in three-dimensional steel structures of substations rely on manual experience. However, this method requires manual handling of multi-dimensional parameters, making it difficult to avoid subjective oversights and objective operational errors. This often results in dimensional deviations in the final breakdown scheme. For example, the breakdown methods are based on the subjective judgments of the processing personnel regarding the panel type. When encountering large rectangular wall panels, some inexperienced processing personnel may rely solely on conventional processing habits, subjectively assuming that a fixed ratio is sufficient. This leads to a lack of unified scientific standards and precise data support for the breakdown scheme. Not only will the dimensions of the broken panels deviate significantly from the design drawings, making it impossible to accurately assemble the panels on-site, requiring repeated adjustments or even reprocessing, severely impacting the construction progress, but unreasonable breakdown can also lead to a significant reduction in material utilization, generating a large amount of scrap material and increasing production costs. Furthermore, dimensional deviations can affect the overall structural stability of the wall panels, posing a potential hazard to the safe operation of the three-dimensional steel structure of the substation. Summary of the Invention

[0004] This invention provides a processing method and system for wall panels used in three-dimensional steel structures of substations, which solves the technical problem of relying on manual experience to process wall panels used in three-dimensional steel structures of substations in existing methods, thereby improving processing efficiency.

[0005] To address the aforementioned technical problems, embodiments of the present invention provide a method and system for processing wall panels for three-dimensional steel structures in substations, the method comprising:

[0006] Acquire initial image data of the wall panels for the three-dimensional steel structure of the substation to be processed;

[0007] The initial image data is sequentially processed by statistical filtering algorithm and voxel lattice downsampling technique for noise reduction and simplification to obtain the first three-dimensional point cloud data;

[0008] The first three-dimensional point cloud data is subjected to spatial rasterization processing to generate a three-dimensional spatial depth map corresponding to the wall panel of the three-dimensional steel structure of the substation.

[0009] Each pixel in the three-dimensional spatial depth map is analyzed, and the depth difference information in the analysis results is used as the basis for occlusion determination of the initial image data to obtain a marked three-dimensional spatial depth map.

[0010] The coordinate system transformation analysis is performed sequentially on the marker information in the marked three-dimensional spatial depth map and the structural information data confirmed by the substation topology to obtain the three-dimensional spatial data to be completed.

[0011] The initial image data is pixel-completed based on the three-dimensional spatial data to be completed to obtain the second image data.

[0012] The third image data, obtained by fusing the second image data and the acquired engineering attribute data of the substation, is input into a splitting strategy model constructed by a deep learning algorithm for analysis and processing to obtain the target splitting strategy.

[0013] Based on the target decomposition strategy, a processing strategy is generated for the wall panels of the three-dimensional steel structure of the substation to be processed.

[0014] Preferably, the initial image data is sequentially subjected to statistical filtering algorithm and voxel grid downsampling technique for noise reduction and simplification to obtain the first three-dimensional point cloud data, including:

[0015] The spatial coordinate information of the initial image data is parsed and extracted to obtain the original three-dimensional point cloud data;

[0016] The statistical filtering algorithm is used to identify and remove noise points from the original 3D point cloud data to obtain denoised 3D point cloud data.

[0017] Based on the denoised 3D point cloud data, the voxel grid downsampling technique is used to perform grid division to obtain rasterized point cloud data.

[0018] The point within each grid that is closest to the grid center is identified as the sampling point;

[0019] The sampling points of each grid are aggregated to obtain the first three-dimensional point cloud data.

[0020] Preferably, the step of performing spatial rasterization processing on the first three-dimensional point cloud data to generate a three-dimensional spatial depth map corresponding to the three-dimensional steel structure wall panel of the substation includes:

[0021] Based on the first three-dimensional point cloud data, the gridded area of ​​the wall panel for the three-dimensional steel structure of the substation is selected;

[0022] The rasterized area is divided according to a preset resolution to obtain a spatial raster structure containing multiple raster units;

[0023] Calculate the average depth value of the point cloud data in each grid cell within the spatial grid structure to obtain the depth information of the corresponding two-dimensional pixel.

[0024] The three-dimensional spatial depth map is generated based on the depth information of the two-dimensional pixels corresponding to each of the grid units.

[0025] Preferably, the step of analyzing each pixel in the three-dimensional spatial depth map and using the depth difference information in the analysis results as the basis for occlusion determination of the initial image data to obtain a labeled three-dimensional spatial depth map includes:

[0026] Extract the depth value of each pixel in the three-dimensional spatial depth map and all adjacent valid pixels within a preset range to obtain a depth value set;

[0027] Calculate the depth difference between the pixel and each of its adjacent valid pixels, and take the average of the absolute values ​​of the differences as the depth fluctuation value of the pixel.

[0028] The pixels whose depth fluctuation values ​​in the three-dimensional spatial depth map exceed a preset threshold are marked as occluded pixels, thus obtaining the marked three-dimensional spatial depth map.

[0029] Preferably, the step of sequentially performing coordinate system transformation analysis on the marker information in the marked three-dimensional spatial depth map and the structural information data confirmed by the substation topology to obtain the three-dimensional spatial data to be completed includes:

[0030] Transform the occlusion region corresponding to the marked information from the image coordinate system to the substation global coordinate system;

[0031] Based on the structural information data, the initial three-dimensional data of the unobstructed region that is consistent with the structural type of the obstructed region is selected;

[0032] By combining the three-dimensional features of the initial three-dimensional data of the unobstructed area with the spatial location of the obstructed area, the three-dimensional spatial data to be completed is generated.

[0033] Preferably, the step of performing completion processing on the initial image data based on the three-dimensional spatial data to be completed to obtain the second image data includes:

[0034] Based on the three-dimensional spatial data to be completed, the reverse perspective transformation technique is used to project the three-dimensional spatial data to be completed from the global coordinate system back to the pixel coordinate system of the initial image data, and to determine the pixel position and texture mapping relationship of the region to be completed in the initial image data.

[0035] By combining the texture features of adjacent unoccluded areas in the initial image data, pixel filling is performed on the area to be filled to obtain the second image data.

[0036] Another aspect of the present invention provides a processing system for wall panels for three-dimensional steel structures in substations, applied in substations where the three-dimensional steel structure wall panels are assembled within the topological structure, comprising:

[0037] The acquisition module is used to acquire the initial image data of the wall panels for the three-dimensional steel structure of the substation to be processed;

[0038] The noise reduction module is used to sequentially apply statistical filtering algorithm and voxel grid downsampling technology to the initial image data for noise reduction and simplification processing to obtain the first three-dimensional point cloud data;

[0039] The depth map module is used to perform spatial rasterization processing on the first three-dimensional point cloud data to generate a three-dimensional spatial depth map corresponding to the three-dimensional steel structure wall panel of the substation.

[0040] The analysis module is used to analyze each pixel in the three-dimensional spatial depth map and use the depth difference information in the analysis results as the basis for occlusion determination of the initial image data to obtain a marked three-dimensional spatial depth map.

[0041] The completion module is used to sequentially perform coordinate system transformation analysis on the marker information in the marked 3D spatial depth map and the structural information data confirmed by the substation topology to obtain the 3D spatial data to be completed.

[0042] The completion module is used to perform pixel completion processing on the initial image data based on the three-dimensional spatial data to be completed, so as to obtain the second image data.

[0043] The strategy module is used to input the third image data obtained by fusing the second image data and the acquired engineering attribute data of the substation into a splitting strategy model constructed by a deep learning algorithm for analysis and processing, so as to obtain the target splitting strategy.

[0044] The generation module is used to generate a processing strategy for the wall panels of the three-dimensional steel structure of the substation to be processed based on the target splitting strategy.

[0045] Preferably, the noise reduction module includes:

[0046] The parsing unit is used to parse and extract the spatial coordinate information of the initial image data to obtain the original three-dimensional point cloud data.

[0047] The noise reduction unit is used to identify and remove noise points from the original 3D point cloud data using the statistical filtering algorithm to obtain denoised 3D point cloud data.

[0048] The partitioning unit is used to perform grid partitioning based on the denoised 3D point cloud data using the voxel grid downsampling technique to obtain rasterized point cloud data.

[0049] A determining unit is used to determine the point within each grid that is closest to the grid center as a sampling point;

[0050] A set unit is used to set the sampling points of each of the grids to obtain the first three-dimensional point cloud data.

[0051] Preferably, the depth map module includes:

[0052] The selected unit is used to select the gridded area of ​​the wall panel for the three-dimensional steel structure of the substation based on the first three-dimensional point cloud data.

[0053] A grid structure unit is used to divide the gridded area according to a preset resolution to obtain a spatial grid structure containing multiple grid units.

[0054] A depth information unit is used to calculate the average depth value of the point cloud data in each grid unit within the spatial grid structure, and to obtain the depth information of the corresponding two-dimensional pixel.

[0055] A generation unit is configured to generate the three-dimensional spatial depth map based on the depth information of the two-dimensional pixels corresponding to each of the grid units.

[0056] Preferably, the analysis module includes:

[0057] The extraction unit is used to extract the depth value of each pixel in the three-dimensional spatial depth map and all adjacent valid pixels within a preset range to obtain a set of depth values.

[0058] The calculation unit is used to calculate the depth difference between the pixel and each of the adjacent effective pixels, and take the average of the absolute values ​​of the differences as the depth fluctuation value of the pixel.

[0059] A marking unit is used to mark the pixels in the three-dimensional spatial depth map whose depth fluctuation value exceeds a preset threshold as occluded pixels, thereby obtaining the marked three-dimensional spatial depth map.

[0060] Compared with the prior art, the beneficial effects of the present invention are at least one of the following:

[0061] This invention acquires initial image data of the wall panel, simplifies it through statistical filtering and voxel grid downsampling to obtain the first 3D point cloud data, then generates a 3D spatial depth map through spatial rasterization. It combines depth difference to mark occlusion information, integrates substation topology information to complete coordinate system transformation and data completion, and finally fuses engineering attribute data and outputs a processing strategy through a deep learning-based splitting strategy model. This multi-step data processing achieves precise optimization of the initial image data of the wall panel, effectively solving problems such as image noise, data loss, and occlusion interference, providing a high-quality data foundation for subsequent processing and analysis. By combining substation topology information, engineering attribute data, and deep learning algorithms, the generated splitting and processing strategies closely adapt to the actual structure and engineering requirements of the substation, significantly improving the scientific rigor, relevance, and feasibility of the processing strategy, and ensuring the accuracy and stability of the wall panel assembly with the substation's 3D steel structure. Attached Figure Description

[0062] Figure 1 This is a flowchart illustrating the processing method of wall panels for three-dimensional steel structures in substations according to one embodiment of the present invention.

[0063] Figure 2 This is a schematic diagram of the processing system for a three-dimensional steel structure wall panel in a substation according to one embodiment of the present invention;

[0064] Figure label:

[0065] The module consists of: 11. Acquisition module; 12. Noise reduction module; 13. Depth map module; 14. Analysis module; 15. To be completed module; 16. Completement module; 17. Strategy module; and 18. Generation module. Detailed Implementation

[0066] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0067] In the description of this application, the terms "first," "second," "third," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined with "first," "second," "third," etc., may explicitly or implicitly include one or more of that feature. In the description of this application, unless otherwise stated, "a plurality of" means two or more.

[0068] In the description of this application, it should be noted that, unless otherwise defined, all technical and scientific terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art. The terminology used in this specification is for the purpose of describing specific embodiments only and is not intended to limit the invention. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0069] The processing of prefabricated wall panels for three-dimensional steel structures in substations is crucial for ensuring efficient substation assembly and shortening the construction cycle, directly impacting project progress and construction quality.

[0070] Current technology relies on manual experience to process wall panels for three-dimensional steel structures in substations. This requires manual handling of multi-dimensional parameters and is prone to problems such as dimensional deviations in the disassembly plan due to subjective oversights and operational errors. For example, disassembly is often based on the subjective judgment of the processing personnel, with large rectangular wall panels frequently disassembled according to fixed proportions. This lack of unified scientific standards and precise data support not only results in significant deviations between the disassembled panels and the design drawings, making precise alignment during on-site assembly difficult and requiring repeated adjustments or even rework, thus affecting construction progress, but also reduces material utilization, increases costs, and, due to dimensional deviations, affects the structural stability of the wall panels, posing potential risks to the safe operation of the substation's steel structure.

[0071] One embodiment of the present invention provides a method for processing wall panels for three-dimensional steel structures in substations. For details, please refer to [link to documentation]. Figure 1 , Figure 1 The diagram shown is a flowchart illustrating a method for processing wall panels for a three-dimensional steel structure in a substation according to one embodiment of the present invention. The method includes:

[0072] S1. Obtain the initial image data of the wall panel for the three-dimensional steel structure of the substation to be processed;

[0073] S2. The initial image data is sequentially processed by statistical filtering algorithm and voxel lattice downsampling technique to reduce noise and simplify the data, and the first three-dimensional point cloud data is obtained.

[0074] S3. Perform spatial rasterization processing on the first three-dimensional point cloud data to generate a three-dimensional spatial depth map corresponding to the wall panel of the three-dimensional steel structure of the substation.

[0075] S4. Analyze each pixel in the 3D spatial depth map, and use the depth difference information in the analysis results as the basis for occlusion judgment marking of the initial image data to obtain the marked 3D spatial depth map.

[0076] S5. Perform coordinate system transformation analysis on the marking information in the marked three-dimensional spatial depth map and the structural information data confirmed by the substation topology to obtain the three-dimensional spatial data to be completed.

[0077] S6. Perform pixel completion processing on the initial image data based on the three-dimensional spatial data to be completed to obtain the second image data;

[0078] S7. The third image data obtained by fusing the second image data and the acquired substation engineering attribute data is input into the splitting strategy model constructed by the deep learning algorithm for analysis and processing to obtain the target splitting strategy.

[0079] S8. A processing strategy for generating wall panels for the three-dimensional steel structure of substations based on a target decomposition strategy.

[0080] In step S1, acquiring initial image data requires high-precision image acquisition equipment adapted to industrial processing scenarios. This typically involves industrial-grade scanning equipment with 3D scanning capabilities, high-resolution industrial cameras paired with multi-angle shooting brackets, etc. The entire wall panel to be processed is scanned comprehensively or continuously photographed from multiple angles at close range. Simultaneously, it's crucial to ensure stable, unobstructed lighting in the acquisition environment to avoid external interference that could lead to missing image information. Regarding the acquired data, it's not simply about collecting photos of the wall panel's appearance; rather, it involves collecting image data that comprehensively reflects the panel's key attributes. This includes the wall panel's 3D geometric data, such as image pixel mapping information corresponding to the overall length, width, and thickness; surface flatness, curvature, and corner shapes; and image features corresponding to the location, size, and number of interfaces and holes used for connection to the substation's 3D steel structure. It also includes surface condition data, such as image traces of scratches, dents, and protrusions, as well as images of the surface material's texture, color, and other appearance characteristics. This data is stored in digital image file format, providing comprehensive and accurate raw data for subsequent processing such as noise reduction and depth analysis using algorithms.

[0081] The spatial coordinate information of the initial image data is analyzed and extracted to obtain the original 3D point cloud data. A statistical filtering algorithm is then used to identify and remove noise points from the original 3D point cloud data, resulting in denoised 3D point cloud data. Based on the denoised 3D point cloud data, voxel grid downsampling technology is used to divide the data into a grid, resulting in rasterized point cloud data. The point closest to the grid center within each grid is determined as the sampling point. The sampling points of each grid are then aggregated to obtain the first 3D point cloud data. After acquiring the initial image data of the wall panels for the 3D steel structure of the substation to be processed, a series of steps are used to transform the original image data into first 3D point cloud data more suitable for subsequent analysis. First, spatial coordinate information containing the spatial location information of the wall panel is extracted from the acquired initial image data, thus obtaining raw 3D point cloud data that intuitively reflects the 3D shape of the wall panel. Then, for noise points that may exist in the raw 3D point cloud data due to interference from the acquisition environment or equipment precision deviations, a statistical filtering algorithm is used to identify and remove them, thereby obtaining denoised 3D point cloud data free of redundant interference. Based on this, voxel grid downsampling technology is used to divide the denoised 3D point cloud data into multiple regular grid cells. Next, within each grid cell, the point closest to the grid center is selected as the sampling point for that grid, thus preserving the most representative geometric information within the grid. Finally, the sampling points of all grids are aggregated to obtain the first 3D point cloud data that retains the core 3D features of the wall panel while simplifying the data scale, laying an efficient and accurate data foundation for the subsequent generation of a 3D spatial depth map.

[0082] Furthermore, based on the first three-dimensional point cloud data, a gridded area for the wall panel of the three-dimensional steel structure of the substation is selected; the gridded area is divided according to a preset resolution to obtain a spatial grid structure containing multiple grid units; the average depth value of the point cloud data in each grid unit within the spatial grid structure is calculated to obtain the depth information of the corresponding two-dimensional pixel; based on the depth information of the two-dimensional pixel corresponding to each grid unit, a three-dimensional spatial depth map is generated. First, based on the first-dimensional point cloud data, the rasterized area of ​​the wall panel requiring depth analysis is precisely selected. Then, according to a pre-set resolution standard, the selected rasterized area is divided into multiple regular raster units, thus constructing a clear spatial raster structure. Specifically, if the wall panel to be processed is large and subsequent precise assessment of the occlusion between the wall panel and the substation steel structure assembly interface is required, the preset resolution can be set higher, for example, dividing each meter length into 500 raster units, i.e., setting the resolution to 500 raster units per meter. In this case, each raster unit is smaller, allowing for more detailed capture of subtle depth changes on the wall panel surface. If the wall panel is smaller and the focus is on overall depth... Instead of focusing on local details, the preset resolution can be appropriately reduced, for example, dividing each meter into two hundred grid units. By using larger grid units, a spatial grid structure can be quickly constructed, improving processing efficiency. Then, for each grid unit in the spatial grid structure, the average depth value of all point cloud data within the unit is calculated. This average depth value will be used as the depth information of the corresponding two-dimensional pixel, thereby accurately representing the depth position of each grid unit in space. Finally, the depth information of the two-dimensional pixels corresponding to all grid units is integrated, transforming the scattered depth data into a three-dimensional spatial depth map that intuitively presents the spatial depth distribution of the wall panel. This provides clear and systematic depth data support for subsequent judgment of occlusion through depth difference.

[0083] Next, the depth values ​​of each pixel in the 3D spatial depth map and all its adjacent valid pixels within a preset range are extracted to obtain a depth value set. The depth difference between the pixel and each of its adjacent valid pixels is calculated, and the average of the absolute values ​​of the differences is taken as the depth fluctuation value of the pixel. Pixels in the 3D spatial depth map whose depth fluctuation values ​​exceed a preset threshold are marked as occluded pixels, resulting in a marked 3D spatial depth map. For each pixel in the 3D spatial depth map, the depth values ​​of the pixel and all its adjacent valid pixels within a predefined range are extracted, and these depth values ​​are integrated to form a corresponding depth value set. Then, the depth difference between the current pixel and each of its adjacent valid pixels in the set is calculated, and the average of the absolute values ​​of all differences is taken as the depth fluctuation value of the pixel. The magnitude of the depth fluctuation value reflects the degree of depth change in the area where the pixel is located. Afterward, a predetermined threshold is set, and pixels in the 3D spatial depth map whose depth fluctuation values ​​exceed the threshold are judged as occluded pixels, because excessively large depth fluctuation values ​​often indicate that there is depth variation caused by object occlusion in the area. For depth abrupt changes, specifically, if the previously set depth map resolution is 500 grid units per meter, and the wall panel to be processed is a thin steel structure wall panel with high flatness requirements, the preset threshold can be set to 3 mm; if the previously set resolution is 200 grid units per meter, and the wall panel is a thick structure wall panel with conventional reinforcing ribs, the preset threshold can be increased to 7 mm, thereby filtering out normal depth fluctuations caused by reinforcing ribs and only marking depth abrupt change pixels that exceed the normal range due to occlusion; finally, all identified occluded pixels are marked to obtain a marked 3D spatial depth map with occlusion information, laying the foundation for subsequent coordinate system transformation analysis combined with the substation topology.

[0084] Next, the occluded area corresponding to the marked information is transformed from the image coordinate system to the substation global coordinate system; based on the structural information data, the initial three-dimensional data of the unoccluded area with the same structural type as the occluded area is selected; combining the three-dimensional features of the initial three-dimensional data of the unoccluded area with the spatial location of the occluded area, the three-dimensional spatial data to be completed is generated. Considering that the marking information of the 3D spatial depth map is recorded based on the image's own coordinate system, while the assembly of the 3D steel structure of the substation requires a globally unified spatial reference, the occluded area corresponding to the marking information is transformed from the image coordinate system to the substation's global coordinate system that matches the overall structural layout of the substation. This ensures that the spatial position of the occluded area accurately corresponds to the assembly scene of the substation's steel structure. Specifically, after completing the initial image data acquisition of the wall panels, to ensure that the subsequent wall panel-related data can accurately match the overall structure of the substation, long-term stable feature points clearly marked on the design drawings within the substation are first selected as reference points. These feature points include the corner points of the land acquisition boundary line, the positioning points of the core building foundations, and the center points of the main equipment foundations. Then, professional surveying equipment is used to measure these reference points with high precision to determine their spatial coordinates, and the coordinate axis direction is determined in conjunction with the substation's main axis. Subsequently, the actual coordinates of the reference points measured on-site are compared and calibrated with the theoretical coordinate system in the design stage to eliminate construction errors, ultimately forming a global coordinate system that can be solidified and entered into the digital management system. This coordinate system provides a unified spatial reference for the subsequent transformation of the initial image data collected in the early stage. For example, it can support the transformation of the occluded area from the image coordinate system to the global coordinate system, while ensuring that the wall panel data can accurately correspond to the spatial position of other steel structures and equipment in the substation, laying a solid foundation for subsequent data completion and processing strategy formulation. Next, based on the structural information data confirmed by the substation topology, unoccluded areas with the same structural type as the occluded area are selected in the three-dimensional steel structure system of the substation, and the initial three-dimensional data of the unoccluded area is extracted. These data have the same structural features as the occluded area and can be used as a reliable reference for completion. Finally, the three-dimensional structural features contained in the initial three-dimensional data of the unoccluded area are fused and adapted with the spatial position of the occluded area that has been transformed to the global coordinate system. By matching the structural shape and connecting the spatial coordinates, three-dimensional spatial data to be completed that can fill the data gaps of the occluded area is generated, providing complete and realistic three-dimensional information support for subsequent pixel completion of the initial image data.

[0085] After generating the 3D spatial data to be completed, which conforms to the global coordinate system of the substation, the pixel completion stage of the initial image data begins to fill in the gaps in image information caused by occlusion. First, based on the 3D spatial data to be completed, a reverse perspective transformation technique is used to project the determined 3D data from the substation's global coordinate system back to the pixel coordinate system corresponding to the initial image. This transformation accurately locates the specific pixel position of the completed area in the initial image, while clarifying the texture mapping relationship between the completed area and the initial image, ensuring that the completed content matches the pixel distribution and texture direction of the original image. Next, referring to the texture features of the unoccluded areas surrounding the completed area in the initial image, such as surface material texture and color distribution patterns, pixel filling is performed on the completed area according to these features, ensuring that the filled pixels are compatible with adjacent pixels. The regions are seamlessly connected without obvious discontinuities. Specifically, based on the size of the completed area and the distribution of surrounding textures, it is divided into corresponding smaller regions. For each smaller region, the matching block with the closest texture and color is selected from the unoccluded area. Then, following the texture mapping direction, the pixel information of the matching block is mapped point by point to the completed area. Gradient fusion processing is applied to the boundary areas to gradually transition the pixel parameters and eliminate the filling boundary, while aligning the pixel contours corresponding to the wall panel structure. Finally, the filled area is compared with the overall image, and pixel parameters are fine-tuned or matching blocks are reselected to address issues such as texture misalignment and color imbalance, until the completed area and the unoccluded part are visually coherent and unified, completing the pixel filling. The final result is a second image data that fully presents the three-dimensional shape and surface features of the wall panel, providing a complete and coherent image foundation for subsequent fusion of engineering attribute data and construction of a splitting strategy model.

[0086] After obtaining the second image data, which fully presents the three-dimensional morphology and surface features of the wall panels used in the three-dimensional steel structure of the substation, the core step of formulating the processing strategy begins. First, the second image data is fused with pre-acquired substation engineering attribute data. This engineering attribute data includes key information such as the assembly dimension requirements of the substation steel structure, the material performance parameters of the wall panels, and the technical standards for construction and installation. The fusion results in a third image data that combines the visual characteristics of the wall panels with the actual engineering requirements. Specifically, the fusion of the second image data and the pre-acquired substation engineering attribute data requires establishing a connection using the substation's global coordinate system as a unified link. First, the pixel positions of each area of ​​the wall panel in the second image are mapped to three-dimensional spatial coordinates in the global coordinate system. Simultaneously, the engineering attribute data is bound to the corresponding wall panel component number and installation position in the global coordinate system during acquisition. Coordinate matching ensures that each pixel area in the second image corresponds to the relevant engineering attribute information. Next, for the complete morphology and surface state of the wall panel structural features in the second image, parameters such as size requirements, material process standards, and installation adaptation requirements from the engineering attributes are matched to clarify the engineering precision and quality standards corresponding to the visual features. Finally, using a single wall panel or its subdivided unit as the basic unit, a structured data body containing a visual data layer and an engineering attribute layer is constructed. The visual data layer retains the pixel matrix texture features and contour coordinates of the image, while the engineering attribute layer embeds parameters such as size, material, strength, and installation. The two are interconnected through data tags, ultimately outputting third image data that combines the complete visual form of the wall panel with the requirements of all-dimensional engineering attributes, preparing it for subsequent input into the subdivision strategy model. Next, the third image data is input into the subdivision strategy model constructed by a deep learning algorithm. This model, by learning the subdivision logic of a large number of similar substation wall panel processing cases, can combine the wall panel structural features and engineering attribute requirements in the third image data to accurately analyze and process the subdivision position, number of subdivisions, and size of the subdivided panels, ultimately outputting a target subdivision strategy adapted to actual processing and assembly. Finally, based on the target subdivision strategy, the specific requirements of the processing steps are further refined, such as the cutting accuracy of each subdivided panel, edge processing methods, and connection interface processing standards, thereby generating a processing strategy for the wall panels to be processed for the three-dimensional steel structure of the substation that can directly guide production, ensuring that the processed wall panels can accurately match the assembly requirements of the three-dimensional steel structure of the substation.

[0087] The training of the splitting strategy model involves constructing a third image dataset and its corresponding target splitting strategy dataset. This dataset includes the basic input data (third image data) and a structured output labeled with the target splitting strategy results, establishing a mapping relationship between input and output. Subsequently, machine learning algorithms are used to train the initial AI model based on this dataset. During training, methods such as transfer learning, hyperparameter optimization, or incremental training are combined to fine-tune the model, enhancing its generalization ability to unknown data. Finally, the optimized model matrix is ​​output. In the deployment phase, users only need to input the partial discharge characteristic data, vibration characteristic data, and voltage characteristic data to be analyzed into the model. The model can then automatically calculate and output the corresponding target splitting strategy results based on its learned intrinsic patterns.

[0088] Another embodiment of the present invention provides an evaluation system for a substation integrated monitoring device; please refer to [link to relevant documentation]. Figure 2 , Figure 2 The diagram shown is a structural schematic of a processing system for wall panels used in three-dimensional steel structures of substations, according to one embodiment of the present invention. The system includes:

[0089] The acquisition module 11 is used to acquire the initial image data of the wall panel for the three-dimensional steel structure of the substation to be processed;

[0090] The noise reduction module 12 is used to perform noise reduction and simplification processing on the initial image data by sequentially applying statistical filtering algorithm and voxel lattice downsampling technology to obtain the first three-dimensional point cloud data;

[0091] The depth map module 13 is used to perform spatial rasterization processing on the first three-dimensional point cloud data to generate a three-dimensional spatial depth map corresponding to the wall panel of the three-dimensional steel structure of the substation.

[0092] Analysis module 14 is used to analyze each pixel in the three-dimensional spatial depth map and use the depth difference information in the analysis results as the basis for occlusion judgment marking of the initial image data to obtain the marked three-dimensional spatial depth map.

[0093] The module 15 to be completed is used to perform coordinate system transformation analysis on the marker information in the marked three-dimensional spatial depth map and the structural information data confirmed by the substation topology to obtain the three-dimensional spatial data to be completed.

[0094] The completion module 16 is used to perform pixel completion processing on the initial image data based on the three-dimensional spatial data to be completed, so as to obtain the second image data;

[0095] Strategy module 17 is used to input the third image data obtained by fusing the second image data and the acquired substation engineering attribute data into the splitting strategy model constructed by the deep learning algorithm for analysis and processing, so as to obtain the target splitting strategy.

[0096] The generation module 18 is used to generate a processing strategy for the wall panels of the three-dimensional steel structure of the substation to be processed based on the target splitting strategy.

[0097] Preferably, the noise reduction module 12 includes:

[0098] The parsing unit is used to parse and extract the spatial coordinate information of the initial image data to obtain the original three-dimensional point cloud data;

[0099] The noise reduction unit is used to identify and remove noise points from the original 3D point cloud data using a statistical filtering algorithm, so as to obtain the denoised 3D point cloud data.

[0100] The partitioning unit is used to divide the denoised 3D point cloud data into grids using voxel lattice downsampling technology to obtain rasterized point cloud data.

[0101] The determination unit is used to determine the point within each grid that is closest to the grid center as the sampling point;

[0102] The set unit is used to aggregate the sampling points of each grid to obtain the first three-dimensional point cloud data.

[0103] Preferably, the depth map module 13 includes:

[0104] Select a cell to select the rasterized area of ​​the substation's three-dimensional steel structure wall panel based on the first three-dimensional point cloud data;

[0105] A grid structure unit is used to divide a gridded area according to a preset resolution to obtain a spatial grid structure containing multiple grid units.

[0106] The depth information unit is used to calculate the average depth value of the point cloud data in each grid cell within the spatial grid structure, and to obtain the depth information of the corresponding two-dimensional pixel.

[0107] The generation unit is used to generate a three-dimensional spatial depth map based on the depth information of the two-dimensional pixels corresponding to each grid unit.

[0108] Preferably, the analysis module 14 includes:

[0109] The extraction unit is used to extract the depth value of each pixel in the three-dimensional spatial depth map and all adjacent valid pixels within a preset range to obtain a set of depth values.

[0110] The calculation unit is used to calculate the depth difference between a pixel and each of its adjacent valid pixels, and take the average of the absolute values ​​of the differences as the depth fluctuation value of the pixel.

[0111] The marking unit is used to mark pixels whose depth fluctuation value in the 3D spatial depth map exceeds a preset threshold as occluded pixels, thereby obtaining a marked 3D spatial depth map.

[0112] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0113] Accordingly, embodiments of the present invention provide a computer-readable storage medium, the computer-readable storage medium including a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform steps in a processing method for wall panels for three-dimensional steel structures in substations as described in the above embodiments, for example... Figure 1 Steps S1 to S8 as described above.

[0114] The present invention provides a method and system for processing wall panels for three-dimensional steel structures in substations, which has the following advantages:

[0115] For substations with specific topologies equipped with 3D steel structure wall panels, this invention starts with initial image data and simplifies the data through a combination of statistical filtering and voxel grid downsampling. A 3D spatial depth map is then constructed using spatial rasterization. Occlusion information is marked based on pixel depth differences, and coordinate system transformation and data completion are performed in conjunction with the substation topology. Finally, engineering attribute data is integrated, and a wall panel processing strategy is generated using a deep learning-based splitting strategy model. This invention employs a progressive data processing flow of noise reduction and simplification, depth analysis, occlusion marking, and data completion, forming a complete image data optimization system. This system can accurately correct defects in the initial data, providing comprehensive and reliable data support for processing strategy formulation. Furthermore, by deeply integrating actual scene information such as substation topology and engineering attributes with deep learning algorithms, the generated splitting and processing strategies are more closely aligned with the substation's assembly requirements, effectively reducing errors between wall panel processing and steel structure assembly, and improving construction efficiency and project quality.

[0116] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A method for processing wall panels for three-dimensional steel structures in substations, characterized in that, Applied to substations where the aforementioned three-dimensional steel structure wall panels are assembled within the topological structure, including: Acquire initial image data of the wall panels for the three-dimensional steel structure of the substation to be processed; The initial image data is sequentially processed by statistical filtering algorithm and voxel grid downsampling technique to reduce noise and simplify the data, thereby obtaining the first three-dimensional point cloud data; The first three-dimensional point cloud data is subjected to spatial rasterization processing to generate a three-dimensional spatial depth map corresponding to the wall panel of the three-dimensional steel structure of the substation. Each pixel in the three-dimensional spatial depth map is analyzed, and the depth difference information in the analysis results is used as the basis for occlusion determination of the initial image data to obtain a marked three-dimensional spatial depth map. The coordinate system transformation analysis is performed sequentially on the marker information in the marked three-dimensional spatial depth map and the structural information data confirmed by the substation topology to obtain the three-dimensional spatial data to be completed. The initial image data is pixel-completed based on the three-dimensional spatial data to be completed to obtain the second image data. The third image data, obtained by fusing the second image data and the acquired engineering attribute data of the substation, is input into a splitting strategy model constructed by a deep learning algorithm for analysis and processing to obtain the target splitting strategy. Based on the target decomposition strategy, a processing strategy is generated for the wall panels of the three-dimensional steel structure of the substation to be processed.

2. The processing method for wall panels used in three-dimensional steel structures of substations as described in claim 1, characterized in that, The initial image data is sequentially subjected to denoising and simplification processing using statistical filtering algorithms and voxel grid downsampling techniques to obtain the first three-dimensional point cloud data, including: The spatial coordinate information of the initial image data is parsed and extracted to obtain the original three-dimensional point cloud data; The statistical filtering algorithm is used to identify and remove noise points from the original 3D point cloud data to obtain denoised 3D point cloud data. Based on the denoised 3D point cloud data, the voxel grid downsampling technique is used to perform grid division to obtain rasterized point cloud data. The point within each grid that is closest to the grid center is identified as the sampling point; The sampling points of each grid are aggregated to obtain the first three-dimensional point cloud data.

3. The processing method for wall panels used in three-dimensional steel structures of substations as described in claim 1, characterized in that, The step of performing spatial rasterization processing on the first three-dimensional point cloud data to generate a three-dimensional spatial depth map corresponding to the three-dimensional steel structure wall panel of the substation includes: Based on the first three-dimensional point cloud data, the gridded area of ​​the wall panel for the three-dimensional steel structure of the substation is selected; The rasterized area is divided according to a preset resolution to obtain a spatial raster structure containing multiple raster units; Calculate the average depth value of the point cloud data in each grid cell within the spatial grid structure to obtain the depth information of the corresponding two-dimensional pixel. The three-dimensional spatial depth map is generated based on the depth information corresponding to the two-dimensional pixels of each of the grid units.

4. The processing method for wall panels used in three-dimensional steel structures of substations as described in claim 1, characterized in that, The step of analyzing each pixel in the three-dimensional spatial depth map and using the depth difference information in the analysis results as the basis for occlusion determination of the initial image data to obtain a labeled three-dimensional spatial depth map includes: Extract the depth value of each pixel in the three-dimensional spatial depth map and all adjacent valid pixels within a preset range to obtain a depth value set; Calculate the depth difference between the pixel and each of its adjacent valid pixels, and take the average of the absolute values ​​of the differences as the depth fluctuation value of the pixel. The pixels whose depth fluctuation values ​​in the three-dimensional spatial depth map exceed a preset threshold are marked as occluded pixels, thus obtaining the marked three-dimensional spatial depth map.

5. The processing method for wall panels used in three-dimensional steel structures of substations as described in claim 1, characterized in that, The coordinate system transformation analysis is performed sequentially on the marker information in the marked three-dimensional spatial depth map and the structural information data confirmed by the substation topology to obtain the three-dimensional spatial data to be completed, including: Transform the occlusion region corresponding to the marked information from the image coordinate system to the substation global coordinate system; Based on the structural information data, select the initial three-dimensional data of the unobstructed region that is consistent with the structural type of the obstructed region; By combining the three-dimensional features of the initial three-dimensional data of the unobstructed area with the spatial location of the obstructed area, the three-dimensional spatial data to be completed is generated.

6. The processing method for wall panels used in three-dimensional steel structures of substations as described in claim 1, characterized in that, The step of completing the initial image data based on the three-dimensional spatial data to be completed, to obtain the second image data, includes: Based on the three-dimensional spatial data to be completed, the reverse perspective transformation technique is used to project the three-dimensional spatial data to be completed from the global coordinate system back to the pixel coordinate system of the initial image data, and to determine the pixel position and texture mapping relationship of the region to be completed in the initial image data. By combining the texture features of adjacent unoccluded areas in the initial image data, pixel filling is performed on the area to be filled to obtain the second image data.

7. A processing system for wall panels used in three-dimensional steel structures of substations, characterized in that, Applied to substations where the aforementioned three-dimensional steel structure wall panels are assembled within the topological structure, including: The acquisition module is used to acquire the initial image data of the wall panels for the three-dimensional steel structure of the substation to be processed; The noise reduction module is used to perform noise reduction and simplification processing on the initial image data by sequentially applying statistical filtering algorithm and voxel grid downsampling technology to obtain the first three-dimensional point cloud data; The depth map module is used to perform spatial rasterization processing on the first three-dimensional point cloud data to generate a three-dimensional spatial depth map corresponding to the three-dimensional steel structure wall panel of the substation. The analysis module is used to analyze each pixel in the three-dimensional spatial depth map and use the depth difference information in the analysis results as the basis for occlusion determination of the initial image data to obtain a marked three-dimensional spatial depth map. The completion module is used to sequentially perform coordinate system transformation analysis on the marker information in the marked 3D spatial depth map and the structural information data confirmed by the substation topology to obtain the 3D spatial data to be completed. The completion module is used to perform pixel completion processing on the initial image data based on the three-dimensional spatial data to be completed, so as to obtain the second image data. The strategy module is used to input the third image data obtained by fusing the second image data and the acquired engineering attribute data of the substation into a splitting strategy model constructed by a deep learning algorithm for analysis and processing, so as to obtain the target splitting strategy. The generation module is used to generate a processing strategy for the wall panels of the three-dimensional steel structure of the substation to be processed based on the target splitting strategy.

8. The processing system for wall panels used in three-dimensional steel structures of substations as described in claim 7, characterized in that, The noise reduction module includes: The parsing unit is used to parse and extract the spatial coordinate information of the initial image data to obtain the original three-dimensional point cloud data. The noise reduction unit is used to identify and remove noise points from the original 3D point cloud data using the statistical filtering algorithm to obtain denoised 3D point cloud data. The partitioning unit is used to perform grid partitioning based on the denoised 3D point cloud data using the voxel grid downsampling technique to obtain rasterized point cloud data. A determining unit is used to determine the point within each grid that is closest to the grid center as a sampling point; A set unit is used to set the sampling points of each of the grids to obtain the first three-dimensional point cloud data.

9. The processing system for wall panels used in three-dimensional steel structures of substations as described in claim 7, characterized in that, The depth map module includes: The selected unit is used to select the gridded area of ​​the wall panel for the three-dimensional steel structure of the substation based on the first three-dimensional point cloud data. A grid structure unit is used to divide the gridded area according to a preset resolution to obtain a spatial grid structure containing multiple grid units. A depth information unit is used to calculate the average depth value of the point cloud data in each grid unit within the spatial grid structure, and to obtain the depth information of the corresponding two-dimensional pixel. A generation unit is configured to generate the three-dimensional spatial depth map based on the depth information corresponding to the two-dimensional pixel points of each of the grid units.

10. The processing system for wall panels used in three-dimensional steel structures of substations as described in claim 7, characterized in that, The analysis module includes: The extraction unit is used to extract the depth value of each pixel in the three-dimensional spatial depth map and all adjacent valid pixels within a preset range to obtain a set of depth values. The calculation unit is used to calculate the depth difference between the pixel and each of the adjacent effective pixels, and take the average of the absolute values ​​of the differences as the depth fluctuation value of the pixel. A marking unit is used to mark the pixels in the three-dimensional spatial depth map whose depth fluctuation value exceeds a preset threshold as occluded pixels, thereby obtaining the marked three-dimensional spatial depth map.

Images