A method, system, device and storage medium for processing and modeling point cloud data of a power distribution device

Abstract

The present invention relates to the technical field of point cloud data modeling, and discloses a method, system, device and storage medium for processing and modeling point cloud data of power distribution equipment, including: collecting point cloud data of power distribution equipment, and constructing a topological relationship graph according to the topological relationship of the power distribution equipment; performing semantic segmentation on the point cloud data based on the topological relationship graph, constructing a component association model, and combining the topological relationship graph to detect missing data; generating a high-precision three-dimensional model based on the component association model. The method of the present invention guides semantic segmentation through the topological relationship graph, improves the accuracy of point cloud data classification, and avoids misclassification problems in traditional methods. At the same time, a component association model is generated based on the segmentation result, comprehensively describing the geometric and functional connection relationships between devices, and realizing the deep integration of device connection information and three-dimensional geometric models.

Description

Technical Field

[0001] The present invention relates to the technical field of point cloud data modeling, and in particular to a method, system, device and storage medium for point cloud data processing and modeling of power distribution equipment. Background Art

[0002] In recent years, with the growing demand for digital transformation in the power industry, point cloud data technology has been gradually applied to the three-dimensional modeling and intelligent analysis of distribution equipment. Point cloud data has the characteristics of high precision and rich geometric information. It can accurately reflect the physical form and structural characteristics of complex distribution equipment and is the core data foundation of digital twin technology. Traditional point cloud acquisition technology mainly relies on equipment such as laser radar (LiDAR), three-dimensional scanners and high-resolution photography. Combined with computer vision and artificial intelligence algorithms, it has been widely used in power grid equipment inspection, condition monitoring and fault prediction. In addition, the introduction of deep learning technology has significantly improved the efficiency of point cloud data processing, especially in semantic segmentation and three-dimensional modeling. Methods such as PointNet, PointNet++ and KPConv have achieved efficient analysis of point cloud data. However, the complex geometric form and diversified structure of distribution equipment, as well as the occlusion, reflection and noise interference in the data acquisition process, make point cloud processing technology still have great challenges in accuracy and robustness.

[0003] In order to further improve the application effect of point cloud data in the field of power equipment, research in recent years has begun to explore the combination of topological structure information of the equipment with point cloud analysis algorithms to make up for the limitations of algorithms based solely on geometric features. At the same time, the rise of generative adversarial network (GAN) technology has provided a new idea for the completion of point cloud data. Through the adversarial learning mechanism between the generator and the discriminator, high-fidelity point clouds can be generated in areas with missing data. However, the integrated application of these technologies still faces many bottlenecks, such as how to build an effective topological model to support point cloud segmentation, how to balance the authenticity of the generated point cloud and the consistency of the equipment structure, and how to achieve efficient three-dimensional modeling in complex scenarios. Summary of the invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the technical problem solved by the present invention is that the existing point cloud data processing technology has many problems in semantic segmentation accuracy, missing data completion and three-dimensional modeling optimization.

[0006] To solve the above technical problems, the present invention provides the following technical solutions: a method for processing and modeling point cloud data of power distribution equipment, comprising: collecting point cloud data of power distribution equipment, and constructing a topological relationship diagram according to the topological relationship of the power distribution equipment; performing semantic segmentation on the point cloud data based on the topological relationship diagram, constructing a component association model, and performing missing data detection in combination with the topological relationship diagram; and generating a high-precision three-dimensional model based on the component association model.

[0007] As a preferred solution of the method for processing and modeling point cloud data of power distribution equipment described in the present invention, the point cloud data of the power distribution equipment includes geometric feature data, spatial relationship data, physical attribute data and functional attribute data.

[0008] As a preferred solution of the method for processing and modeling point cloud data of power distribution equipment described in the present invention, wherein: the construction of a topological relationship graph based on the topological relationship of the power distribution equipment includes extracting the topological relationship data of the power distribution equipment from the distribution network management system, formatting the obtained topological relationship data and converting it into a graph structure; the graph structure is represented as , It is a node in the graph structure, representing a component of a power distribution device. It is the edge in the graph structure, which represents the connection relationship between the components of the power distribution equipment. The geometric coordinates of the point cloud data are aligned with the node coordinates in the topology map through the spatial position, number and name of the equipment. The integrity of the topology map is verified based on the point cloud data, and the errors and missing data in the topology map are corrected to generate an optimized topology relationship map as the input of semantic segmentation.

[0009] As a preferred solution of the method for processing and modeling point cloud data of power distribution equipment described in the present invention, the semantic segmentation includes adding initial geometric characteristics to each point of the point cloud data, aligning the nodes in the topological relationship diagram with the equipment area in the point cloud by position, size and functional attributes, and preliminarily dividing the point cloud area according to the spatial distribution constraints of the equipment connection in the topological diagram; extracting the geometric features of the point cloud data through the point cloud processing model, and adding weights to the points related to the adjacent nodes according to the adjacent nodes in the topological relationship diagram, which is expressed as:

[0010] ;

[0011] in, Indicate point Feature vector after neighborhood enhancement; Indicate point The original feature vector of ; Indicate point The set of neighboring nodes; Indicate point Neighborhood nodes of Indicate point and adjacent points The weight of Indicates points; use graph neural network to extract multi-level features in the topological relationship graph and output high-dimensional feature vectors for each node; embed the high-dimensional feature vectors into the point cloud segmentation model to guide the semantic classification of the point cloud.

[0012] As a preferred solution of the point cloud data processing and modeling method for distribution equipment described in the present invention, the point cloud segmentation model includes: inputting the geometric features of the point cloud data and the high-dimensional feature vector of each node in the topological relationship diagram, and performing joint analysis through feature splicing; using a multi-layer perceptron to classify the fused features, outputting the semantic category of each point, and dividing the point cloud data into different semantic areas.

[0013] As a preferred solution of the method for processing and modeling point cloud data of power distribution equipment described in the present invention, the component association model is constructed as follows: each semantic category corresponds to a segmented area, including all point cloud points belonging to the category, and the spatial position, size and type of each segmented area are calculated; the functional characteristics of the segmented area are supplemented according to the node information of the topological relationship graph matching the segmented area; based on the segmentation results, the point cloud adjacency between different components is checked, and the connection direction between adjacent components is confirmed according to the edge information of the topological relationship graph; the component association model is generated and represented by a graph structure. , A graph structure representing the component association model, Represents a node in the component association model graph structure, Represents an edge in the graph structure of a component association model.

[0014] As a preferred solution of the point cloud data processing and modeling method of the power distribution equipment described in the present invention, the generation of a high-precision three-dimensional model based on the component association model includes generating a component three-dimensional geometry based on the segmented point cloud according to the component association model and the point cloud data, generating connection geometry according to the edge information of the association model, ensuring the accuracy of direction and position, aligning the geometric model, verifying the connection consistency, and binding the equipment and connection properties according to the component association model to generate a high-precision three-dimensional model.

[0015] A distribution equipment point cloud data processing and modeling system using any of the methods described in the present invention, wherein: a data acquisition module collects point cloud data of distribution equipment and performs preprocessing, extracts topological relationship data of distribution equipment from a distribution network management system, and converts it into a graph structure representation; a data processing module extracts geometric features from point cloud data, and semantically classifies point cloud points in combination with node and edge information in a topological relationship graph, and generates a component association model based on the semantic segmentation result; a model construction module generates high-precision three-dimensional geometric bodies based on the component association model, reconstructs a three-dimensional model of each device in combination with the point cloud data, and binds device attributes and functional attributes to the devices and connection geometries in the three-dimensional model according to the component association model.

[0016] A computer device comprises: a memory and a processor; the memory stores a computer program, comprising: the steps of implementing any one of the methods of the present invention when the processor executes the computer program.

[0017] A computer-readable storage medium stores a computer program, comprising: when the computer program is executed by a processor, the steps of implementing any one of the methods of the present invention are implemented.

[0018] Beneficial effects of the present invention: The method of the present invention guides semantic segmentation through a topological relationship diagram, improves the accuracy of point cloud data classification, and avoids the misclassification problem in traditional methods. At the same time, a component association model is generated based on the segmentation results to comprehensively describe the geometric and functional connection relationships between devices, thereby achieving a deep integration of device connection information and three-dimensional geometric models. The present invention can dynamically detect and complete missing data to ensure the integrity and accuracy of modeling. It is suitable for rapid digitization and visual analysis of complex power distribution scenarios, and provides effective support for the intelligent operation and maintenance of power distribution equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0020] Figure 1 An overall flow chart of a method for processing and modeling point cloud data of power distribution equipment provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0021] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0022] Example 1, reference Figure 1 , as an embodiment of the present invention, provides a method for processing and modeling point cloud data of a power distribution device, comprising:

[0023] S1: Collect point cloud data of power distribution equipment and construct a topological relationship diagram based on the topological relationship of the power distribution equipment.

[0024] Furthermore, point cloud data of the power distribution equipment is collected through LiDAR or 3D scanning equipment. The point cloud data contains geometric feature information of the equipment, such as point cloud density, surface curvature, boundary characteristics and spatial dimensions. At the same time, LiDAR equipment can record reflectivity data for analyzing the material characteristics of the equipment. In addition, by collecting different areas of the equipment in segments and combining unified coordinate alignment technology, the integrity and accuracy of the point cloud data are ensured.

[0025] Combine high-resolution photography equipment and infrared thermal imagers to collect texture and thermal characteristic data of the equipment. The photography equipment captures the RGB information of the equipment surface, which is used to merge with the point cloud data to improve the accuracy of equipment recognition; the infrared imager records the temperature distribution of the equipment surface in real time to identify possible overheating areas or hidden dangers. In addition, vibration sensors are used to collect micro-vibration data of the equipment during operation to provide support for dynamic modeling.

[0026] While collecting the above core data, the spatial layout information of the collection scene is analyzed to obtain the spatial relationship data between devices, such as adjacency, connection point coordinates and topological connection weights. Furthermore, the equipment operation status information (such as switch status, historical fault records) and environmental data (such as temperature, humidity, and light intensity) are collected to add rich dimensional information for subsequent model construction.

[0027] Furthermore, a topological diagram of the power distribution equipment is constructed to assist in point cloud segmentation. Through the connection relationship, physical proximity, geometric features and other information between devices, auxiliary constraints are provided for point cloud semantic segmentation to improve the accuracy and efficiency of segmentation.

[0028] By accessing the distribution equipment related information stored in the distribution network management system (such as SCADA system, GIS system or asset management database), the topological relationship data of the distribution equipment is extracted from the distribution network management system. The distribution network management system usually contains static information and operating status data of the equipment, where the static information includes the equipment number, type, spatial coordinates, installation location, physical size, etc., while the operating status data records the connection relationship between the equipment, the real-time operation status (such as the opening and closing position of the circuit breaker) and the current flow direction.

[0029] The acquired topological relationship data is formatted and converted into a graph structure, which is represented as , It is a node in the graph structure, representing a component of a power distribution device. It is the edge in the graph structure, which represents the connection relationship between the components of the power distribution equipment. For each node in the topological data, the geometric and physical attribute information (such as equipment size and location coordinates) is supplemented to correspond to the point cloud data.

[0030] The geometric coordinates of the point cloud data are aligned with the node coordinates in the topology map through the spatial position, number and name of the device, ensuring that the topology map reflects the distribution of the device in the actual space.

[0031] The integrity of the topological map is verified based on point cloud data, errors and missing data in the topological map are corrected, and an optimized topological relationship map is generated as the input for semantic segmentation.

[0032] It should be noted that the topological relationship contained in the distribution network management system may be different from the topological relationship of the current device. Therefore, it is necessary to use the collected point cloud data for comparison and correction. For example, if a node has no corresponding device in the point cloud, it is marked as a potential error node; if there is a connection relationship in the point cloud but the topology map is not displayed, then add the connection.

[0033] S2: Perform semantic segmentation on point cloud data based on the topological relationship graph, build a component association model, and perform missing data detection in combination with the topological relationship graph.

[0034] Furthermore, by combining the topological relationship diagram of the distribution equipment and deep learning technology, semantic segmentation of point cloud data can be achieved, key components (such as transformers, switches, terminals, etc.) can be accurately identified, and association models between components can be automatically generated to improve the accuracy and efficiency of subsequent modeling.

[0035] The collected point cloud data is subjected to noise reduction, normalization and block processing to ensure the quality and computational efficiency of the point cloud data. Then, initial geometric features are added to each point of the point cloud data, and the nodes in the topological relationship diagram are aligned with the device areas in the point cloud by position, size and functional attributes. The point cloud area is preliminarily divided according to the spatial distribution constraints of the device connections in the topological diagram.

[0036] Specifically, the preliminary division of the point cloud area includes using the spatial position attributes of the nodes in the topological relationship diagram (such as transformers, switches, etc.) to find candidate device areas in the point cloud data, and further screening the candidate areas according to the physical size of the equipment (such as the height and width of the transformer).

[0037] The functional attributes of the nodes in the topological relationship graph (such as the number of cable ports of the transformer and the relative connection position of the switch) and the regional features of the point cloud (such as the density of connection points and geometric shape) are used for refined matching. For example, if a node in the topological graph represents a circuit breaker and the edge connected to it indicates that it is adjacent to the transformer, the area adjacent to the transformer point cloud is preferentially selected as the matching area of ​​the circuit breaker.

[0038] The successfully matched point cloud areas are marked as corresponding equipment categories (such as transformers and switches), and their node attributes (position, size, function) are recorded. The point cloud is further partitioned according to the connection relationship between nodes in the topology diagram (such as the adjacency and connection direction of the equipment), and the boundaries of the partitioned areas are adjusted to be consistent with the topological distribution constraints.

[0039] The geometric features of point cloud data are extracted through point cloud processing models (such as PointNet++ or KPConv). Local features (such as surface curvature and density) are extracted through the geometric relationship of local neighborhood points. The global context information of the point cloud is aggregated by the model to extract global features.

[0040] According to the adjacent nodes in the topological relationship graph, the weights of the points related to the adjacent nodes are increased, which can be expressed as:

[0041] ;

[0042] in, Indicate point Feature vector after neighborhood enhancement; Indicate point The original feature vector of ; Indicate point The set of neighboring nodes; Indicate point Neighborhood nodes of Indicate point and adjacent points The weight of Indicates Use graph neural networks to extract multi-level features in the topological graph, output high-dimensional feature vectors for each node, embed the high-dimensional feature vectors into the point cloud segmentation model, and guide the semantic classification of the point cloud.

[0043] It should be noted that direct semantic segmentation in point cloud data may lead to misclassification due to occlusion between devices, acquisition noise or complex scenes. By aligning the topological relationship graph, the possible position, size and adjacency of the device are clarified, providing context constraints for subsequent point cloud segmentation.

[0044] During the alignment process, the point cloud data is divided into possible device areas. The segmentation task no longer needs to process the entire point cloud data, but only performs deep learning processing on the aligned divided areas, which greatly reduces the computational complexity.

[0045] Furthermore, the geometric features of the point cloud data and the high-dimensional feature vector of each node in the topological relationship graph are input, and the point cloud and topological features are combined through feature splicing. For each point cloud point, the point cloud features and topological features are directly spliced ​​to form a unified feature vector.

[0046] A multi-layer perceptron is used to classify the fused features, and each point cloud point is individually passed through a series of MLP layers to extract high-order features. The semantic category probability distribution of each point is output (such as the probability of belonging to a transformer, circuit breaker, or terminal), and the point cloud data is divided into different semantic areas.

[0047] Furthermore, based on the semantic segmentation results and the topological relationship diagram, a component association model is constructed. Each semantic category corresponds to a segmentation area. , contains all point cloud points belonging to this category, and calculates the spatial position, size and type for each segmented area.

[0048] Specifically, calculate the center coordinates of the area:

[0049] ;

[0050] in, Indicates segmentation area The center coordinates of Indicate point Calculate the size (length, width, height) of the bounding box of the region and use the semantic category label of the segmentation result as the type of the region.

[0051] According to the node information of the topological relationship graph matching the segmented area, the functional characteristics of the segmented area are supplemented.

[0052] Based on the segmentation results, check the point cloud adjacency between different components. For each pair of segmented areas, calculate the minimum distance between boundary points. If the minimum distance is less than the preset threshold, the two are considered adjacent, and output the adjacent component pairs and their boundary distances.

[0053] According to the edge information of the topological relationship graph, the connection direction between adjacent components (such as current direction or signal flow direction) is confirmed. For the connection relationship detected in the point cloud, the direction attribute is supplemented. For example, if the segmented terminal is adjacent to the circuit breaker and the topological graph shows that the terminal is the input end of the circuit breaker, the direction attribute is added.

[0054] Generate component association model based on the extracted association information and represent it with graph structure , A graph structure representing the component association model, Represents a node in the component association model graph structure, Represents the edge in the component association model graph structure. Each segmentation area Corresponding to a node , each node Additional equipment category (such as transformer, circuit breaker), equipment center coordinates, equipment bounding box size, and equipment functional information (such as electrical properties, number of connection ports) inherited from the topology diagram.

[0055] Each pair of adjacent partitioned regions generates an edge , indicating the connection relationship between the two components. Each edge Additional connection type, directionality, minimum boundary distance between components, and actual number of contact point cloud points or area properties between segmented regions.

[0056] It should be noted that the component association model is a graph structure, which is a more real-time and accurate topological relationship graph generated based on real-time point cloud data and semantic segmentation results. It not only reflects the actual status and connection relationship of the equipment, but also can dynamically adapt to changes in real-time scenes.

[0057] The node and edge information of the component association model comes from the result of semantic segmentation of the point cloud, which directly reflects the size, position and connection relationship of the equipment in the actual scene. Compared with the topological relationship diagram, the node size and position of the component association model come from the segmented device point cloud area, with higher accuracy. The edge of the component association model not only describes the theoretical connection relationship, but also contains information such as the actual contact area and connection point coordinates.

[0058] Using the component association model for 3D modeling, on the basis of the real-time point cloud modeling, the connection relationship between distribution equipment can be directly established. Compared with the existing 3D model generation that relies only on point cloud data, there is no topological relationship guidance. The device areas in the segmentation results are independent of each other, and it is difficult to capture the functional association or connection relationship between devices. The connection characteristics of the equipment (such as the physical connection between the transformer and the circuit breaker) need additional inference, which increases the complexity and may have errors.

[0059] S3: Generate a high-precision 3D model based on the component association model.

[0060] Furthermore, the component-related model Based on the data, the geometric information, connection relationship and functional attributes of the equipment are combined to build a complete three-dimensional digital model. The specific steps include:

[0061] For each component node , based on its segmented point cloud area and geometric properties, generate an accurate 3D geometric model. For the point cloud in the segmented area, calculate its boundary point set to form a preliminary outline, and use the convex hull algorithm (ConvexHull) or the minimum bounding box to generate the initial bounding box. Use the point cloud surface reconstruction algorithm (such as Poisson reconstruction, Delaunay triangulation) to generate the 3D mesh model of the component. For regular-shaped components (such as rectangular transformers), fit the geometric body (such as cuboid, cylinder) according to the boundary size.

[0062] According to the edge set The connection relationship information in the , builds the connection geometry between the components. The edge attributes in the extract the connection point coordinates (such as the start and end points of the cable connection), and according to the direction attribute of the edge, determine the direction of the connection (such as the direction vector from the output of the transformer to the input of the circuit breaker).

[0063] Generate connection geometry. For cable connections, generate cylinders or curves based on connection points. For rigid connections (such as bolt connections between terminals and devices), generate corresponding fixture geometry. For non-rigid connections (such as flexible cables), generate naturally curved connection geometry based on curve fitting.

[0064] Use the ICP (Iterative Closest Point) algorithm to globally align the component model and the connection model to ensure accurate connections between components. Verify whether the generated 3D model complies with the constraints of the topology diagram: ensure that the connection geometry between components is consistent with the edge relationship in the topology diagram; check whether the connection points match the actual contact points in the segmented point cloud. Remove redundant point clouds or redundant connection geometry in the 3D model to optimize model storage.

[0065] Add functional attributes to each component of the 3D model, including attributes such as component category, equipment number, operating status, and cable connection type, connection direction, current capacity, etc. Export the 3D model and functional attributes in a standard format (such as IFC, glTF) for subsequent visualization or integration into device management systems.

[0066] The present embodiment also provides a distribution equipment point cloud data processing and modeling system, including: a data acquisition module, which collects and preprocesses the point cloud data of the distribution equipment, extracts the topological relationship data of the distribution equipment from the distribution network management system, and converts it into a graph structure representation; a data processing module, which extracts geometric features of the point cloud data, and semantically classifies the point cloud points in combination with the node and edge information in the topological relationship graph, and generates a component association model based on the semantic segmentation results; a model construction module, which generates a high-precision three-dimensional geometric body based on the component association model, reconstructs the three-dimensional model of each device in combination with the point cloud data, and binds device attributes and functional attributes to the devices and connection geometries in the three-dimensional model according to the component association model.

[0067] Example 2: The following is an embodiment of the present invention, which provides a method for processing and modeling point cloud data of power distribution equipment. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0068] In order to verify the effectiveness of the method of constructing a topological relationship diagram and generating a high-precision three-dimensional model based on the point cloud data of distribution equipment, a substation was selected as the test scene. The substation contains multiple typical distribution equipment, including transformers, circuit breakers, terminals, etc., a total of 18 equipment units, with relatively complex spatial layout and connection relationships.

[0069] Use high-precision LiDAR equipment to scan the experimental scene and obtain point cloud data. The point cloud resolution is 1000 points per square meter. Equipped with a high-resolution camera to capture the surface texture information (RGB data) of the device, and an infrared thermal imager to collect the temperature distribution data of the device surface. When collecting point clouds, scan the regions, and integrate the point cloud data of different regions through unified coordinate alignment technology to ensure integrity.

[0070] The topological relationship data extracted from the distribution network management system includes the connection relationship table between devices (device ID pairs and connection types) and the node attribute table (device number, spatial position, size). The topological relationship data is formatted into a graph structure, with nodes representing devices and edges representing connection relationships, and potential errors in the topological graph are corrected in combination with point cloud data. Three-dimensional modeling is performed using the method of the present invention, and some device data is shown in Table 1.

[0071] Table 1 Experimental data table

[0072] ;

[0073] Through the analysis of experimental data, it can be clearly seen that the method of the present invention has significant advantages in point cloud data processing and modeling: the experimental table records the spatial position and boundary size of each device, and the segmentation result is highly matched with the geometric properties of the actual device, with an error of less than 1%. For example, the actual size of transformer T001 is 2.5×1.8×2.0 meters, which is completely consistent with the point cloud segmentation result.

[0074] The component association model effectively captures the connection relationship between devices and adds geometric characteristics such as the minimum boundary distance. It should be noted that the method of the present invention can extract the connection relationship and device attributes between devices when the modeling is completed. Modeling based only on point cloud data requires additional analysis and addition of connection relationships and device attributes.

[0075] The prior art only relies on point cloud data when generating a 3D model, without the guidance of topological relationships. In the experiment, the connection relationship of switch P001 may be difficult to detect automatically due to the point cloud segmentation error, while the present invention accurately identifies its connection with circuit breaker B001 by combining the topological relationship diagram. The prior art has the risk of misclassification or connection omission in complex scenes, while the present invention completely avoids such problems through dynamic correction of the topological relationship diagram.

[0076] If the above functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program code.

[0077] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0078] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0079] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0080] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for processing and modeling point cloud data of power distribution equipment, characterized in that: include: Collect point cloud data of power distribution equipment and construct a topological relationship diagram based on the topological relationship of the power distribution equipment; Perform semantic segmentation on point cloud data based on topological relationship graph, build component association model, and perform missing data detection in combination with topological relationship graph; Generate high-precision 3D models based on component association models; The constructing of a topological relationship graph according to the topological relationship of the power distribution equipment includes extracting the topological relationship data of the power distribution equipment from the power distribution network management system, formatting the obtained topological relationship data and converting it into a graph structure; The graph structure is represented by G=(V, E), where V is a node in the graph structure, representing a power distribution equipment component, and E is an edge in the graph structure, representing a connection relationship between the power distribution equipment components; Align the geometric coordinates of the point cloud data with the node coordinates in the topology map through the spatial position, number and name of the device; Verify the integrity of the topology map based on point cloud data, correct errors and missing data in the topology map, and generate an optimized topology relationship map as input for semantic segmentation; The semantic segmentation includes adding initial geometric characteristics to each point of the point cloud data, aligning the nodes in the topological relationship diagram with the device areas in the point cloud by position, size and functional attributes, and preliminarily dividing the point cloud area according to the spatial distribution constraints of the device connections in the topological diagram; The geometric features of point cloud data are extracted through the point cloud processing model. According to the adjacent nodes in the topological relationship graph, the weights of the points related to the adjacent nodes are increased, which can be expressed as: Among them, f'(p i ) represents point p i The feature vector after neighborhood enhancement; f(p i ) represents point p i The original eigenvector of i The neighborhood node set of point p i Neighborhood nodes of ij Represents point p i and adjacent point p j The weight of ; i represents the i-th point; Use graph neural networks to extract multi-level features in the topological graph and output high-dimensional feature vectors for each node; Embed high-dimensional feature vectors into point cloud segmentation models to guide semantic classification of point clouds.

2. The method for processing and modeling point cloud data of power distribution equipment according to claim 1, characterized in that: The point cloud data of the power distribution equipment includes geometric feature data, spatial relationship data, physical attribute data and functional attribute data.

3. The method for processing and modeling point cloud data of power distribution equipment according to claim 2, characterized in that: The point cloud segmentation model includes inputting geometric features of point cloud data and high-dimensional feature vectors of each node in the topological relationship graph, and performing joint analysis through feature splicing; A multi-layer perceptron is used to classify the fused features, output the semantic category of each point, and divide the point cloud data into different semantic areas.

4. The method for processing and modeling point cloud data of power distribution equipment according to claim 3, characterized in that: The component association model is constructed by corresponding each semantic category to a segmented region, including all point cloud points belonging to the category, and calculating the spatial position, size and type of each segmented region; Supplementing functional characteristics of the segmented area according to node information of the topological relationship graph matching the segmented area; Based on the segmentation results, check the point cloud adjacency between different components, and confirm the connection direction between adjacent components according to the edge information of the topological relationship graph; Generate a component association model and represent it with a graph structure G′=(V′,E′), where G′ represents the graph structure of the component association model, V′ represents a node in the graph structure of the component association model, and E′ represents an edge in the graph structure of the component association model.

5. The method for processing and modeling point cloud data of power distribution equipment according to claim 4, characterized in that: The method of generating a high-precision three-dimensional model based on a component association model includes generating a component three-dimensional geometry based on segmented point clouds according to the component association model and point cloud data, generating connection geometry according to edge information of the association model, ensuring accurate direction and position, aligning geometric models, verifying connection consistency, and binding equipment and connection properties according to the component association model to generate a high-precision three-dimensional model.

6. A distribution equipment point cloud data processing and modeling system using any of the methods of claims 1 to 5, characterized in that: include, The data acquisition module collects and preprocesses the point cloud data of the distribution equipment, extracts the topological relationship data of the distribution equipment from the distribution network management system, and converts it into a graph structure representation; The data processing module extracts geometric features from point cloud data, and combines the node and edge information in the topological relationship graph to semantically classify the point cloud points, and generates a component association model based on the semantic segmentation results; The model building module generates high-precision 3D geometry based on the component association model, reconstructs the 3D model of each device in combination with the point cloud data, and binds device attributes and functional attributes to the devices and connection geometries in the 3D model according to the component association model.

7. A computer device comprising: Memory and processor; The memory stores a computer program, characterized in that when the processor executes the computer program, the steps of the distribution equipment point cloud data processing and modeling method as described in any one of claims 1-5 are implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for processing and modeling point cloud data of power distribution equipment as described in any one of claims 1 to 5 are implemented.

Images