Feature extraction method and device of power distribution network, processor and storage medium
By identifying and extracting the key points and structural features of point clouds of distribution network insulators, the problem of low extraction accuracy in the prior art is solved, efficient processing of complex shapes and surface uneven point cloud data is achieved, and insulator feature extraction is improved accuracy.
Patent Information
- Application Number
- CN202411154003.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-21
- Publication Date
- 2024-12-03
AI Technical Summary
When existing feature extraction algorithms process point cloud data with complex shapes and uneven surfaces, it is difficult to accurately extract the features of distribution network insulators, resulting in low extraction accuracy.
By obtaining point cloud data and background data of insulators in the distribution network, determining the target point cloud data, identifying point cloud key points, and extracting structural features, including boundary values and spatial features, based on these key points, to improve the uniformity of point cloud data. sex.
The feature extraction accuracy of distribution network insulators is improved, the effective processing of complex shapes and surface uneven point cloud data is ensured, and the accuracy of subsequent analysis is enhanced.
Smart Images

Figure CN119068475A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of feature extraction technology, and in particular to a method, device, processor and storage medium for extracting features of a distribution network. Background Art
[0002] At present, with the expansion of the scale and complexity of the distribution network, the current transformer in the substation is an important device for measuring and monitoring current. Among these devices, insulators play a vital role. The current transformer insulator in the substation is an important part to ensure the safety and normal operation of the equipment. Insulators can effectively isolate the high-voltage part of the current transformer from the ground or low-voltage part to prevent accidental leakage of current, thereby protecting the safety of equipment and personnel. Insulators not only have insulation functions, but also bear the load of supporting the current transformer to ensure the stable operation of the equipment. Therefore, feature extraction of insulators through three-dimensional point clouds can improve the monitoring and maintenance efficiency of current transformers in many aspects.
[0003] In related technologies, due to the complexity and variability of the outdoor environment (such as changes in lighting, rain, dust, temperature differences, spatial distortion caused by flashover discharges, etc.), as well as the viewing angle limitations of the equipment, the acquired point cloud data may differ in density and uniformity. When processing point cloud data with complex shapes, uneven surfaces, or noise, existing feature extraction algorithms may extract inaccurate or invalid features, affecting the results of subsequent analysis. Therefore, there is a technical problem of low accuracy in feature extraction of distribution network insulators.
[0004] With respect to the above-mentioned technical problem of low accuracy in feature extraction of insulators in distribution networks, no effective solution has been proposed so far. Summary of the invention
[0005] Embodiments of the present invention provide a method, device, processor and storage medium for extracting features of a distribution network, so as to at least solve the technical problem of low accuracy in extracting features of insulators in the distribution network.
[0006] According to one aspect of an embodiment of the present invention, a method for extracting features of a distribution network is provided. The method may include: acquiring point cloud data of an insulator in the distribution network and background data of the insulator; determining target point cloud data of the insulator based on the point cloud data and the background data, wherein the target point cloud data is used to indicate the geometric profile and position information of the insulator; determining at least one point cloud key point of the target point cloud data based on the target point cloud data, wherein the point cloud key point is located at the edge of the insulator; extracting structural features of the insulator in the distribution network based on the point cloud key point, wherein the structural features are used to indicate the structure of the insulator.
[0007] Optionally, based on the point cloud data and the background data, determining the target point cloud data of the insulator includes: preprocessing the point cloud data and the background data to obtain the target point cloud data.
[0008] Optionally, based on the target point cloud data, determining at least one point cloud key point of the target point cloud data includes: traversing multiple target point cloud data, determining at least one point cloud boundary in the multiple target point cloud data, wherein the point cloud boundary is used to indicate a dividing line or transition area between each target point cloud data; based on the point cloud boundary, determining the point cloud key point of the corresponding target point cloud data.
[0009] Optionally, based on the point cloud boundary, the point cloud key points of the corresponding target point cloud data are determined, including: based on each point cloud boundary, determining the boundary value corresponding to the target point cloud data, wherein the boundary value is used to indicate the target point cloud data with the largest curvature change among the target point cloud data; based on the boundary value, determining the point cloud key points of the target point cloud data.
[0010] Optionally, determining the point cloud key points of the target point cloud data based on the boundary value includes: smoothing and filtering the boundary value; and determining multiple point cloud key points of the target point cloud data based on the smoothed boundary value.
[0011] Optionally, based on the point cloud key points, the structural features of the insulators in the distribution network are extracted, including: determining the spatial features between each point cloud key point and the corresponding neighborhood points, wherein the spatial features are used to indicate the change of the normal relationship between the point cloud key point and the domain point, and the domain point is used to indicate the target point cloud data of the point cloud key point within a predetermined range; based on the spatial features, the structural features of the insulators in the distribution network are determined.
[0012] According to another aspect of an embodiment of the present invention, a feature extraction device for a distribution network is also provided. The device may include: an acquisition unit, used to acquire point cloud data of insulators in the distribution network and background data of the insulators; a first determination unit, used to determine target point cloud data of the insulator based on the point cloud data and the background data, wherein the target point cloud data is used to indicate the geometric contour and position information of the insulator; a second determination unit, used to determine at least one point cloud key point of the target point cloud data based on the target point cloud data, wherein the point cloud key point is located at the edge of the insulator; an extraction unit, used to extract structural features of the insulators in the distribution network based on the point cloud key points, wherein the structural features are used to indicate the structure of the insulator.
[0013] According to another aspect of an embodiment of the present invention, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored program, wherein when the program is executed by a processor, the device where the storage medium is located is controlled to execute the feature extraction method of the distribution network in the embodiment of the present invention.
[0014] According to another aspect of an embodiment of the present invention, a processor is further provided. The processor is used to run a program, wherein the method for extracting characteristics of a distribution network in an embodiment of the present invention is executed when the program is run.
[0015] According to another aspect of the embodiments of the present invention, a computer program product is provided, which includes computer instructions, and when the computer instructions are executed by a processor, the method for extracting characteristics of a power distribution network in the embodiments of the present invention is implemented.
[0016] In an embodiment of the present invention, point cloud data of an insulator in a distribution network and background data of the insulator are obtained; based on the point cloud data and the background data, target point cloud data of the insulator is determined, wherein the target point cloud data is used to indicate the geometric outline and position information of the insulator; based on the target point cloud data, at least one point cloud key point of the target point cloud data is determined, wherein the point cloud key point is located at the edge of the insulator; based on the point cloud key point, the structural features of the insulator in the distribution network are extracted, wherein the structural features are used to indicate the structure of the insulator. That is to say, the present invention extracts the structural features of the insulator from the target point cloud data of the insulator in the distribution network according to the point cloud key point, and achieves the purpose of improving the uniformity of the point cloud data by performing multi-faceted processing on the point cloud data of the insulator, thereby solving the technical problem of low feature extraction accuracy of the insulator in the distribution network and achieving the technical effect of improving the feature extraction accuracy of the insulator in the distribution network. BRIEF DESCRIPTION OF THE DRAWINGS
[0017] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:
[0018] Figure 1 is a flow chart of a method for extracting features of a distribution network according to an embodiment of the present invention;
[0019] Figure 2 is a flow chart of a method for extracting three-dimensional point cloud features of a current transformer insulator based on a substation operation and maintenance robot according to an embodiment of the present invention;
[0020] Figure 3 is a schematic diagram of a filtering principle according to an embodiment of the present invention;
[0021] Figure 4 is a schematic diagram of comparing point cloud data before and after downsampling according to an embodiment of the present invention;
[0022] Figure 5 is a schematic diagram of an influence area of a point histogram calculation of a query point according to an embodiment of the present invention;
[0023] Figure 6is a schematic diagram of a local coordinate system defined according to an embodiment of the present invention;
[0024] Figure 7 According to an embodiment of the present invention, a P q A schematic diagram of the influence range of the k-neighborhood centered on ;
[0025] FIG8 is a schematic diagram of a statistical subinterval according to an embodiment of the present invention;
[0026] Fig. 9 is a flow chart of a method for detecting key points of a current transformer insulator according to an embodiment of the present invention;
[0027] Fig.10 is a schematic diagram of a feature extraction device for a distribution network according to an embodiment of the present invention. DETAILED DESCRIPTION
[0028] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.
[0029] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, functional component or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, functional components or devices.
[0030] According to an embodiment of the present invention, an embodiment of a feature extraction method for a distribution network is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.
[0031] Figure 1 is a flow chart of a method for extracting features of a distribution network according to an embodiment of the present invention. Figure 1 As shown, the method may include the following steps:
[0032] Step S101, obtaining point cloud data of insulators in a distribution network and background data of insulators.
[0033] In the technical solution provided in the above step S101 of the present invention, the point cloud data of the insulator and the background data of the insulator can be collectively referred to as the original point cloud data or three-dimensional point cloud data of the insulator. The insulator can also be referred to as a current transformer insulator.
[0034] In this embodiment, point cloud data of insulators in the distribution network and background data of insulators are obtained. For example, according to the location distribution of current transformers in the substation, by designing an automatic patrol route for current transformer insulators for the substation operation and maintenance robot, a large amount of three-dimensional point cloud data of current transformer insulators at different stages can be regularly obtained. This is only an illustrative example, and the specific method for obtaining three-dimensional point cloud data of insulators is not limited.
[0035] Step S102: determining target point cloud data of the insulator based on the point cloud data and the background data.
[0036] In the technical solution provided in the above step S102 of the present invention, the target point cloud data is used to indicate the geometric profile and position information of the insulator.
[0037] In this embodiment, after the point cloud data of the insulators in the distribution network and the background data of the insulators are acquired in step S101, the target point cloud data of the insulators are determined according to the point cloud data and the background data.
[0038] Optionally, the point cloud data and background data are preprocessed to obtain target point cloud data.
[0039] For example, the point cloud data and background data are pre-processed by denoising, segmentation, and downsampling to obtain target point cloud data.
[0040] For another example, the density threshold of the point cloud data is calculated, and denoising is performed according to the density threshold; the point cloud data and background data are segmented using a radius filter, and the point cloud data is downsampled using a voxel filter.
[0041] Step S103: determining at least one point cloud key point of the target point cloud data based on the target point cloud data.
[0042] In the technical solution provided in the above step S103 of the present invention, the key point of the point cloud is located at the edge of the insulator.
[0043] In this embodiment, after the target point cloud data of the insulator is determined in step S102, at least one point cloud key point of the target point cloud data is determined according to the target point cloud data.
[0044] Optionally, multiple target point cloud data are traversed to determine at least one point cloud boundary among the multiple target point cloud data, wherein the point cloud boundary is used to indicate a dividing line or transition area between each target point cloud data, and the point cloud boundary can also be called a boundary interest value (denoted as I(p)).
[0045] For example, the insulators of the current transformer in the substation mostly have a uniform and regular surface shape. The location with large curvature at a certain point in the point cloud data corresponds to the structural boundary point of different parts. The boundary interest value I(p) can be calculated by the following formula (1):
[0046] I(p)=I1(p)·I2(p) (1)
[0047] Among them, I1(p) can be expressed as the surface change intensity estimation factor, and I2(p) can be expressed as the main direction discrimination factor. p can be expressed as a boundary point, and the calculation method of I1(p) is shown in the following formula (2), and the calculation method of I2 is shown in the following formula (3):
[0048]
[0049]
[0050]
[0051]
[0052] Among them, n is the neighbor element of point p, σ is the spatial scale, and w n is the weight value corresponding to the neighboring element n, and γ is the one-dimensional angle corresponding to point n. For the edge point of the current transformer insulator surface, w n The value is 1, and for other points, w n The value is 1-(1-ρ) 3 , ρ is the curvature value obtained after PCA principal component analysis of p and its surrounding points.
[0053] Optionally, after determining the point cloud boundary points, the boundary values corresponding to the target point cloud data are determined based on the point cloud boundary points, wherein the boundary values are used to indicate the target point cloud data with the largest curvature change among the target point cloud data; and then the point cloud key points of the target point cloud data are determined based on the boundary values.
[0054] Step S104: extracting structural features of insulators in the distribution network based on key points of the point cloud.
[0055] In the technical solution provided in the above step S104 of the present invention, the structural feature is used to indicate the structure of the insulator.
[0056] In this embodiment, after at least one point cloud key point of the target point cloud data is determined in step S103, the structural features of the insulators in the distribution network are extracted according to the point cloud key points.
[0057] Optionally, the spatial features between each point cloud key point and the corresponding neighborhood point are determined, wherein the spatial features are used to indicate the change in the normal relationship between the point cloud key point and the domain point, and the domain point is used to indicate the target point cloud data of the point cloud key point within a predetermined range; thereby, the structural features of the insulators in the distribution network are determined based on the spatial features.
[0058] For example, after obtaining the point cloud key points of the current transformer insulator, the fast point feature histogram (FPFH) calculation method is used to determine the structural characteristics of the insulator through the spatial characteristics between each point cloud key point and the neighboring points.
[0059] It should be noted that the above embodiments can be executed by a feature extraction device for a distribution network.
[0060] In the above steps S101 to S104 of the present invention, the point cloud data of the insulator in the distribution network and the background data of the insulator are obtained; based on the point cloud data and the background data, the target point cloud data of the insulator is determined, wherein the target point cloud data is used to indicate the geometric outline and position information of the insulator; based on the target point cloud data, at least one point cloud key point of the target point cloud data is determined, wherein the point cloud key point is located at the edge of the insulator; based on the point cloud key point, the structural features of the insulator in the distribution network are extracted, wherein the structural features are used to indicate the structure of the insulator. That is to say, the present invention extracts the structural features of the insulator from the target point cloud data of the insulator in the distribution network according to the point cloud key point, and achieves the purpose of improving the uniformity of the point cloud data by performing multi-faceted processing on the point cloud data of the insulator, thereby solving the technical problem of low feature extraction accuracy of the insulator in the distribution network and achieving the technical effect of improving the feature extraction accuracy of the insulator in the distribution network.
[0061] The above method of this embodiment is further introduced below.
[0062] As an optional implementation manner, based on the point cloud data and the background data, the target point cloud data of the insulator is determined, including: preprocessing the point cloud data and the background data to obtain the target point cloud data.
[0063] In this embodiment, the point cloud data and the background data are preprocessed to obtain the target point cloud data, wherein the preprocessing may at least include denoising, segmentation and downsampling.
[0064] For example, the conditional removal filter is used to define the density threshold of the current transformer insulator point cloud. When the point cloud density in the original point cloud is less than the threshold, it is a noise point, and the filter is used to remove the noise point. The radius filter is used to set a radius and point cloud quantity limit with the geometric center of the current transformer insulator as the center of the sphere, and the number of point clouds falling inside the sphere drawn with the given radius is calculated. When the calculated point cloud number is greater than the given limit, the point is retained, and if it is less than the limit, it is removed. The voxel filter is used to create a three-dimensional voxel grid for the original point cloud. In each voxel grid, the centroid of all points in the voxel is used to approximate other points in the voxel. All points in the voxel are finally represented by a centroid point. After processing all voxels, the filtered point cloud is obtained.
[0065] As an optional implementation method, based on the target point cloud data, at least one point cloud key point of the target point cloud data is determined, including: traversing multiple target point cloud data, determining at least one point cloud boundary in the multiple target point cloud data, wherein the point cloud boundary is used to indicate a dividing line or transition area between each target point cloud data; based on the point cloud boundary, determining the point cloud key point of the corresponding target point cloud data.
[0066] In this embodiment, multiple target point cloud data are traversed to determine at least one point cloud boundary among the multiple target point cloud data, thereby determining the point cloud key points of the corresponding target point cloud data according to the point cloud boundary.
[0067] Optionally, the point cloud boundary can be calculated using the above formulas (1) to (5), which will not be described in detail here.
[0068] Optionally, by calculating the point cloud boundaries, it can help quickly identify and distinguish the structures of different parts, so that the data can be classified and processed more efficiently.
[0069] As an optional implementation method, based on the point cloud boundary, the point cloud key points of the corresponding target point cloud data are determined, including: based on each point cloud boundary, determining the boundary value corresponding to the target point cloud data, wherein the boundary value is used to indicate the target point cloud data with the largest curvature change among the target point cloud data; based on the boundary value, determining the point cloud key points of the target point cloud data.
[0070] In this embodiment, the boundary value corresponding to the target point cloud data is determined according to each point cloud boundary, and thus the point cloud key points of the target point cloud data are determined according to the boundary value.
[0071] Optionally, by determining the boundary value, the scope of the point cloud data can be narrowed, and the amount of data to be processed can be reduced, thereby improving processing efficiency.
[0072] As an optional implementation manner, based on the boundary value, determining the point cloud key points of the target point cloud data includes: smoothing and filtering the boundary value, and determining multiple point cloud key points of the target point cloud data based on the smoothed boundary value.
[0073] In this embodiment, the boundary value is smoothed and filtered, and multiple point cloud key points of the target point cloud data are determined according to the smoothed boundary value. For example, the boundary value of the point cloud data is smoothed and filtered using a sliding average or Gaussian filter method. This is only an illustrative example and does not limit the specific method of smoothing and filtering the boundary value.
[0074] Optionally, smoothing filtering can make the boundary values more uniform and continuous, thus improving the overall quality of the image.
[0075] As an optional implementation method, based on point cloud key points, the structural features of insulators in the distribution network are extracted, including: determining the spatial features between each point cloud key point and the corresponding neighborhood points, wherein the spatial features are used to indicate the change of the normal relationship between the point cloud key point and the domain point, and the domain point is used to indicate the target point cloud data of the point cloud key point within a predetermined range; based on the spatial features, the structural features of the insulators in the distribution network are determined.
[0076] In this embodiment, the spatial features between each point cloud key point and the corresponding neighborhood point are determined, and the structural features of the insulators in the distribution network are determined based on the spatial features.
[0077] Optionally, the FPFH calculation method is used to parameterize the spatial difference between the query point and the neighboring points, and form a multidimensional histogram to describe the k-neighborhood geometric properties of the point.
[0078] It should be noted that the above embodiments can be executed by a feature extraction device for a distribution network.
[0079] In this embodiment, the point cloud data of the insulator in the distribution network and the background data of the insulator are obtained; based on the point cloud data and the background data, the target point cloud data of the insulator is determined, wherein the target point cloud data is used to indicate the geometric outline and position information of the insulator; based on the target point cloud data, at least one point cloud key point of the target point cloud data is determined, wherein the point cloud key point is located at the edge of the insulator; based on the point cloud key point, the structural features of the insulator in the distribution network are extracted, wherein the structural features are used to indicate the structure of the insulator. That is to say, the present invention extracts the structural features of the insulator from the target point cloud data of the insulator in the distribution network according to the point cloud key point, and achieves the purpose of improving the uniformity of the point cloud data by performing multi-faceted processing on the point cloud data of the insulator, thereby solving the technical problem of low feature extraction accuracy of the insulator in the distribution network and achieving the technical effect of improving the feature extraction accuracy of the insulator in the distribution network.
[0080] The technical solution of the embodiment of the present invention is illustrated below in conjunction with preferred implementation modes.
[0081] At present, with the expansion of the scale and complexity of the distribution network, the current transformer in the substation is an important device for measuring and monitoring current. Among these devices, insulators play a vital role. The current transformer insulator in the substation is an important part to ensure the safety and normal operation of the equipment. Insulators can effectively isolate the high-voltage part of the current transformer from the ground or low-voltage part to prevent accidental leakage of current, thereby protecting the safety of equipment and personnel. Insulators not only have insulation functions, but also bear the load of supporting the current transformer to ensure the stable operation of the equipment. Therefore, feature extraction of insulators through three-dimensional point clouds can improve the monitoring and maintenance efficiency of current transformers in many aspects.
[0082] In related technologies, due to the complexity and variability of the outdoor environment (such as changes in light, rain, dust, temperature differences, flashover discharges causing spatial distortion, etc.), as well as the viewing angle limitations of the equipment, the acquired point cloud data may have differences in density and uniformity, and the existing feature extraction algorithms may extract inaccurate or invalid features when processing point cloud data with complex shapes, uneven surfaces, or noise, affecting the results of subsequent analysis. Therefore, there is a technical problem of low feature extraction accuracy for distribution network insulators. No effective solution has been proposed so far for the above-mentioned technical problem of low feature extraction accuracy for distribution network insulators.
[0083] However, an embodiment of the present invention proposes a three-dimensional point cloud feature extraction method for current transformer insulators based on a substation operation and maintenance robot. A topological structure is established based on the connection relationship and equipment status information of the distribution network equipment, and a graph embedding model is established based on the nodes in the topological structure. The changes in the equipment status and connection relationship in the power grid are monitored in real time, and the update of the local topological structure is detected. When new equipment is added or the equipment status changes, the graph embedding features in the graph embedding model are normalized with the updated node features, the topological structure is updated again, the recognition model is trained, and the main and auxiliary equipment of the distribution network are identified according to the trained recognition model. By performing multi-faceted processing on the point cloud data of the insulator, the purpose of improving the uniformity of the point cloud data is achieved, thereby solving the technical problem of low feature extraction accuracy of the distribution network insulators and achieving the technical effect of improving the feature extraction accuracy of the distribution network insulators.
[0084] The embodiments of the present invention are further described below.
[0085] Figure 2 The present invention is a flowchart of a method for extracting three-dimensional point cloud features of a current transformer insulator based on a substation operation and maintenance robot according to an embodiment of the present invention. The feature extraction method comprises the following steps:
[0086] Step S201, obtaining three-dimensional point cloud data of current transformer insulators at different stages.
[0087] In this embodiment, according to the location distribution of current transformers in the substation, by designing an automatic patrol route for current transformer insulators for the substation operation and maintenance robot, a large amount of three-dimensional point cloud data of current transformer insulators at different stages can be regularly acquired.
[0088] Step S202: pre-processing the three-dimensional point cloud data.
[0089] In this embodiment, the three-dimensional point cloud data acquired in step S202 is preprocessed.
[0090] Optionally, the original substation current transformer insulator three-dimensional point cloud data volume is huge, and the discrete noise points have a great influence on feature extraction, so the original cloud needs to be preprocessed, including denoising, background segmentation, downsampling, etc.
[0091] Optionally, denoising the point cloud of the current transformer insulator: Due to interference from the acquisition equipment, environment, etc., the original point cloud contains outliers introduced by measurement noise. The density threshold of the current transformer insulator point cloud is defined using a conditional removal filter. When the point cloud density in the original point cloud is less than the threshold, it is a noise point.
[0092] Optionally, point cloud segmentation of current transformer insulator: the initial point cloud of the current transformer insulator contains useless information such as natural background and environmental debris. A radius filter is used to set a radius and point cloud quantity limit with the geometric center of the current transformer insulator as the center of the sphere, and the number of point clouds falling inside the sphere drawn with the given radius is calculated. When the calculated point cloud number is greater than the given limit, the point is retained, and if it is less than the limit, it is removed.
[0093] Optionally, Figure 3 is a schematic diagram of a filtering principle according to an embodiment of the present invention, such as Figure 3 As shown in the figure, if the limit is 1 adjacent point, the square point is deleted, and if the limit is 2 adjacent points, both the square point and the diamond point are deleted.
[0094] Optionally, downsampling of the current transformer insulator point cloud: The original collected point cloud is relatively dense, which makes subsequent normal calculation and analysis difficult. A voxel filter is used to create a three-dimensional voxel grid for the original point cloud (a voxel grid can be understood as a collection of tiny three-dimensional cubes). In each voxel grid, the centroid of all points in the voxel is used to approximate other points in the voxel, so that all points in the voxel are finally represented by a centroid point. After processing all voxels, the filtered point cloud is obtained.
[0095] Optionally, Table 1 is a table showing a change in the number of point clouds before and after preprocessing. The number of point clouds in the time period before downsampling (Frame1) and the time period after downsampling (Frame2) are shown in Table 1. The number of point clouds after preprocessing is less than the number of point clouds before preprocessing.
[0096] Table 1 Changes in the number of point clouds before and after preprocessing
[0097]
[0098] Optionally, Figure 4 is a schematic diagram of comparing point cloud data before and after downsampling according to an embodiment of the present invention. Figure 4 As shown in the figure, after downsampling, the number of points at the same position in the same point cloud data is significantly reduced, but its geometric structure does not change, which greatly improves the rate of subsequent point cloud processing.
[0099] Step S203: performing key point detection on the preprocessed three-dimensional point cloud data.
[0100] In this embodiment, key point detection is performed on the three-dimensional point cloud data pre-processed in step S202.
[0101] Optionally, in the multi-frame point cloud key point extraction of the three-dimensional point cloud of the substation current transformer insulator, considering that the points on the edge of the object are more likely to be key points than the points in other parts, the edge of the point cloud is first extracted. Compared with the obvious change of the edge grayscale of the two-dimensional image, the edge of the point cloud has a clearer physical meaning. For the point cloud, the edge of the scene represents the boundary between the foreground object and the background object.
[0102] Optionally, the edge of the 3D point cloud of the current transformer insulator has a very important feature. If two points on the edge of the point cloud are adjacent in the depth image, but are far apart in the 3D distance, then this is likely to be an edge. Therefore, by traversing each point in the depth image, finding the position where the depth changes suddenly in the nearest area, and performing edge detection, the point cloud boundary can be extracted.
[0103] Optionally, a nonparametric autoregressive model (Noise-Aware Robust Factorization, abbreviated as NARF) algorithm based on Gaussian process extracts edge interest points with stable surface but great neighborhood changes through the point cloud image boundary detection results, focusing on analyzing the object structure. It has the advantages of typicality and rotation-invariance, and is more suitable for current transformer insulator point cloud data with strong edge structure regularity.
[0104] Step S204: extracting features from the three-dimensional point cloud data according to the detected key points.
[0105] In this embodiment, feature extraction is performed on the three-dimensional point cloud data based on the detected key points.
[0106] Optionally, after obtaining the point cloud key points of the current transformer insulator, the FPFH calculation method is used to parameterize the spatial difference between the query point and the neighborhood points, and a multidimensional histogram is formed to describe the k-neighborhood geometric attributes of the point.
[0107] Optionally, a point feature histogram (PFH) describes the geometric features of the pixel object by counting the changes in the normal relationship between the query point and the k-domain and between all points in the k-domain. Figure 5 is a schematic diagram of an influence area of a point histogram calculation of a query point according to an embodiment of the present invention, Figure 5 A query point P is shown in q The PFH calculation influence area, where P1 to P6 represent points outside the influence area, P k1 To P k5 represents the point in the influence area, P q In Figure 5 The center square of the middle circle marks the position, the radius of the circle is r, P q All k domain elements are interconnected in a network, and the PFH histogram is finally obtained by calculating the relationship between any two points in the domain and their normals.
[0108] Optionally, for any point P in the point cloud q Query all neighboring points in its k domain; for point P q Any pair of points P in the k-domain s and P t (s!=t) Estimated normal n s and n t , define a fixed local coordinate system at one of the points to calculate n s and n t The deviation between them can be calculated by the following formula (6):
[0109]
[0110] in, Figure 6 is a schematic diagram of a local coordinate system defined according to an embodiment of the present invention, such as Figure 6 As shown, in Figure 6 In the u, v, w local coordinate system, the following set of angles is used to estimate the normal n in the point cloud s and n t Deviation between:
[0111] α=v·n t (7)
[0112]
[0113]
[0114] Among them, ||P t -P s ||=d represents point P s and P t The Euclidean distance between two points, then P q Any two points P in all k fields s and P t The relationship can be expressed by (α, β, θ, d). All four sets of values in the k domain are statistically put into the histogram to obtain P q feature descriptor.
[0115] Optionally, the fast point feature histogram (FPFH) is improved from PFH. FPFH effectively reduces the computational complexity of the algorithm, but retains most of the recognition characteristics of PFH, and has good robustness and real-time performance. In order to simplify the feature calculation of the histogram, the calculation process is as follows: Simple Point Feature Histograms (SPFH), that is, for each query point P q , calculate a tuple between the point and its neighboring points Calculate FPFH, redetermine the neighborhood k of each point, and use the neighboring SPFH to calculate P q The final FPFH is calculated as follows:
[0116]
[0117] Among them, the weight ω k Query point P in a given metric space q and its neighboring point P k The distance between Figure 7 According to an embodiment of the present invention, a P q A schematic diagram of the influence range of the k-neighborhood centered on Figure 7 As shown, P1 to P 17 And represents the points within the area of influence, Figure 7 Shows SPFH with P q The influence range of the k-neighborhood centered on is .
[0118] Optionally, according to the above principle, FIG8 is a schematic diagram of a statistical subinterval according to an embodiment of the present invention, wherein FIG8(a) is a FPFH feature histogram of a key point of Frame1, and FIG8(b) is a FPFH feature histogram of a key point of Frame2. Frame1 and Frame2 are described by FPFH features, and FIG8 shows a FPFH feature histogram of a set of corresponding key point pairs of Frame1 and Frame2. q Calculate a tuple between the point and its neighboring points To establish a spatial rectangular coordinate system, and Each axis is divided into 11 sub-intervals. Since the dimension of FPFH is 33, the feature histograms are combined to obtain a 33-element feature vector. Then the query point P is calculated. q Search all points (p1, p2, ..., p k )The deviation of the normal angle {(α1, β1, θ1), (α2, β2, θ2), …, (α k , β k ,θ k )}, according to the calculation results, we can convert {(α1, β1, θ1), (α2, β2, θ2), …, (α k , β k ,θ k )} each value is placed in In the 33 intervals divided in the spatial rectangular coordinate system.
[0119] Optionally, in the two broken line graphs on the left of FIG. 8 , the values of the horizontal axis are divided into 33 statistical intervals from 0 to 32, and the values of the vertical axis are the points (p1, p2, ..., p k ). It can be seen from the figure that the FPFH features of a group of corresponding key points in Frame1 and Frame2 are very similar. Although the broken lines of the two figures are not completely consistent, it can be seen that the basic trends of the broken lines are consistent, and the horizontal coordinates corresponding to the peaks of the vertical coordinates are also roughly the same. This shows that the features of the key points selected by Frame1 and Frame2 are extremely similar, which is very beneficial to the accuracy of subsequent point cloud registration and other processing work.
[0120] Fig. 9 1 is a flow chart of a method for detecting key points of a current transformer insulator according to an embodiment of the present invention, the method comprising the following steps:
[0121] Step S901, traverse each depth image point, find the position with depth mutation in the adjacent area, and perform edge detection.
[0122] In this embodiment, in the multi-frame point cloud key point extraction of the three-dimensional point cloud of the substation current transformer insulator, considering that the points on the edge of the object are more likely to be key points than the points in other parts, the point cloud is firstly edge extracted, and each depth image point is traversed to find the position with depth mutation in the surrounding area and perform edge detection.
[0123] Optionally, the edge of a point cloud has a clearer physical meaning than the obvious change in grayscale at the edge of a two-dimensional image. For a point cloud, the edge of a scene represents the boundary between foreground objects and background objects. The edge of a three-dimensional point cloud of a current transformer insulator has a very important feature. If two points on the edge of a point cloud are adjacent in the depth image, but are far apart in three-dimensional distance, then this is likely to be an edge. Therefore, by traversing each point in the depth image, finding the location where the depth suddenly changes in the nearest field, and performing edge detection, the point cloud boundary can be extracted.
[0124] Step S902, determining a coefficient for measuring surface change and a main direction of change according to the surface change of the neighboring area.
[0125] In this embodiment, the coefficient of surface variation can be calculated by comparing the elevation data or color data of adjacent areas. Common methods include calculating statistics such as standard deviation, variance, mean difference, etc. of elevation or color to quantify the degree of surface variation.
[0126] Alternatively, the dominant direction of change can be determined by calculating the gradient or direction of adjacent regions. In elevation data, the dominant direction of the surface can be determined by calculating the slope and aspect. In color data, the dominant direction can be determined by calculating the gradient or histogram.
[0127] Step S903, calculating an interest value based on the main direction, characterizing the difference between this direction and other directions, and the change of the surface at that location.
[0128] In this embodiment, the interest value is calculated based on the main direction to characterize the difference between the direction and other directions and the change of the surface at that location. The interest value can be calculated by processing information such as the curvature and normal vector of the surface.
[0129] Optionally, the boundary interest value may be calculated according to the aforementioned formula (1) to formula (5), which will not be described in detail here.
[0130] Step S904: performing smoothing filtering on the interest value.
[0131] In this embodiment, the smoothing filtering of the interest value refers to adjusting the original interest value through a certain algorithm or method to make it smoother and more stable, which can reduce mutations or fluctuations and improve the stability and reliability of the data.
[0132] Optionally, common smoothing filtering methods include moving average method, exponential smoothing method, weighted average method, etc. These methods can be processed by selecting appropriate methods according to specific data characteristics and requirements.
[0133] Step S905: Perform non-maximum value compression to find the final key point.
[0134] In this embodiment, no maximum value compression is performed to find the final key point.
[0135] Optionally, to perform non-maximum compression to find the final key point, the following steps can be followed: perform non-maximum compression on the original data to obtain a compressed data sequence. Analyze the compressed data sequence to find the extreme value or inflection point. Based on the key points found, important change points or outliers in the data can be determined.
[0136] In this embodiment, a topological structure is established based on the collection of connection relationships and equipment status information of distribution network equipment, and a graph embedding model is established based on the nodes in the topological structure. Changes in the equipment status and connection relationships in the power grid are monitored in real time, and updates to the local topological structure are detected. When new equipment is added or the equipment status changes, the graph embedding features in the graph embedding model are normalized with the updated node features, the topological structure is re-updated, the recognition model is trained, and the primary and secondary equipment of the distribution network are identified based on the trained recognition model. By performing multi-faceted processing on the point cloud data of the insulators, the purpose of improving the uniformity of the point cloud data is achieved, thereby solving the technical problem of low feature extraction accuracy for the insulators in the distribution network and achieving the technical effect of improving the feature extraction accuracy for the insulators in the distribution network.
[0137] According to an embodiment of the present invention, a feature extraction device for a distribution network is also provided. It should be noted that the feature extraction device for a distribution network can be used to execute the feature extraction method for a distribution network in Embodiment 1.
[0138] Fig.10 Schematic diagram of a feature extraction device for a distribution network according to an embodiment of the present invention. Fig.10 As shown, the feature extraction device 1000 of the distribution network may include: an acquisition unit 1001 , a first determination unit 1002 , a second determination unit 1003 and an extraction unit 1004 .
[0139] An acquisition unit 1001 is used to acquire point cloud data and background data of insulators in a distribution network;
[0140] A first determining unit 1002 is used to determine target point cloud data of the insulator based on the point cloud data and the background data, wherein the target point cloud data is used to indicate the geometric profile and position information of the insulator;
[0141] A second determining unit 1003 is used to determine at least one point cloud key point of the target point cloud data based on the target point cloud data, wherein the point cloud key point is located at an edge of the insulator;
[0142] The extraction unit 1004 is used to extract the structural features of the insulators in the distribution network based on the key points of the point cloud, wherein the structural features are used to indicate the structure of the insulators.
[0143] Optionally, the first determining unit 1002 may include: a preprocessing module, configured to preprocess the point cloud data and the background data to obtain target point cloud data.
[0144] Optionally, the second determination unit 1003 may include: a traversal module, used to traverse multiple target point cloud data and determine at least one point cloud boundary in the target point cloud data, wherein the point cloud boundary is used to indicate a dividing line or transition area between each target point cloud data; a first determination module, used to determine the point cloud key points of the corresponding target point cloud data based on the point cloud boundary.
[0145] Optionally, the first determination module may include: a first determination submodule, used to determine the boundary value corresponding to the target point cloud data based on each point cloud boundary, wherein the boundary value is used to indicate the target point cloud data with the largest curvature change among the target point cloud data; a second determination submodule, used to determine the point cloud key points of the target point cloud data based on the boundary value.
[0146] Optionally, the second determination submodule may also be used to perform smoothing filtering on the boundary value; and determine a plurality of point cloud key points of the target point cloud data based on the smoothed boundary value.
[0147] Optionally, the extraction unit 1004 may include: a second determination module, used to determine the spatial features between each point cloud key point and the corresponding neighborhood point, wherein the spatial features are used to indicate the change in the normal relationship between the point cloud key point and the domain point, and the domain point is used to indicate the target point cloud data of the point cloud key point within a predetermined range; a third determination module, used to determine the structural features of the insulators in the distribution network based on the spatial features.
[0148] In this embodiment, the point cloud data of the insulator in the distribution network and the background data of the insulator are obtained; based on the point cloud data and the background data, the target point cloud data of the insulator is determined, wherein the target point cloud data is used to indicate the geometric outline and position information of the insulator; based on the target point cloud data, at least one point cloud key point of the target point cloud data is determined, wherein the point cloud key point is located at the edge of the insulator; based on the point cloud key point, the structural features of the insulator in the distribution network are extracted, wherein the structural features are used to indicate the structure of the insulator. That is to say, the present invention extracts the structural features of the insulator from the target point cloud data of the insulator in the distribution network according to the point cloud key point, and achieves the purpose of improving the uniformity of the point cloud data by performing multi-faceted processing on the point cloud data of the insulator, thereby solving the technical problem of low feature extraction accuracy of the insulator in the distribution network and achieving the technical effect of improving the feature extraction accuracy of the insulator in the distribution network.
[0149] According to an embodiment of the present invention, a computer-readable storage medium is further provided. The storage medium includes a stored program, wherein the program executes the feature extraction method of the distribution network in the method embodiment.
[0150] According to an embodiment of the present invention, a processor is further provided. The processor is used to run a program, wherein the feature extraction method of the distribution network in the method embodiment is executed when the program is run.
[0151] According to an embodiment of the present invention, a computer program product is further provided. The computer program product includes computer instructions. When the computer instructions are executed by a processor, the method for extracting characteristics of a distribution network in the method embodiment is implemented.
[0152] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.
[0153] In the above embodiments of the present invention, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.
[0154] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of units can be a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.
[0155] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed over multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.
[0156] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.
[0157] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent functional component, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software functional component, which 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 methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk, etc., which can store program code.
[0158] The above are only preferred embodiments of the present invention. It should be pointed out that, for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.
Claims
1. A method for extracting features of a distribution network, characterized in that: include: Acquire point cloud data of insulators in a distribution network and background data of the insulators; Determine target point cloud data of the insulator based on the point cloud data and the background data, wherein the target point cloud data is used to indicate the geometric profile and position information of the insulator; Based on the target point cloud data, determining at least one point cloud key point of the target point cloud data, wherein the point cloud key point is located at an edge of the insulator; Based on the point cloud key points, structural features of the insulator in the distribution network are extracted, wherein the structural features are used to indicate the structure of the insulator.
2. The method according to claim 1, characterized in that Determining target point cloud data of the insulator based on the point cloud data and the background data includes: The point cloud data and the background data are preprocessed to obtain the target point cloud data.
3. The method according to claim 1, characterized in that Determining at least one point cloud key point of the target point cloud data based on the target point cloud data includes: Traversing the plurality of target point cloud data, determining at least one point cloud boundary among the plurality of target point cloud data, wherein the point cloud boundary is used to indicate a dividing line or a transition area between the respective target point cloud data; Based on the point cloud boundary, the point cloud key points of the corresponding target point cloud data are determined.
4. The method according to claim 3, characterized in that Determining the point cloud key points of the corresponding target point cloud data based on the point cloud boundary includes: Based on each of the point cloud boundaries, determining a boundary value corresponding to the target point cloud data, wherein the boundary value is used to indicate the target point cloud data having the largest curvature change among the target point cloud data; Based on the boundary value, the point cloud key points of the target point cloud data are determined.
5. The method according to claim 4, characterized in that Determining the point cloud key points of the target point cloud data based on the boundary value includes: Performing smoothing filtering on the boundary value; Based on the smoothed boundary value, a plurality of point cloud key points of the target point cloud data are determined.
6. The method according to claim 1, characterized in that Extracting structural features of the insulator in the distribution network based on the key points of the point cloud includes: Determine the spatial features between each of the point cloud key points and the corresponding neighborhood points, wherein the spatial features are used to indicate the change of the normal relationship between the point cloud key points and the domain points, and the domain points are used to indicate the target point cloud data within a predetermined range of the point cloud key points; Based on the spatial characteristics, the structural characteristics of the insulator in the distribution network are determined.
7. A feature extraction device for a distribution network, characterized in that: include: An acquisition unit, used for acquiring point cloud data of insulators in the distribution network and background data of the insulators; A first determining unit, configured to determine target point cloud data of the insulator based on the point cloud data and the background data, wherein the target point cloud data is used to indicate a geometric profile and position information of the insulator; A second determining unit is used to determine at least one point cloud key point of the target point cloud data based on the target point cloud data, wherein the point cloud key point is located at an edge of the insulator; An extraction unit is used to extract structural features of the insulator in the distribution network based on the point cloud key points, wherein the structural features are used to indicate the structure of the insulator.
8. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored program, wherein when the program is executed by a processor, the device where the storage medium is located is controlled to execute the method according to any one of claims 1 to 6.
9. A processor, characterized in that: The processor is used to run a program, wherein the program executes the method according to any one of claims 1 to 6 when running.
10. A computer program product, characterized in that The computer program product comprises computer instructions, which implement the method according to any one of claims 1 to 6 when executed by a processor.
Citation Information
Patent Citations
A contact network part whole-network 3D reconstruction method based on an NARF and an FPFH
CN107123161A
Scene classification point cloud coarse registration method combining geometric information and luminosity information
CN111009005A
Pole tower insulator extraction method and device based on point cloud data and storage medium
CN115641450A
Registration method and apparatus for 3D image data
US20160210751A1
Cited By
Relay protection parameter self-configuration method and device, program product and medium
CN121307764A
