Sub-conductor extraction method, device and equipment based on adaptive multi-scale spherical neighborhood geometric feature descriptor, and medium

Abstract

The invention belongs to the technical field of electric power inspection, and discloses a sub-conductor extraction method, device and equipment based on adaptive multi-scale spherical neighborhood geometric feature descriptors, and a medium, and the method comprises the steps: obtaining scene point cloud data, and obtaining a non-ground point cloud after filtering; performing voxel-based height feature analysis on the non-ground point cloud, and detecting the position of the split conductor to determine an area; constructing geometric feature descriptors to describe point cloud features in the region, and establishing central position and neighborhood scale adaptive mechanism optimization descriptors; according to an optimization result, sub-conductor candidate points are analyzed and screened through a point cloud distribution direction and a line fitting tangential included angle; dimensionality reduction is carried out by adopting a nonlinear mapping strategy, and the candidate points after dimensionality reduction are segmented to obtain two-dimensional sub-conductor points; and raising the dimension of the two-dimensional sub-conductor point to three dimensions, and extracting a three-dimensional sub-conductor. The method is high in automation degree, high in accuracy and strong in noise immunity, and can be used for finely extracting the sub-conductor in a complex power scene.

Description

Technical Field

[0001] This application relates to the field of power line inspection technology, and in particular to a method, apparatus, equipment and medium for extracting sub-conductors based on an adaptive multi-scale spherical neighborhood geometric feature descriptor. Background Technology

[0002] Split conductors, typically composed of multiple sub-conductors, are core components of high-voltage transmission systems and critical infrastructure for ensuring the stable operation of power grids. In recent years, with the rapid development of my country's power industry, the scale of high-voltage transmission systems has continued to expand, resulting in a large-scale power grid covering the entire country and ranking among the world's leading systems. Transmission lines, exposed to the natural environment for extended periods, are susceptible to factors such as vegetation growth, extreme weather conditions, and line aging. To prevent line faults and reduce the resulting socio-economic losses, power management departments need to conduct regular inspections of transmission corridors to promptly identify and eliminate potential hazards.

[0003] Traditional power grid inspection methods have significant limitations: 1) High-voltage transmission channels typically span thousands of kilometers, requiring a large investment of human resources for manual inspection; 2) These channels often traverse areas with complex terrain, leading to high-risk working environments for inspection personnel and highlighting safety issues; 3) Traditional methods have limited detection accuracy, which can easily result in missed or misjudged hazards, affecting the reliability of power grid safety operation assessments. Summary of the Invention

[0004] To address the aforementioned issues, this application provides a method, apparatus, device, and medium for extracting sub-conductors based on an adaptive multi-scale spherical neighborhood geometric feature descriptor. This overcomes the shortcomings of traditional power line inspection methods, such as low automation, limited detection accuracy, and difficulty in effectively extracting individual sub-conductors. This application features high automation, high detection accuracy, and strong noise resistance, effectively extracting individual sub-conductors and achieving a leap from traditional power line inspection methods to intelligent detection.

[0005] According to the first technical solution of this application, a sub-wire extraction method based on an adaptive multi-scale spherical neighborhood geometric feature descriptor is provided, the method comprising:

[0006] Acquire scene point cloud data;

[0007] The scene point cloud data is filtered to obtain non-ground point clouds;

[0008] Voxel-based height feature analysis is performed on the non-ground point cloud to detect the location of the split conductor and determine the split conductor region.

[0009] Construct geometric feature descriptors to describe the features of the point cloud within the split conductor region;

[0010] Establish an adaptive mechanism between center location and neighborhood scale to optimize geometric feature descriptors;

[0011] Based on the calculation results of the optimized geometric feature descriptor, candidate points for sub-traverses are screened by analyzing the angle between the point cloud distribution direction and the tangent of the line fitting.

[0012] A nonlinear mapping strategy is used to reduce the dimensionality of the candidate points of the sub-trajectories;

[0013] The candidate points of the dimensionality-reduced sub-traverse are segmented to obtain two-dimensional sub-traverse points;

[0014] The two-dimensional sub-traverse points are upgraded to three dimensions to achieve the extraction of three-dimensional sub-traverses.

[0015] Furthermore, voxel-based height feature analysis is performed on the non-ground point cloud to detect the location of the split conductor and determine the split conductor region, including:

[0016] The non-ground point cloud is voxelized, and the point cloud distribution characteristics along the height direction within each voxel are statistically analyzed.

[0017] Based on the point cloud distribution characteristics, a maximum height threshold h is set. max Using a continuous non-empty voxel threshold n, voxels that meet the conditions are selected to detect the location of power towers;

[0018] Based on the detected center location of the power tower in the scene, gp s (x s ,y s ,z s ), where s represents the s-th tower in the scene, and the angle θ between the line connecting the centers of adjacent power towers and the X-axis is calculated using the following formula:

[0019]

[0020] In the formula, x s+1 and y s+1 Let x and y represent the x and y coordinates of the center position of the (s+1)th tower in the scene, respectively; s and y s Let x and y represent the x and y coordinates of the center position of the s-th tower in the scene, respectively.

[0021] Then, the point cloud is rotated using the following formula to align the direction of the line connecting the centers of adjacent power towers with the X-axis of the coordinate system:

[0022]

[0023] In the formula, P(P x ,P y ,P z P represents the point cloud coordinates before rotation. T (P x ,P y ,Pz () are the coordinates of the rotated point cloud;

[0024] Connectivity analysis is used to identify the spatial location of the split conductors in order to determine the region of the split conductors.

[0025] Furthermore, the geometric feature descriptor is represented as:

[0026]

[0027] In the formula, Represents geometric feature descriptors, Let P represent the tensor product, C be the point to be determined, R be the neighborhood radius, and P be the tensor product. l Let N be the l-th point in the neighborhood, and N be the total number of points in the neighborhood. Let ω(·) be the normalized distance, and let ω(·) be the radial distance weighting function, expressed as:

[0028]

[0029] In the formula, t and m are values ​​optimized through experiments, e is a natural constant, and x is the independent variable of the radial distance weighting function.

[0030] Furthermore, an adaptive mechanism for center location and neighborhood scale is established to optimize the geometric feature descriptor, including:

[0031] A weighted mechanism is introduced to differentiate the contribution of each point within the neighborhood; the formula for calculating the weight of each point is as follows:

[0032]

[0033]

[0034] In the formula, Let ρ(P) be the weight of point l. l Let σ be the local density at point l, and G(P) be the Gaussian standard deviation. l () are points l and P MD The Gaussian distance between them, δ, is a parameter used to adjust the degree of influence of local density on the weights; P MD Location is determined by minimizing the distance from each point in the neighborhood to P. MD Distance and determination:

[0035]

[0036] In the formula, l is the index of the point, and N is the number of points;

[0037] P is determined through iterative calculation. MD Position, during iterative calculation, δ is initially set to 1, P MD The initial value is the mean center, and P is continuously updated by adjusting the value of δ.MD Position until P MD The position changes tend to stabilize, so let's take P at this point. MD The final position is C;

[0038] A basic neighborhood is constructed using the average distance between multiple neighboring points as an initial parameter, and the neighborhood radius is dynamically adjusted using an exponential growth strategy. The adjustment function is as follows:

[0039] R = Δ mdn 1.5 q (9)

[0040] Where R is the neighborhood radius, Δ mdn It is the average distance from multiple adjacent points, where q is a constant. The formula for calculating the linearity involved is as follows:

[0041] C L =(λ1-λ2) / λ1,λ1>λ2>λ3 (10)

[0042] In the formula, C L λ1, λ2, and λ3 are the linearity of the point cloud neighborhood; λ1, λ2, and λ3 are the eigenvalues ​​of the point distribution tensor.

[0043] By comparing the linearity at different scales, the neighborhood scale corresponding to the maximum value is selected as the optimal parameter.

[0044] Furthermore, based on the calculation results of the optimized geometric feature descriptor, candidate points for sub-traverses are screened by analyzing the angle between the point cloud distribution direction and the tangent of the line fitting, including:

[0045] The transmission line is approximated as a parabola along its span in space. The tangent direction at a certain location is obtained by parabolic fitting. The actual distribution direction of the point cloud calculated by the optimized geometric feature descriptor is compared with the tangent direction. If the angle between the two is less than a preset threshold, it is judged as a candidate point of the sub-conductor.

[0046] Furthermore, a nonlinear mapping strategy is employed to reduce the dimensionality of the candidate points for the sub-traverse, including:

[0047] Projecting the split conductor structure onto a horizontal plane yields projected line segments. A central baseline is constructed at the midpoint of these projected line segments, and the distance from each sub-conductor projected line segment to this central baseline is calculated using the following formula:

[0048]

[0049] In the formula, d h Let be the distance from the sub-traverse to the center baseline, and a, b, and c be the fitting coefficients of the center baseline, respectively.

[0050] Projecting the split conductors onto a vertical plane along the span of the line yields a projected parabola. A central reference parabola is constructed at the midpoint of this projected parabola, and the distances from the projected parabolas of each sub-conductor to this central reference parabola are calculated.

[0051]

[0052]

[0053] In the formula, d v P is the vertical distance from the sub-traverse to the central reference parabola, where A, B, and D are the fitting coefficients of the reference parabola. x P y Z represents the coordinates of the projection point on the vertical plane along the span of the track; i To represent the Z-coordinate of a point in the sub-traverse point cloud in three-dimensional space; S i The distance in the plane from the origin or reference point of the coordinate system after projecting the sub-traverse point onto a vertical plane along the span of the line.

[0054] Furthermore, the dimensionality-reduced candidate sub-traverse points are segmented to obtain two-dimensional sub-traverse points, including:

[0055] An adaptive multi-density algorithm based on reverse nearest neighbor is used to segment the dimensionality-reduced candidate points of the sub-traverse.

[0056] The reverse nearest neighbor is defined as: if q i It is q j If q is one of the nearest neighbors, then q i It is q j The reverse nearest neighbor;

[0057] The algorithm characterizes local density by the number of reverse nearest neighbors: in dense regions, points with a number of reverse nearest neighbors exceeding a threshold k are defined as core points; in sparse regions, points with a number of reverse nearest neighbors less than k but covered by high-density core points are defined as boundary points, and the rest are noise points.

[0058] By connecting the core points and boundary points of different density regions using a KNN graph, cross-density growth of clusters is achieved, thus completing the segmentation.

[0059] According to the second technical solution of this application, a sub-wire extraction device based on an adaptive multi-scale spherical neighborhood geometric feature descriptor is provided, the device comprising:

[0060] The data acquisition module is configured to acquire scene point cloud data;

[0061] The point cloud filtering module is configured to filter the scene point cloud data to obtain non-ground point clouds;

[0062] The split conductor identification module is configured to perform voxel-based height feature analysis on the non-ground point cloud to detect the location of the split conductor and determine the split conductor region.

[0063] The descriptor building module is configured to build geometric feature descriptors to describe the features of the point cloud within the split conductor region;

[0064] The descriptor optimization module is configured to establish an adaptive mechanism between the center location and the neighborhood scale to optimize the geometric feature descriptor.

[0065] The sub-traverse screening module is configured to screen candidate points for sub-traverses based on the calculation results of the optimized geometric feature descriptors and by analyzing the angle between the point cloud distribution direction and the tangent of the line fitting.

[0066] The sub-traverse dimension reduction module is configured to use a nonlinear mapping strategy to reduce the dimension of the candidate points of the sub-traverse;

[0067] The sub-trajectory segmentation module is configured to segment the dimensionality-reduced candidate points of the sub-trajectory to obtain two-dimensional sub-trajectory points;

[0068] The sub-trajectory dimensionality upscaling module is configured to upscale the two-dimensional sub-trajectory points to three dimensions, thereby enabling the extraction of three-dimensional sub-trajectory points.

[0069] According to the third technical solution of this application, an electronic device is provided, the electronic device comprising: a memory for storing a computer program; and a processor for executing the computer program to implement the method described above.

[0070] According to the fourth technical solution of this application, a non-transitory computer-readable storage medium storing instructions is provided, which, when executed by a processor, performs the method described above.

[0071] The sub-wire extraction methods, apparatuses, devices, and media based on adaptive multi-scale spherical neighborhood geometric feature descriptors according to various schemes in this application have at least the following technical effects:

[0072] This application boasts a high degree of automation and fully leverages the descriptor's advantage in effectively capturing point cloud features, effectively filtering out noise points and ensuring precise extraction of sub-conductors. More importantly, this application realizes the transformation from traditional manual inspection to an automated, high-precision, and high-efficiency intelligent inspection method, significantly improving the intelligence level and management efficiency of high-voltage transmission channel inspection. Attached Figure Description

[0073] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0074] Figure 1 A flowchart illustrating a sub-wire extraction method based on an adaptive multi-scale spherical neighborhood geometric feature descriptor provided in this application embodiment;

[0075] Figure 2 This is a schematic diagram illustrating the type of split wire targeted for extraction in an embodiment of this application;

[0076] Figure 3 A schematic diagram illustrating the construction of an adaptive center position provided in an embodiment of this application;

[0077] Figure 4 A schematic diagram illustrating the construction of an adaptive multi-scale spherical neighborhood provided in an embodiment of this application;

[0078] Figure 5 A schematic diagram of the candidate points for screening sub-wires provided in an embodiment of this application;

[0079] Figure 6 A schematic diagram of the adaptive multi-density algorithm provided in the embodiments of this application;

[0080] Figure 7 The result diagram of fine extraction of sub-wires provided in the embodiments of this application;

[0081] Figure 8 This is a structural diagram of a sub-wire extraction device based on an adaptive multi-scale spherical neighborhood geometric feature descriptor, provided in an embodiment of this application. Detailed Implementation

[0082] To enable those skilled in the art to better understand the technical solution of this application, the application will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0083] One aspect of this application provides a sub-wire extraction method based on an adaptive multi-scale spherical neighborhood geometric feature descriptor. For example... Figure 1 As shown, Figure 1 This is a flowchart of an embodiment of a sub-traverse extraction method based on an adaptive multi-scale spherical neighborhood geometric feature descriptor provided in this application. The sub-traverse extraction method based on the adaptive multi-scale spherical neighborhood geometric feature descriptor includes the following steps S10-S90.

[0084] S10: Acquire scene point cloud data.

[0085] In some embodiments, UAV-borne lidar is used to collect point cloud data F(x) of the power transmission channel scene. i ,y i ,z i ) i＝1,2,...,nWhere x, y, and z are the 3D coordinates of the objects in the scene, n is the number of point clouds collected, and i represents the i-th point. The type of split conductor within the scene is as follows: Figure 2 As shown.

[0086] S20: Filter the scene point cloud data to obtain non-ground point clouds.

[0087] For example, in step S20, a cloth simulation filtering algorithm can be used to perform ground filtering on the point cloud to obtain a non-ground point cloud.

[0088] In some embodiments, the point cloud F(x) of a high-voltage transmission channel scenario i ,y i ,z i ) i＝1,2,…,n Perform ground filtering to obtain the point set F'(x) after removing ground points. j ,y j ,z j ) j＝1,2,…,m , where m is the number of non-ground points after removing ground points, and j represents the j-th point in the point set F'.

[0089] S30: Perform voxel-based height feature analysis on non-ground point clouds to detect the location of split conductors and determine the region of the split conductor.

[0090] In some embodiments, step S30 may specifically detect the position of the split wire in the following manner:

[0091] For non-ground point clouds F'(x j ,y j ,z j ) j＝1,2,…,m Voxelization was performed, and the point cloud distribution characteristics along the height direction within each voxel were statistically analyzed. Based on the significant continuity between power towers and tall vegetation along the elevation direction (manifested as a large number of continuous non-empty voxels), while transmission lines exhibit discontinuous characteristics, a maximum height threshold h was set. max A threshold of n for continuous non-empty voxels is used to filter voxels that meet the criteria for detecting power tower locations. The number of power towers detected in the scene is n. p Its center position is gp s (x s ,y s ,z s ), where s represents the s-th tower in the scene. First, calculate the angle θ between the line connecting the centers of adjacent power towers and the X-axis:

[0092]

[0093] In the formula, x s+1 and y s+1Let x and y represent the x and y coordinates of the center position of the (s+1)th tower in the scene, respectively; s and y s Let x and y represent the x and y coordinates of the center position of the s-th tower in the scene, respectively.

[0094] Subsequently, the point cloud was rotated to align the direction of the line connecting the centers of adjacent power towers with the X-axis of the coordinate system:

[0095]

[0096] In the formula, P(P x ,P y ,P z P represents the point cloud coordinates before rotation. T (P x ,P y ,P z () represents the coordinates of the rotated point cloud; finally, the spatial location of the split traverse is accurately identified through connected component analysis.

[0097] S40: Construct a geometric feature descriptor to describe the features of the point cloud within the split conductor region.

[0098] In some embodiments, to achieve accurate description of the global geometric features of the sub-traverse point cloud, point distribution tensor theory is introduced to construct a geometric feature descriptor. This descriptor is defined as:

[0099]

[0100] In the formula, Represents geometric feature descriptors, Let P represent the tensor product, C be the point to be determined, R be the neighborhood radius, and P be the tensor product. l Let N be the l-th point in the neighborhood, and N be the total number of points in the neighborhood. Let ω(·) be the normalized distance, and ω(·) be the radial distance weighting function, which can be expressed as:

[0101]

[0102] In the formula, the parameters m and t are set to 0.6 and 0.1 respectively, which are values ​​optimized through experiments, e is a natural constant, and x is the independent variable of the radial distance weighting function.

[0103] S50: Establish an adaptive mechanism for center location and neighborhood scale to optimize geometric feature descriptors.

[0104] In some embodiments, a geometric feature descriptor optimized by a center location and neighborhood scale adaptive mechanism is constructed through the following steps:

[0105] 1) The selection of the center position in the geometric feature descriptor significantly affects the feature extraction results. To improve its adaptability in scenarios with non-uniform sampling and data gaps, an adaptive center position selection method is proposed, which introduces a weighting mechanism to differentiate the contribution of each point in the neighborhood. First, the weight value of each point is calculated:

[0106]

[0107]

[0108] In the formula, Let ρ(P) be the weight of point l. l Let σ be the local density at point l, and σ be the Gaussian standard deviation. 2 Set to 0.5, G(P) l () are points l and P MD The Gaussian distance between them, δ, is a parameter used to adjust the degree of influence of local density on the weights; P MD Location is determined by minimizing the distance from each point in the neighborhood to P. MD Distance and determination:

[0109]

[0110] In the formula, l is the index of the point, and N is the number of points.

[0111] P is determined through iterative calculation. MD Position, during iterative calculation, δ is initially set to 1, P MD The initial value is the mean center, and P is continuously updated by adjusting the value of δ. MD Position until P MD The position changes tend to stabilize, so let's take P at this point. MD The final position of position C, such as Figure 3 As shown.

[0112] 2) Using geometric feature descriptors with a fixed neighborhood radius has limitations in sub-traverse point clouds with varying densities and structures. An excessively large neighborhood scale can obscure the linear features of the sub-traverse, while an excessively small scale fails to fully capture its structural information. To address this issue, this paper proposes an adaptive neighborhood scale adjustment scheme. A basic neighborhood is constructed using the average distance between six neighboring points as an initial parameter, and then the neighborhood radius is dynamically expanded using an exponential growth strategy. The adjustment function is as follows:

[0113] R = Δ mdn 1.5 q (9)

[0114] Where R is the neighborhood radius, Δ mdn It is the average distance to 6 adjacent points, and q is a constant. The formula for calculating the linearity involved is as follows:

[0115] C L =(λ1-λ2) / λ1,λ1>λ2>λ3 (10)

[0116] In the formula, C L λ1, λ2, and λ3 represent the linearity of the point cloud neighborhood; λ1, λ2, and λ3 are the eigenvalues ​​of the point distribution tensor. By comparing the linearity at different scales, the neighborhood scale corresponding to the maximum value is selected as the optimal parameter, such as... Figure 4 As shown.

[0117] S60: Based on the calculation results of the optimized geometric feature descriptor, candidate points for sub-traverses are selected by analyzing the angle between the point cloud distribution direction and the tangent of the line fitting.

[0118] In some embodiments, step S60 can specifically screen candidate points for sub-conductors in the following way: The transmission line is approximately parabolically distributed along its span in space. The tangent direction at a certain location is obtained through parabolic fitting, and the tangent direction reflects the linear distribution direction at that location. The actual distribution direction of the point cloud calculated by the adaptive multi-scale spherical neighborhood geometric feature descriptor is compared with the tangent direction at an angle: if the angle between the two is less than a preset threshold, it is judged as a candidate point for a sub-conductor; otherwise, if the angle is too large, it is discarded. Figure 5 As shown.

[0119] S70: A nonlinear mapping strategy is used to reduce the dimensionality of candidate points for sub-traverses.

[0120] In some embodiments, the dimensionality reduction of candidate points using a nonlinear mapping strategy specifically includes: transforming the sub-traverse segmentation problem in three-dimensional space into a two-dimensional processing task through a nonlinear mapping strategy. Specifically, the split traverse structure is projected onto a horizontal plane, and a central baseline is constructed at the midpoint of these projected segments. Subsequently, the distance from each projected sub-traverse segment to this baseline is calculated:

[0121]

[0122] In the formula, d h Let be the distance from the sub-traverse to the central reference line, and let a, b, and c be the fitting coefficients of the central reference line, respectively. To further describe the hierarchical distribution characteristics of the sub-traverses in the vertical direction, the split traverses are projected onto a vertical plane along the span direction of the line, constructing a central reference parabola located at the midpoint of these projected parabolas. Subsequently, the distance from each sub-traverse's projected parabola to this reference parabola is calculated:

[0123]

[0124]

[0125] In the formula, dv P represents the vertical distance from the sub-traverse to the central reference parabola, where A, B, and D are the fitting coefficients of the reference parabola. x P y Z represents the coordinates of the projection point on the vertical plane along the span of the track; i To represent the Z-coordinate of a point in the sub-traverse point cloud in three-dimensional space; S i This refers to the distance in the plane from the origin or reference point of the coordinate system after projecting a sub-traverse point onto a vertical plane along the track span. This is achieved by establishing a coordinate system with d... h d v The new coordinate system, based on the two-dimensional space, can accurately characterize the relative positional relationship of the sub-trajectories in three-dimensional space.

[0126] S80: Segment the candidate points of the sub-traverse after dimensionality reduction to obtain two-dimensional sub-traverse points.

[0127] In some embodiments, an adaptive multi-density algorithm is used to segment the dimensionality-reduced candidate points for sub-traverses. Specifically, to address the problem of significant spatial differences in the density of sub-traverse point clouds, an adaptive multi-density algorithm based on reverse nearest neighbors significantly improves the segmentation difficulty caused by density differences. The reverse nearest neighbor is defined as: if q i It is q j One of the nearest neighbors is called q. j It is q i The algorithm uses the reverse nearest neighbor (RNN) to accurately reflect local density. In densely populated regions, points with more than a threshold k are defined as core points. In sparsely populated regions, points with fewer than k RNNs but covered by core points in high-density regions are defined as boundary points. The remaining points are considered noise points. The algorithm further connects core and boundary points in different density regions using a KNN graph, enabling cross-density growth of clusters and effectively segmenting the dimensionality-reduced candidate point set for sub-trajectories. Figure 6 As shown.

[0128] S90: Upgrade the two-dimensional sub-traverse points to three dimensions to achieve the extraction of three-dimensional sub-traverses.

[0129] In some embodiments, the point cloud of the sub-trajectories in the segmented two-dimensional projection plane is upscaled to three-dimensional space, ultimately achieving fine extraction of the three-dimensional sub-trajectories, such as... Figure 7 As shown.

[0130] Another aspect of this application embodiment provides a sub-wire extraction device based on an adaptive multi-scale spherical neighborhood geometric feature descriptor, such as... Figure 8The diagram shown is a structural diagram of a sub-trajectory extraction device based on an adaptive multi-scale spherical neighborhood geometric feature descriptor provided in an embodiment of this application. The sub-trajectory extraction device based on the adaptive multi-scale spherical neighborhood geometric feature descriptor includes:

[0131] Data acquisition module 801 is configured to acquire scene point cloud data;

[0132] The point cloud filtering module 802 is configured to filter the scene point cloud data to obtain non-ground point clouds.

[0133] The split conductor identification module 803 is configured to perform voxel-based height feature analysis on the non-ground point cloud to detect the location of the split conductor and determine the split conductor region.

[0134] Descriptor building module 804 is configured to build geometric feature descriptors to describe the features of the point cloud within the split conductor region;

[0135] The descriptor optimization module 805 is configured to establish an adaptive mechanism between the center position and the neighborhood scale to optimize the geometric feature descriptor.

[0136] The sub-traverse screening module 806 is configured to screen candidate points of the sub-traverse based on the calculation results of the optimized geometric feature descriptor and by analyzing the angle between the point cloud distribution direction and the tangent of the line fitting.

[0137] The sub-traverse dimension reduction module 807 is configured to use a nonlinear mapping strategy to reduce the dimension of the candidate points of the sub-traverse;

[0138] The sub-trajectory segmentation module 808 is configured to segment the dimensionality-reduced candidate points of the sub-trajectory to obtain two-dimensional sub-trajectory points;

[0139] The sub-trajectory dimensionality upscaling module 809 is configured to upscale the two-dimensional sub-trajectory points to three dimensions, thereby enabling the extraction of three-dimensional sub-trajectory points.

[0140] It should be noted that the sub-wire extraction device based on adaptive multi-scale spherical neighborhood geometric feature descriptor provided in the above embodiments and the sub-wire extraction method based on adaptive multi-scale spherical neighborhood geometric feature descriptor provided in the foregoing embodiments belong to the same concept. The specific way in which each module and unit performs operations has been described in detail in the method embodiments, and will not be repeated here.

[0141] Another aspect of this application provides an electronic device, including: a controller; and a memory for storing one or more programs, which, when executed by the controller, perform the methods described in the various embodiments above.

[0142] Another aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned method. This computer-readable storage medium may be included in the electronic device described in the above embodiments, or it may exist independently and not incorporated into the electronic device.

[0143] It should be noted that the computer-readable medium shown in the embodiments of this application can be a computer-readable signal medium or a computer-readable storage medium, or any combination of the two. A computer-readable storage medium can be, for example, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this application, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying a computer-readable computer program. The transmitted data signal can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. The computer-readable signal medium can also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The computer program contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to wireless, wired, etc., or any suitable combination thereof.

[0144] Another aspect of this application provides a computer program product or computer program that includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the methods provided in the various embodiments described above.

[0145] According to one aspect of the embodiments of this application, a computer system is also provided, including a Central Processing Unit (CPU), which can perform various appropriate actions and processes based on a program stored in read-only memory (ROM) or a program loaded from storage into random access memory (RAM), such as performing the methods described above. Various programs and data required for system operation are also stored in the RAM. The CPU, ROM, and RAM are interconnected via a bus. Input / output (I / O) interfaces are also connected to the bus.

[0146] For example, a computer system includes a Central Processing Unit (CPU), which can perform various appropriate actions and processes based on programs stored in read-only memory (ROM) or loaded from storage into random access memory (RAM), such as executing the methods described in the above embodiments. The RAM also stores various programs and data required for system operation. The CPU, ROM, and RAM are interconnected via a bus. Input / output (I / O) interfaces are also connected to the bus.

[0147] The following components are connected to the I / O interface: input components including keyboards, mice, etc.; output components including cathode ray tubes (CRTs), liquid crystal displays (LCDs), and speakers; storage components including hard drives; and communication components including network interface cards such as LAN (Local Area Network) cards and modems. The communication components perform communication processing via networks such as the Internet. Drives are also connected to the I / O interface as needed. Removable media, such as disks, optical discs, magneto-optical discs, semiconductor memories, etc., are installed on the drive as needed so that computer programs read from them can be installed into the storage components as required.

[0148] Specifically, according to embodiments of this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program including a computer program for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication component, and / or installed from a removable medium. When the computer program is executed by a central processing unit (CPU), it performs various functions defined in the system of this application.

[0149] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. Each block in a flowchart or block diagram may represent a module, segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0150] The module units described in the embodiments of this application can be implemented in software or hardware, and the described units can also be located in a processor. The names of these units do not necessarily limit the specific unit itself.

[0151] The above embodiments are only used to illustrate this application and are not intended to limit this application. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of this application. Therefore, all equivalent technical solutions also fall within the scope of this application, and the patent protection scope of this application should be defined by the claims.

Claims

1. A method for extracting sub-wires based on adaptive multi-scale spherical neighborhood geometric feature descriptors, characterized in that, The method includes: Acquire scene point cloud data; The scene point cloud data is filtered to obtain non-ground point clouds; Voxel-based height feature analysis is performed on the non-ground point cloud to detect the location of the split conductor and determine the split conductor region. Construct geometric feature descriptors to describe the features of the point cloud within the split conductor region; Establish an adaptive mechanism between center location and neighborhood scale to optimize geometric feature descriptors; Based on the calculation results of the optimized geometric feature descriptor, candidate points for sub-traverses are screened by analyzing the angle between the point cloud distribution direction and the tangent of the line fitting. A nonlinear mapping strategy is used to reduce the dimensionality of the candidate points of the sub-trajectories; The candidate points of the sub-traverse are segmented and reduced in dimension to obtain two-dimensional sub-traverse points; The two-dimensional sub-traverse points are upgraded to three dimensions to achieve the extraction of three-dimensional sub-traverses.

2. The method according to claim 1, characterized in that, Voxel-based height feature analysis is performed on the non-ground point cloud to detect the location of the split traverse and determine the split traverse region, including: The non-ground point cloud is voxelized, and the point cloud distribution characteristics along the height direction within each voxel are statistically analyzed. Based on the point cloud distribution characteristics, a maximum height threshold h is set. max Using a continuous non-empty voxel threshold n, voxels that meet the conditions are selected to detect the location of power towers; Based on the detected center location of the power tower in the scene, gp s (x s ,y s ,z s ), where s represents the s-th tower in the scene, and the angle θ between the line connecting the centers of adjacent power towers and the X-axis is calculated using the following formula: In the formula, x s+1 and y s+1 Let x and y represent the x and y coordinates of the center position of the (s+1)th tower in the scene, respectively; s and y s Let x and y represent the x and y coordinates of the center position of the s-th tower in the scene, respectively. Then, the point cloud is rotated using the following formula to align the direction of the line connecting the centers of adjacent power towers with the X-axis of the coordinate system: In the formula, P(P x ,P y ,P z P represents the point cloud coordinates before rotation. T (P x ,P y ,P z () are the coordinates of the rotated point cloud; Connectivity analysis is used to identify the spatial location of the split conductors in order to determine the region of the split conductors.

3. The method according to claim 1, characterized in that, The geometric feature descriptor is represented as follows: In the formula, Represents geometric feature descriptors, Let P represent the tensor product, C be the point to be determined, R be the neighborhood radius, and P be the tensor product. l Let N be the l-th point in the neighborhood, and N be the total number of points in the neighborhood. Let ω(·) be the normalized distance, and let ω(·) be the radial distance weighting function, expressed as: In the formula, t and m are values ​​optimized through experiments, e is a natural constant, and x is the independent variable of the radial distance weighting function.

4. The method according to claim 3, characterized in that, Establish an adaptive mechanism for center location and neighborhood scale to optimize geometric feature descriptors, including: A weighted mechanism is introduced to differentiate the contribution of each point within the neighborhood; the formula for calculating the weight of each point is as follows: In the formula, Let ρ(P) be the weight of point l. l Let σ be the local density at point l, and G(P) be the Gaussian standard deviation. l () are points l and P MD The Gaussian distance between them, δ, is a parameter used to adjust the degree of influence of local density on the weights; P MD Location is determined by minimizing the distance from each point in the neighborhood to P. MD Distance and determination: In the formula, l is the index of the point, and N is the number of points; P is determined through iterative calculation. MD Position, during iterative calculation, δ is initially set to 1, P MD The initial value is the mean center, and P is continuously updated by adjusting the value of δ. MD Position until P MD The positional change tends to stabilize, so let's take P at this point. MD The final position is C; A basic neighborhood is constructed using the average distance between multiple neighboring points as an initial parameter, and the neighborhood radius is dynamically adjusted using an exponential growth strategy. The adjustment function is as follows: R=Δ mdn ·1.5 q (9) Where R is the neighborhood radius, Δ mdn It is the average distance from multiple adjacent points, where q is a constant. The formula for calculating the linearity involved is as follows: C L =(λ1-λ2) / λ1,λ1>λ2>λ3 (10) In the formula, C L λ1, λ2, and λ3 are the linearity of the point cloud neighborhood; λ1, λ2, and λ3 are the eigenvalues ​​of the point distribution tensor. By comparing the linearity at different scales, the neighborhood scale corresponding to the maximum value is selected as the optimal parameter.

5. The method according to claim 1, characterized in that, Based on the calculation results of the optimized geometric feature descriptor, candidate points for the sub-traverse are screened by analyzing the angle between the point cloud distribution direction and the tangent of the line fitting, including: The transmission line is approximated as a parabola along its span in space. The tangent direction at a certain location is obtained by parabolic fitting. The actual distribution direction of the point cloud calculated by the optimized geometric feature descriptor is compared with the tangent direction. If the angle between the two is less than a preset threshold, it is judged as a candidate point of the sub-conductor.

6. The method according to claim 1, characterized in that, The dimensionality reduction of the candidate points of the sub-traverse is performed using a nonlinear mapping strategy, including: Projecting the split conductor structure onto a horizontal plane yields projected line segments. A central baseline is constructed at the midpoint of these projected line segments, and the distance from each sub-conductor projected line segment to this central baseline is calculated using the following formula: In the formula, d h Let be the distance from the sub-traverse to the center baseline, and a, b, and c be the fitting coefficients of the center baseline, respectively. Projecting the split conductors onto a vertical plane along the span of the line yields a projected parabola. A central reference parabola is constructed at the midpoint of this projected parabola, and the distances from the projected parabolas of each sub-conductor to this central reference parabola are calculated. In the formula, d v P is the vertical distance from the sub-traverse to the central reference parabola, where A, B, and D are the fitting coefficients of the reference parabola. x P y Z represents the coordinates of the projection point on the vertical plane along the span of the track; i To represent the Z-coordinate of a point in the sub-traverse point cloud in three-dimensional space; S i The distance in the plane from the origin or reference point of the coordinate system after projecting the sub-traverse point onto a vertical plane along the span of the line.

7. The method according to claim 1, characterized in that, After segmenting and reducing the dimensionality of the candidate sub-traverse points, we obtain two-dimensional sub-traverse points, including: An adaptive multi-density algorithm based on reverse nearest neighbor is used to segment the dimensionality-reduced candidate points of the sub-traverse. The reverse nearest neighbor is defined as: if q i It is q j If q is one of the nearest neighbors, then q i It is q j The reverse nearest neighbor; The algorithm characterizes local density by the number of reverse nearest neighbors: in dense regions, points with a number of reverse nearest neighbors exceeding a threshold k are defined as core points; in sparse regions, points with a number of reverse nearest neighbors less than k but covered by high-density core points are defined as boundary points, and the rest are noise points. By connecting the core points and boundary points of different density regions using a KNN graph, cross-density growth of clusters is achieved, thus completing the segmentation.

8. A sub-wire extraction device based on an adaptive multi-scale spherical neighborhood geometric feature descriptor, characterized in that, The device includes: The data acquisition module is configured to acquire scene point cloud data; The point cloud filtering module is configured to filter the scene point cloud data to obtain non-ground point clouds; The split conductor identification module is configured to perform voxel-based height feature analysis on the non-ground point cloud to detect the location of the split conductor and determine the split conductor region. The descriptor building module is configured to build geometric feature descriptors to describe the features of the point cloud within the split conductor region; The descriptor optimization module is configured to establish an adaptive mechanism between the center location and the neighborhood scale to optimize the geometric feature descriptor. The sub-traverse screening module is configured to screen candidate points for sub-traverses based on the calculation results of the optimized geometric feature descriptors and by analyzing the angle between the point cloud distribution direction and the tangent of the line fitting. The sub-traverse dimension reduction module is configured to use a nonlinear mapping strategy to reduce the dimension of the candidate points of the sub-traverse; The sub-trajectory segmentation module is configured to segment the dimensionality-reduced candidate points of the sub-trajectory to obtain two-dimensional sub-trajectory points; The sub-trajectory dimensionality upscaling module is configured to upscale the two-dimensional sub-trajectory points to three dimensions, thereby enabling the extraction of three-dimensional sub-trajectory points.

9. An electronic device, characterized in that, The electronic device includes: Memory, used to store computer programs; A processor for executing the computer program to implement the method as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium storing instructions, characterized in that, When the instructions are executed by the processor, the method according to any one of claims 1 to 7 is performed.

Images