Method for intelligent analysis of three-dimensional space form of power transmission tower jumper based on machine learning

Abstract

The present application relates to the technical field of power transmission line monitoring, and discloses a kind of intelligent analysis method of transmission tower jumper three-dimensional space form based on machine learning.The method first obtains the multi-source sensor data stream such as laser radar point cloud data, texture image data and microclimate monitoring data of transmission tower jumper area;Based on laser radar point cloud data, jumper three-dimensional space reference model is constructed, other data is fused therein by space-time registration, and fusion three-dimensional field model is generated;From the model, the key point set of jumper and the corresponding spatial curvature distribution are extracted, and the spatial deviation vector of the form of transmission tower jumper and the preset standard form is calculated;Dynamic risk parameter set is constructed in combination with microclimate monitoring data, and multi-level threshold decision mechanism is triggered to generate form anomaly response instruction.The method realizes accurate evaluation and dynamic risk judgment of jumper form through multi-source data fusion and intelligent analysis, which helps to improve the intelligent level of transmission tower jumper monitoring.

Description

Technical Field

[0001] This invention relates to the field of power transmission line monitoring technology, specifically to a machine learning-based intelligent analysis method for the three-dimensional spatial morphology of power transmission tower jumpers. Background Technology

[0002] As a key component connecting different towers of transmission lines, the spatial form of jumpers directly affects the safe and stable operation of the power system. During long-term operation, jumpers are affected by various factors such as the natural environment and mechanical stress, which can easily lead to abnormal conditions such as deformation and displacement. If these abnormalities are not detected and addressed in time, they may cause serious accidents such as line short circuits and power outages, posing a great threat to the reliability of power supply.

[0003] Traditional jumper wire morphology analysis methods largely rely on manual inspection. Workers observe with the naked eye or use simple tools for measurement, which is not only inefficient but also limited by subjective factors and environmental conditions, making it difficult to accurately capture subtle changes in the jumpers. With the development of sensing technology, monitoring methods based on single sensors have emerged, such as using LiDAR to acquire point cloud data for jumper wire morphology analysis. However, these methods cannot integrate multi-dimensional information such as environmental factors, resulting in insufficient comprehensiveness and accuracy of the analysis results.

[0004] Furthermore, existing analytical methods lack effective spatiotemporal registration mechanisms when processing multi-source data, making it difficult to integrate data collected by different sensors into a unified 3D model, thus affecting the comprehensive assessment of jumper morphology. Simultaneously, in risk assessment, decisions are often made using fixed thresholds, which are ill-suited to complex and changing weather conditions and jumper status, resulting in poor timeliness and targeted responses to anomalies. Therefore, an intelligent analytical method is needed that can integrate multi-source data, construct accurate 3D models, and achieve dynamic risk assessment to improve the efficiency of monitoring the operational status of transmission tower jumpers. Summary of the Invention

[0005] To address the problems existing in the prior art, this invention provides a machine learning-based intelligent analysis method for the three-dimensional spatial morphology of transmission tower jumpers. By constructing an accurate fused three-dimensional field model, it achieves dynamic risk assessment of transmission tower jumpers and improves the efficiency of monitoring the operating status of transmission tower jumpers.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] A machine learning-based intelligent analysis method for the three-dimensional spatial morphology of transmission tower jumpers includes:

[0008] Acquire multi-source sensor data streams for the jumper area of ​​the power transmission tower, including: lidar point cloud data, texture image data, and micro-meteorological monitoring data;

[0009] A three-dimensional spatial reference model for jumpers is constructed based on the lidar point cloud data, and the texture image data and the micro-meteorological monitoring data are fused into the three-dimensional spatial reference model for jumpers using a spatiotemporal registration method to generate a fused three-dimensional field model.

[0010] Extract the jumper key point set and corresponding spatial curvature distribution from the fused 3D field model;

[0011] The key point set of the jumper and the corresponding spatial curvature distribution are input into the morphological comparison and analysis unit to calculate the spatial deviation vector between the jumper morphology of the transmission tower and the preset standard morphology of the jumper of the transmission tower.

[0012] A dynamic risk parameter set is constructed based on the spatial deviation vector and the micrometeorological monitoring data;

[0013] Based on the dynamic risk parameter set, a multi-level threshold decision-making mechanism is triggered to generate an abnormal response command for the morphology of the transmission tower jumper.

[0014] Preferably, the generation process of the fused three-dimensional field model includes:

[0015] The lidar point cloud data is subjected to non-uniform voxelization processing to construct an initial three-dimensional spatial reference model;

[0016] The resolution of the texture image data is compressed using a feature-preserving downsampling algorithm to generate an optimized texture feature map;

[0017] A spatiotemporal transformation matrix is ​​established, and the optimized texture feature map is mapped to the corresponding surface region of the initial three-dimensional spatial reference model through the spatiotemporal transformation matrix;

[0018] The micrometeorological monitoring data is decomposed into wind speed gradient tensor and temperature and humidity influencing factors, and the wind speed gradient tensor is encoded into a spatial vector field, and the temperature and humidity influencing factors are encoded into a scalar field.

[0019] The spatial vector field and the scalar field are embedded into the voxel structure of the initial three-dimensional spatial reference model by tensor splicing to form the fused three-dimensional field model.

[0020] Preferably, extracting the jumper key point set and corresponding spatial curvature distribution from the fused 3D field model includes:

[0021] In the fused three-dimensional field model, a jumper main axis search domain is defined. Curvature extreme points are detected along the jumper main axis search domain to identify jumper suspension points, jumper sag lowest points, and spacer connection points as key points of the transmission tower jumper.

[0022] Construct a local neighborhood sphere centered on the key points, and calculate the eigenvalues ​​of the covariance matrix of the point cloud within the local neighborhood sphere;

[0023] Calculate the curvature value and principal orientation angle of each key point based on the eigenvalues ​​of the covariance matrix;

[0024] The three-dimensional coordinates, curvature values, and principal direction angles of the key points are combined to form the jumper key point set, and the mapping relationship between the key points and their corresponding curvature values ​​is constructed as the spatial curvature distribution of the key points.

[0025] Preferably, calculating the spatial deviation vector between the jumper configuration of the transmission tower and the preset standard configuration of the jumper configuration includes:

[0026] Load a preset standard form database, which contains a set of coordinates of standard jumper key points under ideal working conditions;

[0027] By establishing an elastic point set registration algorithm, the key point set of the transmission tower jumper is non-rigidly matched with the coordinate set of the corresponding standard jumper key points. Curvature similarity constraints are introduced in the non-rigid matching process to minimize the difference between the spatial curvature distribution of the transmission tower jumper and the standard morphological curvature distribution.

[0028] Calculate the spatial position offset of each key point of the transmission tower jumper after non-rigid matching, and arrange all spatial position offsets in the order of key point index to generate the spatial deviation vector of the transmission tower jumper.

[0029] Preferably, constructing a dynamic risk parameter set based on the spatial deviation vector and the micrometeorological monitoring data includes:

[0030] Real-time wind speed components, wind direction angle, and temperature and humidity change rates are extracted from the micrometeorological monitoring data.

[0031] The offset statistical characteristics are calculated based on the spatial deviation vector, including: the maximum offset amplitude and the offset direction consistency coefficient;

[0032] The real-time wind speed component is decomposed into a wind pressure load component perpendicular to the jumper axis.

[0033] A multi-parameter coupling function is constructed, and the wind pressure load component, temperature and humidity change rate, maximum offset amplitude and offset direction consistency coefficient are used as input variables of the multi-parameter coupling function to output the dynamic deformation risk index of the transmission tower jumper.

[0034] The dynamic deformation risk index of the transmission tower jumper is combined with the offset statistical characteristics to form the dynamic risk parameter set of the transmission tower jumper.

[0035] Preferably, the generation of a morphological anomaly response command for the transmission tower jumper based on the multi-level threshold decision-making mechanism triggered by the dynamic risk parameter set includes:

[0036] Multiple preset judgment thresholds are included, including: basic warning threshold, medium risk threshold and emergency response threshold;

[0037] The multi-level judgment thresholds are adjusted based on the historical abnormality records of the transmission tower jumpers.

[0038] The dynamic deformation risk index of the transmission tower jumper in the dynamic risk parameter set is compared step by step with the corrected multi-level judgment threshold:

[0039] When the dynamic deformation risk index exceeds the modified basic warning threshold, the first-level monitoring enhancement command is activated.

[0040] When the dynamic deformation risk index exceeds the modified moderate risk threshold, a level-two early warning notification instruction is activated;

[0041] When the dynamic deformation risk index exceeds the modified emergency response threshold, a Level 3 emergency response command is activated.

[0042] Integrate the current activation command status to generate an abnormal response command.

[0043] Preferably, a process for tracing the status of key points of transmission tower jumpers is also provided:

[0044] In the fused 3D field model, a spatiotemporal index table of jumper key points is established to record the spatial coordinates and curvature values ​​of each jumper key point at different timestamps.

[0045] Based on the spatiotemporal index table of jumper key points, a key point motion trajectory function is constructed to calculate the displacement acceleration of key points between adjacent timestamps.

[0046] When the dynamic deformation risk index of the transmission tower jumper in the dynamic risk parameter set exceeds the basic warning threshold, the abnormal key point identifier with displacement acceleration greater than the acceleration threshold is extracted.

[0047] The anomaly key point identifier is associated with the spatial deviation vector.

[0048] Preferably, a decision buffer mechanism is also included:

[0049] A boundary buffer layer with multiple judgment thresholds is set, and the boundary buffer layer is the safe floating range of the dynamic deformation risk index of the transmission tower jumper.

[0050] When the dynamic deformation risk index enters the boundary buffer layer, the triggering of the abnormal shape response command of the transmission tower jumper is delayed;

[0051] The duration of the dynamic deformation risk index within the boundary buffer layer is monitored. If the duration exceeds the preset tolerance period and the dynamic deformation risk index does not fall below the safety boundary, the corresponding level of morphological anomaly response command is forcibly triggered.

[0052] Preferably, a quantification process for the impact of micrometeorological monitoring data is also provided:

[0053] Establish a wind speed-deformation transfer function, and calculate the theoretical wind-induced vibration amplitude based on the real-time wind speed component and jumper tension parameter in the micrometeorological monitoring data;

[0054] Extract the actual vibration amplitude from the spatial deviation vector;

[0055] Calculate the residual between the theoretical wind-induced vibration amplitude and the actual vibration amplitude;

[0056] When the residual exceeds the residual threshold, the temperature and humidity influence factors are weighted and injected into the multi-parameter coupling function to recalculate the dynamic deformation risk index of the transmission tower jumper.

[0057] Preferably, an execution verification process for the abnormal morphological response command of the transmission tower jumper is also provided: the corresponding jumper morphological retest process is started according to the type of the abnormal morphological response command of the transmission tower jumper;

[0058] During the retesting process, the key point set and spatial curvature distribution of the jumper line at the current moment are reacquired.

[0059] Calculate the spatial consistency coefficient between the retested jumper key point set and the original jumper key point set;

[0060] When the spatial consistency coefficient is lower than the verification threshold, a sensor data verification command is triggered;

[0061] When the spatial consistency coefficient meets the verification requirements, it is confirmed that the abnormal morphological response command of the transmission tower jumper is valid.

[0062] Compared with the prior art, the beneficial effects of the present invention are:

[0063] (1) By acquiring multi-source sensor data streams, including lidar point cloud data, texture image data, and micro-meteorological monitoring data, comprehensive collection of jumper morphology and surrounding environmental information was achieved. Compared with analysis methods that rely on a single data source, the introduction of multi-source data can characterize the spatial features and operating environment of transmission tower jumpers from different dimensions, making the subsequent morphological analysis of transmission tower jumpers more rich and hierarchical;

[0064] (2) The three-dimensional spatial benchmark model of jumpers constructed based on lidar point cloud data provides a precise spatial framework for the digital presentation of jumper morphology. By using a spatiotemporal registration method, texture image data and micro-meteorological monitoring data are fused into this benchmark model to generate a fused three-dimensional field model. This breaks down the spatiotemporal barriers between different types of data and realizes the organic integration of multi-source information in a unified three-dimensional space. This integration method not only preserves the characteristics of each data, but also, through the correlation and interaction between data, can uncover potential information that cannot be reflected by a single data, laying the foundation for in-depth analysis of jumper morphology.

[0065] (3) The key point set of jumpers and the corresponding spatial curvature distribution extracted from the fusion three-dimensional field model can accurately capture the key features and changing trends of jumper morphology. The key point set of jumpers and the corresponding spatial curvature distribution are input into the morphology comparison analysis unit to obtain the spatial deviation vector between the jumper morphology of the transmission tower and the preset standard morphology of the jumper of the transmission tower. The degree of abnormality of the jumper morphology is reflected intuitively through the quantified deviation index, avoiding the subjective ambiguity in traditional qualitative analysis, making the morphology assessment more objective and accurate.

[0066] (4) The dynamic risk parameter set constructed based on spatial deviation vector and micro-meteorological monitoring data comprehensively considers the morphological changes of jumpers and environmental influencing factors, enabling the risk assessment to fully reflect the actual operating status of jumpers. According to the multi-level threshold decision-making mechanism triggered by this dynamic risk parameter set, corresponding morphological anomaly response instructions can be generated according to different risk levels. Compared with fixed threshold decision-making, this dynamic decision-making method can flexibly adapt to the state changes of jumpers under different environmental conditions, making the anomaly response more in line with actual needs and enhancing the pertinence of handling jumper anomalies. Attached Figure Description

[0067] Figure 1 This is a schematic diagram illustrating the working principle of the intelligent analysis method for the three-dimensional spatial morphology of transmission tower jumpers based on machine learning in this invention.

[0068] Figure 2 A flowchart generated for integrating a 3D field model;

[0069] Figure 3 A flowchart for calculating the spatial deviation vector between the jumper configuration of a transmission tower and the preset standard configuration of the jumper configuration;

[0070] Figure 4 A flowchart for constructing a dynamic risk parameter set;

[0071] Figure 5 This is a flowchart for tracing the status of key points on the jumper wires of a power transmission tower. Detailed Implementation

[0072] The technical solutions of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0073] Please see Figure 1 This invention provides a machine learning-based intelligent analysis method for the three-dimensional spatial morphology of power transmission tower jumpers, comprising the following steps:

[0074] The multi-source sensor data stream of the jumper area of ​​the transmission tower is acquired. The multi-source sensor data stream includes lidar point cloud data collected in real time by a laser scanner installed on the transmission tower, texture image data collected in real time by a high-resolution camera installed on the transmission tower, and micro-meteorological monitoring data collected in real time by a meteorological sensor installed on the transmission tower.

[0075] The collected lidar point cloud data, texture image data, and micro-meteorological monitoring data are transmitted to the central processing unit via a 5G network. In the central processing unit, a jumper three-dimensional spatial reference model is constructed based on the lidar point cloud data. The texture image data and micro-meteorological monitoring data are fused into the three-dimensional spatial reference model through a spatiotemporal registration method to generate a fused three-dimensional field model.

[0076] Extract the key point set of jumpers and their corresponding spatial curvature distribution from the fused 3D field model;

[0077] Input the key point set of the jumper and the corresponding spatial curvature distribution into the morphological comparison and analysis unit to calculate the spatial deviation vector between the jumper morphology of the transmission tower and the preset standard morphology of the jumper of the transmission tower.

[0078] A dynamic risk parameter set is constructed based on spatial deviation vector and micro-meteorological monitoring data. The dynamic risk parameter set is stored and updated in tensor form to support multi-level decision-making processes.

[0079] Based on the dynamic risk parameter set, a multi-level threshold decision-making mechanism is triggered to generate an abnormal response command for the power transmission jumper configuration.

[0080] The present invention provides a machine learning-based intelligent analysis method for the three-dimensional spatial morphology of transmission tower jumpers. The method is executed on an edge server using a distributed computing framework to ensure low-latency processing.

[0081] In one embodiment of the present invention, a specific scheme is provided for constructing a three-dimensional spatial reference model of jumpers based on lidar point cloud data, and fusing texture image data and micro-meteorological monitoring data into the three-dimensional spatial reference model through a spatiotemporal registration method to generate a fused three-dimensional field model, such as... Figure 2 The specific process is as follows:

[0082] The lidar point cloud data is subjected to non-uniform voxelization processing to construct an initial three-dimensional spatial reference model, specifically:

[0083] Read the lidar point cloud data of the jumper area of ​​the power transmission tower collected by the laser scanner. The lidar point cloud data includes three-dimensional coordinate information and reflection intensity value.

[0084] Based on the distribution density of lidar point cloud data in the jumper area of ​​the transmission tower, the voxel size adjustment rules are set: a voxel side length of 0.05 meters is used in dense point cloud areas such as jumper suspension points and spacers, and a voxel side length of 0.1 meters is used in relatively sparse point cloud areas such as the middle of the jumper sag.

[0085] Statistical analysis is performed on the lidar point cloud data within each voxel. The mean three-dimensional coordinates and mean reflection intensity of all points within each voxel are calculated. Noise points with a mean reflection intensity lower than a preset threshold are removed, and valid point cloud data are retained.

[0086] Using the retained voxels as basic units, an initial three-dimensional spatial reference model is constructed according to the spatial positional relationship. In the initial three-dimensional spatial reference model, each voxel is associated with the mean value of the corresponding point's three-dimensional coordinates and the mean value of the reflection intensity. After non-uniform voxelization processing, the coordinate information refers to the mean value of the three-dimensional coordinates of all lidar points within the voxel.

[0087] The feature-preserving downsampling algorithm is used to compress the resolution of texture image data and generate an optimized texture feature map. Specifically:

[0088] The texture image data of the jumper area of ​​the power transmission tower captured by a high-resolution camera is obtained. The texture image data contains RGB three-channel color information. The key feature areas such as jumper edges and spacer outlines in the texture image data are identified by the edge detection algorithm, and the pixel coordinates in the key feature areas are marked.

[0089] Based on the image resolution compression requirements, target resolution parameters are set. In non-critical feature areas, mean sampling is used to reduce pixel density, while in critical feature areas, neighborhood pixel interpolation is used to preserve pixel details and avoid loss of feature information.

[0090] Color correction and noise filtering are performed on the processed texture image data, the grayscale value range of the RGB channels is adjusted, salt-and-pepper noise and Gaussian noise in the texture image data are removed, and an optimized texture feature map is generated. This optimized texture feature map reduces the amount of data by more than 60% compared with the original texture image data while maintaining key texture features.

[0091] A spatiotemporal transformation matrix is ​​established, which maps the optimized texture feature map to the corresponding surface region of the initial 3D spatial reference model. Specifically:

[0092] The spatial coordinate reference of the initial 3D spatial reference model and the acquisition timestamp of the laser scanner are extracted. The image coordinate reference of the optimized texture feature map and the acquisition timestamp of the high-resolution camera are also extracted. The spatial coordinate reference refers to the 3D coordinate system of the initial 3D spatial reference model. This 3D coordinate system is established with a fixed reference point in the transmission tower jumper area as the origin, forming a 3D rectangular coordinate system composed of X, Y, and Z axes. The position of each voxel in the initial 3D spatial reference model is determined by the coordinate values ​​under this coordinate system, which is the reference for locating the spatial position of the transmission tower jumper. The image coordinate reference refers to the 2D pixel coordinate system of the optimized texture feature map. It is usually with the upper left corner of the optimized texture feature map as the origin, with the horizontal direction to the right as the U axis and the vertical direction downward as the V axis. The position of each pixel in the optimized texture feature map is represented by (U,V) pixel coordinates, which is the reference for locating the jumper texture position in the image.

[0093] Based on the difference in the collection timestamps of the two types of data, the time offset is calculated, and the sampling timestamps of the optimized texture feature map are calibrated to ensure that the two are consistent in the time dimension.

[0094] In the spatial dimension, three fixed feature points are selected within the jumper area of ​​the transmission tower. Their three-dimensional coordinates in the initial three-dimensional spatial reference model and their two-dimensional pixel coordinates in the optimized texture feature map are obtained respectively. Based on these corresponding point coordinates, a set of spatial transformation equations is constructed, and the spatiotemporal transformation matrix is ​​obtained by solving them.

[0095] This spatiotemporal transformation matrix maps each pixel in the optimized texture feature map to the corresponding surface region of the initial three-dimensional spatial reference model, so that the model surface presents texture information consistent with the actual jumper.

[0096] The micrometeorological monitoring data is decomposed into wind speed gradient tensor and temperature and humidity influencing factors. The wind speed gradient tensor is encoded as a spatial vector field, and the temperature and humidity influencing factors are encoded as a scalar field. Specifically:

[0097] Wind speed and temperature / humidity data were separated from micro-meteorological monitoring data collected by meteorological sensors.

[0098] The wind speed data is decomposed into three dimensions to calculate the components of wind speed on the three spatial axes of X, Y, and Z. The wind speed gradient tensor is constructed by combining the spatial location information. The wind speed gradient tensor contains the wind speed magnitude and direction information at different spatial locations. The wind speed gradient tensor is encoded into a spatial vector field. Each vector in the spatial vector field corresponds to the wind speed gradient information at a specific spatial location. The direction of the vector represents the wind speed direction, and the length of the vector represents the wind speed magnitude.

[0099] The temperature and humidity data are quantified and converted into standardized values ​​to construct a temperature and humidity influence factor. This factor reflects the temperature and humidity status of the transmission tower jumper area at different times. The temperature and humidity influence factor is encoded into a scalar field, and each scalar value in the scalar field corresponds to the temperature and humidity information at a specific spatial location and time point.

[0100] The spatial vector field and scalar field are embedded into the voxel structure of the initial three-dimensional spatial reference model through tensor splicing to form a fused three-dimensional field model. Specifically:

[0101] Read the voxel structure data of the initial three-dimensional spatial reference model and determine the spatial coordinate range of each voxel;

[0102] The encoded spatial vector field and scalar field are matched with the voxels of the initial three-dimensional spatial reference model according to the spatial coordinate correspondence.

[0103] For each voxel, the corresponding spatial vector field data and scalar field data are tensor-concatenated to form a composite tensor containing three-dimensional coordinates, reflection intensity, texture features, wind speed gradient, temperature and humidity information;

[0104] By embedding composite tensors into the voxel structure of the initial 3D spatial benchmark model and updating the voxel structure data, each voxel integrates multi-source data information, ultimately forming a fused 3D field model. This 3D field model can intuitively present the 3D spatial morphology, surface texture features, and surrounding micro-meteorological environment of the transmission tower jumpers, providing comprehensive data support for subsequent jumper morphology analysis.

[0105] In one embodiment of the present invention, a specific scheme for extracting the jumper key point set from a fused three-dimensional field model is provided, the process of which is as follows:

[0106] In the fused 3D field model, the main axis search domain of the jumper is first defined based on the spatial distribution characteristics of the jumper wires of the transmission tower: by identifying the connection position of the transmission tower in the fused 3D field model, the starting and ending points of the jumper wires of the transmission tower are determined. Taking the line connecting the two points as the central axis, a cuboid region is expanded outward to form the main axis search domain. The expansion range is set according to the design span and possible deformation range of the jumper wires of the transmission tower to ensure that the complete spatial distribution of the jumper wires of the transmission tower is covered.

[0107] Curvature calculation is performed on the lidar point cloud data within the fused 3D field model along the main axis search domain: A local surface fitting method is used. The lidar point cloud data within the jumper main axis search domain is divided into sliding windows with a fixed step size. A quadratic surface is fitted within each window. When fitting the quadratic surface within each window, the 3D coordinate data of all lidar point cloud data within that sliding window is first obtained. These data all come from the effective point cloud within the jumper main axis search domain of the fused 3D field model. Next, a general quadratic surface equation that accurately describes the local surface curvature characteristics of the transmission tower jumper is selected as the fitting model. This fitting model can adapt to the surface morphology of different parts of the transmission tower jumper. Subsequently, the least squares method is used, with the optimization objective of minimizing the sum of squared distances from each lidar point cloud data within the window to the quadratic surface. The coefficients of each term in the general quadratic surface equation are solved. The optimal coefficients are determined through multiple iterations, and finally, the fitting of the quadratic surface within each window is completed. The curvature value is then solved using the optimal general quadratic surface equation.

[0108] Traverse all windows for curvature values, and mark points with curvature values ​​significantly higher than the surrounding areas as curvature extrema. Among them, extrema points located at both ends of the main axis search domain and close to the transmission tower connection position are determined as jumper suspension points; points located in the middle of the main axis search domain with curvature values ​​of local minimum are determined as jumper sag minimum points; points located between suspension points and sag minimum points with curvature values ​​showing periodic small fluctuations are determined as spacer bar connection points. The jumper suspension points, jumper sag minimum points, and spacer bar connection points of the transmission tower are combined to form the jumper key point set.

[0109] In one embodiment of the present invention, in the voxel structure of the fused three-dimensional field model, each voxel is associated with the average reflection intensity information retained after the LiDAR point cloud data has undergone non-uniform voxelization processing. Since the main body of the transmission tower jumper is mostly made of metal, its reflection intensity of the laser scanner signal is significantly higher than that of surrounding background debris such as vegetation, non-jumper components of the tower, and suspended particulate matter in the air. Before conducting curvature extremum point detection, preliminary screening can be performed based on the average reflection intensity of the voxels: a reflection intensity threshold that conforms to the characteristics of the metal material of the transmission tower jumper is set, and curvature calculation and extremum point detection are performed only on the LiDAR point cloud data within voxels whose average reflection intensity is higher than the reflection intensity threshold. Through this operation, point cloud interference from background debris can be effectively filtered out, ensuring that the identified curvature extremum points all originate from the transmission tower jumper body, laying the foundation for accurately extracting the transmission tower jumper suspension point, the lowest point of jumper sag, and the spacer connection point.

[0110] In one embodiment of the present invention, the surface of the fused three-dimensional field model has been mapped with an optimized texture feature map using a spatiotemporal registration method. The optimized texture feature map fully preserves the texture details of the transmission tower jumpers and related components. The spacer, as a key component connecting and fixing the transmission tower jumpers, has significantly different materials and surface structures from the jumper body, resulting in different texture features. The jumper body's texture is mostly continuous and uniform with a metallic luster, while the spacer's texture exhibits a discontinuous and non-uniform texture. After locating suspected spacer connection points through curvature extremum point detection, the optimized texture features of the suspected area can be extracted and compared with the texture features of the jumper body. If the texture features of the suspected area are significantly different from those of the jumper body and match the texture feature attributes of the spacer, then the point can be further confirmed as a spacer connection point, providing supplementary evidence for the accurate identification of key jumper points and avoiding potential positioning errors that may occur when relying solely on curvature detection.

[0111] In one embodiment of the present invention, a method for determining the spatial curvature distribution corresponding to a set of key points is also provided, the specific process of which is as follows:

[0112] A local neighborhood sphere is constructed with each key point as the center. The radius of the sphere is dynamically adjusted according to the type of key point: the point cloud density around the suspension point and the connection point of the spacer bar is relatively high, and the radius is set to 0.1 meters; the point cloud around the lowest point of the jumper arc is relatively sparse, and the radius is set to 0.15 meters, to ensure that each sphere contains a sufficient amount of lidar point cloud data to support subsequent calculations.

[0113] The covariance matrix of the LiDAR point cloud data within each local neighborhood sphere was calculated using principal component analysis. Then, eigenvalues ​​were obtained by eigenvalue decomposition, resulting in three eigenvalues ​​ordered from largest to smallest. , , Calculate the curvature value of the key point based on these three eigenvalues. Simultaneously extract the largest eigenvalue. The eigenvector direction is converted into the principal direction angle in a spatial coordinate system. The three-dimensional coordinates of the key points, the calculated curvature values, and the principal direction angles are integrated, and a spatial curvature distribution of the transmission tower jumper is formed by establishing a correspondence table between the unique identifier of the key points and the curvature values.

[0114] In one embodiment of the present invention, a calculation process for the spatial deviation vector between the jumper configuration of the transmission tower and the preset standard configuration of the jumper is also provided, such as... Figure 3 It includes the following steps:

[0115] Load the preset standard form database, which stores the standard form data of transmission tower jumpers of different models and under different working conditions. Each standard form data contains the coordinate set of key points of the standard jumper under ideal working conditions of standard temperature 25℃, windless environment and rated tension, and the type and number of key points of the standard jumper are consistent with the key points actually extracted.

[0116] Based on the current analysis of the transmission tower jumper model, design parameters, and installation environment, the corresponding standard jumper key point coordinate set is matched from a pre-set standard form database. Specifically, a flexible point set registration algorithm is used to perform non-rigid matching between the transmission tower jumper key point set and the standard jumper key point coordinate set.

[0117] During the initialization phase, the key point set of the transmission tower jumper is initially translated and rotated with reference to the standard key point coordinate set so that the overall spatial position of the two is roughly aligned.

[0118] During the matching process, a deformation model is constructed using a thin-plate spline interpolation function. By adjusting the control parameters of the interpolation function, the key points of the transmission tower jumpers gradually approach the standard key points. Simultaneously, curvature similarity constraints are introduced. Based on minimizing the spatial positional deviation between the set of key points of the transmission tower jumpers and the set of key points of the standard jumpers, a difference term is added between the curvature values ​​of the key points of the transmission tower jumpers and the curvature values ​​of the standard key points. Then, weight coefficients are set according to the degree of influence of curvature on the jumper shape. For example, the curvature of the lowest point of the jumper sag has a greater impact on the jumper shape, so its difference term weight coefficient is set to a higher value, while the weight coefficient of the spacer bar connection point is set to a lower value. This ensures that the registration process not only achieves spatial position matching but also guarantees the consistency of the curvature distribution characteristics of the two, avoiding deviations in morphological characteristics due to simply pursuing positional matching.

[0119] After completing the non-rigid matching, for each key point of the transmission tower jumper, the difference between its three-dimensional coordinates and the three-dimensional coordinates of the corresponding standard key point is calculated to obtain the spatial position offset of each key point of the transmission tower jumper, including the offset values ​​in the X, Y, and Z axes. According to the index order uniquely identified by the key points in the jumper key point set, the spatial position offsets of all key points are arranged sequentially to form a multi-dimensional vector. This multi-dimensional vector is the spatial deviation vector between the shape of the transmission tower jumper and the preset standard shape of the transmission tower jumper, which can intuitively reflect the degree and directional distribution of deformation of the transmission tower jumper at different positions.

[0120] In one embodiment of the present invention, a specific scheme for constructing a dynamic risk parameter set based on spatial deviation vectors and micrometeorological monitoring data is also provided, such as... Figure 4 The specific process is as follows:

[0121] The wind speed in the micrometeorological monitoring data is decomposed into three orthogonal wind speed components: X, Y, and Z. With due north as the 0-degree reference, the wind direction angle is obtained from the arctangent function of the horizontal wind speed component. The time series of temperature and humidity change rate compared with the historical average is calculated based on the temperature and humidity influencing factors. The sliding window difference algorithm is used, with the window size set to 5 minutes, to output the rate of change of temperature and humidity per minute.

[0122] The statistical characteristics of transmission tower jumper offset are calculated based on spatial deviation vectors, including: maximum offset amplitude and offset direction consistency coefficient. The maximum offset amplitude is determined by traversing the spatial deviation vectors of all key points of the transmission tower jumper and selecting the spatial deviation vector with the maximum Euclidean distance. The offset direction consistency coefficient is obtained based on vector field divergence analysis. Using the set of key points extracted from the fused 3D field model, the spatial offset of each key point is converted into a 3D direction vector. These 3D direction vectors together constitute a vector field describing the overall offset trend of the transmission tower jumper. The direction of each 3D direction vector corresponds to the offset direction of the key point, and the magnitude is uniformly normalized to 1 to eliminate the influence of the offset amplitude. For each key point of the transmission tower jumper, a predetermined number of neighboring key points are selected. The dot product of the direction vectors of the key point and its neighboring key points is calculated to quantify the degree of coordination of the offset directions within a local area. The closer the dot product value is to 1, the more consistent the offset directions of the key point and its neighboring key points are; a dot product value close to 0 indicates a significant difference in direction. Based on the above local dot product calculation results, a weighted average is used to obtain the local orientation consistency index of each transmission tower jumper key point. The weight allocation is related to the curvature value of the transmission tower jumper key point. The larger the curvature value, the more sensitive the transmission tower jumper key point is to shape changes, and the higher the weight. Subsequently, the local orientation consistency indices of all key points are globally integrated to finally generate the transmission tower jumper offset direction consistency coefficient, with a value range of [0,1]. The closer the value is to 1, the more uniform the overall offset direction.

[0123] The wind speed component is decomposed into wind pressure load components perpendicular to the axis of the transmission tower jumper: the wind speed components in the X and Y directions are the components perpendicular to the axis of the transmission tower jumper. Since the magnitude of the wind pressure load is directly related to the wind speed component perpendicular to the surface of the object, for the transmission tower jumper, the direction mainly affected by wind load is the plane perpendicular to its own axis, i.e., the XY plane. Therefore, it is necessary to calculate the resultant wind speed in this plane using the X and Y components. The magnitude of the resultant wind speed directly determines the intensity of the wind pressure load. Combining this with the aerodynamic principle that wind pressure is proportional to the square of wind speed, the wind pressure load component perpendicular to the axis of the transmission tower jumper is obtained, thereby quantifying the actual load impact of wind on the transmission tower jumper.

[0124] A multi-parameter coupling function is constructed, using wind pressure load components, temperature and humidity change rates, maximum offset amplitude, and offset direction consistency coefficient as input variables. The output is a dynamic deformation risk index for the transmission tower jumper. The construction process of the multi-parameter coupling function is as follows:

[0125]

[0126] In the formula, The dynamic deformation risk index of the output transmission tower jumper is dimensionless. This represents the wind pressure load component, with units of Newtons per square meter. This represents the rate of change in temperature and humidity and is dimensionless. This represents the maximum offset magnitude and is dimensionless. This represents the consistency coefficient in the offset direction. This indicates the maximum permissible wind pressure threshold for this type of transmission tower jumper; This represents the temperature sensitivity adjustment coefficient, set to 1.5; This indicates that the hyperbolic tangent function compresses the range of influence of temperature and humidity. express Weighting factors express Weighting factors express Weighting factors express The weighting factor; the formula is set by... and The nonlinear transformation is used to amplify the effects of abnormal offset and directional dispersion of transmission tower jumpers, and to... Stored as floating-point numbers, with precision retained to three decimal places;

[0127] The dynamic deformation risk index of transmission tower jumpers is combined with the statistical characteristics of transmission tower jumper offset to form a dynamic risk parameter set for transmission tower jumpers.

[0128] In one embodiment of the present invention, during the conversion of wind speed components into wind pressure load components perpendicular to the jumper axis, the process first relies on the spatial vector field corresponding to the wind speed gradient tensor encoded in the fused three-dimensional field model. This spatial vector field is the carrier of the wind speed spatial distribution information provided by the wind speed gradient tensor, containing wind speed magnitude and direction data at different spatial locations in the transmission tower jumper area. Based on the specific spatial coordinates of each part of the transmission tower jumper in the fused three-dimensional field model, the real-time wind speed data of the corresponding location is accurately extracted from the aforementioned spatial vector field, thereby obtaining the real-time wind speed components of different parts of the transmission tower jumper, ensuring that the wind speed data of each part matches its actual wind environment. Next, combined with the determined direction of the jumper main axis in the fused three-dimensional field model, the axial direction of each part of the transmission tower jumper is determined, and the real-time wind speed components extracted from each part are vector decomposed to separate the wind speed components parallel to the jumper axis and the wind speed components perpendicular to the jumper axis. Since the effect of wind pressure load is mainly related to the wind speed perpendicular to the jumper axis, the wind speed components perpendicular to the jumper axis of each part are screened out. Then, based on the correlation characteristics between wind pressure load and wind speed perpendicular to the object axis, the real-time wind speed components perpendicular to the jumper axis of each part are converted into wind pressure load data of the corresponding part. Finally, the wind pressure load data perpendicular to the axis of all parts of the jumper are integrated to obtain the complete wind pressure load components perpendicular to the jumper axis.

[0129] In one embodiment of the present invention, the multi-level threshold decision mechanism adopts a hierarchical triggering architecture, and the initial value setting of the multi-level judgment threshold is based on engineering specifications: basic warning threshold. moderate risk threshold Emergency response threshold Threshold adaptive adjustment relies on a historical anomaly database: querying risk event records of similar transmission tower jumpers over the past 30 days, and calculating the peak risk index for each risk event. Perform a Gaussian-weighted average to generate a new threshold:

[0130]

[0131] in: express , or , Indicates the new threshold , or , This represents the historical forgetting factor, set to 0.7. This represents the severity weight of risk event i.

[0132] The new threshold is automatically revised every 24 hours. Dynamic deformation risk index of transmission tower jumpers. The comparison with the new threshold follows a step-by-step logic: when At that time, the Level 1 monitoring enhancement command is activated, which increases the scanning frequency of the laser scanner from 10Hz to 30Hz and the image sampling rate from 1fps to 5fps; when When the time comes, the Level II early warning notification command is activated, generating a JSON-formatted alarm message containing the key locations of transmission tower jumpers, offset vectors, and real-time meteorological data, which is then pushed to the regional monitoring center via the MQTT protocol; when At that time, the Level 3 emergency response command is activated, triggering the pre-tensioning procedure of the transmission tower jumper tension adjustment device, and simultaneously sending an SMS alarm to the maintenance personnel's mobile terminal.

[0133] In one embodiment of the present invention, a key point status traceability process for transmission tower jumpers is also provided, such as... Figure 5 The specific process is as follows:

[0134] A spatiotemporal index table of key points for transmission tower jumpers was established within the fused 3D field model. This index table is a relational data structure containing six core fields: a timestamp accurate to milliseconds, a key point ID represented by a unique identifier, the 3D spatial coordinates of the key point, the curvature value of the key point, displacement acceleration, and anomaly marker. Table 1 shows the key point state segments of a certain transmission tower jumper at three consecutive time points.

[0135] Table 1 Spatiotemporal Status Table of Key Jumper Points

[0136]

[0137] At time 082003, the initial states of keypoints KP1001 and KP1002 are recorded, with anomaly markers of 0, indicating a normal state. As time progresses, a keypoint motion trajectory function is constructed based on a spatiotemporal index table. This function uses a piecewise cubic Hermitian interpolation algorithm to connect keypoints with the same ID at adjacent timestamps. Taking keypoint KP1002 as an example: based on the coordinates (128.563, 384.219, 98.761) at time 082003 and (128.592, 384.201, 98.732) at time 082013, the displacement change between the two points is calculated and divided by the time interval of 10 seconds to obtain the instantaneous velocity; then, combined with the data at time 082023, the rate of change of velocity is calculated, ultimately deriving the displacement acceleration value.

[0138] When the dynamic deformation risk index reaches 0.65, exceeding the basic warning threshold of 0.6, the spatiotemporal index table is retrieved. An acceleration threshold of 0.8 m / s² is set, and the acceleration records of all key points before and after the risk exceedance time are scanned. As shown in Table 1, KP1002's acceleration of 0.81 m / s² at time 082013 exceeds the threshold, and its ID is added to the abnormal key point set. The abnormal key point set is stored using a hash table and associated with the current spatial deviation vector: in the vector data structure describing the offset of each key point, an abnormal marker bit is added for KP1002, and its acceleration exceedance value of 1.05 m / s² is recorded. When the risk parameter set is transmitted to the decision module, the abnormal key point information is transmitted synchronously with the offset, allowing risk analysis to focus on specific problem areas.

[0139] The decision buffer mechanism relies on the defined boundary buffer layer. A safe floating range [0.6, 0.63] is set based on a baseline warning threshold of 0.6. When the dynamic deformation risk index enters this range, a delayed response procedure is triggered: the generation of the morphological anomaly response command is suspended, a 500ms timer is started, and the risk index is added to the circular buffer monitoring queue. A state tracker is set within the buffer to continuously record the index's change trajectory and dwell time.

[0140] Consider a real-world scenario: In strong winds, the risk index first reaches 0.605 at 08:30:15, entering a buffer zone. Timing begins, recording index changes: it rises to 0.618 at 08:30:17, falls back to 0.602 at 08:30:20, and rises again to 0.625 at 08:30:23. When the dwell time reaches 8 minutes, less than the preset tolerance period of 10 minutes, the index remains at 0.622 at 08:38:15. At this point, a mandatory response logic is activated: based on the current index value of 0.622 exceeding the basic threshold but not reaching the moderate risk threshold, a Level 1 monitoring enhancement command is activated. This command immediately increases the lidar scanning frequency to 30Hz and simultaneously appends all key point status change data recorded during the buffer period to the response message.

[0141] In another scenario, the risk index entered the buffer layer at 09:15:00, but dropped back to 0.592 at 09:18:30. The buffer timer was then cleared without triggering any commands; the fluctuation event was only recorded in the log. The buffer mechanism uses a state machine to implement state transitions: initially in the "monitoring state," entering the buffer layer transitions to the "delayed state," failing to drop back within the timeout transitions to the "forced response state," and returning to the "monitoring state" upon falling back to the safe zone. State transition data is written to an independent register for analysis by the fault diagnosis module. The entire process does not interrupt the main business process; monitoring and timing operations are performed only in a background thread.

[0142] In one embodiment of the present invention, a quantification process for the impact of micrometeorological monitoring data is also provided:

[0143] A wind speed-deformation transfer function is established, which is derived based on the vibration equation of overhead conductors: inputting the three-dimensional vector of real-time wind speed components, calling the pre-stored jumper material parameter library to obtain the tension coefficient, unit mass and length; the calculation process integrates the Strocha number model in aerodynamics to simulate the vortex shedding frequency of the conductor wake under specific wind speeds; and deriving the theoretical wind-induced vibration amplitude by combining the natural frequency characteristics of the transmission tower jumper.

[0144] The standard deviation of the Z-axis offset of all key points is taken from the spatial deviation vector to eliminate the influence of static deviations such as temperature expansion. After median filtering, the pure dynamic vibration component is derived as the actual vibration amplitude.

[0145] The residual between the theoretical wind-induced vibration amplitude and the actual vibration amplitude is calculated using the relative error rate.

[0146] When the residual exceeds the preset residual threshold, it is determined that temperature and humidity factors significantly interfere with deformation. The current temperature and humidity influencing factors are extracted from the micro-meteorological monitoring data, and weighting is applied to the factor. The weighting value is adjusted non-linearly according to the residual. When the residual is in the range of 15%-30%, a linear weight is taken, and when it is higher than 30%, an exponential weight curve is switched. At the same time, the temperature change trend is monitored. If there is a sudden rise or fall within 24 hours, an additional sensitivity boost is applied.

[0147] The compensated temperature and humidity factor is injected as an independent variable into the multi-parameter coupling function, replacing the original temperature and humidity change rate parameter.

[0148] The scale of all input parameters of the multi-parameter coupled function is reset, the weight allocator dynamically adjusts the influence ratio of temperature and humidity factors according to the residual, and the final output revised dynamic deformation risk index covers the original value and updates the dynamic risk parameter set.

[0149] In one embodiment of the present invention, the abnormal morphological response command of the transmission tower jumper executes a closed-loop verification process:

[0150] After generating the disposal command, the morphological retesting process of the transmission tower jumper is initiated: for Level 1 monitoring enhancement commands, a single retest is performed after 10 seconds of frequency-enhanced sampling; for Level 2 warning commands and above, a high-precision scanning mode is immediately invoked for retesting. During the retest, lidar point cloud and texture image data are simultaneously acquired, and the key point set of the new jumper is quickly extracted through fusion of the 3D field model.

[0151] The comparison between the old and new keypoint sets employs a spatial consistency evaluation algorithm: a weighted proximity mapping relationship is constructed between the two sets of point sets, and a spatial correlation index is calculated based on the point-pair spacing. This spatial correlation index integrates three metrics, including: the mean positional deviation characterizes the overall morphological change, the maximum offset reflects the degree of local abrupt change, and the Hausdorf distance of the point sets evaluates the boundary consistency.

[0152] When the spatial consistency index falls below the verification threshold, a sensor data verification command is activated. The verification command is executed in three levels: The initial verification re-acquires the original lidar spot data and verifies signal distortion through echo intensity analysis; the intermediate verification activates an infrared thermal imager to measure the temperature of the jumper connection hardware and verify the rationality of the temperature distribution; the advanced verification calls upon the redundant sensor group from the weather station for data cross-verification. If any verification level detects data inconsistencies, the response command is paused and marked as requiring manual intervention.

[0153] After verification, the aforementioned retesting process is repeated. When the spatial consistency index meets the standard, the original morphological anomaly response command is confirmed to be valid: for emergency response commands, the execution completion time and retested morphological parameters are recorded; for early warning commands, an alarm confirmation code is sent to the monitoring center. All verification data is written to an independent audit log and stored in isolation from the main business data stream.

[0154] After each command verification, the deviation between the actual change in the shape of the transmission tower jumper and the expected response is analyzed: if the reduction in the spatial deviation vector after handling does not meet the expected target, the trigger threshold for this type of command is automatically lowered; if retesting shows over-response, the judgment threshold is raised and the buffer zone is widened. Adjustment parameters are pushed to edge computing nodes in real time through a distributed configuration center to ensure synchronized updates of the entire network policy. This mechanism continuously optimizes the system's response accuracy and avoids invalid alarms interfering with the operation and maintenance process.

[0155] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A machine learning-based intelligent analysis method for the three-dimensional spatial morphology of transmission tower jumpers, characterized in that, include: Acquire multi-source sensor data streams for the jumper area of ​​the power transmission tower, including: lidar point cloud data, texture image data, and micro-meteorological monitoring data; A three-dimensional spatial reference model for jumpers is constructed based on the lidar point cloud data, and the texture image data and the micro-meteorological monitoring data are fused into the three-dimensional spatial reference model for jumpers using a spatiotemporal registration method to generate a fused three-dimensional field model. Extract the jumper key point set and corresponding spatial curvature distribution from the fused 3D field model; The key point set of the jumper and its corresponding spatial curvature distribution are input into the morphological comparison and analysis unit to calculate the spatial deviation vector between the jumper morphology and the preset standard morphology of the transmission tower jumper, including: Load a preset standard form database, which contains a set of coordinates of standard jumper key points under ideal working conditions; By establishing an elastic point set registration algorithm, the jumper key point set is non-rigidly matched with the corresponding standard jumper key point coordinate set. Curvature similarity constraints are introduced in the non-rigid matching process to minimize the difference between the spatial curvature distribution of the transmission tower jumper and the standard curvature distribution. Calculate the spatial position offset of each key point of the transmission tower jumper after non-rigid matching, and arrange all spatial position offsets in the order of key point index to generate the spatial deviation vector of the transmission tower jumper. A dynamic risk parameter set is constructed based on the spatial deviation vector and the micrometeorological monitoring data, including: Real-time wind speed components, wind direction angle, and temperature and humidity change rates are extracted from the micrometeorological monitoring data. The offset statistical characteristics are calculated based on the spatial deviation vector, including: the maximum offset amplitude and the offset direction consistency coefficient; The real-time wind speed component is decomposed into a wind pressure load component perpendicular to the jumper axis. A multi-parameter coupling function is constructed, and the wind pressure load component, temperature and humidity change rate, maximum offset amplitude and offset direction consistency coefficient are used as input variables of the multi-parameter coupling function to output the dynamic deformation risk index of the transmission tower jumper. The dynamic deformation risk index of the transmission tower jumper is combined with the offset statistical characteristics to form the dynamic risk parameter set of the transmission tower jumper; Based on the dynamic risk parameter set, a multi-level threshold decision-making mechanism is triggered to generate an abnormal response command for the morphology of the transmission tower jumper.

2. The intelligent analysis method for the three-dimensional spatial morphology of transmission tower jumpers based on machine learning according to claim 1, characterized in that, The generation process of the fused three-dimensional field model includes: The lidar point cloud data is subjected to non-uniform voxelization processing to construct an initial three-dimensional spatial reference model; The resolution of the texture image data is compressed using a feature-preserving downsampling algorithm to generate an optimized texture feature map; A spatiotemporal transformation matrix is ​​established, and the optimized texture feature map is mapped to the corresponding surface region of the initial three-dimensional spatial reference model through the spatiotemporal transformation matrix; The micrometeorological monitoring data is decomposed into wind speed gradient tensor and temperature and humidity influencing factors, and the wind speed gradient tensor is encoded into a spatial vector field, and the temperature and humidity influencing factors are encoded into a scalar field. The spatial vector field and the scalar field are embedded into the voxel structure of the initial three-dimensional spatial reference model by tensor splicing to form the fused three-dimensional field model.

3. The intelligent analysis method for the three-dimensional spatial morphology of transmission tower jumpers based on machine learning according to claim 1, characterized in that, Extracting the jumper key point set and corresponding spatial curvature distribution from the fused 3D field model includes: In the fused three-dimensional field model, a jumper main axis search domain is defined. Curvature extreme points are detected along the jumper main axis search domain to identify jumper suspension points, jumper sag lowest points, and spacer connection points as key points of the transmission tower jumper. Construct a local neighborhood sphere centered on the key points, and calculate the eigenvalues ​​of the covariance matrix of the point cloud within the local neighborhood sphere; Calculate the curvature value and principal orientation angle of each key point based on the eigenvalues ​​of the covariance matrix; The three-dimensional coordinates, curvature values, and principal direction angles of the key points are combined to form the jumper key point set, and the mapping relationship between the key points and their corresponding curvature values ​​is constructed as the spatial curvature distribution of the key points.

4. The intelligent analysis method for the three-dimensional spatial morphology of transmission tower jumpers based on machine learning according to claim 1, characterized in that, The dynamic risk parameter set triggers a multi-level threshold decision-making mechanism to generate abnormal transmission tower jumper morphology response instructions, including: Multiple preset judgment thresholds are included, including: basic warning threshold, medium risk threshold and emergency response threshold; The multi-level judgment thresholds are adjusted based on the historical abnormality records of the transmission tower jumpers. The dynamic deformation risk index of the transmission tower jumper in the dynamic risk parameter set is compared step by step with the corrected multi-level judgment threshold: When the dynamic deformation risk index exceeds the modified basic warning threshold, the first-level monitoring enhancement command is activated. When the dynamic deformation risk index exceeds the modified moderate risk threshold, a level-two early warning notification instruction is activated; When the dynamic deformation risk index exceeds the modified emergency response threshold, a Level 3 emergency response command is activated. Integrate the current activation command status to generate an abnormal response command.

5. The intelligent analysis method for the three-dimensional spatial morphology of transmission tower jumpers based on machine learning according to claim 2, characterized in that, It also provides the process for tracing the status of key points on transmission tower jumpers: In the fused 3D field model, a spatiotemporal index table of jumper key points is established to record the spatial coordinates and curvature values ​​of each jumper key point at different timestamps. Based on the spatiotemporal index table of jumper key points, a key point motion trajectory function is constructed to calculate the displacement acceleration of key points between adjacent timestamps. When the dynamic deformation risk index of the transmission tower jumper in the dynamic risk parameter set exceeds the basic warning threshold, the abnormal key point identifier with displacement acceleration greater than the acceleration threshold is extracted. The anomaly key point identifier is associated with the spatial deviation vector.

6. The intelligent analysis method for the three-dimensional spatial morphology of transmission tower jumpers based on machine learning according to claim 5, characterized in that, It also includes a decision buffer mechanism: A boundary buffer layer with multiple judgment thresholds is set, and the boundary buffer layer is the safe floating range of the dynamic deformation risk index of the transmission tower jumper. When the dynamic deformation risk index enters the boundary buffer layer, the triggering of the abnormal shape response command of the transmission tower jumper is delayed; The duration of the dynamic deformation risk index within the boundary buffer layer is monitored. If the duration exceeds the preset tolerance period and the dynamic deformation risk index does not fall below the safety boundary, the corresponding level of morphological anomaly response command is forcibly triggered.

7. The intelligent analysis method for the three-dimensional spatial morphology of transmission tower jumpers based on machine learning according to claim 6, characterized in that, It also provides a quantification process for the impact of micrometeorological monitoring data: Establish a wind speed-deformation transfer function, and calculate the theoretical wind-induced vibration amplitude based on the real-time wind speed component and jumper tension parameter in the micrometeorological monitoring data; Extract the actual vibration amplitude from the spatial deviation vector; Calculate the residual between the theoretical wind-induced vibration amplitude and the actual vibration amplitude; When the residual exceeds the residual threshold, the temperature and humidity influence factors are weighted and injected into the multi-parameter coupling function to recalculate the dynamic deformation risk index of the transmission tower jumper.

8. The intelligent analysis method for the three-dimensional spatial morphology of transmission tower jumpers based on machine learning according to claim 7, characterized in that, It also provides the execution verification process of the abnormal morphological response command of the transmission tower jumper: start the corresponding jumper morphological retest process according to the type of the abnormal morphological response command of the transmission tower jumper; During the retesting process, the key point set and spatial curvature distribution of the jumper line at the current moment are reacquired. Calculate the spatial consistency coefficient between the retested jumper key point set and the original jumper key point set; When the spatial consistency coefficient is lower than the verification threshold, a sensor data verification command is triggered; When the spatial consistency coefficient meets the verification requirements, it is confirmed that the abnormal morphological response command of the transmission tower jumper is valid.

Images