Power transmission and distribution line suspended foreign matter detection method and system based on image processing and deep learning

Abstract

The invention provides a power transmission and distribution line suspended foreign matter detection method and system based on image processing and deep learning. Wherein a historical point cloud set on the surface of the power transmission tower is acquired, a deformation gradient map is generated through dynamic registration of curvature differences of adjacent point clouds, a super-threshold region is marked as a deformation correlation region, and foreign matter contour parameters are extracted; performing time-frequency decomposition on the node strain signal, retaining a steady-state component, and mapping the steady-state component to an association region to form a strain map; calculating the displacement offset of the deformation association interval to construct a deformation propagation network, and extracting load over-threshold nodes to construct a stress diffusion network; and superposing the network, fusing the displacement and the load value in the superposition area to generate a risk index, and outputting early warning in combination with the foreign matter coordinates and the risk index. According to the technical scheme provided by the invention, the cross-modal dynamic analysis of deformation propagation and stress diffusion paths is fused, and the early warning efficiency of foreign matter attachment and structure abnormity risks of the power transmission tower is improved.

Description

Technical Field

[0001] The present application relates to the technical field of detection of foreign objects hanging on power transmission and distribution lines, and in particular to a method and system for detecting foreign objects hanging on power transmission and distribution lines based on image processing and deep learning. Background Art

[0002] In the complex fault scenario where the transmission tower suffers from abnormal structural stress due to hanging foreign objects, it is necessary to achieve: synchronous dynamic perception of the shape of foreign objects and structural deformation on the tower surface, eliminating the error in foreign object identification caused by local occlusion and light interference in the inspection image; establish a correlation model between the attachment position of foreign objects and the dynamic stress distribution of the tower body, and distinguish between short-term deformation caused by the deadweight of foreign objects and permanent deformation caused by structural damage; construct a cross-modal abnormal propagation path inference mechanism to predict the cascading impact of local deformation on the overall mechanical stability of the transmission tower, and provide a precise positioning basis for live fault elimination.

[0003] The mainstream solution adopts a multimodal temporal feature alignment detection framework: based on the improved Mask R-CNN, the drone inspection image is used to segment the contours of foreign objects, and a vibration sensor network is simultaneously deployed to collect the acceleration spectrum of key nodes of the tower body; the image segmentation results and the vibration frequency domain features are fused through the temporal cross-attention mechanism to construct a foreign object-deformation association map; the graph neural network (GNN) is used to model the abnormal mechanical transmission path, and the finite element simulation pre-training model is combined to output the structural risk heat map.

[0004] Image segmentation models have low accuracy in detecting the boundaries of translucent and reflective foreign objects (such as plastic bags and fiberglass kite strings), resulting in distorted contour parameter extraction. Modeling the correlation between vibration spectrum characteristics and actual structural deformation under strong wind interference is difficult, and environmental noise can be misinterpreted as stress anomalies. Graph neural networks rely on simulation pre-training data, making it difficult to generalize to unseen complex fault modes (such as localized arc deformation caused by foreign objects wrapped around insulators). This results in the prediction confidence of risk heat maps being lower than actual O&M requirements. Summary of the Invention

[0005] The present application provides a method and system for detecting foreign objects hanging on power transmission and distribution lines based on image processing and deep learning, which is used to solve the problems in the existing technology of low efficiency of manual inspection, insufficient accuracy of foreign object detection in complex environments, and difficulty in integrating dynamic deformation and stress data to achieve cross-modal risk warning.

[0006] In a first aspect, the present application provides a method for detecting foreign objects hanging on power transmission and distribution lines based on image processing and deep learning, comprising:

[0007] Obtaining a surface point cloud set and a historical point cloud set of the transmission tower, dynamically registering the surface point cloud set and the historical point cloud set based on the curvature difference of adjacent point clouds to generate a deformation gradient map, wherein point cloud regions where the curvature difference exceeds a set range during the dynamic registration process are marked as deformation-related regions of the deformation gradient map, and simultaneously identifying the contour of a surface foreign body based on the curvature difference to obtain geometric parameters of the contour of the surface foreign body;

[0008] Strain signals are collected at the connection nodes of the transmission towers, the strain signals are subjected to time-frequency domain energy decomposition, steady-state components are retained, and the steady-state components are mapped to positions corresponding to deformation-related regions to generate a strain map.

[0009] Calculating displacement offsets of adjacent deformation-associated regions in the deformation gradient map according to preset spatial coordinates in the deformation gradient map, and connecting the displacement offsets to form a deformation propagation path network;

[0010] Obtaining load change values ​​of connected nodes in the strain map, and associating nodes whose load change amplitudes exceed a preset threshold to construct a stress diffusion path network;

[0011] spatially superimposing the deformation propagation path network and the stress diffusion path network, and fusing the displacement offset and the load change value in the overlapping area of ​​the spatial superposition to generate a risk indicator;

[0012] The coordinates of the foreign body attachment position are determined according to the geometric parameters of the surface foreign body profile, and a transmission tower abnormality warning instruction including the coordinates of the foreign body attachment position is generated in combination with the risk index.

[0013] Optionally, the deformation propagation path network and the stress diffusion path network are spatially superimposed, and the displacement offset and the load change value are fused in the overlapping area of ​​the spatial superposition to generate a risk indicator, including:

[0014] Determining an overlapping area of ​​the deformation propagation path network and the stress diffusion path network on a preset spatial coordinate, wherein a condition for determining the overlapping area is that a range of spatial coordinates corresponding to a displacement offset in the deformation propagation path network overlaps a range of spatial coordinates corresponding to a load change value in the stress diffusion path network;

[0015] Calculating a fusion weight factor for each superposition region according to a change degree of the displacement offset and an increase or decrease trend of the load change value in the superposition region;

[0016] The displacement offset and load change value of each coordinate point in the superposition area are superimposed and calculated according to the corresponding fusion weight factor to obtain the risk index of each coordinate point.

[0017] Optionally, the surface point cloud set and the historical point cloud set are dynamically registered based on the curvature difference between adjacent point clouds to generate a deformation gradient map, wherein point cloud areas where the curvature difference exceeds a set range during the dynamic registration process are marked as deformation-related areas of the deformation gradient map, including:

[0018] Extracting three-dimensional geometric features of adjacent point clouds in the surface point cloud set and the historical point cloud set, and calculating the curvature difference of corresponding areas between the adjacent point clouds, wherein the adjacent point clouds are determined by the nearest neighbor relationship of spatial coordinates;

[0019] Setting a dynamic adjustment threshold according to the distribution range of the curvature difference, and marking the point cloud area where the curvature difference exceeds the dynamic adjustment threshold as the initial deformation area;

[0020] Dynamically registering the correspondence between the surface point cloud set and the historical point cloud set based on a changing trend of the curvature difference between adjacent point clouds in the initial deformation area, so that the curvature difference in the initial deformation area converges to a dynamically adjusted threshold;

[0021] Calculate the curvature difference variation of each point cloud region based on the corresponding relationship after dynamic registration, and accumulate the variation along a preset three-dimensional spatial direction to generate a deformation gradient map;

[0022] A spatial continuity verification is performed on the region in the gradient value distribution of the deformation gradient map that continuously exceeds the dynamic adjustment threshold, and the region that meets the distribution condition of spatial continuity is marked as a deformation association region.

[0023] Optionally, according to the correspondence after dynamic registration, the change in curvature difference of each point cloud region is calculated, and the change is accumulated along a preset three-dimensional space direction to generate a deformation gradient map, including:

[0024] Mapping the surface point cloud set to the three-dimensional geometric features of each point cloud in the historical point cloud set according to the corresponding relationship after dynamic registration, and determining the point cloud area where the curvature difference changes in the corresponding relationship;

[0025] Calculating the change in curvature difference of each point cloud region before and after dynamic registration, and accumulating the change in curvature difference of each point cloud region along a preset three-dimensional spatial direction, where the three-dimensional spatial direction is determined by the geometric feature distribution of the surface point cloud set;

[0026] The accumulated curvature difference variation is matched with the preset three-dimensional spatial coordinates in each point cloud area to generate a deformation gradient map.

[0027] Optionally, a strain signal is collected at a connection node of a transmission tower, a time-frequency domain energy decomposition is performed on the strain signal, a steady-state component is retained, and the steady-state component is mapped to a position corresponding to a deformation association area to generate a strain map, including:

[0028] Strain signals at the connection nodes of the transmission towers are collected and divided into multiple time segments according to a preset time length. The signal in each time segment is divided into multiple continuous frequency intervals, and the energy value of each combination of time segment and frequency interval is calculated;

[0029] Filter the units whose energy value fluctuation amplitude over time is lower than the preset fluctuation threshold as steady-state components;

[0030] According to the range of the spatial coordinates of the deformation association area, the energy value corresponding to the steady-state component of each connection node is mapped to the spatial coordinates of the corresponding deformation association area;

[0031] The energy values ​​of the steady-state components of multiple connection nodes mapped to the same deformation association area are spatially weighted fused to obtain an energy fusion value, and a strain map is generated according to the corresponding relationship between the energy fusion value and the spatial coordinates.

[0032] Optionally, obtaining load change values ​​of connected nodes in the strain map and associating nodes whose load change amplitudes exceed a preset threshold to construct a stress diffusion path network includes:

[0033] Extracting a load change value of each connection node in the strain spectrum, wherein the load change value is derived from an energy value corresponding to a steady-state component in the strain spectrum;

[0034] Calculating a dynamic fluctuation threshold of a change amplitude of the load change value based on distribution characteristics of the load change values ​​of all connected nodes, wherein the dynamic fluctuation threshold is determined by the sum of the median and the standard deviation of the change amplitude;

[0035] Screening the connection nodes whose change amplitude exceeds the dynamic fluctuation threshold and marking these nodes as key nodes;

[0036] Analyze the spatial proximity relationship between the key nodes, and determine the association direction between the key nodes based on the increase and decrease trends of the coordinate distances and change amplitudes of the key nodes in the preset three-dimensional space;

[0037] The key nodes are connected in sequence according to the association direction to form a continuous path, and for the fracture area in the continuous path caused by the coordinate distance exceeding the preset range, the fracture area is supplemented by the increase or decrease trend of the change amplitude of the load change value of the adjacent key nodes to generate a stress diffusion path network, and the adjacent key nodes are determined by the spatial proximity relationship between the key nodes.

[0038] Optionally, determining the coordinates of the foreign object attachment position according to the geometric parameters of the surface foreign object profile, and generating a transmission tower abnormality warning instruction containing the coordinates of the foreign object attachment position in combination with the risk index, includes:

[0039] Extracting boundary coordinates from the geometric parameters of the surface foreign body contour, and using the boundary coordinates as initial position coordinates of the foreign body attachment area;

[0040] Matching the initial position coordinates with the spatial coordinates of the deformation association area according to the spatial coordinate range of the deformation association area, and eliminating the initial position coordinates that do not fall within the spatial coordinate range to determine the foreign body attachment position coordinates;

[0041] Determine the value of the risk index corresponding to each foreign body attachment location coordinate, and divide the dynamic safety threshold according to the distribution range of the value of the risk index of all foreign body attachment location coordinates;

[0042] Marking the coordinates of the foreign body attachment location where the value of the risk indicator exceeds the dynamic safety threshold as high-risk coordinates;

[0043] Combining the location information of the high-risk coordinates with the corresponding risk index value, a transmission tower abnormality warning instruction including the coordinates of the foreign object attachment location and the risk index is generated.

[0044] In a second aspect, the present application provides a system for detecting foreign objects hanging on power transmission and distribution lines based on image processing and deep learning, comprising:

[0045] an acquisition module, configured to acquire a surface point cloud set and a historical point cloud set of the transmission tower, dynamically register the surface point cloud set and the historical point cloud set based on the curvature difference between adjacent point clouds to generate a deformation gradient map, wherein point cloud regions where the curvature difference exceeds a set range during the dynamic registration process are marked as deformation-related regions of the deformation gradient map, and simultaneously identify the contour of a surface foreign body based on the curvature difference to obtain geometric parameters of the contour of the surface foreign body;

[0046] An acquisition module is used to collect strain signals at the connection nodes of the transmission tower, perform time-frequency domain energy decomposition on the strain signals, retain steady-state components, and map the steady-state components to positions corresponding to deformation-related areas to generate a strain map;

[0047] A calculation module, configured to calculate displacement offsets of adjacent deformation-associated areas in the deformation gradient map according to preset spatial coordinates in the deformation gradient map, and connect the displacement offsets to form a deformation propagation path network;

[0048] an association module, configured to obtain load change values ​​of connected nodes in the strain map, and associate nodes whose load change amplitudes exceed a preset threshold value to construct a stress diffusion path network;

[0049] a superposition module for spatially superimposing the deformation propagation path network and the stress diffusion path network, and fusing the displacement offset and the load change value in the overlapping area of ​​the spatial superposition to generate a risk indicator;

[0050] A generation module is used to determine the coordinates of the foreign body attachment position based on the geometric parameters of the surface foreign body profile, and generate a transmission tower abnormality warning instruction containing the coordinates of the foreign body attachment position in combination with the risk index.

[0051] In a third aspect, an embodiment of the present application provides a computing device comprising a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a method for detecting foreign objects hanging on transmission and distribution lines based on image processing and deep learning as described in the first aspect above.

[0052] In a fourth aspect, an embodiment of the present application provides a computer storage medium storing a computer program. When the computer program is executed by a computer, it implements a method for detecting foreign objects hanging on power transmission and distribution lines based on image processing and deep learning as described in the first aspect.

[0053] The present application obtains a surface point cloud set and a historical point cloud set of a transmission tower, and dynamically aligns the surface point cloud set and the historical point cloud set based on the curvature difference of adjacent point clouds to generate a deformation gradient map, wherein the point cloud area whose curvature difference exceeds a set range in the dynamic alignment process is marked as a deformation association area of ​​the deformation gradient map, and at the same time identifies the surface foreign body contour based on the curvature difference to obtain the geometric parameters of the surface foreign body contour, which can achieve high-precision positioning of the deformation area on the surface of the transmission tower and automatic extraction of foreign body attachment features; by collecting strain signals at the connection nodes of the transmission tower, retaining the steady-state components after time-frequency domain energy decomposition of the strain signals, and mapping the steady-state components to the positions corresponding to the deformation association areas to generate strain maps, it can eliminate environmental noise interference and establish a spatial mapping relationship between structural deformation and mechanical response; by calculating the displacement offsets of adjacent deformation association areas in the deformation gradient map according to the preset spatial coordinates in the deformation gradient map, and The displacement offsets are connected to form a deformation propagation path network, which can quantify the propagation path and cumulative effect of structural deformation and reveal the potential structural damage expansion mode; by obtaining the load change values ​​of the connected nodes in the strain map, the nodes whose load change amplitude exceeds the preset threshold are associated to construct a stress diffusion path network, which can identify the stress concentration phenomenon and its diffusion path of key nodes; by spatially superimposing the deformation propagation path network and the stress diffusion path network, and fusing the displacement offsets and load change values ​​in the overlapping area of ​​spatial superposition to generate a risk index, it is possible to achieve a coordinated assessment of structural deformation and stress state and improve the accuracy of abnormal risk judgment; by determining the coordinates of the foreign body attachment position based on the geometric parameters of the surface foreign body contour, and generating a transmission tower abnormality warning instruction containing the foreign body attachment position coordinates in combination with the risk index, it is possible to comprehensively integrate the structural state and the influence of foreign bodies to achieve intelligent warning of the transmission tower operation state.

[0054] Furthermore, by precisely matching the spatial coordinate ranges of the deformation propagation path network and the stress diffusion path network, the spatial coupling areas of structural deformation and stress concentration phenomena can be automatically identified, providing an accurate spatial benchmark for risk positioning; based on the dynamic weight calculation of the displacement offset change degree and the increase and decrease trend of the load change value, the differentiated fusion of structural deformation characteristics and mechanical response characteristics is achieved, and the danger levels of different areas are effectively distinguished; by fusing weight factors to perform weighted superposition calculations on the displacement offset and load change values, the limitations of single parameter evaluation are broken through, and a composite risk assessment model that takes into account both the degree of structural deformation and the intensity of stress concentration is constructed, which significantly improves the accuracy and reliability of transmission tower structural anomaly judgment, and provides an intelligent decision-making basis with both spatial precision and quantitative depth for the safe operation and maintenance of power facilities.

[0055] These and other aspects of the present application will become more readily apparent from the description of the following embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0057] Figure 1 A flowchart of a method for detecting foreign objects hanging on power transmission and distribution lines based on image processing and deep learning is shown in the present application;

[0058] Figure 2 A schematic diagram of the structure of a system for detecting foreign objects hanging on power transmission and distribution lines based on image processing and deep learning is shown in the present application;

[0059] Figure 3 A schematic structural diagram of a computing device provided by the present application is shown. DETAILED DESCRIPTION

[0060] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0061] In some of the processes described in the specification and claims of this application and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this document or may be executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish between different operations, and the serial numbers themselves do not represent any order of execution. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this document are used to distinguish different messages, devices, modules, etc., and do not represent a sequential order, nor do they limit "first" and "second" to being different types.

[0062] Researchers have found that existing abnormal warning methods for transmission towers rely on single deformation monitoring or static stress analysis, making it difficult to dynamically correlate the spatiotemporal evolution of surface deformation gradients and stress diffusion paths, resulting in large deviations in the positioning of foreign matter attachment, insufficient coordination between deformation propagation and load changes, and delayed risk warnings. Based on this, a dynamic warning method for abnormal transmission towers is provided. This method can achieve multi-dimensional and precise positioning of abnormal risks through the fusion of deformation-stress path networks and collaborative analysis of foreign matter geometric parameters. The technical solution of this application can be applied to transmission tower structural health monitoring, foreign matter attachment warning, and dynamic disaster risk assessment scenarios.

[0063] The entire R&D process embodies the collaborative mechanism of multi-source data dynamic registration and risk path network modeling, aiming to overcome the defects of existing solutions such as the separation of deformation and stress data, the disconnection between foreign body positioning and risk propagation, and the static nature of early warning indicators. Through the joint processing of point cloud dynamic registration and strain time-frequency domain decomposition, the traditional method's reliance on a single data source or static parameters is broken through; based on the superposition and fusion of deformation propagation paths and stress diffusion paths, the complexity of dynamic correlation modeling between deformation gradients and load changes is resolved; combined with the spatial collaborative analysis of foreign body geometric parameters and risk indicators, the adaptation deviation of foreign body positioning and risk warning is eliminated; finally, through the joint decision-making of dynamic path networks and foreign body position coordinates, the real-time, accuracy, and system robustness of abnormal warnings are achieved in multiple dimensions. This method forms a full-link closed-loop optimization from data registration to risk modeling, significantly enhancing the intelligent management capabilities of transmission tower structural anomalies in complex environments.

[0064] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative efforts are within the scope of protection of this application.

[0065] Figure 1 The present invention provides a flowchart of a method for detecting foreign objects hanging on a power transmission and distribution line based on image processing and deep learning, as shown in FIG. Figure 1 As shown, the method includes:

[0066] 101. Obtain a surface point cloud set and a historical point cloud set of the transmission tower, and dynamically register the surface point cloud set and the historical point cloud set based on the curvature difference between adjacent point clouds to generate a deformation gradient map. During the dynamic registration process, point cloud regions where the curvature difference exceeds a set range are marked as deformation-related regions of the deformation gradient map. Simultaneously, the contour of a surface foreign body is identified based on the curvature difference to obtain geometric parameters of the contour of the surface foreign body.

[0067] In this step, the surface point cloud set refers to the data set of surface points on the transmission tower acquired through 3D scanning. The historical point cloud set refers to the comparative point cloud data collected from the transmission tower at historical time points. Curvature difference refers to the change in the degree of surface curvature between adjacent point cloud areas. Dynamic registration refers to the process of spatially aligning point cloud data at different time points. The deformation gradient map refers to a visualization model that reflects the degree and distribution of surface deformation on the transmission tower. The deformation association area refers to the key focus area where the deformation exceeds the set threshold. The surface foreign body contour refers to the external features of foreign objects attached to the surface of the transmission tower. Geometric parameters refer to quantitative indicators that describe the shape, size, and other characteristics of foreign objects.

[0068] In an embodiment of the present application, first, a surface point cloud set of the current state of the transmission tower is obtained by a three-dimensional laser scanner, and at the same time, a historical point cloud set recorded during historical detection is retrieved from the database. Secondly, an improved ICP (iterative closest point) registration algorithm is used to perform preliminary alignment on the two sets of point clouds, and the curvature difference of adjacent point cloud areas is calculated in real time during the registration process. Next, a reasonable threshold range for the curvature difference is set. When it is detected that the curvature difference of a certain area exceeds the set range, the area is marked as a potential deformation area and highlighted as a deformation-related area in the three-dimensional model. Then, a cluster analysis is performed on the abnormal point cloud clusters that have not been successfully matched, and the contour features of these point cloud clusters are identified by an edge detection algorithm, and the geometric shape parameters of the surface foreign body contour (such as area, perimeter, convex hull, etc.) are extracted. Finally, the spatial position information of all deformation-related areas and the geometric parameters of foreign bodies are integrated to generate a visual deformation gradient map containing deformation gradients and foreign body distribution, providing basic data for subsequent analysis.

[0069] For example, in a coastal 500kV high-voltage transmission corridor, the operations and maintenance team used a drone equipped with a lidar to scan transmission towers after a strong typhoon passed through. This scanned the current surface point cloud and dynamically aligned it with the historical point cloud set before the typhoon. Analysis revealed that the curvature difference in the crossarm area in the middle section of the tower exceeded a set threshold. The dynamic alignment algorithm marked this area as a deformation-related region in the deformation gradient map, with the direction of the sudden change in curvature consistent with the dominant wind direction of the typhoon. Simultaneously, a band of abnormal curvature fluctuations was detected near the insulator string at the top of the tower. Three-dimensional contour reconstruction identified the surface contours of foreign objects—areas entangled with the debris of multiple fishing nets swept up by the typhoon. Geometric parameters indicated that the maximum entanglement depth of the foreign objects reached 15 centimeters, and they exhibited a spiral wrapping characteristic from top to bottom. The system automatically associates and stores the coordinates of the deformation-related region with the foreign object contour parameters.

[0070] 102. Collect strain signals at the connection nodes of the transmission tower, perform time-frequency domain energy decomposition on the strain signals, retain steady-state components, and map the steady-state components to positions corresponding to deformation-related regions to generate a strain map;

[0071] In this step, connection nodes refer to key connection points in the transmission tower structure. Strain signals refer to the measured physical quantities that reflect the stress and deformation of the structure. Time-frequency energy decomposition refers to the method of analyzing the signal in the time and frequency dimensions. Steady-state components refer to the stable and continuous energy components in the signal. Strain maps refer to the visualization model that reflects the strain distribution characteristics of the structure. Mapping refers to the process of assigning data to spatial locations.

[0072] In the embodiments of the present application, first, high-precision strain sensors are deployed at the key connection nodes of the power transmission tower to collect real-time structural strain signals. Second, the original strain signals are processed by wavelet packet transformation to decompose the energy distribution of different frequency bands, and the threshold filtering is used to retain the steady-state components reflecting the long-term deformation of the structure. Then, the position coordinates of the deformation correlation area corresponding to each sensor are determined according to the three-dimensional model of the power transmission tower, and the processed steady-state component values are mapped to the corresponding area. Finally, the strain data of all deformation correlation areas are integrated to generate a visual strain map, which displays the strain distribution of each part of the power transmission tower in the form of a heat map.

[0073] In the above example, the pre-deployed fiber grating sensors continuously collect strain signals at the connection nodes of the power transmission tower corresponding to the marked deformation correlation area. For the continuous monitoring data of 72 hours after the typhoon, the transient components caused by wind speed fluctuations are filtered out through time-frequency energy decomposition, and the steady-state components reflecting the cumulative damage of the structure are retained. The steady-state components are mapped to the deformation correlation area of the deformation gradient map according to the spatial position, generating a strain map covering two-thirds of the height of the tower body. The map shows that the strain value at the connection between the cross arm and the main material increases by three times compared to the historical average, and completely coincides with the area with the largest curvature difference in the deformation gradient map. The strain map is transmitted to the cloud analysis platform in real time, triggering the subsequent deformation propagation path analysis task.

[0074] 103、According to the preset spatial coordinates in the deformation gradient map, the displacement offset between adjacent deformation correlation areas in the deformation gradient map is calculated, and the displacement offset is connected to form a deformation propagation path network;

[0075] In this step, the spatial coordinates refer to the position positioning data in the three-dimensional space. The displacement offset refers to the position change of the deformation correlation area. The deformation propagation path network refers to the path relationship model of the deformation transmission in the structure. The connection refers to the association relationship between different deformation areas. The calculation refers to the process of obtaining characteristic parameters through mathematical operation.

[0076] In the embodiments of the present application, first, based on the deformation gradient map generated in step 102, the center position of each deformation correlation area is located according to the preset spatial coordinate system. Second, the displacement offset between adjacent deformation correlation areas is calculated, including distance and direction information. Then, the topological connection relationship of the displacement offset is constructed by using the graph theory algorithm, and the offset with continuous change characteristics is connected to form a deformation propagation path network. Finally, the path optimization algorithm is used to eliminate abnormal connections and retain the main propagation path, forming a network model reflecting the deformation transmission law of the structure.

[0077] Continuing with the above example, based on the spatial coordinates of the deformation gradient map, the displacement offsets of the crossarm deformation-related areas and adjacent tower sections were calculated. Analysis revealed that the displacement offsets propagate downward along the main members, reaching a point of maximum displacement accumulation at the diagonal member connection node 20 meters above the ground. Using a graph theory algorithm, the displacement offset directions of each deformation-related area were vector-connected, constructing a deformation propagation path network extending from the foreign object entanglement area at the top of the tower to the tower base. This network reveals the conduction characteristics of deformation energy within the tower: the eccentric load caused by the entanglement of the fishing net is transmitted through the crossarm-main member node, resulting in multiple levels of displacement offsets in the middle and lower parts of the tower, forming a propagation path structure similar to a tree-like bifurcation.

[0078] 104. Obtain load change values ​​of connected nodes in the strain map, and associate nodes whose load change amplitudes exceed a preset threshold to construct a stress diffusion path network;

[0079] In this step, the load change value refers to the change in the force applied to the connected nodes. The preset threshold refers to the pre-set abnormality judgment standard. The stress diffusion path network refers to the relationship model of the stress transmission paths in the structure. Construction refers to the process of establishing a network model through analysis. Correlation refers to the establishment of mechanical relationships between different nodes.

[0080] In this embodiment, first, time series data of load changes at each connected node is extracted from the strain map obtained in step 102. Next, a preset threshold based on material properties is set to identify abnormal nodes whose load changes exceed the threshold. Next, spatial correlation analysis is used to determine the degree of association between abnormal nodes, and pairs of nodes with strong correlations are connected. Finally, a stress diffusion path network is constructed that reflects the characteristics of abnormal load propagation. This network reveals potential stress concentration areas in the transmission tower structure.

[0081] Continuing with the above example, combined with strain map data, the load change values ​​of the transmission tower connection nodes were extracted. The load change values ​​of the two bolt connection nodes located on the west side of the crossarm increased sharply within twelve hours, exceeding the preset safety threshold by twice. Using a stress diffusion tracking algorithm, these two nodes were linked to the lower diagonal member nodes and the foundation flange nodes, constructing a stress diffusion path network running through the west column of the tower. Comparison with the deformation propagation path network revealed that the stress diffusion path spatially overlapped with the deformation propagation path in the middle of the tower. The corresponding displacement offset and load change values ​​in this overlapping area increased simultaneously, indicating possible plastic deformation of the structure.

[0082] 105. Spatially superimpose the deformation propagation path network and the stress diffusion path network, and fuse the displacement offset and the load change value in the overlapping area of ​​the spatial superposition to generate a risk indicator;

[0083] In this step, spatial superposition refers to the fusion analysis of different networks in the spatial dimension. Overlapping areas refer to the spatial overlap of different networks. Fusion refers to the process of integrating and calculating different parameters. Risk indicators refer to comprehensive parameters used to quantitatively assess the safety status of the structure. Generation refers to the process of obtaining the final result through calculation.

[0084] In the embodiment of the present application, first, the spatial coordinates of the deformation propagation path network of step 103 and the stress diffusion path network of step 104 are aligned to achieve accurate spatial superposition. Secondly, in the overlapping area, a weighted fusion algorithm is used to fuse the displacement offset and the load change value to calculate the comprehensive risk coefficient. Then, the risk level classification standard is set according to the mechanical parameters of the material, and the fused characteristic value is converted into a risk index. Finally, a three-dimensional risk distribution map containing the risk location and level is generated to provide a quantitative basis for subsequent warnings.

[0085] Continuing with the above example, the deformation propagation path network and the stress diffusion path network were spatially superimposed, revealing a three-meter-long overlapping region in the middle of the tower. Within this region, the displacement trend (average daily increase of one centimeter) and the load change (increase of ten kilonewtons per hour) were combined to generate a comprehensive risk index. The risk index model indicated that the probability of structural failure in this region had reached the warning threshold, and the risk propagation was pointing toward weak points in the tower foundation. The system automatically labeled this area a "high-risk transmission zone" and incorporated the geometric parameters of the fishing net entanglement zone (entanglement depth and wrapping angle) into the risk transmission coefficient calculation.

[0086] 106. Determine the coordinates of the foreign object attachment position based on the geometric parameters of the surface foreign object profile, and generate a transmission tower abnormality warning instruction including the coordinates of the foreign object attachment position in combination with the risk index.

[0087] In this step, the coordinates of the foreign object attachment location refer to the specific spatial location of the surface foreign object on the transmission tower. The abnormal warning instruction refers to the risk warning signal issued based on the analysis results. Combination refers to the analysis method that associates different parameters. Determination refers to the process of reaching a clear conclusion through calculation.

[0088] In the embodiment of the present application, first, a high-definition image of a foreign object on the surface of a transmission tower is acquired through drone aerial photography. An image segmentation algorithm is then used to extract the foreign object's outline and calculate its geometric parameters. Secondly, based on the principles of photogrammetry, the two-dimensional image coordinates are converted into the coordinates of the attachment location within the three-dimensional model of the transmission tower. Next, the foreign object's location is spatially matched with the risk indicator in step 105 to determine the risk level of the location where the foreign object is attached. Finally, the location coordinates, risk level, and timestamp information are integrated to generate a transmission tower abnormality warning instruction containing a specific location description and risk warning, which is then pushed to the operation and maintenance management system to trigger the disposal process.

[0089] Based on the geometric parameters of the surface foreign body's profile, the coordinates of the fishing net debris's attachment point were determined to be the second node east of the tower's top insulator string. Combined with the risk indicator for the high-risk conductive zone, the system generated a transmission tower abnormality warning instruction containing these coordinates, clearly indicating that the eccentric load caused by the foreign body's attachment had caused a coupled deformation-stress transmission mechanism to form within the tower. Following this instruction, operations and maintenance personnel prioritized removing the fishing net debris from the tower and reinforcing the high-risk conductive zone. Post-processing monitoring data showed a 70% decrease in the rate of displacement, and the loads at key nodes in the stress diffusion path network returned to safe thresholds, validating the effectiveness of the warning instruction in protecting the transmission tower structure.

[0090] In summary, steps 101 to 106 implement multimodal data fusion and intelligent early warning for detecting foreign objects hanging on transmission and distribution lines. By combining dynamic registration of the transmission tower surface point cloud with time-frequency analysis of strain signals, a dual detection mechanism of deformation gradient maps and strain spectra is constructed, effectively improving the comprehensiveness and accuracy of foreign object detection. By calculating the spatial superposition of the deformation propagation path network and the stress diffusion path network, fusing displacement offsets and load change values ​​to generate risk indicators, and combining the geometric parameters of the foreign object contour to determine the attachment location, a closed-loop early warning system from foreign object identification to risk assessment is implemented, providing intelligent decision-making support for the safe operation and maintenance of transmission towers.

[0091] In some embodiments, step 105 includes spatially superimposing the deformation propagation path network and the stress diffusion path network, and fusing the displacement offset and load change value in the overlapping region of the spatial superposition to generate a risk indicator, including:

[0092] 201. Determine an overlapping area of ​​the deformation propagation path network and the stress diffusion path network on a preset spatial coordinate. A condition for determining the overlapping area is that a range of spatial coordinates corresponding to a displacement offset in the deformation propagation path network overlaps a range of spatial coordinates corresponding to a load change value in the stress diffusion path network.

[0093] In step 201, the overlapping area refers to the spatially overlapping portion of different networks. The judgment condition refers to the logical rules for determining whether the area overlaps. Range overlap refers to the intersection of different parameters in the spatial coverage area. Existence refers to the inclusion relationship confirmed through calculation.

[0094] In an embodiment of the present application, first, the deformation propagation path network generated in step 103 and the stress diffusion path network generated in step 104 are imported into a unified three-dimensional coordinate system. Secondly, the spatial grid division technology is used to discretize the transmission tower structure into a number of cubic units, and a spatial index is established for each unit. Then, the units with the same spatial index in the two networks are retrieved through the spatial query algorithm, and the areas where these units are located are determined to be superimposed areas. Then, the judgment conditions of these areas are verified: check whether the coordinate range corresponding to the displacement offset recorded in the deformation propagation path network has at least 30% overlap with the coordinate range of the load change value in the stress diffusion path network. Finally, the spatial boundary coordinate set of all superimposed areas that meet the conditions is output to provide a basis for regional delimitation for weight calculation.

[0095] 202. Calculate a fusion weight factor for each superimposed region based on a change degree of the displacement offset and an increase or decrease trend of the load change value in the superimposed region;

[0096] In step 202, the degree of change refers to the amplitude of the displacement fluctuation. The increasing or decreasing trend refers to the direction of change in the load change value. The fusion weight factor refers to the adjustment coefficient that reflects the importance of different parameters. Calculation refers to the process of obtaining characteristic parameters through mathematical operations.

[0097] In an embodiment of the present application, first, in each determined superposition area, the time series data of the displacement offset recorded in step 103 is extracted, and its standard deviation is calculated as a quantitative indicator of the degree of change. Secondly, the time series data of the load change value in the same area is obtained from the data of step 104, and the trend analysis algorithm is used to determine its increase or decrease trend (such as linear rise, fluctuation, etc.). Then, a weight calculation model is established, the degree of change of the displacement offset is normalized to a numerical value in the range of 0-1, and the load change trend is converted into a trend coefficient (an upward trend is 1, a downward trend is 0.5, etc.). Then, the two coefficients are synthesized into a fusion weight factor through a weighted sum formula, and the contribution ratio of the two types of parameters is set according to engineering experience (such as displacement weight 0.6, load weight 0.4). Finally, a fusion weight factor value that comprehensively reflects the deformation and stress characteristics is output for each superposition area.

[0098] 203. The displacement offset and load change value of each coordinate point in the superposition area are superimposed and calculated according to the corresponding fusion weight factor to obtain the risk index of each coordinate point.

[0099] In step 203, coordinate points refer to specific locations in space. Overlay calculation refers to the mathematical combination of different parameters. Risk indicators refer to comprehensive parameters used to quantitatively assess the safety status of a structure. Obtaining refers to the process of obtaining a result through calculation.

[0100] In an embodiment of the present application, first, a high-density three-dimensional coordinate point grid (spacing 5 cm) is established in each superposition area delineated in step 201. Secondly, the displacement offset and load change value corresponding to each coordinate point are calculated by a three-dimensional interpolation algorithm. Then, according to the superposition area to which the coordinate point belongs, the fusion weight factor calculated in step 202 is assigned. Then, a linear weighted method is used to superimpose the displacement offset and load change value of each point according to the weight factor to obtain the initial risk value of the point. Finally, abnormal fluctuations are eliminated through spatial smoothing, and the processed risk value is normalized to a risk index in the range of 0-100, and the risk distribution is intuitively displayed in a gradient color in the three-dimensional model, completing the complete transformation process from raw data to risk assessment.

[0101] Here's a specific example:

[0102] In the intelligent inspection system for power transmission and distribution lines, multi-source data fusion technology enables precise location and risk assessment of suspended foreign objects. For a plastic film hanging hazard discovered during a drone inspection of a mountainous line section, the system first determined the conductor deformation propagation path caused by the foreign object hanging point within a 3D point cloud model (step 201). It then identified the overlapping region where the conductor sag section overlapped with the stress diffusion path of the adjacent tower. This region was characterized by the spatial overlap between the most significant conductor deformation location and the tension monitoring anomaly section. Based on the displacement monitoring data of the conductor nodes within this overlapping region (step 202), the system calculated the synergy index between the deformation gradient and the tension change, generating a fusion weight factor reflecting the mechanical influence of the plastic film hanging point. Finally, the deformation offset and tension anomaly values ​​for each conductor node were dynamically weighted (step 203), generating a heat map showing high-risk indicators 15 meters downstream of the hanging point. When new inspection data exhibited similar characteristics to the overlapping region, the system automatically triggered a red alert and generated a remediation plan, guiding operations and maintenance personnel to prioritize addressing the hazard in that section. This case optimized the deformation-stress coupling analysis algorithm, improved the early identification capability of thin film foreign matter, and formed an intelligent inspection closed loop from multi-source data analysis to risk disposal.

[0103] In summary, steps 201 to 203 achieve dynamic risk quantification and precise fusion of deformation and stress path networks. By spatially superimposing the deformation propagation path network and the stress diffusion path network, the displacement offset and load change values ​​are fused in the overlapping area, and a risk indicator calculation model based on spatial correlation is constructed. This method innovatively introduces a fusion weight factor to dynamically adjust the contribution of different areas according to the degree of displacement change and the load increase and decrease trend, solving the limitations of independent analysis of deformation and stress data in traditional detection. By fusion of coordinate point-level data within the superposition area, a refined quantitative expression of risk indicators is achieved, providing multi-dimensional, high-precision data support for the safety assessment of transmission tower structures, and significantly improving the spatial resolution and dynamic response capabilities of risk prediction.

[0104] In some embodiments, the surface point cloud set and the historical point cloud set are dynamically registered based on the curvature difference of adjacent point clouds to generate a deformation gradient map, as described in step 101, wherein the point cloud region with a curvature difference exceeding a set range during dynamic registration is marked as a deformation correlation area of the deformation gradient map, including:

[0105] 301、extracting the three-dimensional geometric features of adjacent point clouds in the surface point cloud set and the historical point cloud set, calculating the curvature difference of the corresponding regions between the adjacent point clouds, the adjacent point clouds being determined by the near-neighbor relationship of spatial coordinates;

[0106] In step 301, the three-dimensional geometric features refer to geometric attribute parameters describing the spatial form of the point cloud. The corresponding region refers to a region with the same spatial position in different point cloud sets. The near-neighbor relationship refers to the neighborhood relationship of point clouds determined by spatial distance. The calculation refers to the process of obtaining feature parameters through mathematical operation.

[0107] In the embodiments of the present application, first, the surface point cloud set of the current detection period of the transmission tower is obtained through the three-dimensional laser scanning system, and the historical point cloud set of the reference state is retrieved from the historical database. Second, the spatial topological relationship of the point cloud is established by using the k-d tree spatial index algorithm, and the corresponding adjacent point cloud regions in the two point cloud sets are determined based on the near-neighbor relationship of spatial coordinates. Then, the three-dimensional geometric features including normal vector, curvature and surface roughness are extracted for each adjacent point cloud region, and the curvature difference value between the corresponding regions is calculated. Finally, the calculated curvature difference value is organized and stored according to the spatial coordinates to form a curvature difference distribution map covering the surface of the transmission tower.

[0108] 302、according to the distribution range of the curvature difference, set a dynamically adjusted threshold, and mark the point cloud region with a curvature difference exceeding the dynamically adjusted threshold as an initial deformation area;

[0109] In step 302, the dynamically adjusted threshold refers to a judgment standard that changes adaptively according to data characteristics. The initial deformation area refers to a point cloud region that is preliminarily judged to have deformation. The marking refers to the operation of assigning an identifier to specific data. The distribution range refers to the variation interval of the feature parameter in space.

[0110] In the embodiment of the present application, first, all the curvature difference values ​​calculated in step 301 are statistically analyzed and their probability density distribution curves are drawn. Secondly, based on the statistical 3σ principle and engineering experience, an initial dynamic adjustment threshold is set in the tail area of ​​the distribution curve. Then, a spatial clustering algorithm is used to identify all point cloud areas where the curvature difference exceeds the dynamic adjustment threshold, and these areas are marked as initial deformation areas. Finally, a morphological expansion process is performed on the initial deformation area to ensure that the potential deformation area is completely covered, and the spatial position and curvature difference value of each initial deformation area are recorded.

[0111] 303. Based on a changing trend of curvature differences between adjacent point clouds in the initial deformation region, dynamically align the correspondence between the surface point cloud set and the historical point cloud set, so that the curvature difference in the initial deformation region converges to a dynamic adjustment threshold.

[0112] In step 303, the trend of change refers to how the curvature difference changes over time or space. Convergence refers to the process by which a parameter gradually approaches its target value. Correspondence refers to the spatial mapping relationship between different point cloud sets. Dynamic registration refers to the method of establishing point cloud correspondences through iterative optimization.

[0113] In the embodiment of the present application, first, within the initial deformation zone determined in step 302, the trend of change of the curvature difference of adjacent point clouds over time is analyzed to determine the direction of deformation development. Secondly, a non-rigid ICP registration algorithm is used, with the minimization of curvature difference as the objective function, to dynamically align the correspondence between the current surface point cloud set and the historical point cloud set. Then, the spatial correspondence of the point clouds is continuously adjusted through iterative optimization, so that the curvature difference in the initial deformation zone gradually converges to the dynamically adjusted threshold. Finally, the displacement vector field of the point cloud during the registration process is recorded to provide basic data for subsequent deformation gradient calculations.

[0114] 304. Calculate the curvature difference variation of each point cloud region based on the corresponding relationship after dynamic registration, and accumulate the variation along a preset three-dimensional spatial direction to generate a deformation gradient map.

[0115] In step 304, the change refers to the quantitative value of the difference between the parameters before and after. Accumulation refers to the operation of adding values ​​along a specific direction. The preset three-dimensional spatial direction refers to the pre-set X / Y / Z analysis direction. Generation refers to the process of obtaining the final result through calculation.

[0116] In the embodiment of the present application, first, based on the correspondence after dynamic registration in step 303, the change in curvature difference of each point cloud area before and after registration is calculated. Secondly, the space is divided into a regular voxel grid, and the change in curvature difference is accumulated in each voxel along the three preset three-dimensional spatial directions of X, Y, and Z. Then, the discrete accumulation results are smoothed by the cubic spline interpolation algorithm to generate a continuous gradient distribution field. Finally, the gradient components in the three directions are synthesized into a deformation gradient map in vector form, and the size and direction of the deformation are intuitively displayed using color and arrows.

[0117] 305. Perform spatial continuity verification on regions in the deformation gradient map where the gradient value distribution continuously exceeds the dynamic adjustment threshold, and mark regions that meet the spatial continuity distribution condition as deformation-related regions.

[0118] In step 305, gradient numerical distribution refers to the quantitative spatial distribution characteristics of deformation gradients. Spatial continuity verification refers to the process of determining whether regions are spatially connected. Distribution conditions refer to the geometric criteria for determining regional continuity. Marking refers to the operation of identifying regions that meet the conditions.

[0119] In the embodiment of the present application, first, the deformation gradient map generated in step 304 is threshold segmented to extract all areas where the gradient values ​​continuously exceed the dynamically adjusted threshold. Secondly, a three-dimensional connected domain analysis algorithm is used to verify the spatial continuity of these areas and eliminate isolated noise points. Then, the geometric features such as the volume, surface area and shape factor of each connected domain are calculated, and the deformation areas that meet the engineering significance are screened out based on the preset spatial continuity distribution conditions. Finally, these areas are marked as deformation-related areas in the three-dimensional model, and their characteristic parameters such as spatial range, gradient size and direction are recorded to provide a basis for subsequent risk assessment.

[0120] Here's a specific example:

[0121] In the intelligent deformation monitoring system for power transmission and distribution lines, three-dimensional point cloud dynamic analysis technology enables accurate identification of abnormal deformation of conductors. For drone inspection data of a certain cross-river line section, the system first extracts the three-dimensional geometric features of the current surface point cloud set and the historical benchmark point cloud (step 301). By calculating the curvature difference of adjacent point clouds in the area where the conductor sag is lowest, it is found that this area has a characteristic bimodal distribution pattern. Based on the distribution characteristics of the curvature difference (step 302), the system sets a dynamic adjustment threshold and marks the middle section of the conductor where the curvature difference exceeds the threshold as the initial deformation zone. By analyzing the curvature change trend of adjacent point clouds in the initial deformation zone (step 303), the system dynamically aligns the correspondence between the current and historical point clouds, so that the curvature difference in the middle section gradually converges to a reasonable range. Based on the aligned correspondence (step 304), the system calculates the cumulative change of the curvature difference of each point cloud area along the conductor axis, generating a deformation gradient map showing that there is a significant gradient mutation in the middle section. The gradient map's persistently exceeded areas are spatially verified (step 305), ultimately marking the 15-meter section of the conductor's middle section, which meets the continuity criteria, as a deformation-related zone. When new inspection data exhibits similar gradient mutation characteristics, the system automatically triggers an abnormal conductor deformation warning, directing operations and maintenance personnel to focus on investigating potential external force damage hazards in this section. This case optimizes the dynamic threshold adjustment algorithm, increasing the sensitivity of detecting early conductor deformation characteristics and forming an intelligent monitoring closed loop from point cloud difference analysis to hazard warning.

[0122] In summary, steps 301 to 305 realize dynamic point cloud registration and intelligent identification of deformation areas based on curvature differences. By calculating the curvature difference between the surface point cloud and the historical point cloud, dynamically adjusting the registration threshold and marking the initial deformation area, adaptive registration of point cloud data is achieved. Based on the convergence mechanism of the curvature difference change trend, the point cloud correspondence is optimized to ensure the stability and accuracy of the registration process. The deformation gradient map is generated by accumulating the curvature difference changes in the three-dimensional space direction, and the deformation association area is determined in combination with spatial continuity verification, which significantly improves the robustness of foreign body contour detection. This technical solution breaks through the limitations of traditional static registration methods, provides high-precision deformation benchmark data for subsequent strain signal mapping and risk analysis, and enhances the adaptability of foreign body detection in complex environments.

[0123] In some embodiments, as described in step 304, calculating the change in curvature difference of each point cloud region based on the correspondence after dynamic registration, and accumulating the change along a preset three-dimensional spatial direction to generate a deformation gradient map includes:

[0124] 401. Mapping the surface point cloud set and the three-dimensional geometric features of each point cloud in the historical point cloud set according to the corresponding relationship after dynamic registration, and determining the point cloud area where the curvature difference changes in the corresponding relationship;

[0125] In step 401, mapping refers to establishing spatial correspondences between different point cloud sets. Determination refers to the process of reaching a clear conclusion through calculation. Curvature difference refers to the change in curvature parameters at different time points. A point cloud region refers to a set of spatial points with the same characteristics.

[0126] In an embodiment of the present application, first, based on the correspondence after the dynamic registration completed in step 303, a point-to-point mapping relationship table is established between the current surface point cloud set and the historical point cloud set. Secondly, the three-dimensional geometric features (including curvature, normal vector, etc.) of each point in the two sets of point clouds are accurately aligned through a feature matching algorithm. Then, a differential calculation method is used to compare the curvature values ​​of the corresponding point clouds before and after registration point by point to identify point cloud areas with obvious curvature differences. Finally, a spatial clustering algorithm is used to cluster these abnormally changing point clouds into continuous areas, and the spatial boundaries and change intensity are marked for each area.

[0127] 402. Calculate the change in curvature difference of each point cloud region before and after dynamic registration, and accumulate the change in curvature difference of each point cloud region along a preset three-dimensional spatial direction, where the three-dimensional spatial direction is determined by the geometric feature distribution of the surface point cloud set.

[0128] In step 402, "before and after" dynamic registration refers to the time points before and after the optimization of the correspondence relationship. "Preset 3D spatial direction" refers to the pre-set analysis direction. "Geometric feature distribution" refers to the spatial distribution of the morphological features of the point cloud. "Accumulation" refers to the operation of numerical superposition along a specific direction.

[0129] In an embodiment of the present application, first, in each point cloud area determined in step 401, the curvature difference data of all points before and after dynamic alignment are extracted. Secondly, based on the geometric feature distribution of the surface point cloud set (such as the main curvature direction), the three preset three-dimensional spatial directions of X, Y, and Z are determined as the accumulation reference axes. Then, the curvature difference change of each point is vector-decomposed along the determined three-dimensional spatial direction, and the component values ​​in the three directions are calculated respectively. Finally, a spatial grid accumulation algorithm is used to accumulate the component values ​​in each direction within its corresponding spatial grid unit to obtain the total distribution of curvature difference changes in the three directions.

[0130] 403. Correspond the accumulated curvature difference variation to the preset three-dimensional space coordinates in each point cloud region to generate a deformation gradient map.

[0131] In step 403, 3D spatial coordinates refer to the positional data of the point cloud in 3D space. Correspondence refers to establishing a relationship between parameters and spatial positions. Generation refers to the process of obtaining the final result through calculation. The deformation gradient map refers to a visualization model that reflects the spatial distribution of deformation.

[0132] In the embodiments of the present application, first, the accumulated results of the change amount of the curvature difference in the three directions calculated in step 402 are normalized to eliminate dimensional differences. Second, a three-dimensional spatial coordinate grid system consistent with the spatial distribution of the point cloud is established, and the accumulated value of each grid element is accurately corresponded to the coordinate position. Then, the discrete grid data is converted into a continuous gradient field by a three-dimensional interpolation algorithm to generate a vector-type deformation gradient map containing size and direction information. Finally, the visualization rendering technology is used to superimpose the deformation gradient map on the three-dimensional model of the transmission tower, and the gradient intensity is represented by the color depth and the deformation trend is represented by the arrow direction, completing the visualization conversion from point cloud data to deformation features.

[0133] The following is a specific example:

[0134] In the intelligent deformation monitoring system of transmission and distribution lines, the three-dimensional point cloud dynamic analysis technology realizes the accurate positioning of the abnormal deformation of the conductor. For the unmanned aerial vehicle inspection data of a line section in a mountainous area, the system first maps the features of the current surface point cloud set and the historical reference point cloud based on the corresponding relationship after dynamic registration (step 401) to accurately locate the point cloud area of the curvature difference change at the conductor sag. By calculating the curvature difference change amount of each point cloud area before and after registration (step 402), the system performs three-dimensional space accumulation along the axial and radial directions of the conductor to find that the 15-meter area in the middle section has a characteristic gradient accumulation effect. The accumulated results are corresponded to the point cloud spatial coordinates (step 403) to generate a deformation gradient map showing that there is a significant gradient mutation in the middle section of the conductor. When the new inspection data presents similar gradient characteristics in the same section, the system automatically associates the strong wind record in the meteorological monitoring data to judge that it is a conductor deformation risk caused by wind deviation, guiding the operation and maintenance personnel to check the fittings in this section, and confirming that there is a bolt loosening hidden danger on site. This case optimizes the three-dimensional gradient accumulation algorithm and improves the early warning capability of the conductor mechanical deformation, forming an intelligent diagnosis closed loop from point cloud dynamic analysis to defect positioning.

[0135] In summary, steps 401 to 403 realize the gradient representation and stereoscopic analysis of the curvature difference in the three-dimensional space. Through the point cloud mapping relationship after dynamic registration, the curvature difference change amount is calculated and accumulated along the three-dimensional spatial direction to construct a spatial quantization model of the deformation gradient map. This method innovatively accumulates the curvature change amount along the three-dimensional direction determined by the geometric feature distribution, breaks through the limitations of traditional two-dimensional gradient calculation, and realizes the stereoscopic representation of the deformation features. Through the accurate correspondence of the three-dimensional spatial coordinates, the deformation gradient map can fully reflect the spatial deformation distribution characteristics of the surface structure of the transmission tower, providing a high-resolution gradient data basis for the generation of the deformation propagation path network, and significantly improving the dimension and accuracy of the deformation propagation path analysis.

[0136] In some embodiments, as described in step 102, collecting strain signals at the connection nodes of the transmission towers, performing time-frequency domain energy decomposition on the strain signals to retain steady-state components, and mapping the steady-state components to positions corresponding to deformation-related regions to generate a strain map may include:

[0137] 501. Collect a strain signal at a connection node of a transmission tower, cut the strain signal into multiple time segments according to a preset time length, divide the signal in each time segment into multiple continuous frequency intervals, and calculate the energy value of each combination of the time segment and the frequency interval;

[0138] In step 501, the strain signal refers to the measured deformation data of the transmission tower connection node under stress. The preset time length refers to the time window size of the signal analysis. The time segment refers to the signal segment divided by time. The frequency interval refers to the sub-band range of the signal frequency domain analysis. The energy value refers to the energy distribution value of the signal in the time-frequency domain. The calculation refers to the process of obtaining characteristic parameters through mathematical operations.

[0139] In the embodiment of the present application, first, high-precision strain sensors are installed at the key connection nodes of the transmission tower to collect structural strain signals in real time at a fixed sampling frequency. Secondly, the sliding window technology is used to cut the continuous strain signal into multiple time segments according to a preset time length (such as 5 minutes), ensuring that each segment contains a complete load cycle. Then, the wavelet packet transform algorithm is applied to each time segment to decompose the time domain signal into multiple continuous frequency intervals (such as 0-10Hz, 10-20Hz, etc.). Finally, the energy value of each time segment and frequency interval combination is calculated to form a three-dimensional feature matrix of time-frequency-energy, which lays the data foundation for the extraction of steady-state components.

[0140] 502. Filtering the units whose energy value fluctuation amplitude over time is lower than a preset fluctuation threshold as steady-state components;

[0141] In step 502, the fluctuation amplitude refers to the range of energy fluctuations over time. The preset fluctuation threshold refers to the critical value for determining signal stability. The steady-state component refers to the stable signal component with small energy fluctuations. Screening refers to the process of extracting target data based on conditions. A unit refers to a combination of a time segment and a frequency bin.

[0142] In the embodiments of the present application, firstly, the standard deviation of each frequency interval in the continuous time segment is calculated as the fluctuation amplitude index. Secondly, the frequency intervals with a fluctuation amplitude lower than the preset fluctuation threshold (for example, 0.3 for steel and 0.5 for concrete) are screened out based on the material characteristics. Then, the stability of these intervals is verified through time correlation analysis, and the accidental low fluctuation intervals are removed. Finally, the frequency intervals meeting the stability requirements are marked as steady-state components, and their energy values and time distribution characteristics are recorded to provide reliable data for subsequent spatial mapping.

[0143] 503. According to the range of spatial coordinates of the deformation associated area, the energy value of the steady-state component corresponding to each connection node is mapped to the spatial coordinates of the corresponding deformation associated area;

[0144] In step 503, spatial coordinates refer to position positioning data in three-dimensional space. Mapping refers to the process of corresponding data to spatial positions. Deformation associated area refers to the key area of interest where the deformation exceeds the set threshold. Correspondence refers to the matching association between different data. Connection node refers to the key connection part in the power tower structure.

[0145] In the embodiments of the present application, firstly, the spatial coordinate range of the deformation associated area determined in step 305 is obtained from the three-dimensional modeling system to establish the spatial index of the deformation area. Secondly, the spatial coordinates of each connection node are determined according to the sensor installation position, and the spatial distance from each connection node to each deformation associated area is calculated. Then, the nearest neighbor matching algorithm is used to assign the energy value of the steady-state component of each connection node to the nearest deformation associated area. Finally, the energy value is accurately mapped to the spatial coordinate system of the deformation associated area through coordinate transformation to establish the energy-space correspondence table, ensuring accurate association between the mechanical response and the deformation area.

[0146] 504. The energy values of the steady-state components of multiple connection nodes mapped to the same deformation associated area are spatially weighted and fused to obtain an energy fusion value, and a strain map is generated according to the correspondence between the energy fusion value and the spatial coordinates.

[0147] In step 504, spatially weighted fusion refers to a numerical integration method considering the spatial position relationship. Energy fusion value refers to the integrated steady-state energy value. Strain map refers to a visual model reflecting the strain distribution characteristics of the structure. Generation refers to the process of obtaining the final result through calculation. The same deformation associated area refers to the area with overlapping spatial coordinates.

[0148] In an embodiment of the present application, first, for the steady-state component energy values ​​of multiple connection nodes mapped to the same deformation association zone, the spatial weight coefficient is calculated according to the distance between the node and the deformation zone (the closer the distance, the greater the weight). Secondly, a weighted average algorithm is used to fuse the energy values ​​of multiple nodes according to the spatial weight to obtain an energy fusion value that characterizes the overall strain characteristics of the deformation zone. Then, the discrete fusion value is expanded into a continuous spatial distribution field through the Kriging interpolation algorithm. Finally, the energy fusion value is associated with the spatial coordinates of the deformation association zone to generate a strain map that uses a color gradient to represent the strain intensity, where the red area represents the high strain area and the blue represents the low strain area, providing an intuitive basis for structural safety assessment.

[0149] Here's a specific example:

[0150] In the intelligent monitoring system for power transmission and distribution lines, multi-source data fusion technology enables accurate diagnosis of tower structural abnormalities. For the online monitoring data of a cross-regional transmission line, the system first collects strain signals from the key connection nodes of the tension tower (step 501), divides the 24-hour continuous signal into minute-level segments, and performs frequency domain energy analysis to identify the characteristic low-frequency energy accumulation at the connection between the crossarm and the main material. By screening the steady-state components with energy fluctuations below the safety threshold (step 502), the system locks on the abnormal energy distribution pattern of the connection node on the southwest side of the tower body. Based on the coordinate range of the deformation association area determined by the previous three-dimensional point cloud analysis (step 503), the system maps the steady-state energy value of each node to the tower body structure model and finds that there is an abnormal energy concentration phenomenon in the southwest diagonal material connection area. The energy values ​​of multiple connection nodes mapped to this area are spatially weighted fused (step 504) to generate a strain map showing the presence of high energy values ​​at the diagonal material connection. When new monitoring data shows similar energy distribution at the same location, the system automatically links it to high wind records from meteorological monitoring, identifies a risk of loose structural bolts, and directs maintenance personnel to conduct on-site inspections to confirm the presence of missing nuts. This case study optimizes the strain-energy spatial mapping algorithm, enhancing the ability to early-warn hidden tower defects and forming an intelligent diagnostic closed loop from multi-source signal analysis to structural hazard location.

[0151] In summary, steps 501 to 504 achieve the precise extraction and spatial energy mapping of the steady-state components of the strain signal. The steady-state components with fluctuation amplitudes below the threshold are screened by time-frequency domain energy decomposition technology, effectively filtering out the influence of environmental noise and transient interference. Based on the spatial coordinate range of the deformation association area, the energy values ​​of the steady-state components are mapped to the corresponding areas and spatially weighted fusion is performed to construct a strain map generation mechanism. This method enhances the spatial continuity expression of strain data through the spatial fusion of multi-connected node data, providing highly reliable energy distribution data for the construction of stress diffusion path networks. This technical solution realizes the precise transformation of strain signals from temporal characteristics to spatial distribution, providing a high-fidelity energy distribution benchmark for structural stress analysis.

[0152] In some embodiments, step 104 includes obtaining load change values ​​of connected nodes in the strain map and associating nodes whose load change amplitudes exceed a preset threshold to construct a stress diffusion path network, including:

[0153] 601. Extracting a load change value of each connection node in the strain map, where the load change value is derived from an energy value corresponding to a steady-state component in the strain map;

[0154] In step 601, the load change value refers to the change in the force applied to the connection node. Extraction refers to the analytical method used to obtain specific information from data. Source refers to the basis for generating the data. Strain map refers to a visual model that reflects the strain distribution characteristics of the structure. Steady-state component refers to a stable signal component with minimal energy fluctuation.

[0155] In the embodiment of the present application, first, the load change value of each connection node is extracted from the strain map generated in step 504. These values ​​are directly derived from the energy values ​​corresponding to the verified steady-state components in the strain map. Secondly, a time series database is established for each connection node to record the fluctuations of its load change value in different time periods. Then, a data cleaning algorithm is used to remove abnormal fluctuation data and retain the effective load change value that reflects the true structural response. Finally, the processed load change value is organized and stored according to the node number and spatial coordinates to provide standardized input data for subsequent threshold analysis.

[0156] 602. Calculate a dynamic fluctuation threshold of the load change amplitude based on the distribution characteristics of the load change values ​​of all connected nodes, where the dynamic fluctuation threshold is determined by the sum of the median and the standard deviation of the load change amplitude.

[0157] In step 602, distribution characteristics refer to the spatial or statistical distribution patterns of parameters. Dynamic fluctuation thresholds refer to adaptively changing abnormality judgment criteria. Median refers to the middle value of a data set. Standard deviation measures the degree of data dispersion. Determination refers to the process of reaching a clear conclusion through calculation.

[0158] In the embodiment of the present application, first, the load change values ​​of all connected nodes are statistically analyzed, and the probability distribution curve of their variation range is drawn. Secondly, the median of the variation range is calculated as the reference value, and the fluctuation range is determined in combination with the standard deviation. Then, a robust statistical method is adopted to eliminate the influence of outliers, and the sum of the median and the standard deviation is used as the calculation basis of the dynamic fluctuation threshold. Then, the threshold is coefficient-adjusted according to the structural safety level (such as an important node multiplied by a safety factor of 1.2). Finally, a dynamic fluctuation threshold table applicable to different types of nodes is output to provide a quantitative standard for key node identification.

[0159] 603. Filter the connection nodes whose change amplitude exceeds the dynamic fluctuation threshold and mark these nodes as key nodes;

[0160] In step 603, key nodes refer to important connection points where load changes significantly. Labeling refers to the process of assigning a label to specific data. Screening refers to the process of extracting target data based on conditions. Variation amplitude refers to the intensity range of parameter fluctuations. Exceeding the index value is greater than the threshold.

[0161] In an embodiment of the present application, first, the amplitude of the load change value of each connection node is compared with the dynamic fluctuation threshold determined in step 602. Secondly, the threshold screening algorithm is used to identify all nodes whose amplitude of change exceeds the threshold. These nodes reflect abnormal stress concentration phenomena. Then, based on the node importance assessment model (taking into account the position criticality and historical damage records), the screened nodes are secondary verified. Finally, the confirmed abnormal nodes are marked as key nodes in the three-dimensional model, and their spatial coordinates, amplitude of change, and specific values ​​exceeding the threshold are recorded to form a list of key nodes.

[0162] 604. Analyze the spatial proximity relationship between the key nodes, and determine the association direction between the key nodes based on the increase and decrease trends of the coordinate distances and change amplitudes of the key nodes in the preset three-dimensional space;

[0163] In step 604, spatial proximity refers to the spatial distance between nodes. Coordinate distance refers to the three-dimensional distance between nodes. Increase / decrease trend refers to the upward or downward trend of parameter changes. Correlation direction refers to the main direction of mechanical transmission between nodes. Analysis refers to the process of in-depth analysis of data.

[0164] In an embodiment of the present application, first, a spatial proximity relationship network of key nodes is established, and the coordinate distance of each pair of nodes in a preset three-dimensional space is calculated. Secondly, the temporal correlation of the change amplitude of the load change value between adjacent key nodes is analyzed to determine whether their increase and decrease trends are synchronized. Then, the stress transfer direction is determined by vector analysis, and the node pairs with the same trend and the closest spatial distance are marked as potential propagation paths. Finally, the two factors of spatial distance and trend similarity are combined to determine the association direction between each pair of key nodes (such as the positive association from node A to node B) to provide directional guidance for path construction.

[0165] 605. Connect the key nodes in sequence according to the association direction to form a continuous path, and for the fracture area in the continuous path caused by the coordinate distance exceeding the preset range, complete the fracture area through the increase and decrease trend of the load change value of the adjacent key nodes to generate a stress diffusion path network, wherein the adjacent key nodes are determined by the spatial proximity relationship between the key nodes.

[0166] In step 605, a continuous path refers to the coherent connection between nodes. A broken region refers to the discontinuous portion of a path. Completion refers to the operation of supplementing missing data. Generation refers to the process of obtaining the final result through calculation. The stress diffusion path network refers to the path relationship model of stress transmission in the structure.

[0167] In the embodiment of the present application, first, according to the association direction determined in step 604, a path tracing algorithm is used to connect the key nodes in sequence to form an initial continuous path. Secondly, the fracture area caused by the large node spacing in the path is detected, and virtual nodes are added in the fracture interval by the spatial interpolation method. Then, according to the increase and decrease trend of the change amplitude of the load change value of the adjacent key nodes, the stress propagation direction of the fracture area is determined, and the missing path segment is completed. Finally, redundant connections are eliminated through topology optimization, and a stress diffusion path network that reflects the actual diffusion law of stress is constructed. The network displays the transfer path and intensity change of stress from high-value areas to low-value areas in the form of a directed graph.

[0168] Here's a specific example:

[0169] In the structural health monitoring system for power transmission and distribution lines, stress propagation path analysis technology enables precise localization of mechanical anomalies in towers. Based on long-term monitoring data from a 500kV transmission line tension tower, the system first extracts the load change values ​​for each connection node from the strain map (step 601). Significant energy concentration is detected at the diagonal connection on the southwest side of the tower. By analyzing the distribution characteristics of load changes at all nodes (step 602), the system calculates a dynamic fluctuation threshold and identifies five key nodes exceeding the threshold (step 603). These nodes exhibit a spatial distribution pattern extending from the tower base to the crossarm. Based on the spatial coordinate analysis of the key nodes (step 604), the system determines the associated direction of load changes propagating upward along the southwest diagonal. The key nodes are connected along this direction (step 605). After completing the path breakage caused by missing monitoring points, a complete path network is generated, showing the stress diffusion from the tower base through the diagonal to the crossarm. When new monitoring data reproduced similar stress diffusion patterns along the same path, the system automatically linked it to meteorological monitoring records of icing, identifying a risk of abnormal structural stress. Operations and maintenance personnel were instructed to conduct a comprehensive inspection of the southwest diagonal material, where microcracks were discovered. This case optimized the stress path completion algorithm, enhancing the ability to warn of hidden tower damage and forming an intelligent monitoring closed loop from mechanical feature analysis to structural defect diagnosis.

[0170] In summary, steps 601 to 605 achieve intelligent construction of stress diffusion paths and path continuity assurance based on dynamic thresholds. The dynamic fluctuation threshold is calculated based on the distribution characteristics of the load change value, key nodes are screened and their spatial proximity is analyzed, and an adaptive generation model of the stress diffusion path network is constructed. This method innovatively combines the coordinate distance of key nodes with the trend of change amplitude to determine the correlation direction, and uses the data of neighboring nodes to complete the path break area, solving the problem of broken links caused by missing data in traditional path construction. This technical solution realizes the integrity and continuity expression of the stress diffusion path, provides accurate stress propagation path data for the fusion calculation of risk indicators, and significantly improves the reliability and spatial coverage capability of stress diffusion path analysis.

[0171] In some embodiments, as described in step 106, determining the coordinates of the foreign object attachment location based on the geometric parameters of the surface foreign object profile, and generating a transmission tower abnormality warning instruction including the coordinates of the foreign object attachment location in combination with the risk indicator, includes:

[0172] 701. Extracting boundary coordinates from the geometric parameters of the surface foreign body contour, and using the boundary coordinates as initial position coordinates of the foreign body attachment area;

[0173] In step 701, boundary coordinates refer to the spatial location data describing the edge of the foreign object's contour. Geometric parameters refer to a set of metrics that quantitatively describe the foreign object's shape characteristics. Initial position coordinates refer to the preliminary determination of the foreign object's attachment location. Extraction refers to an analytical method for obtaining specific information from data. The foreign object attachment area refers to the area of ​​the transmission tower surface covered by foreign objects.

[0174] In the embodiment of the present application, first, the three-dimensional coordinates of the contour boundary points are extracted from the surface foreign body contour data obtained in step 101. These coordinates constitute a set of boundary coordinates that describe the shape of the foreign body. Secondly, the convex hull algorithm is used to calculate the minimum bounding box of the foreign body contour, and the coordinates of the four corner points of the bounding box are extracted as the initial position coordinates. Then, the initial position coordinates are unified into the overall coordinate system of the transmission tower through coordinate transformation to ensure that the same reference system is used for subsequent deformation analysis. Finally, the initial position coordinates are deduplicated and sorted to form standardized foreign body position description data in preparation for precise matching.

[0175] 702. Match the initial position coordinates with the spatial coordinates of the deformation association area according to the spatial coordinate range of the deformation association area, and remove the initial position coordinates that do not fall within the spatial coordinate range to determine the foreign matter attachment position coordinates;

[0176] In step 702, position matching refers to the process of aligning position data in different coordinate systems. Elimination refers to the removal of data items that do not meet the requirements. Determination refers to the process of reaching a clear conclusion through calculation. The spatial coordinate range refers to the boundary range of the deformation association area in three-dimensional space. The foreign object attachment location coordinates refer to the final confirmed spatial location of the foreign object.

[0177] In the embodiment of the present application, first, the spatial coordinate range data of the deformation association area determined in step 305 is called to establish a spatial index structure of the deformation area. Secondly, the spatial relationship between the initial position coordinates obtained in step 701 and the deformation association area range is judged, and the ray method is used to determine whether each coordinate point falls within the polygon range. Then, all initial position coordinates that fail the inclusion test are eliminated, and the coordinate points that are completely within the deformation association area are retained. Finally, a spatial cluster analysis is performed on the filtered coordinate points, and adjacent coordinate points are merged to finally determine the precise set of foreign body attachment position coordinates.

[0178] 703. Determine the value of the risk index corresponding to each foreign object attachment location coordinate, and divide the dynamic safety threshold according to the distribution range of the value of the risk index of all foreign object attachment location coordinates;

[0179] In step 703, the risk indicator refers to a quantitative parameter used to assess the structural safety status. The distribution range refers to the statistical distribution interval of the parameter value. The dynamic safety threshold refers to the safety limit that changes adaptively based on data characteristics. Classification refers to the operation of classifying data according to standards. The value refers to the specific quantitative value of the risk indicator.

[0180] In an embodiment of the present application, first, the risk value corresponding to the coordinates of each foreign body attachment location is queried from the risk index database generated in step 105. Secondly, the risk values ​​of all location points are statistically analyzed and a cumulative distribution curve is drawn. Then, based on the engineering safety standards, key quantiles (such as the top 10% high-risk quantiles) are selected on the distribution curve as the benchmark value of the dynamic safety threshold. Finally, the threshold is dynamically adjusted considering seasonal factors and structural importance coefficients to generate dynamic safety threshold standards that adapt to different working conditions, providing a basis for risk classification.

[0181] 704. Mark the coordinates of the foreign matter attachment location where the value of the risk index exceeds the dynamic safety threshold as high-risk coordinates;

[0182] In step 704, labeling refers to the process of assigning a label to specific data. High-risk coordinates refer to spatial locations where risks exceed safety standards. Exceeding the threshold is the judgment condition. The safety threshold is the critical value for determining the risk level. Determination refers to the process of reaching a clear conclusion through analysis.

[0183] In the embodiment of the present application, first, the value of the risk index of each foreign body attachment location coordinate is compared with the dynamic safety threshold determined in step 703. Secondly, all coordinate points that exceed the safety threshold are identified through a threshold screening algorithm. These points represent locations with higher structural risks. Then, the exceeding coordinate points are sorted and graded based on the size of the risk value, and the coordinate points with a risk value exceeding 1.5 times the threshold are marked as particularly high-risk points. Finally, the risk level of high-risk coordinates is marked with different colors in the three-dimensional model to form an intuitive risk distribution heat map.

[0184] 705. Combining the location information of the high-risk coordinates with the corresponding risk index value, generate a transmission tower abnormality warning instruction including the coordinates of the foreign object attachment location and the risk index.

[0185] In step 705, location information refers to spatial coordinates and related attribute data. Combination refers to the analysis method that associates different parameters. Abnormal warning instructions refer to the alarm information issued based on risk analysis. Generation refers to the process of obtaining the final result through calculation. Transmission tower abnormality warning instructions refer to the alarm content that includes the specific location and severity of the risk.

[0186] In the embodiment of the present application, first, the spatial location information of all high-risk coordinates marked in step 704 and the corresponding risk indicator values ​​are integrated. Secondly, the warning information is classified according to the risk level, and a standardized warning entry containing precise GPS coordinates, risk values ​​and on-site photos is generated. Then, a complete transmission tower abnormality warning instruction is automatically generated through the warning information template, including location description, risk level, recommended disposal measures and other content. Finally, the warning instruction is pushed to the operation and maintenance management system, and the positioning and navigation information is sent to the on-site inspection personnel through the mobile terminal, realizing closed-loop management from risk identification to disposal response.

[0187] Here's a specific example:

[0188] In the intelligent inspection system for power transmission and distribution lines, precise foreign object risk assessment technology enables graded early warnings for hanging foreign objects. For a plastic film hanging defect discovered during a drone inspection of a mountainous line, the system first extracted the boundary coordinates of the foreign object's outline (step 701) and determined its initial attachment position in the middle of the conductor. Combining the deformation association zone determined by previous deformation analysis (step 702), the system matched the initial coordinates with the 15-meter deformation zone in the middle of the conductor to accurately locate the actual attachment location of the foreign object. By calculating the wind-induced vibration response parameters of each attachment point (step 703), the system established a dynamic safety threshold and discovered a significant overshoot at 8 meters in the middle of the conductor. After marking this location as a high-risk coordinate (step 704), the system generated an early warning instruction containing the precise GPS coordinates and risk level (step 705). When a similar hanging object is discovered during a new inspection, the system automatically links it to meteorological warning information, identifying a risk of conductor dancing in windy weather and guiding operations and maintenance personnel to prioritize addressing potential hazards in this section. This case optimized the risk coordinate mapping algorithm, improved the assessment accuracy of the hazard level of suspended foreign objects, and formed an intelligent early warning closed loop from foreign object positioning to risk classification.

[0189] In summary, steps 701 to 705 realize the intelligent matching and early warning decision-making of the foreign body attachment position and risk indicators. The attachment position is determined by spatial matching of the foreign body contour boundary coordinates and the deformation association area, and combined with the dynamic safety threshold division of the risk indicator, the accurate marking of high-risk coordinates is achieved. This method innovatively introduces a dynamic safety threshold mechanism, adaptively adjusts the warning standard according to the distribution range of the risk indicator value, and solves the problem of insufficient adaptability of the traditional fixed threshold. By fusing position information and risk indicator values ​​to generate abnormal warning instructions, a closed-loop response mechanism from detection to warning is constructed, which significantly improves the timeliness and accuracy of the safe operation and maintenance of transmission towers, provides intelligent protection for the stable operation of the power system, and realizes the leap of risk warning from single-location alarm to multi-dimensional data collaborative decision-making.

[0190] Figure 2The present invention provides a schematic diagram of a system for detecting foreign objects hanging on power transmission and distribution lines based on image processing and deep learning. Figure 2 As shown, the system includes:

[0191] an acquisition module 21 for acquiring a surface point cloud set and a historical point cloud set of the transmission tower, dynamically registering the surface point cloud set and the historical point cloud set based on the curvature difference between adjacent point clouds to generate a deformation gradient map, wherein point cloud regions where the curvature difference exceeds a set range during the dynamic registration process are marked as deformation-related regions of the deformation gradient map, and identifying the contour of a surface foreign body based on the curvature difference to obtain geometric parameters of the contour of the surface foreign body;

[0192] An acquisition module 22 is configured to acquire strain signals at the connection nodes of the transmission tower, perform time-frequency domain energy decomposition on the strain signals, retain steady-state components, and map the steady-state components to positions corresponding to deformation-related regions to generate a strain map.

[0193] A calculation module 23 is used to calculate the displacement offsets of adjacent deformation-related areas in the deformation gradient map according to the preset spatial coordinates in the deformation gradient map, and connect the displacement offsets to form a deformation propagation path network;

[0194] An association module 24 is configured to obtain load change values ​​of connected nodes in the strain map and associate nodes whose load change amplitudes exceed a preset threshold value to construct a stress diffusion path network;

[0195] A superposition module 25 is configured to spatially superimpose the deformation propagation path network and the stress diffusion path network, and fuse the displacement offset and the load change value in the overlapping region of the spatial superposition to generate a risk indicator;

[0196] The generating module 26 is configured to determine the coordinates of the foreign body attachment position according to the geometric parameters of the surface foreign body profile, and generate a transmission tower abnormality warning instruction including the coordinates of the foreign body attachment position in combination with the risk index.

[0197] Figure 2 The above-mentioned system for detecting foreign objects hanging on power transmission and distribution lines based on image processing and deep learning can be performed Figure 1 The implementation principles and technical effects of the method for detecting foreign objects hanging from power transmission and distribution lines based on image processing and deep learning, as described in the illustrated embodiment, will not be elaborated upon. The specific manner in which each module and unit performs operations in the aforementioned embodiment of the system for detecting foreign objects hanging from power transmission and distribution lines based on image processing and deep learning has been described in detail in the related embodiments of the method and will not be elaborated upon here.

[0198] In one possible design, Figure 2The embodiment shown is a power transmission and distribution line hanging foreign body detection system based on image processing and deep learning, which can be implemented as a computing device, such as Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32;

[0199] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 32 .

[0200] The processing component 32 is used for the above Figure 1 The embodiment provides a method for detecting foreign objects hanging on power transmission and distribution lines based on image processing and deep learning.

[0201] The processing component 32 may include one or more processors to execute computer instructions to perform all or part of the steps in the above method. Of course, the processing component may also be implemented as one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above method.

[0202] The storage component 31 is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile memory device, or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk, or optical disk.

[0203] Of course, a computing device may also include other components, such as input / output interfaces, display components, communication components, etc.

[0204] The input / output interface provides an interface between the processing component and the peripheral interface module, which can be an output device, an input device, etc.

[0205] The communication component is configured to facilitate, among other things, wired or wireless communications between the computing device and other devices.

[0206] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. In this case, the computing device can refer to a cloud server, and the above-mentioned processing components, storage components, etc. can be basic server resources rented or purchased from the cloud computing platform.

[0207] The present application also provides a computer storage medium storing a computer program, wherein the computer program can achieve the above-mentioned Figure 1 The illustrated embodiment provides a method for detecting foreign objects hanging on power transmission and distribution lines based on image processing and deep learning.

[0208] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0209] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0210] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, or of course, by hardware. Based on this understanding, the essence of the above technical solution or the part that contributes to the existing technology can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or certain parts of the embodiments.

[0211] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for detecting foreign objects hanging on power transmission and distribution lines based on image processing and deep learning, characterized in that: include: Obtaining a surface point cloud set and a historical point cloud set of the transmission tower, dynamically registering the surface point cloud set and the historical point cloud set based on the curvature difference of adjacent point clouds to generate a deformation gradient map, wherein point cloud regions where the curvature difference exceeds a set range during the dynamic registration process are marked as deformation-related regions of the deformation gradient map, and simultaneously identifying the contour of a surface foreign body based on the curvature difference to obtain geometric parameters of the contour of the surface foreign body; Strain signals are collected at the connection nodes of the transmission towers, the strain signals are subjected to time-frequency domain energy decomposition, steady-state components are retained, and the steady-state components are mapped to positions corresponding to deformation-related regions to generate a strain map. Calculating displacement offsets of adjacent deformation-associated regions in the deformation gradient map according to preset spatial coordinates in the deformation gradient map, and connecting the displacement offsets to form a deformation propagation path network; Obtaining load change values ​​of connected nodes in the strain map, and associating nodes whose load change amplitudes exceed a preset threshold to construct a stress diffusion path network; spatially superimposing the deformation propagation path network and the stress diffusion path network, and fusing the displacement offset and the load change value in the overlapping area of ​​the spatial superposition to generate a risk indicator; The coordinates of the foreign body attachment position are determined according to the geometric parameters of the surface foreign body profile, and a transmission tower abnormality warning instruction including the coordinates of the foreign body attachment position is generated in combination with the risk index.

2. The method according to claim 1, characterized in that The deformation propagation path network and the stress diffusion path network are spatially superimposed, and the displacement offset and load change value are fused in the overlapping area of ​​the spatial superposition to generate a risk indicator, including: Determining an overlapping area of ​​the deformation propagation path network and the stress diffusion path network on a preset spatial coordinate, wherein a condition for determining the overlapping area is that a range of spatial coordinates corresponding to a displacement offset in the deformation propagation path network overlaps a range of spatial coordinates corresponding to a load change value in the stress diffusion path network; Calculating a fusion weight factor for each superposition region according to a change degree of the displacement offset and an increase or decrease trend of the load change value in the superposition region; The displacement offset and load change value of each coordinate point in the superposition area are superimposed and calculated according to the corresponding fusion weight factor to obtain the risk index of each coordinate point.

3. The method according to claim 1, characterized in that Dynamically registering the surface point cloud set and the historical point cloud set based on the curvature difference of adjacent point clouds to generate a deformation gradient map, wherein point cloud regions where the curvature difference exceeds a set range during the dynamic registration process are marked as deformation-related regions of the deformation gradient map, including: Extracting three-dimensional geometric features of adjacent point clouds in the surface point cloud set and the historical point cloud set, and calculating the curvature difference of corresponding areas between the adjacent point clouds, wherein the adjacent point clouds are determined by the nearest neighbor relationship of spatial coordinates; Setting a dynamic adjustment threshold according to the distribution range of the curvature difference, and marking the point cloud area where the curvature difference exceeds the dynamic adjustment threshold as the initial deformation area; Dynamically registering the correspondence between the surface point cloud set and the historical point cloud set based on a changing trend of the curvature difference between adjacent point clouds in the initial deformation area, so that the curvature difference in the initial deformation area converges to a dynamically adjusted threshold; Calculate the curvature difference variation of each point cloud region based on the corresponding relationship after dynamic registration, and accumulate the variation along a preset three-dimensional spatial direction to generate a deformation gradient map; A spatial continuity verification is performed on the region in the gradient value distribution of the deformation gradient map that continuously exceeds the dynamic adjustment threshold, and the region that meets the distribution condition of spatial continuity is marked as a deformation association region.

4. The method according to claim 3, wherein According to the correspondence after dynamic registration, the change in curvature difference of each point cloud area is calculated, and the change is accumulated along the preset three-dimensional space direction to generate a deformation gradient map, including: Mapping the surface point cloud set to the three-dimensional geometric features of each point cloud in the historical point cloud set according to the corresponding relationship after dynamic registration, and determining the point cloud area where the curvature difference changes in the corresponding relationship; Calculating the change in curvature difference of each point cloud region before and after dynamic registration, and accumulating the change in curvature difference of each point cloud region along a preset three-dimensional spatial direction, where the three-dimensional spatial direction is determined by the geometric feature distribution of the surface point cloud set; The accumulated curvature difference variation is matched with the preset three-dimensional spatial coordinates in each point cloud area to generate a deformation gradient map.

5. The method according to claim 1, wherein Strain signals are collected at the connection nodes of the transmission towers. After performing time-frequency domain energy decomposition on the strain signals, steady-state components are retained. The steady-state components are mapped to positions corresponding to deformation-related areas to generate a strain map, including: Strain signals at the connection nodes of the transmission towers are collected and divided into multiple time segments according to a preset time length. The signal in each time segment is divided into multiple continuous frequency intervals, and the energy value of each combination of time segment and frequency interval is calculated; Filter the units whose energy value fluctuation amplitude over time is lower than the preset fluctuation threshold as steady-state components; According to the range of the spatial coordinates of the deformation association area, the energy value corresponding to the steady-state component of each connection node is mapped to the spatial coordinates of the corresponding deformation association area; The energy values ​​of the steady-state components of multiple connection nodes mapped to the same deformation association area are spatially weighted fused to obtain an energy fusion value, and a strain map is generated according to the corresponding relationship between the energy fusion value and the spatial coordinates.

6. The method according to claim 1, wherein Obtaining load change values ​​of connected nodes in the strain map, and associating nodes whose load change amplitudes exceed a preset threshold to construct a stress diffusion path network, including: Extracting a load change value of each connection node in the strain spectrum, wherein the load change value is derived from an energy value corresponding to a steady-state component in the strain spectrum; Calculating a dynamic fluctuation threshold of a change amplitude of the load change value based on distribution characteristics of the load change values ​​of all connected nodes, wherein the dynamic fluctuation threshold is determined by the sum of the median and the standard deviation of the change amplitude; Screening the connection nodes whose change amplitude exceeds the dynamic fluctuation threshold and marking these nodes as key nodes; Analyze the spatial proximity relationship between the key nodes, and determine the association direction between the key nodes based on the increase and decrease trends of the coordinate distances and change amplitudes of the key nodes in the preset three-dimensional space; The key nodes are connected in sequence according to the association direction to form a continuous path, and for the fracture area in the continuous path caused by the coordinate distance exceeding the preset range, the fracture area is supplemented by the increase or decrease trend of the change amplitude of the load change value of the adjacent key nodes to generate a stress diffusion path network, and the adjacent key nodes are determined by the spatial proximity relationship between the key nodes.

7. The method according to claim 1, wherein Determining the coordinates of the foreign object attachment position according to the geometric parameters of the surface foreign object profile, and generating a transmission tower abnormality warning instruction containing the coordinates of the foreign object attachment position in combination with the risk index, including: Extracting boundary coordinates from the geometric parameters of the surface foreign body contour, and using the boundary coordinates as initial position coordinates of the foreign body attachment area; Matching the initial position coordinates with the spatial coordinates of the deformation association area according to the spatial coordinate range of the deformation association area, and eliminating the initial position coordinates that do not fall within the spatial coordinate range to determine the foreign body attachment position coordinates; Determine the value of the risk index corresponding to each foreign body attachment location coordinate, and divide the dynamic safety threshold according to the distribution range of the value of the risk index of all foreign body attachment location coordinates; Marking the coordinates of the foreign body attachment location where the value of the risk indicator exceeds the dynamic safety threshold as high-risk coordinates; Combining the location information of the high-risk coordinates with the corresponding risk index value, a transmission tower abnormality warning instruction including the coordinates of the foreign object attachment location and the risk index is generated.

8. A system for detecting foreign objects hanging on power transmission and distribution lines based on image processing and deep learning, characterized in that: include: an acquisition module, configured to acquire a surface point cloud set and a historical point cloud set of the transmission tower, dynamically register the surface point cloud set and the historical point cloud set based on the curvature difference between adjacent point clouds to generate a deformation gradient map, wherein point cloud regions where the curvature difference exceeds a set range during the dynamic registration process are marked as deformation-related regions of the deformation gradient map, and simultaneously identify the contour of a surface foreign body based on the curvature difference to obtain geometric parameters of the contour of the surface foreign body; An acquisition module is used to collect strain signals at the connection nodes of the transmission tower, perform time-frequency domain energy decomposition on the strain signals, retain steady-state components, and map the steady-state components to positions corresponding to deformation-related areas to generate a strain map; A calculation module, configured to calculate displacement offsets of adjacent deformation-associated areas in the deformation gradient map according to preset spatial coordinates in the deformation gradient map, and connect the displacement offsets to form a deformation propagation path network; an association module, configured to obtain load change values ​​of connected nodes in the strain map, and associate nodes whose load change amplitudes exceed a preset threshold value to construct a stress diffusion path network; a superposition module for spatially superimposing the deformation propagation path network and the stress diffusion path network, and fusing the displacement offset and the load change value in the overlapping area of ​​the spatial superposition to generate a risk indicator; A generation module is used to determine the coordinates of the foreign body attachment position based on the geometric parameters of the surface foreign body profile, and generate a transmission tower abnormality warning instruction containing the coordinates of the foreign body attachment position in combination with the risk index.

9. A computing device, characterized in that It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a method for detecting foreign objects hanging on a power transmission and distribution line based on image processing and deep learning as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that A computer program is stored, and when the computer program is executed by a computer, the method for detecting foreign objects hanging on a power transmission and distribution line based on image processing and deep learning as described in any one of claims 1 to 7 is implemented.

Images