Method and system for monitoring potential safety hazards along iron tower based on unmanned aerial vehicle

Abstract

The invention provides a method and system for monitoring potential safety hazards along an iron tower based on an unmanned aerial vehicle, and relates to the technical field of safety monitoring. Establishing a layered acquisition strategy, and dynamically distributing acquisition density and frequency according to a deformation sensitive area; a multi-view depth image is collected to construct a real-time point cloud model, and deformation feature points are extracted; kalman filtering dynamic tracking is carried out on the key monitoring feature points, and early warning is generated when the key monitoring feature points exceed a threshold value. According to the invention, the monitoring precision and efficiency are improved, and early warning of potential safety hazards of the iron tower is realized.

Description

Technical Field

[0001] The present invention relates to safety monitoring technology, and in particular to a method and system for monitoring safety hazards along a tower based on an unmanned aerial vehicle (UAV). Background Art

[0002] As a vital component of transmission networks, the structural safety and stability of steel towers are directly linked to the reliable operation of the system. Traditional tower inspections rely primarily on manual labor, requiring inspectors to climb high into the air to inspect the tower structure. This method is not only inefficient but also poses significant safety risks. With the rapid development of drone technology, drone-based tower inspections have gradually become a focus of industry attention. Drones equipped with high-definition cameras or other sensing equipment can observe and collect data from multiple angles, providing more comprehensive information support for tower safety assessments.

[0003] Existing drone inspections typically operate in a standalone mode, lacking a mechanism for multiple drones to work together. Single-drone inspections struggle to quickly capture comprehensive tower information, resulting in low inspection efficiency and the potential for missing safety hazards in key areas, impacting the comprehensiveness and accuracy of inspections.

[0004] Existing drone inspection methods often use a unified data collection strategy, ignoring the differences in structural characteristics and stress conditions across different parts of the tower. This approach fails to focus on monitoring deformation-sensitive areas of the tower, resulting in blind spots or redundancy in data collection, making it difficult to accurately capture key tower deformation information.

[0005] Existing technologies primarily focus on visible anomalies on the tower surface, such as static defects like loosening, shedding, or corrosion. They lack dynamic monitoring and early warning mechanisms for tower structural deformation. This inability to effectively identify and track the development trends of structural deformation makes it difficult to provide early warnings of potential safety hazards, thus reducing the preventive and proactive nature of inspections. Summary of the Invention

[0006] The embodiments of the present invention provide a method and system for monitoring safety hazards along a tower based on a drone, which can solve the problems in the prior art.

[0007] A first aspect of an embodiment of the present invention provides a method for monitoring safety hazards along a tower using a drone, comprising: Acquire three-dimensional model data and historical deformation data of the target tower, and determine the collaborative inspection paths and collection locations of multiple drones based on the three-dimensional model data; control the multiple drones to fly to corresponding collection locations along the collaborative inspection paths, wherein each drone is equipped with a depth camera; A layered acquisition strategy is established based on the structural stress analysis results of the target tower, the target tower is divided into multiple deformation-sensitive areas, and the acquisition density and frequency of the drone are dynamically allocated according to the stress characteristics of each deformation-sensitive area; and multiple drones are controlled according to the layered acquisition strategy to synchronously acquire multi-view depth images of the target tower; Constructing a real-time three-dimensional point cloud model of the target tower based on the multi-view depth image, registering the real-time three-dimensional point cloud model with the three-dimensional model data of the target tower, extracting deformation feature points, and hierarchically screening the deformation feature points using a dual-threshold iteration method, and determining key monitoring feature points based on the screening results; The key monitoring feature points are dynamically tracked using a Kalman filter algorithm to obtain a motion trajectory of the deformation feature points. When the displacement of the motion trajectory exceeds a preset evaluation threshold, an early warning message is generated.

[0008] and determining the collaborative inspection paths and collection locations of the multiple drones based on the three-dimensional model data; and controlling the multiple drones to fly to the corresponding collection locations along the collaborative inspection paths, including: A target space feature model is constructed based on the three-dimensional model data, and the structural features of the tower are analyzed according to the target space feature model; inspection points are determined based on the structural features, and the inspection points are mapped to a three-dimensional space to form a collection position set, and a collaborative inspection path for multiple drones is generated according to the collection position set; Building a distributed communication network for the plurality of drones, generating a communication topology matrix based on the spatial distribution of the plurality of drones; obtaining a position vector, a velocity vector, and obstacle perception information of each drone through the distributed communication network, and sharing the position vector, the velocity vector, and the obstacle perception information in real time among the plurality of drones; constructing a target point gravitational potential field term based on the obstacle perception information, constructing an obstacle repulsive potential field term and a collaborative obstacle avoidance potential field term based on the position vector, and weightedly combining the target point gravitational potential field term, the obstacle repulsive potential field term, and the collaborative obstacle avoidance potential field term to form a comprehensive potential field function; Calculate the gradient of the comprehensive potential field function, multiply the gradient by the step size factor to obtain a path adjustment amount, dynamically optimize the collaborative inspection path according to the path adjustment amount, obtain an optimized inspection path that meets the safety distance requirements between drones and the communication network connectivity requirements, and control multiple drones to fly to the corresponding collection positions along the optimized inspection path.

[0009] Constructing a real-time three-dimensional point cloud model of the target tower based on the multi-view depth image, registering the real-time three-dimensional point cloud model with the three-dimensional model data of the target tower, and extracting deformation feature points includes: Calculate the stress distribution field under different tower heights and different wind speed conditions to generate a stress distribution map; construct a three-dimensional stress change curve based on the stress distribution map, calculate the stress gradient in the three-dimensional stress change curve, and identify the area with the largest stress gradient as the key monitoring area; Arrange a plurality of depth cameras in the key monitoring area, and extract structural feature points in the multi-view depth image through the depth cameras; Fusing the spatial position information of the structural feature points with the stress distribution map, establishing a correspondence between the feature point positions and the stress distribution, and assigning weight coefficients to the structural feature points according to the correspondence; Registering the multi-view depth images based on the weight coefficients, increasing the registration accuracy requirement in areas where the stress is greater than a preset registration threshold, and generating a real-time 3D point cloud model; comparing structural feature points in the real-time 3D point cloud model with corresponding positions in the standard 3D model, and calculating the displacement direction and displacement amplitude of each structural feature point; Connectivity analysis is performed on the structural feature points according to the displacement direction and the displacement amplitude, a deformation associated point group is identified, and feature points in the deformation associated point group that do not conform to the stress distribution law are eliminated to obtain final deformation feature points.

[0010] Connectivity analysis is performed on the structural feature points according to the displacement direction and the displacement amplitude to identify a deformation associated point group. Feature points that do not conform to the stress distribution law are removed from the deformation associated point group. The final deformation feature points obtained include: constructing a displacement feature vector based on the displacement direction and the displacement amplitude, calculating the Euclidean distance of the displacement feature vector and the spatial position distance between the structural feature points, and constructing a feature association matrix using the Euclidean distance of the displacement feature vector and the spatial position distance; calculating the connectivity strength of the structural feature points according to the feature association matrix, and clustering the structural feature points based on the connectivity strength to obtain a plurality of deformation association point groups; Calculating the displacement feature vector mean and the displacement feature vector deviation of the structural feature points in the deformation associated point group, and constructing a deformation distribution feature with the displacement feature vector mean and the displacement feature vector deviation; Obtain stress distribution data of the iron tower, calculate stress values ​​and stress gradients at structural feature points based on the stress distribution data, and construct stress distribution features using the stress values ​​and stress gradients; input the stress distribution features and the deformation distribution features into a scoring function, and screen the structural feature points based on the output value of the scoring function to obtain final deformation feature points.

[0011] The deformation feature points are graded and screened using a dual-threshold iteration method. The key monitoring feature points are determined based on the screening results, including: Constructing an initial threshold and a target threshold, using the initial threshold to perform a first screening of deformation feature points to obtain candidate feature points, calculating the deformation amplitude distribution of the candidate feature points, and dynamically adjusting the target threshold according to the deformation amplitude distribution; The target threshold is used as an iterative target to gradually screen the candidate feature points. When the screening results are stable, the candidate feature points that are greater than the target threshold are determined as key monitoring feature points.

[0012] The key monitoring feature points are dynamically tracked using the Kalman filter algorithm to obtain the motion trajectory of the deformation feature points. When the displacement of the motion trajectory exceeds a preset assessment threshold, an early warning message is generated, including: Obtaining the three-dimensional coordinates and displacement speed of the key monitoring feature point, constructing a state vector of the key monitoring feature point using the three-dimensional coordinates and the displacement speed, and performing historical state sampling on the key monitoring feature point based on the state vector to obtain a historical state sampling sequence; Constructing a spatiotemporal sequence matrix based on the historical state sampling sequence, analyzing the spatiotemporal sequence matrix using a spatiotemporal feature extraction operator to obtain evolutionary pattern features of the key monitoring feature points, and using the evolutionary pattern features as correction parameters of the Kalman filter algorithm; Inputting the state vector into the Kalman filter algorithm for state prediction, correcting the state prediction result based on the evolution pattern characteristics to obtain a predicted state vector, acquiring real-time observation data of the key monitoring feature point, and performing weighted fusion of the predicted state vector and the real-time observation data to obtain the motion trajectory of the key monitoring feature point; The displacement between adjacent sampling points in the motion trajectory is calculated, the displacement is statistically analyzed to obtain a standard deviation and a mean of the displacement, a preset evaluation threshold is constructed based on the standard deviation and the mean, and a warning message is generated when the displacement is greater than the preset evaluation threshold.

[0013] A spatiotemporal sequence matrix is ​​constructed based on the historical state sampling sequence, and the spatiotemporal feature extraction operator is used to analyze the spatiotemporal sequence matrix to obtain the evolution pattern features of the key monitoring feature points, including: Slidingly segmenting the historical state sampling sequence according to a preset time window length, and arranging the state vectors in each time window in time sequence to construct a spatiotemporal sequence matrix; Calculating the time correlation of the state vector based on the space-time sequence matrix, characterizing the time correlation by the covariance of the state vector and its mean to obtain a time correlation feature, calculating the position difference and velocity difference of adjacent state vectors in the space-time sequence matrix, and combining the position difference and the velocity difference to obtain a spatial difference feature; The time-related features and the spatial differential features are weightedly fused according to preset weights to obtain spatiotemporal fusion features; singular value decomposition is performed on the spatiotemporal fusion features to obtain eigenvector values ​​and eigenvalues, the eigenvector values ​​and the proportion of the eigenvalues ​​to the total eigenvalues ​​are calculated to obtain feature contributions, the feature contributions are sorted from large to small, and the eigenvectors whose cumulative feature contributions reach the preset contribution threshold are selected to construct evolutionary pattern features, which are used to characterize the evolutionary pattern features of the key monitoring feature points.

[0014] The method further comprises: Deploy a drone airport on the tower, which is used for parking, charging and communication of drones; Determining an inspection path based on a preset inspection strategy, wherein the inspection strategy includes a leapfrog inspection path, wherein the leapfrog inspection path is a flight path in which a drone takes off from a drone airport at a current tower, flies to a drone airport at an adjacent tower for parking and charging, and then flies to a drone airport at the next tower; The drone is controlled to inspect the area around the tower according to the inspection path, and image data of the area around the tower is collected; the image data is analyzed to identify illegal reclamation, illegal grazing, fire hazards and abnormal animal intrusion in the area around the tower; when a safety hazard is detected, a hazard alarm message is generated and sent to a monitoring terminal.

[0015] A second aspect of an embodiment of the present invention provides a safety hazard monitoring system along a tower based on a drone, comprising: The first unit is configured to obtain three-dimensional model data and historical deformation data of a target tower, and determine a collaborative inspection path and collection locations for multiple drones based on the three-dimensional model data; and control the multiple drones to fly to corresponding collection locations along the collaborative inspection path, wherein each drone is equipped with a depth camera; The second unit is configured to establish a layered acquisition strategy based on the structural stress analysis results of the target tower, divide the target tower into multiple deformation-sensitive areas, and dynamically allocate the acquisition density and frequency of the drone according to the stress characteristics of each deformation-sensitive area; and control the multiple drones to synchronously acquire multi-view depth images of the target tower according to the layered acquisition strategy; The third unit is used to construct a real-time three-dimensional point cloud model of the target tower based on the multi-view depth image, align the real-time three-dimensional point cloud model with the three-dimensional model data of the target tower, extract deformation feature points, and grade the deformation feature points using a double-threshold iteration method, and determine key monitoring feature points based on the screening results; The fourth unit is used to dynamically track the key monitoring feature points using a Kalman filter algorithm to obtain a motion trajectory of the deformation feature points, and generate early warning information when the displacement of the motion trajectory exceeds a preset evaluation threshold.

[0016] According to a third aspect of an embodiment of the present invention, an electronic device is provided, including: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0017] According to a fourth aspect of an embodiment of the present invention, a computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the method described above is implemented.

[0018] The beneficial effects of this application are as follows: The present invention formulates a collaborative inspection route based on the three-dimensional model data and historical deformation data of the tower, and establishes a layered collection strategy in combination with structural force analysis, so that multiple drones can dynamically adjust the collection density and frequency according to the force characteristics of the tower's deformation-sensitive areas, thereby achieving accurate monitoring of safety hazards of the tower and significantly improving monitoring efficiency and resource utilization.

[0019] The present invention adopts the collaborative operation of multiple UAVs to obtain multi-perspective depth images, and combines a dual-threshold iterative hierarchical screening algorithm to screen deformation feature points, thereby realizing comprehensive monitoring of key structural parts of the tower, being able to timely detect minor deformations and potential safety hazards, and greatly improving the accuracy and reliability of monitoring.

[0020] The present invention uses the Kalman filter algorithm to dynamically track key monitoring feature points, obtains the motion trajectory of the deformation feature points and compares it with the preset evaluation threshold, thereby building a complete safety early warning mechanism that can provide early warning of abnormal deformation of the tower, effectively prevent the occurrence of tower safety accidents, and ensure the safe and stable operation of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is a flow chart of a method for monitoring safety hazards along a tower using a drone according to an embodiment of the present invention; Figure 2 Schematic diagram of the effect of stress gradient analysis on deformation feature point identification according to an embodiment of the present invention; Figure 3 This is a flowchart of dynamic tracking and early warning of key monitoring feature points in an embodiment of the present invention. DETAILED DESCRIPTION

[0022] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0023] The technical solution of the present invention is described in detail below with reference to specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0024] Figure 1 FIG. 1 is a flow chart of a method for monitoring safety hazards along a tower using a drone according to an embodiment of the present invention. Figure 1 As shown, the method includes: Acquire three-dimensional model data and historical deformation data of the target tower, and determine the collaborative inspection paths and collection locations of multiple drones based on the three-dimensional model data; control the multiple drones to fly to corresponding collection locations along the collaborative inspection paths, wherein each drone is equipped with a depth camera; A layered acquisition strategy is established based on the structural stress analysis results of the target tower, the target tower is divided into multiple deformation-sensitive areas, and the acquisition density and frequency of the drone are dynamically allocated according to the stress characteristics of each deformation-sensitive area; and multiple drones are controlled according to the layered acquisition strategy to synchronously acquire multi-view depth images of the target tower; Constructing a real-time three-dimensional point cloud model of the target tower based on the multi-view depth image, registering the real-time three-dimensional point cloud model with the three-dimensional model data of the target tower, extracting deformation feature points, and hierarchically screening the deformation feature points using a dual-threshold iteration method, and determining key monitoring feature points based on the screening results; The key monitoring feature points are dynamically tracked using a Kalman filter algorithm to obtain a motion trajectory of the deformation feature points. When the displacement of the motion trajectory exceeds a preset evaluation threshold, an early warning message is generated.

[0025] In an optional embodiment, determining the collaborative inspection paths and collection locations of multiple drones based on the three-dimensional model data; and controlling the multiple drones to fly to corresponding collection locations along the collaborative inspection paths includes: A target space feature model is constructed based on the three-dimensional model data, and the structural features of the tower are analyzed according to the target space feature model; inspection points are determined based on the structural features, and the inspection points are mapped to a three-dimensional space to form a collection position set, and a collaborative inspection path for multiple drones is generated according to the collection position set; Building a distributed communication network for the plurality of drones, generating a communication topology matrix based on the spatial distribution of the plurality of drones; obtaining a position vector, a velocity vector, and obstacle perception information of each drone through the distributed communication network, and sharing the position vector, the velocity vector, and the obstacle perception information in real time among the plurality of drones; constructing a target point gravitational potential field term based on the obstacle perception information, constructing an obstacle repulsive potential field term and a collaborative obstacle avoidance potential field term based on the position vector, and weightedly combining the target point gravitational potential field term, the obstacle repulsive potential field term, and the collaborative obstacle avoidance potential field term to form a comprehensive potential field function; Calculate the gradient of the comprehensive potential field function, multiply the gradient by the step size factor to obtain a path adjustment amount, dynamically optimize the collaborative inspection path according to the path adjustment amount, obtain an optimized inspection path that meets the safety distance requirements between drones and the communication network connectivity requirements, and control multiple drones to fly to the corresponding collection positions along the optimized inspection path.

[0026] For the processing and analysis of 3D model data, the 3D model data of the towers is imported into the processing system. The towers include various types, such as power towers and communication towers. The octree segmentation algorithm is used to spatially partition the model, and the size of each leaf node is set to 0.5 m × 0.5 m × 0.5 m. By traversing and merging the leaf nodes, the system extracts the spatial outline of the tower, including the geometric shape and position information of components such as the main structure, crossarms, and insulator strings. The system organizes this information into a structured target spatial feature model, which contains the main structure point set P main , cross arm point set P cross and the set of attached equipment points P accessory .

[0027] During the analysis of the tower structure characteristics, a structural analysis algorithm based on graph theory is used to identify key nodes. In the specific implementation, the tower model is converted into an undirected graph G = (V, E), where the vertex V represents the structural node and the edge E represents the structural component. The system calculates the degree centrality C of each node. d (v) and betweenness centrality C b(v) Degree centrality reflects the number of connections of a node, and betweenness centrality reflects the importance of a node in the structure. For angle steel towers, the degree centrality of the main column node is usually greater than 4, and the betweenness centrality value is usually above 0.15; for important connection points, the degree centrality is usually greater than 3, and the betweenness centrality value is usually above 0.08. The system determines the key node set of the structure based on these index values. nodes .

[0028] When determining the inspection point, the system sets the key nodes nodes Applying density weighted algorithm, the system sets the criticality threshold T key For nodes with a criticality exceeding the threshold, 6 inspection points with different observation angles are generated around them; for nodes with a criticality between 0.4 and 0.7, 4 inspection points with different observation angles are generated; for nodes with a criticality below 0.4, 2 inspection points with different observation angles are generated. The distance between the inspection point and the structure surface is dynamically adjusted according to the size of the component, with a minimum distance of 3 meters and a maximum distance of 10 meters. The system maps these inspection points into three-dimensional space to form a collection location set P collection , each acquisition position contains coordinates (x, y, z) and the camera orientation vector (dx, dy, dz).

[0029] When generating collaborative inspection paths, the system first uses the K-means clustering algorithm to cluster the collection location set P collection Divide into k subsets, k is equal to the number of drones. For the common 4 drones collaborative inspection scenario, the initial clustering center points are set at the upper, middle, lower part of the tower and the auxiliary equipment area, and the number of clustering iterations is set to 20. After clustering is completed, the system obtains 4 subsets P sub 1,P sub 2,P sub 3,P sub 4, respectively assigned to 4 drones.

[0030] In each subset, the system uses the TSP algorithm to generate the initial inspection path. In the specific implementation, the system uses the 2-opt optimization algorithm, starting from a random initial path, and trying to obtain a shorter path by exchanging two edges in the path, iterating the optimization until the path cannot be further shortened or the maximum number of iterations is 100. sub In the case of 1 containing 20 collection locations, the total length of the optimized path can usually be reduced by about 25%. The system generates an initial inspection path for each drone. initiali , contains the complete waypoint sequence from the starting position to each collection position and then back to the starting position.

[0031] When building a distributed communication network, the system configures a 2.4GHz wireless communication module for each drone, with a maximum transmission rate of 20Mbps and an effective communication distance of 200 meters. The system calculates the communication topology matrix T based on the initial position of the drone. comm , the element T in the matrix comm [i][j] represents the communication status between UAV i and UAV j. A value of 1 indicates direct communication is possible, and a value of 0 indicates indirect communication is not possible. The system uses TDMA time division multiple access, allocating 20ms time slices to each UAV, and broadcasting position and perception information within their respective time slices.

[0032] The position vector of the UAV is represented by three-dimensional coordinates, such as the position vector P1=(x1, y1, z1) of UAV1; the velocity vector is represented by the velocity components in three directions, such as the velocity vector V1=(vx1, vy1, vz1) of UAV1. The obstacle perception information is represented by a three-dimensional cylindrical model, with each obstacle represented by the center point coordinates (x obs ,y obs ,z obs ) and radius r obs Each drone packages this information in a predefined data packet format and broadcasts it on the network. Other drones receive the information, parse it, and store it in their local environment perception database.

[0033] When constructing the artificial potential field, the system first calculates the gravitational potential field term of the target point based on the acquisition position. For the distance d from the current position Pi of the drone i to the target point Pgoal goal , the gravitational force U attr According to d goal 2 The inverse calculation of the gravitational coefficient k attr Set to 0.5. goal When the distance is less than 1 meter, the gravity value is set to a constant value to prevent the system from being unstable due to excessive gravity when approaching the target point.

[0034] The obstacle repulsion potential field term is calculated based on the obstacle perception information. For the distance d from the current position Pi of the drone i to the center point of the obstacle, obs , when d obs Less than the safety threshold d safe (set to 15 meters), the system calculates the repulsive force U rep , repulsion and (1 / d obs -1 / d safe ) 2 Proportional to the repulsion coefficient k rep Set to 0.3. When multiple obstacles exist at the same time, the system calculates the repulsive force generated by each obstacle, and then performs vector superposition to obtain the composite repulsive force.

[0035] The cooperative obstacle avoidance potential field term is calculated based on the relative position relationship between the UAVs. For the distance d between UAV i and UAV j, uav , when d uav Less than the safety distance d min (set to 10 meters), a mutual repulsion force U is generated cooprep , the magnitude of the force is related to (1 / d uav -1 / d min ) 2 Proportional to the mutual repulsion coefficient k cooprep Set to 0.2; when d uav Greater than the communication distance d comm (set to 200 meters), an attractive force U is generated coopattr , the magnitude of the force is related to (d uav -d comm ) 2 Proportional to the attraction coefficient k coopattr Set to 0.1.

[0036] The system combines these three potential fields into a comprehensive potential field function U total =w1×U attr +w2×U rep +w3×U coop , weight coefficients w1=0.5, w2=0.3, w3=0.2. The system calculates U total Gradient about position grad U , get the direction and magnitude of the potential force at the current position. U Multiplying by the step size factor α (set to 0.8), we get the path adjustment △P=α×grad U .

[0037] Apply the path adjustment amount to optimize the initial inspection path and adjust the initial path initiali Each waypoint P in waypoint , calculate the adjusted new position P new =P waypoint -△P. To ensure the smoothness of the path, the system uses the cubic spline interpolation algorithm to smooth the adjusted waypoint sequence, and the maximum curvature between the control points is set to 0.2 to generate a continuous and smooth optimized inspection path. optimizedi .

[0038] During actual tower inspections, the system recalculates the environmental potential field and path adjustments at a 10Hz frequency, enabling dynamic path optimization. When environmental changes are detected, such as the appearance of new obstacles or shifts in communication topology, the system immediately adjusts the corresponding potential field parameters and recalculates the path. This allows the system to respond to various flight environment changes in real time, ensuring the safe and coordinated operation of multiple drone systems.

[0039] For the inspection of 45-meter-tall angle steel towers on 220kV transmission lines, the system deployed four drones, each with a maximum flight speed of 5m / s and a hovering accuracy of ±0.3m. The system planned an inspection path for each drone, encompassing approximately 20 data points, with an average flight distance of approximately 300 meters. A complete inspection took approximately eight minutes. This method enabled the system to collaborate with multiple drones, effectively improving the efficiency and comprehensiveness of tower inspections and providing a reliable data foundation for subsequent safety hazard monitoring.

[0040] In an optional embodiment, constructing a real-time 3D point cloud model of the target tower based on the multi-view depth image, registering the real-time 3D point cloud model with the 3D model data of the target tower, and extracting deformation feature points includes: Calculate the stress distribution field under different tower heights and different wind speed conditions to generate a stress distribution map; construct a three-dimensional stress change curve based on the stress distribution map, calculate the stress gradient in the three-dimensional stress change curve, and identify the area with the largest stress gradient as the key monitoring area; Arrange a plurality of depth cameras in the key monitoring area, and extract structural feature points in the multi-view depth image through the depth cameras; Fusing the spatial position information of the structural feature points with the stress distribution map, establishing a correspondence between the feature point positions and the stress distribution, and assigning weight coefficients to the structural feature points according to the correspondence; Registering the multi-view depth images based on the weight coefficients, increasing the registration accuracy requirement in areas where the stress is greater than a preset registration threshold, and generating a real-time 3D point cloud model; comparing structural feature points in the real-time 3D point cloud model with corresponding positions in the standard 3D model, and calculating the displacement direction and displacement amplitude of each structural feature point; Connectivity analysis is performed on the structural feature points according to the displacement direction and the displacement amplitude, a deformation associated point group is identified, and feature points in the deformation associated point group that do not conform to the stress distribution law are eliminated to obtain final deformation feature points.

[0041] In the method of constructing a real-time three-dimensional point cloud model of the target tower based on multi-view depth images, stress distribution calculation is a key link. This method uses finite element analysis technology to establish a refined model of the tower. The tower includes various types, such as power towers, communication towers, etc. The model contains main structural components such as the main pole tower body, crossarms, and diagonal braces. Taking the 220kV angle steel tower as an example, the tower body height is 45 meters, the bottom width is 12 meters, and the top width is 2 meters. The material properties of the tower are considered in the model, such as the elastic modulus of Q345 steel is 210GPa, the Poisson's ratio is 0.3, and the density is 7850kg / m 3 In the finite element model, the tower is divided into 12,000 grid cells, with an average size of 0.3 meters per cell, ensuring a balance between calculation accuracy and efficiency.

[0042] The wind loads on the tower vary significantly at different heights, and the relationship between wind speed and height is described using an exponential distribution model. A baseline wind speed of 10 m / s is set at 10 m above ground level, and wind speeds at different heights are calculated based on this value. For example, at 20 m, the wind speed is approximately 11.5 m / s; at 30 m, it is approximately 12.8 m / s; and at 45 m, it is approximately 14.3 m / s. Based on these wind speed values, the wind loads acting on various parts of the tower are calculated. The wind load calculations take into account the structure's windward area, wind pressure coefficient, and height variations. At the top crossarm at 45 m, the wind pressure per unit area reaches 1.02 kPa; at 30 m, it is approximately 0.82 kPa; and at 15 m, it is approximately 0.62 kPa.

[0043] Wind load data is input into finite element analysis software to calculate the stress distribution of the tower under different wind speed conditions. The calculation results are output as a stress distribution map, with color coding to indicate the stress magnitude at different locations. For example, at a baseline wind speed of 10 m / s, the maximum stress at the tower's base connection point can reach 120 MPa, the stress at the middle diagonal brace connection point is approximately 80 MPa, and the stress at the top crossarm connection point is approximately 65 MPa. When the wind speed increases to 15 m / s, the maximum stress at the bottom connection point increases to 180 MPa, the stress at the middle diagonal brace connection point increases to 120 MPa, and the stress at the top crossarm connection point increases to 95 MPa.

[0044] Based on stress distribution maps under multiple wind speed conditions, a three-dimensional stress variation curve was constructed. This curve uses tower height, horizontal position, and wind speed as the three dimensions, with stress as the dependent variable. For example, within the 0-10 meter height range of the northeast main column of the tower, an increase in wind speed from 5 m / s to 20 m / s causes stress to increase from 50 MPa to 200 MPa. Meanwhile, within the same height range of 20-30 meters, the same wind speed change causes stress variations ranging from 30 MPa to 120 MPa. Spatial stress gradients were calculated within the three-dimensional stress variation curve to identify areas with the highest stress gradients. Calculations revealed that stress gradients at the base of the tower, at the connections between the four main columns and the foundation, at the connections between the main columns and the diagonal braces, and at the connections between the crossarms and the main structure, were significantly higher than those in other areas. These areas were designated as key monitoring areas.

[0045] Depth cameras are deployed in designated key monitoring areas. These cameras operate using structured light or time-of-flight principles, with a depth resolution of 2mm, a 60-degree field of view, and a working distance of 3-10 meters. Taking a structured light camera as an example, the structured light pattern it emits is a pseudo-random dot matrix with a density of 1,200 dots per square meter and an acquisition frequency of 30 frames per second. Twelve depth cameras are deployed on a standard iron tower: four are located in the base connection area, four in the middle diagonal brace connection area, and four in the top crossarm connection area. The installation location of each camera has been carefully designed to ensure that there are no blind spots in the monitoring area. The fields of view of adjacent cameras overlap by 30%, facilitating subsequent image stitching.

[0046] The multi-view depth images captured by the depth camera have a resolution of 640×480, with each pixel containing both a depth value and corresponding 3D coordinate information. A two-stage approach is used to extract structural feature points from depth images: edge detection and corner extraction. Edge detection uses a gradient algorithm with an adaptive threshold. The threshold is dynamically adjusted based on the depth change rate of the image region and is generally set to 1.5 times the standard deviation of the regional depth. For the tower angle steel structure, the angle steel contours can be extracted during the edge detection stage. Corner extraction uses the Harris corner detection algorithm with a response threshold set to 0.01 and a non-maximum suppression window size of 5×5 pixels. Typically, 200–300 structural feature points can be extracted from a single depth image frame. These feature points are primarily located in key structural locations, such as angle steel joints and bolt locations.

[0047] The spatial location information of the extracted structural feature points is fused with the stress distribution map to establish a correspondence between the feature point locations and the stress distribution. During the fusion process, the 3D coordinates of the feature points are first converted to the same reference coordinate system as the stress distribution map, with a conversion accuracy of better than 5 mm. Subsequently, the stress value at each feature point is calculated using trilinear interpolation. For example, the feature point extracted at the connection between the main column and the foundation at the base of the tower corresponds to a stress value of approximately 120 MPa; the feature point extracted at the connection between the central brace corresponds to a stress value of approximately 80 MPa. Based on the stress values ​​at these locations, a weight coefficient is assigned to each feature point. The weight coefficient is calculated using a normalization process: the stress value is divided by the maximum stress value and then multiplied by an adjustment factor of 0.8 to ensure that the weight distribution is between 0.1 and 0.8. Feature points with larger stress values ​​have higher weight coefficients and are more important in the subsequent registration process.

[0048] Multi-view depth image registration is a key step in constructing a complete 3D point cloud model. The registration process utilizes a weighted iterative closest point algorithm with 50 iterations and a convergence threshold of 0.001 meters. Within each iteration, the matching weights of feature points are dynamically adjusted based on the aforementioned weight coefficients. In areas where stress exceeds the preset registration threshold (set to 80 MPa), the registration accuracy requirement is increased by lowering the matching distance threshold from the standard 0.05 meters to 0.02 meters, ensuring higher registration accuracy in high-stress areas. After registration, the point cloud data from multiple viewpoints is merged to form a real-time 3D point cloud model covering the entire tower. The model point cloud density reaches 5,000 points per square meter in key monitoring areas and 2,000 points per square meter in general areas.

[0049] Comparing the real-time 3D point cloud model with a standard 3D model is the core step in extracting deformable feature points. The standard 3D model serves as the tower's design model or a benchmark model in its undeformed state, with a point cloud density of 10,000 points per square meter. The comparison process begins with a coarse registration, using bounding box alignment along the three principal axes to align the real-time point cloud model and the standard model in the global coordinate system, achieving a registration error of less than 0.1 meter. Subsequently, a fine registration is performed, using a local feature descriptor matching method based on principal component analysis to precisely align local regions, reducing the registration error to within 0.01 meter. After the comparison is complete, the displacement vector between each structural feature point in the real-time point cloud model and the corresponding point in the standard model is calculated, including the displacement direction and magnitude. For example, at a connection point at the base of the tower, the measured displacement direction is horizontal outward, with a displacement magnitude of 15 mm; at a diagonal brace connection in the middle, the measured displacement direction is vertical downward, with a displacement magnitude of 12 mm.

[0050] Connectivity analysis of displacement data is used to identify deformation-related point groups. Connectivity analysis is based on two metrics: spatial distance and displacement similarity. The spatial distance threshold is set to 0.5 meters, the displacement direction similarity threshold is set to 15 degrees, and the displacement amplitude similarity threshold is set to 30%. When the spatial distance between two feature points is less than the threshold, and the differences in displacement direction and amplitude are both less than their respective thresholds, the two points are considered to be deformed. In this way, all associated feature points are aggregated into deformation-related point groups. In actual cases, the base of the tower forms a deformation-related point group containing 15-20 feature points, indicating that the area has undergone overall deformation; while the central diagonal bracing area forms a deformation-related point group containing 8-12 feature points, indicating that the local structure has deformed.

[0051] The deformation-related point group contains outliers caused by measurement errors or environmental interference. The displacement patterns of these points do not conform to the stress distribution pattern. These outliers are eliminated by comparing the actual displacement of the feature point with the theoretical displacement predicted based on stress analysis. The theoretical displacement is calculated through material mechanics, taking into account stress values, material properties, and structural constraints. When the difference between the actual and theoretical displacements of a feature point exceeds 50%, or the angle between the displacement direction and the main stress direction exceeds 30 degrees, the feature point is marked as an outlier and eliminated. After outlier elimination, the resulting set of deformation feature points is more reliable and can accurately reflect the actual deformation state of the tower.

[0052] In a case study of monitoring a 220kV transmission line tower, the system extracted 2,500 initial structural feature points. Connectivity analysis led to 35 deformation-related point groups, totaling 1,200 feature points. Anomaly elimination removed 180 feature points that did not conform to stress distribution patterns, ultimately resulting in 1,020 deformation feature points. These deformation feature points were primarily located in the bottom connection area (420 points), the middle brace connection area (380 points), and the top crossarm connection area (220 points), providing a reliable data foundation for subsequent deformation tracking and early warning.

[0053] Figure 2This is a schematic diagram of the effect of stress gradient analysis on the identification of deformation feature points according to an embodiment of the present invention. The figure shows a comparison of the relationship between stress value and deformation displacement under two different methods. The triangular solid line mark in the figure represents "this technical solution", which adopts a deformation feature evolution pattern extraction algorithm based on a spatiotemporal sequence matrix to achieve accurate measurement through the extraction of effective feature points; the circular dotted line mark represents the "stress-free gradient analysis" method, which is a traditional deformation analysis method containing noise points. From the data trend, when the stress value increases from 10MPa to 50MPa, both methods show a trend that the deformation displacement increases with the increase of stress. Specifically, the deformation displacement of this technical solution is about 5mm when the stress is 10MPa. As the stress increases to 50MPa, the deformation displacement gradually increases to about 20mm, showing a relatively stable linear growth trend as a whole. The stress-free gradient analysis method produces a deformation displacement of approximately 3 mm at an initial stress of 10 MPa, increasing to approximately 13 mm at 50 MPa. However, its growth curve exhibits significant fluctuations, with localized fluctuations occurring at multiple stress points, such as 25 MPa, 35 MPa, and 45 MPa. Comparing the two methods, it can be seen that not only does the present technical solution produce larger overall deformation displacement measurements than the traditional method, but the growth trend is also more stable, with no significant fluctuations. This demonstrates that the deformation feature evolution pattern extraction algorithm based on the spatiotemporal sequence matrix effectively eliminates the interference of noise points by extracting effective feature points, thereby improving the accuracy and reliability of deformation measurements. This figure intuitively reflects the superiority of the present technical solution in stress-deformation evolution analysis.

[0054] In an optional embodiment, connectivity analysis is performed on the structural feature points according to the displacement direction and the displacement amplitude to identify a deformation associated point group, and feature points in the deformation associated point group that do not conform to the stress distribution law are removed to obtain the final deformation feature points including: constructing a displacement feature vector based on the displacement direction and the displacement amplitude, calculating the Euclidean distance of the displacement feature vector and the spatial position distance between the structural feature points, and constructing a feature association matrix using the Euclidean distance of the displacement feature vector and the spatial position distance; calculating the connectivity strength of the structural feature points according to the feature association matrix, and clustering the structural feature points based on the connectivity strength to obtain a plurality of deformation association point groups; Calculating the displacement feature vector mean and the displacement feature vector deviation of the structural feature points in the deformation associated point group, and constructing a deformation distribution feature with the displacement feature vector mean and the displacement feature vector deviation; Obtain stress distribution data of the iron tower, calculate stress values ​​and stress gradients at structural feature points based on the stress distribution data, and construct stress distribution features using the stress values ​​and stress gradients; input the stress distribution features and the deformation distribution features into a scoring function, and screen the structural feature points based on the output value of the scoring function to obtain final deformation feature points.

[0055] The connectivity analysis and deformation feature point screening method for structural feature points is based on the construction of displacement feature vectors and the calculation of Euclidean distances. This method extracts displacement direction and magnitude information for each structural feature point. The displacement direction is represented as a three-dimensional unit vector consisting of three components: x, y, and z. The displacement magnitude is a scalar value expressed in millimeters. The displacement feature vector is a four-dimensional vector formed by combining the displacement direction and magnitude. For example, for feature point A detected in the bottom connection area of ​​a 220kV transmission tower, its displacement direction is (0.6, 0.3, 0.74) and its magnitude is 15mm. Its displacement feature vector is (0.6, 0.3, 0.74, 15). Feature point B in the central area has a displacement direction of (0.5, 0.2, 0.84) and a magnitude of 12mm. Its displacement feature vector is (0.5, 0.2, 0.84, 12).

[0056] In order to calculate the similarity of displacement characteristics between structural feature points, the system uses weighted Euclidean distance measurement. For the displacement direction component, a weight coefficient of 0.7 is assigned; for the displacement amplitude component, a weight coefficient of 0.3 is assigned after standardization. The standardization process is to divide the original displacement amplitude by the maximum displacement amplitude in the monitoring area (such as 25mm) so that its value range is between 0 and 1. Taking the above two feature points as an example, the weighted Euclidean distance calculation result of their displacement feature vectors is 0.18, indicating that the displacement characteristics of these two points are highly similar. The spatial position distance is calculated using three-dimensional Euclidean distance. The coordinates of feature point A are (2.5, 1.8, 3.2) meters, and the coordinates of feature point B are (2.8, 2.1, 7.5) meters. The calculated spatial distance is 4.38 meters.

[0057] When constructing the feature association matrix, the system comprehensively considers the displacement feature vector distance and the spatial position distance, and calculates the correlation between each pair of all structural feature points identified in the monitoring area (usually 1000-2000). The correlation calculation uses an exponential decay function with a displacement feature vector distance parameter of 0.3 and a spatial position distance parameter of 5.0, indicating that when the displacement feature vector distance exceeds 0.3 or the spatial position distance exceeds 5.0 meters, the correlation decreases significantly. For feature points A and B, the calculated correlation is 0.65, which is stored in the corresponding position of the feature association matrix. The dimension of the final feature association matrix is ​​the same as the number of feature points. It is a symmetric matrix with all diagonal elements being 1.

[0058] The connectivity strength of the structural feature points is calculated based on the feature association matrix. The connectivity strength reflects the degree of association between the feature point and the surrounding points. For each feature point, its correlation with all other points is extracted, and the correlation threshold is set to 0.5. The number of points with a correlation greater than the threshold and the total correlation are counted. The connectivity strength of a feature point is equal to the total correlation divided by the number of associated points. In actual monitoring, the connectivity strength of feature points in the bottom area is usually between 0.6-0.8, between 0.5-0.7 in the middle area, and between 0.4-0.6 in the top area. Areas with high connectivity strength indicate strong consistency of deformation features and the presence of overall structural deformation.

[0059] A density peak clustering algorithm is used to cluster structural feature points. This algorithm defines cluster centers based on connectivity strength and local density. Local density is calculated based on the number of points within a 5-meter radius of the feature point, with a connectivity threshold of 0.6. The algorithm identifies points with high connectivity strength and high local density as cluster centers and then assigns other points to the nearest cluster center, forming multiple deformation-related point groups. In a typical tower monitoring scenario, the algorithm can identify 30-50 deformation-related point groups, each containing 10-30 feature points. For example, four point groups are formed in the bottom main column connection area, corresponding to the connections between the four main columns and the foundation; 8-12 point groups are formed in the middle area, corresponding to the connections between the diagonal braces and the main columns; and 6-8 point groups are formed in the top area, corresponding to the connections between the crossarms and the main columns.

[0060] After the deformation-associated point group is formed, the system calculates the mean displacement eigenvector and the displacement eigenvector deviation of the structural feature points within each point group. The mean displacement eigenvector is obtained by calculating the arithmetic mean of the displacement eigenvectors of all feature points in the point group, representing the overall deformation trend of the point group. The displacement eigenvector deviation calculates the Euclidean distance between each feature point in the point group and the mean, and takes the average of these distances to represent the degree of deformation consistency within the point group. Taking a deformation-associated point group at the bottom as an example, which contains 18 feature points, the calculated mean displacement eigenvector is (0.58, 0.25, 0.77, 14.5) and the displacement eigenvector deviation is 0.12, indicating that the deformation pattern of the feature points in this point group is relatively consistent. The mean displacement eigenvector and the displacement eigenvector deviation are combined to form the deformation distribution feature, which is used for subsequent comparison with the stress distribution feature.

[0061] Finite element analysis (FEM) is used to obtain stress distribution data for the tower. A refined tower model is established, and the stress distribution at each tower location is calculated, taking into account actual load conditions (such as wind load and conductor tension). Stress calculations utilize linear statics analysis, with the results expressed as stress tensors. For each structural feature point, the system extracts the stress value (in MPa) and stress gradient (in MPa / m) at that location. The stress value indicates the magnitude of the mechanical stress at that point, while the stress gradient indicates the severity of the stress changes around that point. Typical stress values ​​at the base of a tower can reach 120 MPa, with a stress gradient of approximately 25 MPa / m; stress values ​​at the central brace connection are approximately 80 MPa, with a stress gradient of approximately 15 MPa / m; and stress values ​​at the top crossarm connection are approximately 60 MPa, with a stress gradient of approximately 10 MPa / m.

[0062] The stress distribution characteristics consist of stress values ​​and stress gradients. To make the stress distribution characteristics comparable with the deformation distribution characteristics, the stress values ​​and stress gradients are normalized. The stress value is divided by the maximum stress value (e.g., 150 MPa), and the stress gradient is divided by the maximum stress gradient (e.g., 30 MPa / m), so that their values ​​range from 0 to 1. The normalized stress distribution characteristics are used to evaluate the consistency of deformation and stress. In theory, deformation should be consistent with the stress distribution. For example, areas with greater stress should exhibit greater deformation, and the deformation direction should be consistent with the main stress direction; areas with greater stress gradients exhibit localized deformation concentration or discontinuity.

[0063] The scoring function is used to assess whether the deformation of structural feature points conforms to the stress distribution law. The scoring function design considers the following three aspects: the correlation between the deformation amplitude and the stress value, the consistency of the deformation direction and the principal stress direction, and the matching degree of the deformation distribution and the stress gradient. For the correlation between the deformation amplitude and the stress value, the correlation coefficient between the two is calculated, with a weight of 0.4; for the consistency between the deformation direction and the principal stress direction, the cosine value of the angle between the two directions is calculated, with a weight of 0.4; for the matching degree of the deformation distribution and the stress gradient, the ratio of the deformation eigenvector deviation to the normalized stress gradient is calculated, with a weight of 0.2. The final score is obtained by weighting the sum of these three indicators, ranging from 0 to 1, with the closer to 1, the more consistent the deformation is with the stress distribution law.

[0064] Taking a deformation-related point group at the bottom as an example, the mean displacement eigenvectors for this point group are (0.58, 0.25, 0.77, 14.5), the deviation of the displacement eigenvectors is 0.12, the normalized stress value of the corresponding region is 0.8, the normalized stress gradient is 0.85, and the principal stress directions are (0.6, 0.2, 0.78). The correlation coefficient between the calculated deformation amplitude and stress value is 0.92, the cosine of the angle between the deformation direction and the principal stress direction is 0.98, and the ratio of the deformation eigenvector deviation to the normalized stress gradient is 0.14. Based on the scoring function, the score for this point group is 0.92 × 0.4 + 0.98 × 0.4 + (1 - 0.14) × 0.2 = 0.93, indicating that the deformation characteristics of this point group are highly consistent with the stress distribution.

[0065] Based on the scoring function, the structural feature points in the deformation-related point group are screened, and feature points that do not conform to the stress distribution law are eliminated. The specific screening criteria are as follows: feature points with a score lower than 0.7 are directly eliminated; feature points with a score between 0.7 and 0.8 are also eliminated if the Euclidean distance between their displacement feature vector and the point group mean is greater than 0.2; feature points with a score between 0.8 and 0.9 are only eliminated if the Euclidean distance between their displacement feature vector and the point group mean is greater than 0.3; all feature points with a score greater than 0.9 are retained. In this way, the system can effectively eliminate unreliable feature points caused by measurement errors, environmental interference or local anomalies, and retain feature points that truly reflect structural deformation.

[0066] In a monitoring case study of a 220kV transmission tower, the system initially identified 1,500 structural feature points. Through connectivity analysis, 40 deformation-related point groups were formed, totaling 1,200 feature points. After screening using a scoring function, 210 feature points that did not conform to stress distribution patterns were eliminated, resulting in a final set of 990 deformation feature points. These deformation feature points are distributed across key areas of the tower and accurately reflect its overall deformation state, providing a reliable data foundation for subsequent dynamic tracking and safety warnings. High-quality deformation feature points significantly improve the accuracy and reliability of deformation monitoring, reduce false alarms and missed alarms, and provide strong support for the safe operation of transmission lines.

[0067] In an optional embodiment, the deformation feature points are graded and screened using a dual-threshold iteration method, and the key monitoring feature points are determined according to the screening results, including: Constructing an initial threshold and a target threshold, using the initial threshold to perform a first screening of deformation feature points to obtain candidate feature points, calculating the deformation amplitude distribution of the candidate feature points, and dynamically adjusting the target threshold according to the deformation amplitude distribution; The target threshold is used as an iterative target to gradually screen the candidate feature points. When the screening results are stable, the candidate feature points that are greater than the target threshold are determined as key monitoring feature points.

[0068] The method for hierarchical screening of deformation feature points and identification of key monitoring feature points adopts a dual-threshold iterative strategy to achieve accurate identification and dynamic tracking of the monitored objects. This method first constructs initial threshold and target threshold parameters. The initial threshold setting is based on the statistical analysis results of historical monitoring data. For 220kV transmission tower deformation monitoring, the system sets the initial threshold for deformation amplitude to 5 mm. This value was obtained by analyzing the historical deformation data of 50 towers of the same model and corresponds to the upper limit of the minimal deformation of the tower structure under normal conditions. The target threshold is initially set to 12 mm. This value is determined based on the safety margin in the tower design specifications and represents the deformation amplitude that requires special attention. The interval between the two thresholds is the gray area, which requires iterative screening to determine the final classification.

[0069] When the initial threshold is used to perform the first screening of deformation feature points, the system evaluates the deformation amplitude of each deformation feature point. The deformation amplitude is calculated using the Euclidean distance metric, with the distance between the feature point's current position and the reference position serving as the deformation amplitude value. In a specific transmission tower monitoring example, the system screened out 990 deformation feature points with a deformation amplitude greater than 5 mm, resulting in a total of 350 candidate feature points. These candidate feature points are primarily located in key stress-bearing areas of the tower, such as the bottom main column connection area (120 points), the middle diagonal brace connection area (150 points), and the top crossarm connection area (80 points).

[0070] After the first screening is completed, the system calculates the deformation amplitude distribution characteristics of the candidate feature points, including the mean, standard deviation, quantiles and other statistical quantities of the deformation amplitude. In the above monitoring example, the average deformation amplitude of the 350 candidate feature points is 8.3 mm, the standard deviation is 2.5 mm, the minimum value is 5.1 mm, the maximum value is 16.8 mm, the 25% quantile is 6.2 mm, the 50% quantile (median) is 7.8 mm, the 75% quantile is 10.1 mm, and the 95% quantile is 13.5 mm. The deformation amplitude distribution shows a right-skewed characteristic, indicating that the deformation amplitudes of most candidate feature points are concentrated in the smaller value area, and a few points have larger deformation amplitudes.

[0071] Based on the deformation amplitude distribution characteristics, the system dynamically adjusts the target threshold. This adjustment method takes into account the importance of the monitored object, the dispersion of the deformation distribution, and the required sensitivity for anomaly detection. For important transmission towers, the target threshold adjustment is conservative to improve monitoring sensitivity. For cases with large deformation dispersion, the target threshold adjustment needs to consider the overall distribution shape. For scenarios requiring high sensitivity, the target threshold can be appropriately lowered. The specific adjustment calculation uses an adaptive algorithm that considers the mean and standard deviation of the deformation amplitude and introduces a safety factor. In this example, the adjusted target threshold is 9.5 mm, calculated as the mean plus the standard deviation multiplied by the safety factor (0.5): 8.3 + 2.5 × 0.5 = 9.5 mm.

[0072] After the target threshold is determined, the system begins the iterative screening process. The iterative screening adopts a progressive threshold adjustment strategy. The current threshold is initially set to the initial threshold (5 mm). In each iteration, the current threshold is increased by one step until the target threshold (9.5 mm) is reached. The step size is set to 0.5 mm, and a total of 9 iterations are required to complete the screening process. In each iteration, the system uses the current threshold to screen candidate feature points and counts the number of feature points that meet the conditions. The first iteration uses a threshold of 5.5 mm to screen out 342 feature points; the second iteration uses a threshold of 6 mm to screen out 330 feature points; and so on. The last iteration uses a threshold of 9.5 mm to screen out 156 feature points.

[0073] During the iterative screening process, the system monitors the stability of the screening results. Stability is assessed based on the rate of change between two consecutive iterations, calculated as the absolute difference between the number of retained points in the current iteration and the number of retained points in the previous iteration divided by the number of retained points in the previous iteration. When the rate of change falls below a preset threshold (typically 5%), the screening results are considered stable. In this example, the seventh iteration (threshold 8.5 mm) screened 188 feature points, and the eighth iteration (threshold 9 mm) screened 170 feature points, with a rate of change of 9.6%. The eighth and ninth iterations (threshold 9.5 mm, 156 feature points) showed a rate of change of 8.2%, still not meeting the stability condition. The system continued with the tenth iteration (threshold 10 mm), screening 145 feature points, with the rate of change dropping to 7.1%. The eleventh iteration (threshold 10.5 mm) screened 138 feature points, with a rate of change of 4.8%, meeting the stability condition, and the iterative process terminated.

[0074] After the screening results stabilize, the system identifies candidate feature points with a value greater than the final threshold (10.5 mm) as key monitoring feature points. In this example, a total of 138 key monitoring feature points were finally determined, distributed in various key areas of the tower. Among them, there were 42 in the bottom main column connection area, 58 in the middle diagonal brace connection area, and 38 in the top crossarm connection area. The average deformation amplitude of these feature points was 12.3 mm, the minimum was 10.6 mm, the maximum was 16.8 mm, and the standard deviation was 1.8 mm. The deformation directions were mainly concentrated in the horizontal outward and vertical downward directions.

[0075] To validate the screening results, the system conducted spatial distribution and deformation pattern analysis on key monitoring feature points. Spatial distribution analysis revealed that key monitoring feature points exhibited a clustered distribution within the tower structure, primarily concentrated at highly stressed connection points, such as the connection between the main column and the foundation, the connection between the main column and the diagonal brace, and the connection between the crossarm and the main structure. This distribution closely matched the tower's stress characteristics, validating the screening results. Deformation pattern analysis revealed that key monitoring feature points within the same area exhibited similar deformation directions and magnitudes, indicating that the deformation was global in nature, rather than localized anomalies caused by random noise.

[0076] The system also evaluated the temporal stability of key monitoring points, collecting data three times daily for 30 consecutive days and analyzing their deformation changes at different time points. The results showed that the deformation amplitude of the majority of key monitoring points (approximately 85%) did not change by more than 2 mm over 30 days, demonstrating good temporal stability. The deformation amplitude of approximately 12% of the points showed a slow growth trend, at a rate of approximately 0.1 mm / day. The deformation amplitude of approximately 3% of the points fluctuated significantly, significantly affected by environmental factors (such as temperature and wind speed), requiring a comprehensive analysis based on environmental parameters.

[0077] The advantage of the dual-threshold iterative screening method lies in its ability to adapt to different monitoring objects and environmental conditions, dynamically adjusting screening criteria to find the most appropriate threshold. Compared with the fixed threshold method, this method can reduce the false alarm rate by approximately 30% and the missed alarm rate by 25%, significantly improving the accuracy and reliability of monitoring. The key monitoring feature points ultimately identified can accurately reflect the critical deformation states of various tower types, such as power towers and communication towers. This provides a reliable data foundation for subsequent dynamic tracking and early warning information generation, effectively supporting the timely detection and resolution of safety hazards along the tower lines.

[0078] In an optional embodiment, the key monitoring feature points are dynamically tracked using a Kalman filter algorithm to obtain a motion trajectory of the deformation feature points. When the displacement of the motion trajectory exceeds a preset evaluation threshold, generating warning information includes: Obtaining the three-dimensional coordinates and displacement speed of the key monitoring feature point, constructing a state vector of the key monitoring feature point using the three-dimensional coordinates and the displacement speed, and performing historical state sampling on the key monitoring feature point based on the state vector to obtain a historical state sampling sequence; Constructing a spatiotemporal sequence matrix based on the historical state sampling sequence, analyzing the spatiotemporal sequence matrix using a spatiotemporal feature extraction operator to obtain evolutionary pattern features of the key monitoring feature points, and using the evolutionary pattern features as correction parameters of the Kalman filter algorithm; Inputting the state vector into the Kalman filter algorithm for state prediction, correcting the state prediction result based on the evolution pattern characteristics to obtain a predicted state vector, acquiring real-time observation data of the key monitoring feature point, and performing weighted fusion of the predicted state vector and the real-time observation data to obtain the motion trajectory of the key monitoring feature point; The displacement between adjacent sampling points in the motion trajectory is calculated, the displacement is statistically analyzed to obtain a standard deviation and a mean of the displacement, a preset evaluation threshold is constructed based on the standard deviation and the mean, and a warning message is generated when the displacement is greater than the preset evaluation threshold.

[0079] like Figure 3 As shown, the method includes: The dynamic tracking and early warning information generation method for key monitoring feature points is based on the improvement and application of the Kalman filter algorithm. This method first obtains the three-dimensional coordinates and displacement velocity information of key monitoring feature points. The three-dimensional coordinates consist of three components (x, y, and z) in meters; the displacement velocity also consists of three directional components (vx, vy, and vz) in millimeters per day. For example, the three-dimensional coordinates of the main column connection point at the bottom of a 220kV transmission tower are (3.25, 2.60, 1.85) meters, and the displacement velocity is (0.35, 0.22, -0.18) mm / day, indicating that the point has an outward horizontal displacement trend and a downward vertical displacement trend. The three-dimensional coordinates and displacement velocity combine to form a state vector consisting of six components: (x, y, z, vx, vy, vz). This state vector comprehensively describes the feature point's position and motion in space and serves as the basic data for subsequent dynamic tracking.

[0080] Historical state sampling is carried out on key monitoring feature points. The state of the feature points is sampled at fixed time intervals (usually 24 hours), and their three-dimensional coordinates and displacement speed are recorded to form a historical state sampling sequence. The sampling time is usually 30 days, generating 30 sets of state data. For the above feature points, the coordinate change range within 30 days is x: (3.25-3.36) meters, y: (2.60-2.67) meters, z: (1.85-1.80) meters; the speed change range is vx: (0.30-0.42) mm / day, vy: (0.18-0.25) mm / day, vz: (-0.15 to -0.20) mm / day. These data show that the point has continuous small displacement during the monitoring period, and the displacement direction is relatively stable without sudden changes.

[0081] When constructing a spatiotemporal sequence matrix from a historical state sampling sequence, the system treats each set of state data as a row, resulting in a matrix with 30 rows and 6 columns. Each column of this matrix represents the temporal evolution of a state component, and each row represents the complete state of a feature point at a specific moment. The spatiotemporal sequence matrix captures the temporal evolution and spatial distribution characteristics of the feature point's motion, providing the data foundation for extracting evolutionary pattern features. The system then applies spatiotemporal feature extraction operators to this matrix, including trend analysis, periodicity analysis, and correlation analysis. Trend analysis uses linear regression to calculate the changing trends of each state component; periodicity analysis employs the autocorrelation function to detect periodic patterns in state changes; and correlation analysis calculates the correlation coefficients between different state components to identify dependencies between state variables.

[0082] Analysis of the characteristic points reveals that the x and y coordinates exhibit a linear growth trend, with growth rates of 0.37 mm / day and 0.23 mm / day, respectively, which generally coincide with the velocity components vx and vy. The z coordinate exhibits a linear downward trend, with a rate of decrease of 0.17 mm / day, which generally coincides with the velocity component vz. Correlation analysis between state variables reveals a strong positive correlation between the x and y coordinates (correlation coefficient 0.85), indicating consistency in horizontal displacement. A negative correlation exists between x, y, and z (correlation coefficients of -0.72 and -0.68, respectively), indicating that vertical displacement increases with increasing horizontal displacement (negative in the downward direction). These analysis results collectively constitute the evolutionary pattern characteristics of the characteristic points, including trend parameters, periodicity parameters, and correlation parameters, which are used to refine the subsequent Kalman filter algorithm.

[0083] The application of the Kalman filter algorithm begins with the state prediction stage, predicting the state at the next moment based on the current state vector and the state transition model. The state transition model uses a uniform motion model, assuming that the displacement velocity remains constant over a short period of time. For the state vector (x, y, z, vx, vy, vz), the predicted state at the next moment (interval Δt = 1 day) is (x + vx × Δt, y + vy × Δt, z + vz × Δt, vx, vy, vz). For example, if the current state is (3.30, 2.64, 1.82, 0.38, 0.23, -0.19), the predicted state at the next moment is (3.304, 2.642, 1.818, 0.38, 0.23, -0.19). This prediction is based on a simplified motion assumption and needs to be modified based on the characteristics of the evolutionary pattern.

[0084] The state prediction results are corrected based on the evolutionary pattern characteristics. This correction process takes into account the historical trends, periodicity, and correlation constraints of the feature points. For trend analysis, the system compares the predicted displacement velocity with the historical trend. If the deviation exceeds 20%, the velocity component is adjusted to align with the historical trend. For periodicity, the system examines the current time point's position within the cycle and adjusts the predicted state to conform to the periodic variation pattern. For correlation constraints, the system ensures that the correlation between different state components conforms to the results of historical data analysis. In the above example, the predicted vx value (0.38) is very close to the historical trend (0.37) and does not require adjustment. However, considering the strong correlation between x and y, the system fine-tunes the vy value to 0.24 to bring its ratio to vx closer to the historical data. The corrected predicted state vector is (3.304, 2.642, 1.818, 0.38, 0.24, -0.19).

[0085] Real-time observation data of key monitoring feature points is obtained using a depth camera with a capture frequency of 30 frames per second and a spatial resolution of 2 mm. The acquired depth images are processed by a feature extraction algorithm to obtain the real-time 3D coordinates of the feature points. Due to measurement noise, real-time observation data often contains random errors. To reduce the impact of noise, the system averages 30 consecutive frames of data to obtain more reliable observation results. At a certain moment, the real-time observed coordinates of the feature point are (3.306, 2.645, 1.816), which differs slightly from the predicted coordinates.

[0086] The system calculates the optimal weight coefficient based on the uncertainty of the predicted state and the noise level of the observed data. The uncertainty of the predicted state is represented by the state covariance matrix, which is initially set as a diagonal matrix. The diagonal elements are the variance of the coordinate components (0.0001 meters). 2 ) and the variance of the velocity component (0.0004 (mm / day) 2); The observation noise is represented by the observation covariance matrix, which is set as a diagonal matrix, and the diagonal elements are the variance of the measurement error (0.0009 m 2 Based on these parameters, the system calculates the Kalman gain, which is used to determine the weights of the predicted and observed values. In the above example, the calculated Kalman gain of the position component is approximately 0.1, indicating that the fusion result relies primarily on the predicted value (weight 0.9), supplemented by fine-tuning the observed value (weight 0.1). The fused state estimate is (3.305, 2.644, 1.817, 0.39, 0.25, -0.19).

[0087] By continuously applying the Kalman filter algorithm, the system obtains a state sequence of the feature point changing over time, namely the motion trajectory. This trajectory includes the position changes and velocity changes of the feature point in three-dimensional space, reflecting the motion pattern of the feature point. For the above-mentioned feature point, 30 days of monitoring formed a trajectory containing 30 points, showing that the point continued to move outward in the horizontal direction and continued to sink in the vertical direction, with a total displacement of approximately 12 mm. The system calculates the displacement between adjacent sampling points in the trajectory and obtains 29 displacement data. Under normal circumstances, these displacements should be relatively stable, reflecting the stable motion pattern of the feature point. Statistical analysis of the 29 displacement data showed a mean displacement of 0.42 mm / day and a standard deviation of 0.07 mm / day, indicating that the displacement rate of the feature point is relatively stable.

[0088] Based on the statistical characteristics of the displacement, a preset assessment threshold is constructed, and an adaptive threshold strategy is adopted to dynamically adjust the threshold according to the mean and standard deviation of the displacement. The specific calculation method is: the preset assessment threshold is equal to the mean of the displacement plus the standard deviation multiplied by the safety factor. The safety factor is set according to the importance and risk level of the monitored object, and is usually in the range of 3-5. For important transmission towers, the safety factor is set to 4, and the calculated preset assessment threshold is 0.42 + 0.07 × 4 = 0.70 mm / day. This means that when the displacement on a certain day exceeds 0.70 mm / day, the system considers the movement of the feature point to be abnormal and needs to generate an early warning information.

[0089] The abnormal detection of displacement adopts a sliding window strategy, and each detection takes into account the displacement data of the last three days to reduce the influence of accidental factors. When the average displacement for three consecutive days exceeds the preset assessment threshold, the system generates an early warning message. The early warning information contains the feature point ID, location information, abnormal displacement, historical displacement trend, degree of abnormality, etc. The degree of abnormality is assessed based on the proportion of the displacement exceeding the threshold, and is divided into slight abnormality (exceeding 0-20%), moderate abnormality (exceeding 20-50%) and severe abnormality (exceeding more than 50%). In one monitoring, the system detected that the average displacement of a feature point at the bottom for three consecutive days was 0.85 mm / day, which exceeded the preset threshold of 0.70 mm / day by about 21.4%. It was assessed as a moderate abnormality and generated a corresponding early warning message.

[0090] After an early warning is generated, the system adopts different handling strategies based on the severity of the anomaly. For minor anomalies, the system increases the frequency of monitoring in that area from once a day to three times a day. For moderate anomalies, the system initiates a deep inspection process, using higher-precision sensors to perform a detailed scan of the anomaly area and notifying operations and maintenance personnel. For severe anomalies, the system immediately sends an emergency alert to the operations and maintenance center, recommending on-site inspections and emergency response. Through this hierarchical response mechanism, the system can take appropriate measures based on the severity of the anomaly, ensuring that safety hazards are addressed promptly while avoiding the waste of resources caused by excessive responses.

[0091] In practical applications, this method has been successfully applied to monitoring towers on multiple transmission lines. In one case, the system continuously tracked 138 key monitoring points for 90 days, detecting 12 anomalies: 8 minor, 3 moderate, and 1 severe. Field verification confirmed that 10 of these anomalies corresponded to actual structural deformation, while 2 were temporary anomalies caused by environmental factors (such as strong winds and temperature fluctuations). This method significantly improves the accuracy and reliability of monitoring safety hazards along towers, providing strong support for the safe operation of power grids, inspections around towers, and monitoring of the tower structures themselves.

[0092] In an optional embodiment, a spatiotemporal sequence matrix is ​​constructed based on the historical state sampling sequence, and the spatiotemporal feature extraction operator is used to analyze the spatiotemporal sequence matrix to obtain the evolution pattern characteristics of the key monitoring feature points, including: Slidingly segmenting the historical state sampling sequence according to a preset time window length, and arranging the state vectors in each time window in time sequence to construct a spatiotemporal sequence matrix; Calculating the time correlation of the state vector based on the space-time sequence matrix, characterizing the time correlation by the covariance of the state vector and its mean to obtain a time correlation feature, calculating the position difference and velocity difference of adjacent state vectors in the space-time sequence matrix, and combining the position difference and the velocity difference to obtain a spatial difference feature; The time-related features and the spatial differential features are weightedly fused according to preset weights to obtain spatiotemporal fusion features; singular value decomposition is performed on the spatiotemporal fusion features to obtain eigenvector values ​​and eigenvalues, the eigenvector values ​​and the proportion of the eigenvalues ​​to the total eigenvalues ​​are calculated to obtain feature contributions, the feature contributions are sorted from large to small, and the eigenvectors whose cumulative feature contributions reach the preset contribution threshold are selected to construct evolutionary pattern features, which are used to characterize the evolutionary pattern features of the key monitoring feature points.

[0093] The historical state sampling sequence is segmented into sliding windows, with a preset time window length of 10 days and a sliding step of 1 day. For 30 days of monitoring data, 21 time windows are obtained, each containing 10 consecutive state vectors. For example, the state vector of the main column connection point at the bottom of a 220kV transmission tower consists of six parameters: position coordinates (x, y, z) and velocity components (vx, vy, vz). Within the first time window, the state vector sequence for this feature point is: (3.250, 2.600, 1.850, 0.350, 0.220, -0.180) on day 1; (3.254, 2.602, 1.848, 0.355, 0.218, -0.182) on day 2; and so on, up to (3.280, 2.620, 1.835, 0.370, 0.230, -0.190) on day 10. Arrange these 10 state vectors in time sequence to construct a space-time sequence matrix with 10 rows and 6 columns. Each row in the matrix corresponds to the state at a time point, and each column corresponds to the sequence of changes of a state component over time.

[0094] After constructing the spatiotemporal sequence matrix, the system calculates the temporal correlation of the state vector. The calculation process first calculates the mean of each column of the spatiotemporal sequence matrix to obtain the average value of the state vector. For the aforementioned feature point, within the first time window, the mean of the position coordinates is (3.267, 2.611, 1.842), and the mean of the velocity component is (0.363, 0.224, -0.186). The covariance of each state vector with the mean is then calculated, resulting in a 6×6 covariance matrix. The diagonal elements of this matrix represent the variance of each state component, while the off-diagonal elements represent the covariance between different state components. In actual calculations, the diagonal elements of the covariance matrix are: x-coordinate variance 0.0001, y-coordinate variance 0.00005, z-coordinate variance 0.00003, vx velocity variance 0.00007, vy velocity variance 0.00004, and vz velocity variance 0.00002. Among the off-diagonal elements, the covariance between x and y is 0.00006, indicating a positive correlation; the covariance between x and z is -0.00005, indicating a negative correlation. These covariance values ​​together constitute the temporal correlation feature, describing the correlation pattern of the components of the state vector over time.

[0095] Calculating the position and velocity differences between adjacent state vectors in the space-time sequence matrix is ​​a key step in obtaining spatial differential features. For 10 consecutive state vectors, the system calculates 9 position differences and 9 velocity differences. Position differences calculate the difference between the position coordinates of two consecutive days, while velocity differences calculate the difference between the velocity components of two consecutive days. For example, the position difference between day 1 and day 2 is (0.004, 0.002, -0.002), and the velocity difference is (0.005, -0.002, -0.002); the position difference between day 2 and day 3 is (0.003, 0.003, -0.002), and the velocity difference is (0.002, 0.003, -0.001). Statistical characteristics are calculated for all differential values, including the mean difference, the variance of the difference, and the autocorrelation coefficient of the difference series.

[0096] In the above example, the mean of the position differences is (0.0033, 0.0022, -0.0017), indicating that the feature point has a slight upward trend in the x and y directions and a slight downward trend in the z direction. The mean of the velocity differences is (0.0022, 0.0013, -0.0011), indicating relatively stable velocity changes. The variance of the position differences is (0.000001, 0.0000008, 0.0000005), and the variance of the velocity differences is (0.000002, 0.0000015, 0.0000007). These small variances indicate that the feature point's motion is very stable. The autocorrelation coefficients of the difference series are 0.85, 0.72, and 0.61, respectively, within the lag range of 1-3 days, indicating strong continuity and predictability in the short term. These statistical features together constitute the spatial differential features, which describe the motion pattern of the feature point in space.

[0097] The weighted fusion of temporal correlation features and spatial differential features is a crucial step in constructing comprehensive spatiotemporal fusion features. The system sets fusion weights based on the characteristics of the monitored object and application requirements. Typically, the weight for temporal correlation features is 0.6, and the weight for spatial differential features is 0.4. Temporal correlation features contain 6 × 6 = 36 covariance values, while spatial differential features contain 18 statistics for position differentials and velocity differentials. These two features have different dimensions and require normalization to make them comparable. Normalization involves dividing each feature value by the maximum value for that feature type, ensuring that all feature values ​​fall within the range of 0–1.

[0098] After normalization, the two features are concatenated according to preset weights to form a 54-dimensional spatiotemporal fusion feature vector. For the above feature points, the normalized temporal correlation feature has a maximum value of 0.95 (corresponding to the x-coordinate autocorrelation) and a minimum value of 0.08 (corresponding to the cross-correlation between x- and z-velocity). The normalized spatial difference feature has a maximum value of 0.88 (corresponding to the x-position difference mean) and a minimum value of 0.12 (corresponding to the z-velocity difference variance). These 54 eigenvalues ​​together constitute the spatiotemporal fusion feature, comprehensively describing the spatiotemporal variation of the feature points.

[0099] Performing singular value decomposition on spatiotemporal fusion features is an effective method for dimensionality reduction and extracting key characteristic patterns. The system organizes the 54-dimensional spatiotemporal fusion feature vectors into a matrix and performs singular value decomposition on this matrix to obtain eigenvectors and corresponding eigenvalues. The singular value decomposition results in 54 eigenvectors and their corresponding eigenvalues, which represent different patterns of variation in the data. The size of the eigenvalue indicates the importance of the corresponding eigenvector; the larger the eigenvalue, the greater its contribution to the original data. In actual calculations, the top 10 eigenvalues ​​are 5.8, 4.2, 3.1, 2.6, 1.9, 1.5, 1.2, 0.9, 0.7, and 0.5, respectively; the remaining eigenvalues ​​are all less than 0.5. The system calculates the proportion of each eigenvalue to the total sum of eigenvalues ​​to determine the feature contribution. The contributions of the top 10 eigenvalues ​​are 22.6%, 16.3%, 12.1%, 10.1%, 7.4%, 5.8%, 4.7%, 3.5%, 2.7%, and 1.9% respectively, with a cumulative contribution of 87.1%.

[0100] The feature contributions are sorted from largest to smallest, with a preset contribution threshold of 85%. Eigenvectors whose cumulative contributions reach the threshold are selected to construct the evolutionary pattern signature. In the above example, the cumulative contribution of the first nine eigenvectors is 85.2%, exceeding the preset threshold of 85%, and therefore the first nine eigenvectors are selected as key features. These nine eigenvectors represent different change patterns: the first eigenvector primarily represents the synchronous change pattern of the x and y coordinates, with corresponding coefficients of 0.42 and 0.38, respectively; the second eigenvector primarily represents the change pattern of the z coordinate, with a corresponding coefficient of 0.56; the third eigenvector primarily represents the synchronous change pattern of the vx and vy velocities, with corresponding coefficients of 0.45 and 0.40, respectively; the fourth through ninth eigenvectors represent more complex change patterns, such as the interaction between different state components. These nine eigenvectors and their corresponding eigenvalues ​​constitute the evolutionary pattern signature, which is used to characterize the spatiotemporal variation patterns of key monitoring feature points.

[0101] The extracted evolution pattern features are used to adjust the parameters of the Kalman filter, primarily the state transition matrix and the process noise covariance matrix. The state transition matrix describes how the system state evolves from one moment to the next. Traditional Kalman filters typically assume a uniform motion model, but the evolution pattern features provide a more precise representation of state transitions. For example, the first eigenvector indicates that the x and y coordinates change synchronously. The system adjusts the corresponding elements in the state transition matrix accordingly so that the prediction reflects this synchronization. The process noise covariance matrix describes the uncertainty of the state prediction. The eigenvalues ​​in the evolution pattern features indicate the stability of different change patterns. Patterns with larger eigenvalues ​​are more stable and should be set to a smaller process noise; patterns with smaller eigenvalues ​​are less stable and should be set to a larger process noise. In this way, the Kalman filter algorithm can fully utilize the evolution pattern features extracted from historical data to improve the accuracy and stability of predictions.

[0102] In actual applications, the system extracts evolutionary pattern features for each key monitoring feature point and updates them regularly. Evolutionary pattern features are usually updated every 10 days to ensure that the features can promptly reflect changes in the motion patterns of the feature points. For sudden changes, the system sets a trigger mechanism. When the deviation between the observed data and the predicted value exceeds the preset threshold, the recalculation of the evolutionary pattern features is immediately triggered. For example, when the position deviation of an observation exceeds 5 mm or the speed deviation exceeds 0.2 mm / day, the system considers that the motion pattern of the feature point has changed and the evolutionary pattern features need to be re-extracted. This dynamic update mechanism enables the system to adapt to changes in the motion patterns of feature points and maintain the accuracy and reliability of tracking.

[0103] The evolutionary pattern features extracted using this method can effectively characterize the motion patterns of key monitoring feature points, providing a reliable basis for subsequent dynamic tracking and early warning assessments. Compared with traditional methods, this method fully utilizes the spatiotemporal correlation information contained in historical data, more accurately predicting the motion trajectories of feature points, improving the sensitivity and accuracy of anomaly detection, and providing technical support for the timely detection and resolution of tower safety hazards. In practical safety monitoring of multiple towers, this method has been demonstrated to reduce deformation prediction errors by approximately 35%, effectively reducing both false alarm and missed alarm rates, and significantly improving the reliability and practicality of the monitoring system.

[0104] In an optional embodiment, the method further includes: Deploy a drone airport on the tower, which is used for parking, charging and communication of drones; Determining an inspection path based on a preset inspection strategy, wherein the inspection strategy includes a leapfrog inspection path, wherein the leapfrog inspection path is a flight path in which a drone takes off from a drone airport at a current tower, flies to a drone airport at an adjacent tower for parking and charging, and then flies to a drone airport at the next tower; The drone is controlled to inspect the area around the tower according to the inspection path, and image data of the area around the tower is collected; the image data is analyzed to identify illegal reclamation, illegal grazing, fire hazards and abnormal animal intrusion in the area around the tower; when a safety hazard is detected, a hazard alarm message is generated and sent to a monitoring terminal.

[0105] The drone airport deployed on the tower consists of a charging pod, a communication module, and a protective cover. The foldable charging pod automatically unfolds when the drone lands and closes after landing to protect the drone. An inductive charging coil is installed within the pod, automatically starting charging once the drone has landed accurately. The communication module includes a 5G communication unit and a short-range communication unit. The 5G communication unit is used for data transmission with a remote monitoring center, while the short-range communication unit is used for communication and control with the drone. The protective cover is made of waterproof and breathable material, providing protection against rain, dust, and bird interference.

[0106] The drone airport is fixed to the crossarm atop the tower, 35-40 meters above the ground. Each drone airport is equipped with a quadcopter drone with a 60-minute flight time and a maximum speed of 20 meters per second. The drone is equipped with a 4K high-definition visible light camera and an infrared thermal imaging camera. The visible light camera has a 120-degree field of view, while the infrared camera has a 60-degree field of view. Both cameras support 30x optical zoom.

[0107] Inspection routes are designed based on the distribution of towers. The typical distance between adjacent towers is 300-500 meters. Considering the drone's endurance, a single inspection is limited to four towers. After taking off from the current tower, the drone flies to the next tower at a set altitude of 45-50 meters and a speed of 15 meters per second. Upon reaching the target tower, the drone orbits the tower within a 100-meter radius, capturing images of the surrounding environment. After completing the acquisition, the drone lands at the tower's designated drone airport for a 15-minute recharge. After recharging, it continues to the next tower, creating a leapfrog inspection pattern.

[0108] During inspections of the area surrounding the tower, the drone uses a spiral descent scanning pattern, starting at an altitude of 50 meters and collecting images in 5-meter increments until it reaches 25 meters. Both visible light and infrared cameras operate simultaneously during the acquisition process. The visible light camera is primarily used for daytime inspections, with an image resolution of 3840x2160 pixels; the infrared camera is primarily used for nighttime inspections, with a resolution of 640x512 pixels.

[0109] The system intelligently analyzes and processes collected image data. For illegal reclamation, it detects newly added cultivated land within the image and triggers an alarm if it detects more than 100 square meters of newly added cultivated land within 50 meters of the tower. For illegal grazing, the system identifies livestock in the image and issues an early warning if any livestock activity is detected within 30 meters of the tower. Fire hazard monitoring primarily relies on infrared cameras, which trigger an alarm if an abnormal temperature point exceeds 150 degrees Celsius or an abnormal temperature area exceeds 2 square meters. Animal intrusion is identified when a large wild animal is detected within 20 meters of the tower.

[0110] Alarm information is divided into three levels: red alerts indicate serious hazards requiring immediate attention, such as fires or large-scale illegal land reclamation; yellow alerts indicate potential hazards requiring attention, such as animal intrusions or small-scale illegal grazing; and green alerts indicate minor anomalies that require recording. Alert information includes the type of hazard, location coordinates, on-site images, and time of discovery, and is transmitted in real time to the monitoring center via the 5G network. Upon receipt of the alert, the monitoring center activates the appropriate emergency response mechanism based on the alert level.

[0111] The system is programmed to conduct four regular inspections daily: at 6:00 AM, 12:00 PM, 6:00 PM, and 12:00 PM, each lasting 90 minutes. Furthermore, the system automatically increases the inspection frequency in exceptional weather conditions, such as strong winds or heavy rain. When wind speeds exceed 12 meters per second or rainfall exceeds 50 millimeters per hour, the drone automatically returns to the airport for standby. This ensures round-the-clock, automated monitoring of safety hazards around the towers.

[0112] A second aspect of an embodiment of the present invention provides a safety hazard monitoring system along a tower based on a drone, comprising: The first unit is configured to obtain three-dimensional model data and historical deformation data of a target tower, and determine a collaborative inspection path and collection locations for multiple drones based on the three-dimensional model data; and control the multiple drones to fly to corresponding collection locations along the collaborative inspection path, wherein each drone is equipped with a depth camera; The second unit is configured to establish a layered acquisition strategy based on the structural stress analysis results of the target tower, divide the target tower into multiple deformation-sensitive areas, and dynamically allocate the acquisition density and frequency of the drone according to the stress characteristics of each deformation-sensitive area; and control the multiple drones to synchronously acquire multi-view depth images of the target tower according to the layered acquisition strategy; The third unit is used to construct a real-time three-dimensional point cloud model of the target tower based on the multi-view depth image, align the real-time three-dimensional point cloud model with the three-dimensional model data of the target tower, extract deformation feature points, and grade the deformation feature points using a double-threshold iteration method, and determine key monitoring feature points based on the screening results; The fourth unit is used to dynamically track the key monitoring feature points using a Kalman filter algorithm to obtain a motion trajectory of the deformation feature points, and generate early warning information when the displacement of the motion trajectory exceeds a preset evaluation threshold.

[0113] According to a third aspect of an embodiment of the present invention, an electronic device is provided, including: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0114] According to a fourth aspect of an embodiment of the present invention, a computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the method described above is implemented.

[0115] The present invention may be a method, an apparatus, a system and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for executing various aspects of the present invention.

[0116] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for monitoring safety hazards along the tower based on drones, characterized in that: include: Acquire three-dimensional model data and historical deformation data of a target tower, which may include multiple types of towers, and determine collaborative inspection paths and collection locations for multiple drones based on the three-dimensional model data; control the multiple drones to fly to corresponding collection locations along the collaborative inspection paths, wherein each drone is equipped with a depth camera; A layered acquisition strategy is established based on the structural stress analysis results of the target tower, the target tower is divided into multiple deformation-sensitive areas, and the acquisition density and frequency of the drone are dynamically allocated according to the stress characteristics of each deformation-sensitive area; and multiple drones are controlled according to the layered acquisition strategy to synchronously acquire multi-view depth images of the target tower; Constructing a real-time three-dimensional point cloud model of the target tower based on the multi-view depth image, registering the real-time three-dimensional point cloud model with the three-dimensional model data of the target tower, extracting deformation feature points, and hierarchically screening the deformation feature points using a dual-threshold iteration method, and determining key monitoring feature points based on the screening results; The key monitoring feature points are dynamically tracked using a Kalman filter algorithm to obtain a motion trajectory of the deformation feature points. When the displacement of the motion trajectory exceeds a preset evaluation threshold, an early warning message is generated.

2. The method according to claim 1, characterized in that and determining the collaborative inspection paths and collection locations of the multiple drones based on the three-dimensional model data; and controlling the multiple drones to fly to the corresponding collection locations along the collaborative inspection paths, including: A target space feature model is constructed based on the three-dimensional model data, and the structural features of the tower are analyzed according to the target space feature model; inspection points are determined based on the structural features, and the inspection points are mapped to a three-dimensional space to form a collection position set, and a collaborative inspection path for multiple drones is generated according to the collection position set; Building a distributed communication network for the plurality of drones, generating a communication topology matrix based on the spatial distribution of the plurality of drones; obtaining a position vector, a velocity vector, and obstacle perception information of each drone through the distributed communication network, and sharing the position vector, the velocity vector, and the obstacle perception information in real time among the plurality of drones; constructing a target point gravitational potential field term based on the obstacle perception information, constructing an obstacle repulsive potential field term and a collaborative obstacle avoidance potential field term based on the position vector, and weightedly combining the target point gravitational potential field term, the obstacle repulsive potential field term, and the collaborative obstacle avoidance potential field term to form a comprehensive potential field function; Calculate the gradient of the comprehensive potential field function, multiply the gradient by the step size factor to obtain a path adjustment amount, dynamically optimize the collaborative inspection path according to the path adjustment amount, obtain an optimized inspection path that meets the safety distance requirements between drones and the communication network connectivity requirements, and control multiple drones to fly to the corresponding collection positions along the optimized inspection path.

3. The method according to claim 1, characterized in that Constructing a real-time three-dimensional point cloud model of the target tower based on the multi-view depth image, registering the real-time three-dimensional point cloud model with the three-dimensional model data of the target tower, and extracting deformation feature points includes: Calculate the stress distribution field under different tower heights and different wind speed conditions to generate a stress distribution map; construct a three-dimensional stress change curve based on the stress distribution map, calculate the stress gradient in the three-dimensional stress change curve, and identify the area with the largest stress gradient as the key monitoring area; Arrange a plurality of depth cameras in the key monitoring area, and extract structural feature points in the multi-view depth image through the depth cameras; Fusing the spatial position information of the structural feature points with the stress distribution map, establishing a correspondence between the feature point positions and the stress distribution, and assigning weight coefficients to the structural feature points according to the correspondence; Registering the multi-view depth images based on the weight coefficients, increasing the registration accuracy requirement in areas where the stress is greater than a preset registration threshold, and generating a real-time 3D point cloud model; comparing structural feature points in the real-time 3D point cloud model with corresponding positions in the standard 3D model, and calculating the displacement direction and displacement amplitude of each structural feature point; Connectivity analysis is performed on the structural feature points according to the displacement direction and the displacement amplitude, a deformation associated point group is identified, and feature points in the deformation associated point group that do not conform to the stress distribution law are eliminated to obtain final deformation feature points.

4. The method according to claim 3, characterized in that Connectivity analysis is performed on the structural feature points according to the displacement direction and the displacement amplitude to identify a deformation associated point group. Feature points that do not conform to the stress distribution law are removed from the deformation associated point group. The final deformation feature points obtained include: constructing a displacement feature vector based on the displacement direction and the displacement amplitude, calculating the Euclidean distance of the displacement feature vector and the spatial position distance between the structural feature points, and constructing a feature association matrix using the Euclidean distance of the displacement feature vector and the spatial position distance; calculating the connectivity strength of the structural feature points according to the feature association matrix, and clustering the structural feature points based on the connectivity strength to obtain a plurality of deformation association point groups; Calculating the displacement feature vector mean and the displacement feature vector deviation of the structural feature points in the deformation associated point group, and constructing a deformation distribution feature with the displacement feature vector mean and the displacement feature vector deviation; Obtain stress distribution data of the iron tower, calculate stress values ​​and stress gradients at structural feature points based on the stress distribution data, and construct stress distribution features using the stress values ​​and stress gradients; input the stress distribution features and the deformation distribution features into a scoring function, and screen the structural feature points based on the output value of the scoring function to obtain final deformation feature points.

5. The method according to claim 1, wherein The deformation feature points are graded and screened using a dual-threshold iteration method. The key monitoring feature points are determined based on the screening results, including: Constructing an initial threshold and a target threshold, using the initial threshold to perform a first screening of deformation feature points to obtain candidate feature points, calculating the deformation amplitude distribution of the candidate feature points, and dynamically adjusting the target threshold according to the deformation amplitude distribution; The target threshold is used as an iterative target to gradually screen the candidate feature points. When the screening results are stable, the candidate feature points that are greater than the target threshold are determined as key monitoring feature points.

6. The method according to claim 1, characterized in that The key monitoring feature points are dynamically tracked using the Kalman filter algorithm to obtain the motion trajectory of the deformation feature points. When the displacement of the motion trajectory exceeds a preset assessment threshold, an early warning message is generated, including: Obtaining the three-dimensional coordinates and displacement speed of the key monitoring feature point, constructing a state vector of the key monitoring feature point using the three-dimensional coordinates and the displacement speed, and performing historical state sampling on the key monitoring feature point based on the state vector to obtain a historical state sampling sequence; Constructing a spatiotemporal sequence matrix based on the historical state sampling sequence, analyzing the spatiotemporal sequence matrix using a spatiotemporal feature extraction operator to obtain evolutionary pattern features of the key monitoring feature points, and using the evolutionary pattern features as correction parameters of the Kalman filter algorithm; Inputting the state vector into the Kalman filter algorithm for state prediction, correcting the state prediction result based on the evolution pattern characteristics to obtain a predicted state vector, acquiring real-time observation data of the key monitoring feature point, and performing weighted fusion of the predicted state vector and the real-time observation data to obtain the motion trajectory of the key monitoring feature point; The displacement between adjacent sampling points in the motion trajectory is calculated, the displacement is statistically analyzed to obtain a standard deviation and a mean of the displacement, a preset evaluation threshold is constructed based on the standard deviation and the mean, and a warning message is generated when the displacement is greater than the preset evaluation threshold.

7. The method according to claim 6, characterized in that A spatiotemporal sequence matrix is ​​constructed based on the historical state sampling sequence, and the spatiotemporal feature extraction operator is used to analyze the spatiotemporal sequence matrix to obtain the evolution pattern features of the key monitoring feature points, including: Slidingly segmenting the historical state sampling sequence according to a preset time window length, and arranging the state vectors in each time window in time sequence to construct a spatiotemporal sequence matrix; Calculating the time correlation of the state vector based on the space-time sequence matrix, characterizing the time correlation by the covariance of the state vector and its mean to obtain a time correlation feature, calculating the position difference and velocity difference of adjacent state vectors in the space-time sequence matrix, and combining the position difference and the velocity difference to obtain a spatial difference feature; The time-related features and the spatial differential features are weightedly fused according to preset weights to obtain spatiotemporal fusion features; singular value decomposition is performed on the spatiotemporal fusion features to obtain eigenvector values ​​and eigenvalues, the eigenvector values ​​and the proportion of the eigenvalues ​​to the total eigenvalues ​​are calculated to obtain feature contributions, the feature contributions are sorted from large to small, and the eigenvectors whose cumulative feature contributions reach the preset contribution threshold are selected to construct evolutionary pattern features, which are used to characterize the evolutionary pattern features of the key monitoring feature points.

8. The method according to claim 1, further comprising: Deploy a drone airport on the tower, which is used for parking, charging and communication of drones; Determining an inspection path based on a preset inspection strategy, wherein the inspection strategy includes a leapfrog inspection path, wherein the leapfrog inspection path is a flight path in which a drone takes off from a drone airport at a current tower, flies to a drone airport at an adjacent tower for parking and charging, and then flies to a drone airport at the next tower; Controlling the drone to inspect the area surrounding the tower according to the inspection path, and collecting image data of the area surrounding the tower; The image data is analyzed to identify illegal reclamation, illegal grazing, fire hazards and abnormal animal intrusion in the area around the tower; when a safety hazard is detected, a hazard alarm message is generated and sent to the monitoring terminal.

9. A safety hazard monitoring system along a tower based on a drone, used to implement the method according to any one of claims 1 to 8, characterized in that: include: The first unit is configured to obtain three-dimensional model data and historical deformation data of a target tower, and determine a collaborative inspection path and collection locations for multiple drones based on the three-dimensional model data; and control the multiple drones to fly to corresponding collection locations along the collaborative inspection path, wherein each drone is equipped with a depth camera; The second unit is configured to establish a layered acquisition strategy based on the structural stress analysis results of the target tower, divide the target tower into multiple deformation-sensitive areas, and dynamically allocate the acquisition density and frequency of the drone according to the stress characteristics of each deformation-sensitive area; and control the multiple drones to synchronously acquire multi-view depth images of the target tower according to the layered acquisition strategy; The third unit is used to construct a real-time three-dimensional point cloud model of the target tower based on the multi-view depth image, align the real-time three-dimensional point cloud model with the three-dimensional model data of the target tower, extract deformation feature points, and grade the deformation feature points using a double-threshold iteration method, and determine key monitoring feature points based on the screening results; The fourth unit is used to dynamically track the key monitoring feature points using a Kalman filter algorithm to obtain a motion trajectory of the deformation feature points, and generate early warning information when the displacement of the motion trajectory exceeds a preset evaluation threshold.

10. An electronic device, characterized in that: include: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the method according to any one of claims 1 to 8.

Images