A Method and System for Joint Detection of Tower and Line Status and Safe Distance of Three-Span Transmission Lines Based on Multi-Source Information Fusion

By using multi-source information fusion technology, the problem of isolated information processing in fault detection of three-span transmission lines has been solved. It enables accurate construction of tower-line topology and dynamic correction of safety distance in complex environments, providing continuous and reliable fault detection results and supporting rapid emergency response.

CN123111201APending Publication Date: 2026-11-17STATE GRID HEBEI ELECTRIC POWER CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611300878.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-26
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

In existing technologies, fault detection methods for three-span transmission lines fail to effectively integrate multi-source heterogeneous information, leading to difficulties in constructing tower-line topology relationships in complex environments, easy confusion in attribution of faulty conductors, inaccurate assessment of safe distances, and a lack of continuity and reliability, making it difficult to provide rapid and accurate emergency detection results in harsh environments.

Method used

By using a multi-source information fusion method, multi-source detection information and prior information collected by UAVs are obtained, and spatiotemporal unified and adaptive fusion is performed to establish a unified spatiotemporal relationship between the tower line and the three-span object. The tower line status and spatial morphology are jointly identified, and the safety distance is dynamically corrected based on uncertainty information to achieve continuous judgment of intrusion risk.

Benefits of technology

It significantly improves the robustness and stability of tower tilt and conductor anomaly identification, avoids misjudgment and missed detection of detection results, provides accurate and reliable safety distance assessment, and supports rapid emergency response to tower collapse and line breakage faults.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123111201A_ABST
    Figure CN123111201A_ABST
Patent Text Reader

Abstract

This invention discloses a method and system for joint detection of tower and conductor status and safe distance in three-span transmission lines based on multi-source information fusion, relating to the fields of UAV intelligent inspection and multi-source information fusion technology. The method acquires multi-source detection information and prior information collected by UAVs; it unifies the multi-source information spatiotemporally, generates data quality and uncertainty information, and adaptively fuses it to establish a unified spatiotemporal relationship; it extracts tower and conductor status and spatial morphology through joint identification; it calculates the minimum distance based on the spatial morphology and the safety clearance of the three-span object, and dynamically corrects it based on uncertainty information to obtain an effective safe distance; and it jointly determines the intrusion risk level based on the effective safe distance and the tower and conductor status change trend. This invention jointly detects tower status, conductor status, and safe distance, improving the accuracy and reliability of emergency detection of tower collapse and line breakage faults in three-span transmission lines under complex environments.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the fields of intelligent inspection by unmanned aerial vehicles (UAVs) and multi-source information fusion technology, and in particular to a method and system for joint detection of the status and safe distance of three-span transmission line towers based on multi-source information fusion. Background Technology

[0002] Overhead transmission lines are crucial infrastructure for power systems to transmit large volumes of electricity over long distances. As the power grid continues to expand, transmission lines inevitably cross high-speed railways, conventional railways, highways, and important transmission corridors, creating "three-cross" sections that affect power grid safety, traffic safety, and public safety. Affected by factors such as strong winds, blizzards, icing, freezing rain, severe convective weather, geological subsidence, and external damage, three-cross lines may experience faults such as tower foundation instability, tower tilting, tower head or crossarm deformation, abnormal conductor sag, broken wires, and multiple wire overlap. After these faults occur, tilted towers, deformed components, and faulty conductors may gradually approach or directly encroach upon the safety clearances of railway contact networks, highway traffic areas, crossed transmission lines, and other important facilities, leading to secondary accidents such as railway shutdowns, highway closures, discharges from adjacent lines, electric shocks, and widespread power outages.

[0003] Currently, emergency fault detection on three-span railway lines mainly relies on manual inspections, conventional drone photography, distance measuring equipment measurements, and on-site assessment by personnel. In mountainous areas, valleys, areas with traffic disruptions, and adverse conditions such as rain, snow, and strong winds, these methods suffer from problems such as slow personnel arrival times, limited detection range, reliance on experience for spatial location judgment, and long preparation times for emergency response. Especially when tower collapse components obstruct the view, multiple conductors overlap, and there is interference from railway contact networks, road facilities, and surrounding terrain features, it is difficult to quickly determine the tower deformation state, the identity of the faulty conductor, and its actual spatial relationship with the three-span object.

[0004] In recent years, with the development of technologies such as UAV platforms, machine vision, LiDAR, laser ranging, satellite positioning, and inertial measurement, multi-source sensing methods have been gradually applied to power transmission line tower tilt detection, conductor sag measurement, broken strand and wire identification, and safe distance detection. Existing tower tilt detection methods typically utilize UAVs to acquire visible light images or LiDAR point clouds, obtaining the tower area through target detection, semantic segmentation, point cloud clustering, key point extraction, principal component analysis, or random sampling consistency fitting. The degree of tower tilt is then calculated based on the offset between the tower top and base or the angle between the tower's central axis and the vertical direction. Existing conductor detection methods typically employ edge detection, curve fitting, pixel-level segmentation, or catenary fitting to extract the conductor contour and three-dimensional spatial curve, thereby identifying conductor breaks, sag, overlaps, and abnormal sag. Regarding safe distance detection, some methods utilize LiDAR point clouds or distance measurement information to obtain the spatial coordinates of the conductor and three-span objects, then calculate the clearance distance using the point-to-plane distance or the shortest distance between spatial curves, and compare it with a preset safe distance threshold.

[0005] However, tower collapses and line breaks in three-span power lines are characterized by complex target structures, abrupt changes in spatial relationships, significant overlapping and obstruction, and continuous state changes. Existing methods typically employ independent algorithms to process visible light images, lidar point clouds, and distance measurement information. There is a lack of unified data organization and result transmission mechanisms between tower tilt detection, conductor anomaly identification, and safe distance measurement. Furthermore, these methods fail to fully integrate UAV pose information, prior line information, and continuous temporal information to jointly constrain the tower-line topology and spatial geometry, easily leading to inconsistencies between tower identification results and conductor suspension relationships, incorrect attribution of overlapping conductors, and difficulty in matching detection results from different times. Simultaneously, under conditions of rain, snow, fog, haze, backlighting, and strong wind disturbances, visible light images may become blurred and reflective, lidar point clouds may become sparse and missing, distance measurements may fluctuate, and UAV positioning and attitude information may drift. Existing methods lack unified modeling of data quality and uncertainty information, fail to adjust the weight of different information sources according to the reliability of the data, and fail to use continuous time-series information to distinguish between real tower line changes and apparent changes caused by the movement of the UAV itself. This leads to misjudgment of tower status, missed detection of faulty conductors, and distortion of safety distances in complex environments, making it difficult to provide continuous and reliable detection results for rapid emergency assessment and subsequent handling of tower collapse and line breakage faults. Summary of the Invention

[0006] To address these issues, this invention provides a method and system for joint detection of tower and line status and safety distance of three-span transmission lines based on multi-source information fusion. This method solves the technical problems in the prior art where tower tilt detection, conductor anomaly identification, and safety distance measurement are independent of each other, and the failure to effectively integrate multi-source heterogeneous information and fully utilize temporal correlation and data quality uncertainties leads to difficulties in constructing tower and line topology relationships in complex environments, easy confusion in attribution of faulty conductors, inaccurate safety distance assessment, and a lack of continuity and reliability in emergency risk determination.

[0007] To address the aforementioned technical problems, embodiments of the present invention provide a method for joint detection of tower and line status and safe distance of a three-span transmission line based on multi-source information fusion, comprising: Acquire multi-source detection information and prior information collected by drones; The multi-source detection information is spatiotemporally unified to generate data quality and uncertainty information, and adaptive fusion is performed based on the data quality and uncertainty information to establish a unified spatiotemporal relationship between the tower line and the three-span object; Based on the unified spatiotemporal relationship, the tower line status and spatial morphology are extracted through joint identification. The minimum distance is calculated based on the spatial morphology and the safety clearance of the three-span object, and the minimum distance is dynamically corrected based on the uncertainty information to obtain an effective safety distance; The risk level of intrusion is jointly determined based on the effective safe distance and the changing trend of the tower line status.

[0008] Preferably, the step of spatiotemporally unifying the multi-source detection information to generate data quality and uncertainty information, and adaptively fusing the data quality and uncertainty information, includes: Based on a unified detection time, the position, velocity and attitude of the UAV between adjacent sampling times are interpolated, and the visible light image, lidar point cloud and distance measurement information are compensated to the unified detection time. Point-level time correction is performed on the lidar point cloud, and each point in the same frame of point cloud is compensated to the unified detection time according to the scanning time of each point and the UAV pose change. The compensated data is transformed from the coordinate systems of each sensor to a unified world coordinate system, and the image pixels are restored to world coordinates by combining the camera intrinsic parameters and distance measurement results, so that the image target matches the point cloud target space. For dynamic targets such as guide wires, the spatial position changes between continuous detection times are used to estimate the motion components of the guide wires, and the platform's own motion components are eliminated based on the UAV's pose changes. The overall quality scores of each type of data are evaluated, and corresponding uncertainty information is generated based on measurement error, registration residual, identification confidence, and temporal consistency. Based on the comprehensive quality score and uncertainty information, adaptive fusion weights for various types of information are determined, and features from different information sources are combined in a weighted manner. When the overall quality score of any information source is lower than the preset failure threshold, the information source will not participate in the target spatial location calculation at the current moment, but its detection results and quality flags will be retained for subsequent cross-frame verification.

[0009] Preferably, the step of extracting tower line status and spatial morphology through joint identification based on the unified spatiotemporal relationship includes: Image features, point cloud features, distance features, motion features, prior features, and temporal features obtained after spatiotemporal unification and adaptive fusion are input into a pre-established joint recognition model; The joint identification model synchronously outputs the tower body area, key points of the tower body structure, conductor area, conductor instance, conductor endpoint, conductor suspension point, three-span object type, three-span object spatial boundary, and the position confidence and position uncertainty of each target; When multiple conductors overlap, conductors are obscured, conductors break and their positions change abruptly, or there are linear interference targets in the background, conductor instances are associated by comprehensively considering conductor endpoints, suspension points, extension directions, phase arrangement, crossarm levels, historical topological relationships, and cross-frame motion trajectories. For normal conductors, the phase and span are determined based on the suspension points at both ends, the direction of extension, and the design arrangement sequence; for faulty conductors that have broken off from the suspension points at one or both ends, their original phase, original connecting tower, and original suspension position are restored based on historical topology, break location, spatial continuity, and cross-frame trajectory. Towers, key structural points, crossarms, insulator strings, suspension points, conductors, and three-span objects are represented as target nodes, and support, suspension, connection, adjacency, crossing, and spatial proximity relationships are represented as associated edges to construct a tower-line topology map. The node association score is calculated based on spatial location, extension direction, structural priors, and cross-frame consistency. When the association score exceeds a preset threshold, an associated edge is established.

[0010] Preferably, the step of calculating the minimum distance based on the spatial shape and the safety clearance of the three-span object, and dynamically correcting the minimum distance based on the uncertainty information to obtain an effective safety distance, includes: Based on the type of object spanning three sections and its spatial boundaries, and combined with prior information, a corresponding three-dimensional safety clearance model is established: for railway objects, a railway safety clearance is established based on the rail surface, overhead contact line, vehicle operating space, and corresponding safety distances; for highway objects, a highway safety clearance is established based on the road surface, road boundaries, vehicle clearance, and corresponding safety distances; for crossed power transmission lines, a dynamic safety clearance is established based on the spatial curve of the crossed conductor, its normal swing range, voltage level, and electrical safety distance. Search along the spatial curve of the conductor to find the point closest to the safety clearance, and obtain the minimum geometric distance between the conductor and the three-span object; search for the minimum spatial distance between the outer contour of the tower and the safety clearance; record the conductor point or tower point at which the minimum distance is obtained and the corresponding safety clearance point as the most dangerous point pair; Based on the spatial curve of the conductor at continuous time, estimate the velocity and acceleration of the conductor at each position, predict the position of the conductor within a short time range, and take the minimum distance between the conductor curve and the safety clearance within the prediction time as the minimum distance for dynamic prediction. Based on the direction of the connection between the most dangerous points, the uncertainty of traverse reconstruction, the uncertainty of the three-span object boundary, the UAV pose error, the registration error, and the distance measurement error are projected onto this direction to calculate the uncertainty of the minimum distance; The effective safe distance of the conductor is obtained by jointly correcting the dynamic predicted minimum distance, the uncertainty of the minimum distance, and the conductor swing compensation amount. Compare the effective safety distance of the conductor with the safety distance of the tower outline, take the smaller one as the control safety distance, and calculate the safety distance margin in combination with the required safety distance.

[0011] Preferably, the step of jointly determining the intrusion risk level based on the effective safety distance and the changing trend of the tower line status includes: Based on the constructed tower-line topology diagram, the tower status, conductor status, and safety distance status are correlated; For the same tower and its connecting conductors, analyze the consistency of the tower tilt direction, suspension point displacement direction, conductor swing direction, most dangerous point direction, and three-span object direction; when the tower continues to tilt towards the three-span object direction and the conductor simultaneously shows displacement, sag increase, or abnormal swing, and the effective safety distance continues to decrease, it is determined to be an associated fault on the same topology chain; Tower risk values ​​are constructed based on tower tilt angle, offset, and degree of bending; conductor risk values ​​are constructed based on conductor breakage, sag, splicing, abnormal sag, and degree of sway; and safety distance risk values ​​are constructed based on effective safety distance, safety margin, and distance change trends. The joint risk value is calculated using the tower risk value, conductor risk value, safety distance risk value, and association enhancement value determined based on topological association and directional consistency. Based on the uncertainties of each information source and the uncertainties of tower, conductor, and safety distance detection, a comprehensive detection uncertainty is generated, and the comprehensive detection reliability is determined accordingly. The joint risk value within the sliding time window is smoothed, and the final risk level is determined by combining the risk change rate, duration, and safety margin: when it is below the first threshold, it is a normal state; when there is an abnormal change but it is not close to the safety limit, it is a state of concern; when there is a broken line, a drop, or the effective safety distance is close to the required value and continues to decrease, it is an alarm state; when the safety margin is less than zero or has violated the safety limit, it is a severe state and an intrusion alarm is immediately output.

[0012] Preferably, the multi-source detection information includes visible light image information, lidar point cloud information, distance measurement information, and UAV position and motion status information; the prior information includes prior information of transmission lines and three-span objects, continuous time series and historical reference information; the prior information of transmission lines and three-span objects includes tower type, tower design height, tower structure dimensions, crossarm level, insulator string position, conductor arrangement, conductor phase, span, design sag, suspension point design position, three-span object type, historical spatial range of three-span objects, and safety distance requirements; the continuous time series and historical reference information includes images, point clouds, distances, poses, target recognition and spatial reconstruction results before the current moment, as well as the tower centerline, tower top position, conductor spatial curve, normal sag range, normal swing range, and safety distance range of the same object under historical normal conditions.

[0013] Preferably, the data quality and uncertainty information includes: the sharpness, contrast, exposure, rain, snow, and fog interference, occlusion ratio, and recognition confidence index of visible light images; the point cloud density, target area integrity, echo stability, outlier ratio, and local registration residual index of lidar point clouds; the continuous fluctuation and abnormal jump ratio index of distance measurements; and the positioning accuracy, attitude accuracy, pose continuity, and motion stability index of the UAV's position and motion state. The comprehensive quality score of various types of information at the detection time is the weighted sum of its various quality indicators; the normalization uncertainty of various types of information is determined based on measurement error, spatial registration residual, target recognition confidence, and continuous temporal consistency. When image quality degrades, reduce the weight of image features and increase the weight of point cloud, distance, and temporal features; when point clouds are sparse or missing, reduce the weight of point cloud features and increase the weight of image contour, structural prior, and historical curves; when the UAV pose accuracy decreases, increase the uncertainty of the target's absolute position and use conservative corrections in the safe distance calculation.

[0014] Preferably, the spatial morphology includes the spatial outline of the tower and the spatial curve of the conductor; the tower-line status includes the tower status and the conductor status. The status of the tower is identified in the following ways: Extract the effective 3D point cloud of the tower based on the tower body area and key structural points; Divide the tower into multiple height levels along the vertical direction of the scene, calculate the center point of the tower in each level, and form a sequence of center points; A robust linear fit was performed on the center point sequence to obtain the main central axis of the tower. The overall tilt angle is determined based on the angle between the main central axis and the vertical direction of the scene, the tilt direction is determined based on the horizontal component of the main central axis, and the horizontal offset of the tower top relative to the tower base is calculated based on the effective detection height. Calculate the lateral deviation of the center point of each level relative to the main central axis; when most levels are offset along the same direction and the tilt angle exceeds the threshold, it is determined to be an overall tilt; when only some continuous levels have increased lateral deviation while other levels are still close to the main central axis, it is determined to be a local bend. The trend of change is determined by the tilt angle, direction, offset, and deviation within the sliding time window; when the apparent tilt direction is consistent with the direction of UAV movement but the key points in the world coordinates are stable, it is determined to be an apparent change caused by the movement of UAV. For the conductor suspension points that are obstructed, their position uncertainty is restored and increased based on the key points of the crossarm structure, the direction of inclination, the designed suspension position, and the historical topology. The actual spatial outline of the tilted tower is constructed based on the tower body point cloud, key structural points, main central axis, and local bending positions.

[0015] Preferably, the spatial morphology includes the spatial outline of the tower and the spatial curve of the conductor; the tower-line status includes the tower status and the conductor status. The conductor status is identified in the following ways: The image regions, point clouds, endpoints, suspension points, previous time curves, and historical normal curves of traverses belonging to the same span, phase, and historical suspension relationship are aggregated into a multi-source observation set for a single traverse. The current connection status is determined based on the spatial relationship between the conductor endpoint and the suspension point; For double-ended suspended traverses, the two suspension points are used as endpoint constraints, and the vertical curve is initialized using a catenary shape. For single-ended suspended traverses, the still-connected end is used as the fixed end, and the drooping section is reconstructed based on point cloud, image contour, free end position, and temporal trajectory. For traverses that have detached at both ends or whose towers have collapsed, piecewise curves and local smoothing constraints are used for reconstruction. The reconstruction of traverse curves is constrained by the spatial observation points, the connection of suspension points, the continuity of curves, the phase arrangement relationship, and the temporal consistency. When a section of the conductor lacks observation points due to obstruction, the observation points are supplemented based on the positions on both sides, tangent direction, curvature, suspension relationship, and previous curves; when the same area cannot be restored at multiple consecutive moments and a reasonable connection cannot be formed on both sides, it is identified as a suspected broken line or broken strand location. The maximum sag is calculated based on the spatial chord between the two suspension points of the double-ended suspended conductor; the wind deflection and sag deviation are obtained based on the lateral and vertical deviations of the current curve from historical or design benchmarks; and overlapping and multi-line overlap anomalies are identified based on the minimum distance between adjacent conductors, their intersection positions, and durations. Calculate the dynamic swing amplitude based on the spatial curves of the same conductor at different times within a sliding time window; Abnormal states are identified based on conductor curves, endpoint connection status, suspension point connection status, spatial continuity, and temporal characteristics: a broken line is defined as a continuous interruption of the curve that cannot be reasonably connected; drooping is defined as one or both ends detaching from the original suspension point and moving significantly downward; single-end suspension is defined as a situation where only one suspension end remains and the conductor moves significantly around that end; overlapping or multiple-line overlap is defined as a situation where different conductors continuously approach, cross, or overlap and the minimum phase-to-phase distance is below the threshold; sag is defined as a situation where the maximum sag deviates from the historical or design value by more than the threshold; and abnormal oscillation is defined as a situation where the swing amplitude, speed, or spatial range exceeds the normal range and shows a continuous trend of approaching the safety limit.

[0016] This invention also provides a joint detection system for the status and safe distance of three-span transmission line towers based on multi-source information fusion, comprising: The information acquisition module is used to acquire multi-source detection information and prior information collected by the UAV; The spatiotemporal unification and adaptive fusion module is used to perform spatiotemporal unification on the multi-source detection information, generate data quality and uncertainty information, and perform adaptive fusion based on the data quality and uncertainty information to establish a unified spatiotemporal relationship between the tower line and the three-span object. The joint identification module is used to extract the tower line status and spatial morphology based on the unified spatiotemporal relationship through joint identification; The safety distance calculation and correction module is used to calculate the minimum distance based on the spatial shape and the safety clearance of the three-span object, and to dynamically correct the minimum distance based on the uncertainty information to obtain an effective safety distance; The joint determination and output module is used to jointly determine the intrusion risk level and output it based on the effective safe distance and the changing trend of the tower line status.

[0017] As can be seen from the above technical solutions, this invention application has the following beneficial effects: First, this invention constructs a unified evaluation system for data quality and uncertainty information by unifying the spatiotemporal information of multi-source detection information and adaptively fusing quality perception. It can dynamically adjust the weight of each information source in feature fusion according to quality indicators such as image clarity, point cloud integrity, and pose accuracy. This effectively solves the problem of detection results failing due to the degradation of a single data source in complex environments such as rain, snow, fog, haze, strong wind disturbance, and overlapping occlusions. It significantly improves the robustness and stability of pole tilt and conductor anomaly identification under harsh working conditions.

[0018] Second, this invention extracts key points of the tower structure, conductor instances, suspension points, and spatial boundaries of three-span objects simultaneously through a joint identification model, and constructs a tower line topology map with "tower-crossarm-insulator string-suspension point-conductor-three-span object" as the main line. This enables tower tilt detection, conductor anomaly identification, and safe distance calculation to achieve mutual verification through topological constraints and temporal correlation within the same spatiotemporal framework. This effectively avoids the technical defects in existing technologies, such as confusion of conductor attribution, inconsistency of tower line topology relationships, and difficulty in corresponding cross-frame detection results caused by the independent processing of each detection task.

[0019] Third, this invention introduces a dynamic prediction and uncertainty correction mechanism in the calculation of safe distance. It estimates the speed and acceleration of the conductor based on the continuous time-series trajectory to predict the short-term swing range. It also projects the uncertainty of conductor reconstruction, the uncertainty of three-span object boundary identification, the UAV pose error, the coordinate registration error, and the distance measurement error onto the most dangerous point to jointly correct the direction, thereby obtaining an effective safe distance with credibility evaluation. It combines the continuous change trend of tower status, conductor status, and effective safe distance to make a joint risk judgment, avoiding the problems of false high or low distance and false alarms caused by single-frame measurement fluctuations or instantaneous swings. It provides an accurate and reliable spatial basis for rapid emergency assessment and wire cutting of tower collapse and wire breakage faults. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly described below. Referring to the drawings will make the features and advantages of the present invention clearer. The drawings are illustrative and should not be construed as limiting the present invention in any way. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1 This is a flowchart of a method for joint detection of tower and line status and safe distance of a three-span transmission line based on multi-source information fusion, provided by an embodiment of the present invention. Figure 2 This is a schematic diagram illustrating the process of time synchronization, coordinate transformation, motion compensation, data quality evaluation, and uncertainty generation of multi-source detection information in an embodiment of the present invention; Figure 3 This is a schematic diagram of the quality-sensing spatiotemporal multi-task joint identification model, conductor instance association, and tower-line topology construction in an embodiment of the present invention; Figure 4 This is a schematic diagram of three-dimensional spatial reconstruction and abnormal state identification of conductors based on suspension relationship, spatial continuity and continuous temporal information in an embodiment of the present invention; Figure 5This is a schematic diagram illustrating the construction of a three-span object safety boundary, the search for the most dangerous point pair, the dynamic prediction distance calculation, and the correction of the effective safety distance in an embodiment of the present invention. Figure 6 This is a schematic diagram illustrating the continuous joint determination of tower status, conductor status, safety distance status, change trend, and comprehensive detection reliability in an embodiment of the present invention; Figure 7 This is a block diagram of a joint detection system for the status and safe distance of a three-span transmission line tower based on multi-source information fusion, provided by an embodiment of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Example 1: To address the technical problems in existing technologies where tower tilt detection, conductor anomaly identification, and safety distance measurement are independent and fail to effectively integrate multi-source heterogeneous information and fully utilize temporal correlations and data quality uncertainties, leading to difficulties in constructing tower-line topology relationships in complex environments, easy confusion regarding faulty conductor attribution, inaccurate safety distance assessments, and a lack of continuity and reliability in emergency risk determination, this invention proposes a joint detection method for tower-line status and safety distance of three-span transmission lines based on multi-source information fusion. Figure 1 As shown, the method includes the following steps: S1: Acquire multi-source detection information and prior information collected by the UAV; S2: Spatiotemporally unify multi-source detection information to generate data quality and uncertainty information, and adaptively fuse the data quality and uncertainty information to establish a unified spatiotemporal relationship between the tower line and the three-span object; S3: Based on a unified spatiotemporal relationship, tower line status and spatial morphology are extracted through joint identification; S4: Calculate the minimum distance based on the spatial morphology and the safety clearance of the three-span object, and dynamically correct the minimum distance based on uncertainty information to obtain an effective safety distance; S5: Based on the changing trends of the effective safe distance and tower line status, jointly determine the level of intrusion risk.

[0023] In step S1, the time corresponding to the preset detection period is... As the first A unified detection time is used to obtain the multi-source detection information corresponding to that detection time. The multi-source detection information includes visible light image information. LiDAR point cloud information Distance measurement information And information on the location and movement of the drone Simultaneously, prior information on transmission lines and three-span objects is acquired. and continuous time series and historical benchmark information .

[0024] No. The set of multi-source detection information at a unified detection moment is represented as: , in, This provides information on the quality and uncertainty of the data generated subsequently.

[0025] The prior information for the transmission line and the three-span object includes tower type, tower design height, tower structure dimensions, crossarm level, insulator string position, conductor arrangement, conductor phase, span, design sag, suspension point design position, three-span object type, historical spatial range of the three-span object, and corresponding safety distance requirements. The continuous temporal and historical reference information includes images, point clouds, distances, UAV poses, target recognition and spatial reconstruction results before the current detection time, as well as the tower centerline, tower top position, conductor three-dimensional spatial curve, normal sag range, normal swing range, and safety distance range of the same tower, the same span, and the same three-span object under historical normal conditions.

[0026] In step S2, as Figure 2 As shown, the process involves time synchronization, spatial coordinate transformation, motion compensation, data quality evaluation, and uncertainty generation of multi-source detection information, followed by adaptive fusion based on data quality and uncertainty information. Specifically, this includes the following sub-steps: Since image, point cloud, distance measurement, and UAV pose data have different sampling frequencies and data output times, a unified detection time is needed. Based on the baseline, the position, velocity and attitude of the UAV between adjacent sampling times are interpolated, and the visible light image, lidar point cloud and distance measurement information are compensated to the unified detection time according to the actual acquisition time of various data.

[0027] Point-level time correction is performed on the lidar point cloud. Based on the scanning time corresponding to each point and the position and attitude changes of the UAV during the scanning process, the points in the same frame of the point cloud are compensated to a unified detection time to reduce the stretching, misalignment and bending of the point cloud caused by the flight motion of the UAV.

[0028] The compensated lidar points are transformed from the lidar coordinate system to the unified world coordinate system, and their homogeneous coordinate relationship is expressed as: , in, Let be the homogeneous coordinates of the point in the unified world coordinate system; Let be the homogeneous coordinates of the point in the lidar coordinate system; To standardize the transformation matrix from the UAV body coordinate system to the world coordinate system at the time of detection; This is the extrinsic transformation matrix from the lidar coordinate system to the UAV body coordinate system.

[0029] For target pixels in visible light images, their world coordinates are recovered by combining camera intrinsic parameters, distance measurement results, camera extrinsic parameters, and UAV pose, so that key points of the tower, wire endpoints, suspension points, and three-span object boundaries in the image can be spatially matched with corresponding targets in the lidar point cloud.

[0030] For dynamic targets such as guide wires, the motion components of the guide wires are initially estimated by utilizing the candidate regions, point sets, and local spatial position changes of the guide wires between continuous detection times. Then, the motion components of the platform itself are eliminated based on the UAV pose changes, thereby distinguishing between the apparent displacement caused by the UAV motion and the spatial changes caused by the actual swaying of the guide wires.

[0031] After completing time synchronization, coordinate transformation, and motion compensation, the effectiveness of various data types at the current detection time is evaluated. Image quality evaluation indicators include image sharpness, contrast, exposure, rain, snow, and fog interference, occlusion ratio, and recognition confidence; point cloud quality evaluation indicators include point cloud density, target region integrity, echo stability, outlier ratio, and local registration residual; distance measurement quality evaluation indicators include continuous measurement fluctuation and abnormal jump ratio; and UAV position and motion state quality evaluation indicators include positioning accuracy, attitude accuracy, pose continuity, and motion stability.

[0032] No. Class information at the detection time The overall quality score is expressed as: , in, For the first The overall quality score of the information category; For this type of information Quality indicators; The weights for the corresponding quality indicators; This refers to the number of quality indicators included in this type of information.

[0033] Uncertainty is generated based on measurement error, spatial registration residual, target recognition confidence, and continuous temporal consistency. The uncertainty increases when data quality is low, registration residual is large, recognition confidence is low, or cross-frame variations are unstable.

[0034] Taking into account both the overall quality score and data uncertainty, determine the first... Adaptive weights of class information in subsequent multi-source feature fusion: , in, For the first Adaptive fusion weights for class information; For the first Normalization uncertainty of class information; The number of information categories participating in the integration; This is the quality adjustment coefficient; This is the uncertainty suppression coefficient.

[0035] When images are affected by rain, snow, fog, haze, backlighting, or occlusion, the role of image features in spatial location determination is reduced, while the role of point cloud, distance, and continuous temporal features is increased. When point clouds are sparse, missing, or have an increase in outliers, the weight of point cloud features is reduced, while the role of image contours, suspension relationships, prior knowledge of line structures, and historical spatial curves is increased. When the pose accuracy of the UAV decreases, the uncertainty of the target's absolute position is increased, and a more conservative correction is used in subsequent safe distance calculations.

[0036] When the quality score of a certain type of information is lower than the preset failure threshold, this type of information will not directly participate in the calculation of the target spatial location at the current detection time, but its original detection results and quality indicators will be retained for subsequent cross-frame verification and data recovery.

[0037] The quality-perceived fusion feature at the current detection moment is represented as: , in, The quality perception fusion features at the current detection moment; For the first Feature vectors corresponding to class information; This refers to the adaptive fusion weights for this type of information.

[0038] Through the above processing, a unified time reference, unified world coordinates, quality adaptive weights, and location uncertainty information are obtained, and a unified spatiotemporal relationship is established between the tower, the conductor, and the three-span object.

[0039] In step S3, as Figure 3As shown, image features, point cloud features, distance features, UAV motion features, prior features of power transmission lines and three-span objects, and continuous temporal features obtained after spatiotemporal unification and adaptive fusion are input into a pre-established spatiotemporal multi-task joint recognition model. The spatiotemporal multi-task joint recognition model includes four processing parts: multimodal feature encoding, quality-aware feature fusion, temporal feature aggregation, and multi-task result output.

[0040] Multimodal feature encoding is used to extract texture, edge and contour features from images, spatial coordinates, local geometry, density and continuity features from point clouds, target depth features from distance measurements, pose and motion compensation features from UAV position and motion states, and prior and historical baseline features of line structure.

[0041] Quality-perceived feature fusion is based on the adaptive fusion weights obtained in step S2. Features from different sources are weighted and combined. Temporal feature aggregation maps the fused features at the current detection time to the target features, spatial location, and topological relationships at previous detection times, ensuring that the same tower, the same conductor, and the same three-span object maintain a stable identity at different detection times.

[0042] The multi-task result output section synchronously outputs the tower body area, key points of the tower body structure, tower top and tower base positions, crossarm level, insulator string position, conductor area, conductor instance, conductor endpoint, conductor suspension point, three-span object type, three-span object spatial boundary, confidence level of each target position, and uncertainty of each target position.

[0043] To enable all tasks to share information on tower structure and spatial relationships, a joint loss function is used to constrain the spatiotemporal multi-task joint identification model: , in, Losses are segmented for the tower body, conductors, and three-span area; Detecting damage at key points of the tower structure; Detect loss at the conductor endpoints and suspension points; This is due to the loss in tower-line topology. Loss for identifying the boundary of a three-span object; This results in cross-frame timing consistency loss. Estimate the loss for location uncertainty; to This corresponds to the task weight.

[0044] When multiple conductors overlap, conductors are obscured by tower components, the spatial position changes abruptly after a conductor breaks, or linear targets such as railway catenary, road guardrails, and building edges exist in the background, the conductor identity is not only determined based on the local linear appearance, but also by comprehensively considering the conductor endpoints, suspension points, extension direction, phase arrangement, crossarm level, historical topological relationships, and cross-frame motion trajectories to associate conductor instances.

[0045] For normal conductors, the phase and span are determined based on the suspension points at both ends, the direction of conductor extension, and the designed arrangement sequence. For faulty conductors that have detached from the suspension points at one or both ends, the original phase, the original connecting tower, and the original suspension position are restored based on the historical topology before the fault occurred, the current break location, the spatial continuity of the conductor, and the cross-frame motion trajectory. For railway catenaries, road guardrails, and other background linear objects, since they do not meet the connection relationship between the transmission conductor and the tower crossarm and suspension point, no suspension topology is established with the target transmission conductor.

[0046] The tower, key structural points of the tower body, crossarms, insulator strings, suspension points, conductors, and three-span objects are represented as target nodes, and the relationships of support, suspension, connection, adjacency, crossing, and spatial proximity are represented as associated edges to construct a tower-line topology diagram: , in, For the first Tower topology diagram at each detection time; For the target node set; This is the set of edges connecting the target nodes.

[0047] For any two target nodes and Node association scores are calculated based on spatial location, extension direction, structural priors, and cross-frame consistency. , in, Score based on spatial location association; The score is related to the direction of extension; The score is the prior matching score for the line structure. The score represents the continuous temporal correlation. , , , The corresponding weights are assigned. When a node's association score exceeds a preset association threshold, an association edge is established between the corresponding nodes. When multiple association metrics between nodes conflict, the confidence of the association edge is reduced, and the topological relationship from previous time steps and the target position from subsequent time steps are used for re-verification.

[0048] Furthermore, the spatial morphology includes the spatial outline of the tower and the spatial curve of the conductor, and the tower-line status includes the tower status and the conductor status.

[0049] The status of the tower is identified in the following ways, such as Figure 3 As shown: First, based on the tower area and key points of the tower structure obtained in step S3, the effective three-dimensional point cloud of the target tower is extracted, and the points corresponding to conductors, insulator strings, crossarm accessories and background features are removed.

[0050] The tower is divided into multiple height levels along the vertical direction of the scene. For each height level, weights are assigned to each tower point within that level based on point cloud quality, tower segmentation confidence, and outlier severity, and the tower center point for that height level is calculated. , in, For the first The center point of the tower body at each height level; For the first in this level One tower body point; The quality weights for the corresponding points; This represents the number of valid tower points within this height level.

[0051] The center points of each height level are arranged in order of height to form a sequence of tower body level center points. A robust linear fit is then performed on this center point sequence to obtain the main central axis of the tower body. The objective function for fitting the tower body central axis is expressed as: , in, This serves as a reference point on the central axis of the tower. The direction vector of the central axis of the tower; For the first Quality weights at each high level; A robust loss function; The effective number of height levels.

[0052] By minimizing the fitting error corresponding to the above formula, the optimal central axis reference point of the tower is obtained. and the optimal central axis direction vector .

[0053] Let the vertical unit vector of the scene be... The overall tilt angle of the tower is expressed as: , in, It is the angle between the main central axis of the tower and the vertical direction of the scene.

[0054] The tower tilt direction is determined based on the horizontal components of the tower's central axis in the direction along the line and in the transverse direction, and the horizontal offset of the tower top relative to the tower base is calculated in conjunction with the effective detection height of the tower.

[0055] To distinguish between overall tower tilt and local bending, the lateral deviation of the center point of each height level relative to the main central axis of the tower is calculated: , in, For the first The lateral deviation of the center point of each height level from the main central axis of the tower.

[0056] When the center points of most height levels shift in the same direction relative to their historical normal positions, and the angle between the main central axis of the tower body and the vertical direction exceeds a preset threshold, the tower is judged to have tilted as a whole. When only some continuous height levels show a significant increase in lateral deviation, while the tower base and other height levels remain close to the main central axis, the tower is judged to have bent locally. When the relative distance, angle, or hierarchical relationship between key structural points such as the tower head and crossarms deviates significantly from the historical normal benchmark, the tower head or crossarms is judged to have undergone local structural deformation.

[0057] The tilt angle, tilt direction, tower top offset, and local deviation within a sliding time window are used to determine the trend of tower status changes. When the tower tilt angle remains stable during continuous detection, a stable tilt state is output; when the tilt angle, tower top offset, or local bending degree continuously increases, a continuously unstable state is output; when the apparent tilt direction in the image or point cloud is consistent with the direction of UAV movement, but the position of the tower key points in the unified world coordinate system remains stable, the change is determined to be an apparent change caused by the UAV's own movement or attitude change.

[0058] For the location of the conductor suspension point, the actual three-dimensional coordinates with higher identification confidence and lower location uncertainty obtained in step S3 are preferred. When the suspension point is blocked by tower components, insulator strings, or other conductors, the suspension point location is restored based on the key points of the crossarm structure, the tower tilt direction, the designed suspension location, and historical topology, while increasing the uncertainty of the restored location.

[0059] The actual spatial outline of the tilted tower is constructed based on the point cloud of the tower area, key points of the tower structure, the main central axis, and local bending locations. The actual spatial outline includes the three-dimensional envelope of the tower body, tower head, crossarms, and significantly deformed components in a unified world coordinate system.

[0060] The state of the conductor is identified in the following ways, such as Figure 4 As shown: Based on the tower-line topology established in step S3, the image regions, point clouds, endpoints, suspension points, previous time-series conductor curves, and historical normal conductor curves belonging to the same span, phase, and historical suspension relationship are aggregated to form a multi-source observation set for a single conductor.

[0061] First, the current connection state of the conductor is determined based on the spatial relationship between the conductor endpoints and suspension points. For conductors where both ends are connected to their corresponding suspension points, it is determined to be a double-end suspension state; for conductors where only one end is connected to the original suspension point, it is determined to be a single-end suspension candidate state; for conductors where both ends are separated from the original suspension points, it is determined to be a disengagement candidate state; for cases where there is a partial spatial interruption in the conductor but the endpoint state is not yet clear, it is determined to be a candidate state of obstruction, broken strand, or broken wire by combining continuous time sequence information.

[0062] Establish a local coordinate system for the traverse with the span direction, horizontal direction, and vertical direction as coordinate axes. The wire at the detection time The three-dimensional space curve is represented as: , in, For the first The three-dimensional curve of a conductor; The origin of the local coordinate system of the traverse; It is a unit vector along the span direction; This is the unit vector in the horizontal direction of the line. is the unit vector in the vertical direction; These are the curve parameters along the direction of the conductor.

[0063] For traverses with normal suspension at both ends, the actual or modified suspension points at both ends are used as endpoint constraints, and the vertical curve of the traverse is initialized using a catenary model. For drooping traverses with suspension at one end, the end still connected to the suspension point is used as the fixed end, and the drooping section is reconstructed based on the current traverse point cloud, image contour, free end position, and continuous temporal trajectory. For traverses that have detached at both ends or have experienced severe displacement after the tower collapse, they are not forced to meet the complete catenary model, but are reconstructed using piecewise spatial curves and local smoothing constraints.

[0064] The reconstruction of the traverse curve simultaneously considers the spatial observation points, the connection relationship of the suspension points, the continuity of the curve, the phase arrangement relationship, and the consistency of continuous temporal sequence. Its objective function is expressed as: , in, To belong to the The first wire One space observation point; The mass weights for the corresponding observation points; This refers to the constraint terms between the conductor endpoint and the suspension point; These are constraints on the continuity and smoothness of the conductor curve; Constraints on the arrangement of conductors in different phases; These are continuous timing constraint terms; , , , These are the corresponding constraint coefficients.

[0065] When a local area of ​​a conductor lacks observation points due to obstruction by tower components, insulator strings, other conductors, or three-span objects, the missing segment is supplemented based on the conductor positions on both sides of the missing area, tangent direction, local curvature, suspension relationship, and conductor curves from previous times. If the same area cannot be recovered at multiple consecutive detection times, and the conductor curves on both sides of the missing segment cannot form a reasonable connection that satisfies smoothness and length constraints, the area is identified as a suspected broken conductor or broken strand location, rather than as a regular obstruction area.

[0066] After obtaining the three-dimensional spatial curve of the conductor, the conductor sag is calculated based on the spatial chord between the actual suspension points at both ends of the double-ended conductor. The maximum sag of a conductor is expressed as: , in, For the first The maximum sag of the conductor; The height of the chord between the suspension points at both ends of the conductor; This refers to the actual height of the conductor. This corresponds to the conductor span length.

[0067] Spatially register the current conductor's 3D curve with historical normal curves or design reference curves. Based on the deviations between the two in the horizontal and vertical directions, obtain the conductor's lateral wind deflection and vertical sag deviation, respectively. Identify conductor overlaps, multiple line intersections, and phase spacing anomalies based on the minimum distance, intersection location, and duration between adjacent conductor spatial curves.

[0068] Within the sliding time window, the dynamic swing amplitude of the conductor is calculated based on the three-dimensional spatial curves of the same conductor at different detection times: , in, For the conductor at the parameter position The dynamic swing amplitude at that location; For sliding time windows; This is the average spatial curve of the conductor within the sliding time window.

[0069] Abnormal conductor states are identified based on the conductor's spatial curve, endpoint connection status, suspension point connection status, spatial continuity, and temporal motion characteristics: When a conductor curve experiences a continuous spatial interruption, and the interrupted area remains stable across multiple detection times and cannot be connected by a reasonable continuous curve, it is identified as a conductor breakage; when one or both ends of the conductor detach from the original suspension point and move significantly downwards towards the ground or a three-span object, it is identified as a conductor sag; when the conductor retains only one suspension end and exhibits significant movement around that end, it is identified as a single-end suspension anomaly; when different conductor curves continuously approach, cross, or overlap, and the minimum phase distance is below a preset threshold, it is identified as a conductor overlap or multi-line intersection anomaly; when the maximum sag deviates from the historical normal value or design value beyond a preset threshold, it is identified as a sag anomaly; when the swing amplitude, swing speed, or swing spatial range within the sliding time window exceeds the normal range and shows a trend of continuously approaching the safety limit of a three-span object, it is identified as abnormal swing.

[0070] In step S4, as Figure 5 As shown, the minimum distance is calculated based on the spatial morphology and the safety clearance of the three-span object, and the minimum distance is dynamically corrected based on uncertainty information to obtain an effective safety distance. Specifically, this includes the following sub-steps: Based on the three-span object type, spatial boundary, and location uncertainty obtained in step S3, a corresponding three-dimensional safety boundary model is established by combining the prior information of the three-span object.

[0071] For railway three-way crossings, a three-dimensional safety clearance is established based on the rail surface, railway catenary, vehicle operating space, line centerline, and corresponding safety distance requirements. For highway three-way crossings, a three-dimensional safety clearance is established based on the road surface, road boundary, vehicle clearance, road centerline, and corresponding safety distance requirements. For crossed power transmission lines, a dynamic three-dimensional safety clearance is established based on the three-dimensional spatial curve of the crossed conductor, normal swing range, voltage level, and specified electrical safety distance.

[0072] When the local boundary of a three-span object cannot be fully identified due to occlusion or missing point cloud data, the complete boundary is restored using the identified boundary, prior object structure, continuous temporal boundary, and historical normal position, while increasing the uncertainty of the boundary position of the corresponding missing region. The three-dimensional safety clearance of a three-span object is represented as follows: .

[0073] Along the first The 3D spatial curve of the root traverse searches for the point closest to the safety clearance of the three-span object, thus obtaining the minimum geometric distance between the current traverse and the three-span object: , in, For the first root wire and the first The minimum geometric distance between three-span object safety limits; Located within the range of conductor curve parameters; Located in the three-span object safety clearance Inside.

[0074] Based on the actual spatial outline of the tilted tower obtained in step S3, search for the minimum spatial distance between the tower's spatial outline and the safety clearance of the three-span object: , in, For a tilted tower, it is a spatial point within the actual spatial outline of the tower. For the first Spatial points within the safety clearance of a three-span object; It is the minimum distance between the actual spatial outline of the tower and the safety clearance of the three-span object.

[0075] Record the conductor point or tower point at which the minimum distance is achieved, along with the corresponding safety clearance point, and designate these as the most dangerous point pair. The most dangerous point pair is used to determine the specific risk object, the location of the intrusion, and the direction of approach.

[0076] Because the conductor may sway continuously due to wind, the geometric distance at the current detection moment cannot fully reflect the risk of intrusion within a short period. Therefore, based on the three-dimensional spatial curve of the continuous conductor obtained in step S3, the velocity and acceleration of the conductor at each location are estimated, and the position of the conductor within a short time range is predicted: , in, For predicting time The spatial position of the conductor afterward; For the conductor at the parameter position The speed of motion at that location; For the conductor at the parameter position The acceleration of motion at that point.

[0077] Within the preset prediction time range, calculate the minimum distance between the predicted guide curve and the safety clearance of the three-span object, and take the minimum value among all prediction times as the dynamic prediction minimum distance.

[0078] Based on the direction of the connection between the most dangerous points, the uncertainties of 3D reconstruction of the traverse, identification of the three-span object boundary, UAV pose error, coordinate registration error, and distance measurement error are projected onto the direction of minimum distance. The uncertainty of the minimum distance is expressed as: , in, The unit vector for the direction of the line connecting the most dangerous point; The spatial position covariance of the conductor; The covariance of the boundary positions of the three-span object; The covariance of the drone's pose error; Covariance of coordinate registration error; This represents the variance of the distance measurement error.

[0079] By jointly correcting the dynamic prediction minimum distance, the uncertainty of the minimum distance, and the conductor sway compensation amount that is not yet fully covered by short-time prediction, the effective safe distance of the conductor is obtained: , in, The corrected effective safety distance for the conductor; The minimum distance between the conductor and the safety clearance within the predicted time range; This is a correction factor corresponding to the preset confidence level; This is the compensation amount for conductor oscillation.

[0080] Compare the effective safety distance of the conductor and the spatial outline distance of the tilted tower, and take the smaller one as the first. The control safety distance corresponding to each three-span object: , in, For the tower or conductor relative to the first Control safety distance for three-span objects.

[0081] Let the first The required safe distance for a three-way cross-category object is The safety margin is then expressed as: , in, For the first Safety distance margin corresponding to each three-span object.

[0082] When the safety margin is greater than zero, it means that after considering the dynamic swing of the conductor and the uncertainty of multi-source measurement, the tower and conductor have not yet intruded into the safety clearance of the three-span object; when the safety margin is equal to zero, it means that the tower or conductor is in a critical safety state; when the safety margin is less than zero, it means that the tower component or faulty conductor has intruded into the safety clearance, or is about to intrude into the safety clearance based on the short-term movement trend.

[0083] In step S5, as Figure 6 As shown, based on the changing trends of the effective safety distance and the tower line status, the intrusion risk level is jointly determined, which specifically includes the following sub-steps: Based on the tower-line topology diagram established in step S3 The tower status, conductor status, and safe distance status obtained in step S3 are correlated accordingly with each other.

[0084] For the same tower and its connecting conductors, analyze the consistency between the tower tilt direction, the tower top offset direction, the suspension point displacement direction, the conductor drooping or swinging direction, the direction of the most dangerous point, and the direction of the three-span object.

[0085] When a tower continuously tilts towards a three-span object, and the conductor connected to the tower simultaneously exhibits suspension point displacement, increased sag, lateral offset, drooping, or abnormal swaying, while the effective safety distance continuously decreases within the sliding time window, it is determined that the tower anomaly, conductor anomaly, and decreased safety distance belong to the same topological chain of associated faults, and the joint risk level is increased.

[0086] When a tower is tilted, but the corresponding conductor connection status, spatial curve, and safety distance remain stable during continuous monitoring, the tower tilt is recorded as an independent anomaly and is not directly judged as a serious risk of intrusion.

[0087] When the tower is in normal condition, but the conductor exhibits spatial breakage, end detachment, single-end suspension, drooping, lap splicing, or abnormal swinging, the conductor fault risk is determined based on the spatial relationship between the faulty conductor and the safety clearance of the three-span object.

[0088] When the safe distance suddenly decreases at a certain detection moment, but the tower status, conductor curve and the position of the most dangerous point are discontinuous at the previous and subsequent detection moments, the data quality and uncertainty information generated in step S2 are used to determine whether the change is caused by pose drift, registration error or distance measurement fluctuation, so as to reduce the impact of single frame outliers on the final risk level.

[0089] The risk value of the tower is constructed based on the overall tilt angle of the tower, the offset of the tower top, the degree of local bending, and the trend of tilt change. Conductor risk values ​​are constructed based on the degree of conductor breakage, sag, single-end suspension, lap splicing, abnormal sag, and abnormal sway. ; Construct a safety distance risk value based on the effective safety distance, safety distance margin, and distance change trend. .

[0090] Calculate the joint risk value using three types of risks, the degree of correlation between tower and line topology, and continuous changing trends: , in, For the first The combined risk value at each detection moment; This represents the risk value of the tower's condition. This represents the risk value of the conductor's condition. Risk value for safe distance; The correlation enhancement value is determined based on the tower line topology, consistency of anomaly direction, and continuous change trend. , , , For the corresponding weights; For risk bias parameters; This is a function that maps the joint risk value to the interval between 0 and 1.

[0091] Specifically, when tower anomalies, conductor anomalies, and insufficient safety distances occur on the same tower-line topology chain, and the direction and trend of the anomalies are consistent, the correlation enhancement value is [value missing]. Increase; when there is no topological correspondence between the three types of anomalies, or when the anomaly direction and change trend are inconsistent, the correlation enhancement value decreases.

[0092] Based on the uncertainties of each information source, tower condition detection uncertainty, conductor condition detection uncertainty, and safety distance detection uncertainty, a comprehensive detection uncertainty is generated. Based on this, the reliability of the comprehensive testing was determined: , in, The overall reliability of the detection at the current detection moment; To comprehensively detect uncertainty.

[0093] When the image, point cloud, distance, and pose data are of high quality, the detection results from different data sources are consistent, and the towers, conductors, and safety distances change steadily over continuous detection time, the overall detection reliability is high. When the image and point cloud are severely degraded, the UAV pose drift is large, the object boundary recognition is unstable, or the conclusions from different data sources conflict with each other, the overall detection reliability decreases.

[0094] To suppress frequent changes in risk level caused by instantaneous conductor swing, instantaneous fluctuations in distance measurement, or misidentification of targets in a single frame, the joint risk value is smoothed within a sliding time window, and the final risk level is determined by comprehensively considering the rate of risk change, duration, and safety distance margin.

[0095] When the tower, conductor, and safety distance are all within the normal range, and the combined risk value is lower than the first risk threshold, it is considered to be in a normal state.

[0096] When there is an abnormal change in at least one of the following: tower tilt, conductor spatial shape, or safety distance, but the object does not approach the safety clearance of a three-span object, it is determined to be in a state of concern.

[0097] An alarm state is determined when a conductor is broken, drooping, suspended at one end, spliced, or swings abnormally, or when the effective safety distance is close to the required safety distance and continues to decrease.

[0098] When the safety distance margin is less than zero, or when a tilted tower component or a faulty conductor has encroached on the railway, highway, or crossed the safety clearance of a transmission line, it is determined to be a serious condition and an intrusion alarm is immediately output, without being limited by the number of times the sliding time window can be used.

[0099] The final output includes the abnormal location of the tower, the overall tilt angle of the tower, the tilt direction, the offset of the tower top, the location of local bending and the trend of tilt change; the phase of the faulty conductor, the original suspension relationship, the current connection status, the three-dimensional spatial curve, the location of the broken wire, the drooping section, the overlapping position, the maximum sag, the lateral wind deflection and the range of abnormal swing; the type of the three-span object, the three-dimensional safety clearance, the most dangerous point pair, the current geometric distance, the dynamically predicted distance, the effective safety distance and the safety distance margin; as well as the risk transmission relationship, the reliability of the comprehensive detection, the intrusion status and the emergency risk level.

[0100] As can be seen from the above technical solution, this invention proposes a joint detection method for tower and line status and safety distance of three-span transmission lines based on multi-source information fusion. This method unifies multi-source data spatiotemporal unification, quality and uncertainty modeling, spatiotemporal multi-task joint identification, tower and line topology reasoning, tower tilt detection, conductor anomaly identification, three-dimensional safety distance calculation, and continuous risk assessment within the same algorithm framework. The tower status detection results are used to determine the actual spatial outline of the tower and assist in restoring the conductor suspension point position under obstruction conditions. The conductor topology and three-dimensional spatial curves are used to determine the identity of the faulty conductor, abnormal sections, and dynamic movement range. The actual spatial outline of the tower and the three-dimensional spatial curves of the conductor jointly participate in the calculation of the three-span safety distance. The tower status, conductor status, and safety distance results are then cross-verified through tower and line topology, continuous time series, and historical benchmark information. This can reduce misjudgments of tower status, incorrect attribution of faulty conductors, missed detection of abnormal conductors, and distortion of safety distances caused by complex backgrounds, overlapping lines, partial obstruction, drone movement, degradation of sensing data, and dynamic swing of conductors. It can improve the accuracy, continuity, and reliability of emergency detection of tower collapse and line breakage faults in three-span transmission lines, and provide detection basis for confirming the location of encroachment, rapid fault assessment, and subsequent emergency line cutting.

[0101] Example 2: This invention provides a joint detection system for the status and safe distance of three-span transmission line towers based on multi-source information fusion. This system is used to implement the joint detection method for the status and safe distance of three-span transmission line towers based on multi-source information fusion described in Embodiment 1 above. Figure 7 As shown, it specifically includes: The information acquisition module 100 is used to acquire multi-source detection information and prior information collected by the UAV. The multi-source detection information includes visible light image information, lidar point cloud information, distance measurement information, and UAV position and motion status information; the prior information includes prior information on power transmission lines and three-span objects, as well as continuous time series and historical reference information.

[0102] The spatiotemporal unification and adaptive fusion module 200 is used to perform spatiotemporal unification on the multi-source detection information, generate data quality and uncertainty information, and perform adaptive fusion based on the data quality and uncertainty information to establish a unified spatiotemporal relationship between the tower line and the three-span object. Specifically, this module is used to: synchronize and compensate the multi-source data based on a unified detection time; convert the data from each sensor to a unified world coordinate system; eliminate the platform's own motion components for dynamic targets such as conductors; evaluate the comprehensive quality score of each type of data and generate corresponding uncertainty information; determine the adaptive fusion weights of each type of information based on the comprehensive quality score and uncertainty information; and perform weighted combination of features from different information sources.

[0103] The joint identification module 300 is used to extract the tower line status and spatial morphology based on the unified spatiotemporal relationship through joint identification. Specifically, this module is used to: input various features obtained after spatiotemporal unification and adaptive fusion into a pre-established joint identification model, and simultaneously output the tower body area, key points of the tower body structure, conductor area, conductor instances, conductor endpoints, conductor suspension points, three-span object types, three-span object spatial boundaries, and the position confidence and position uncertainty of each target; when multiple conductors overlap, conductors are obscured, the position changes abruptly after a break, or there are linear interference targets in the background, the module comprehensively associates conductor instances based on conductor endpoints, suspension points, extension direction, phase arrangement, crossarm level, historical topology relationship, and cross-frame motion trajectory; and represents towers, key structural points, crossarms, insulator strings, suspension points, conductors, and three-span objects as target nodes to construct a tower line topology map.

[0104] The joint identification module 300 is also used to identify the status of the tower and the conductor. The tower status identification includes: extracting the effective three-dimensional point cloud of the tower based on the tower area and key structural points; dividing the tower into multiple height levels along the vertical direction of the scene and calculating the center point of each level; performing robust straight-line fitting on the center point sequence to obtain the main central axis of the tower; determining the overall tilt angle based on the angle between the main central axis and the vertical direction of the scene; determining the tilt direction based on the horizontal component of the main central axis; calculating the horizontal offset of the tower top relative to the tower base based on the effective detection height; calculating the lateral deviation of the center point of each level relative to the main central axis to distinguish between overall tilt and local bending; and constructing the actual spatial outline of the tilted tower based on the tower point cloud, key structural points, main central axis, and local bending positions. Conductor status identification includes: aggregating multi-source observations of conductors belonging to the same span, phase, and historical suspension relationship into a single conductor observation set; determining the current connection status based on the spatial relationship between the conductor endpoints and suspension points; reconstructing the three-dimensional spatial curve of the conductor using corresponding reconstruction strategies based on different connection statuses; and identifying states such as broken conductors, sag, single-end suspension, lap joints, abnormal sag, and abnormal oscillation based on conductor curves, endpoint connection status, suspension point connection status, spatial continuity, and temporal characteristics.

[0105] The safety distance calculation and correction module 400 is used to calculate the minimum distance based on the spatial morphology and the safety clearance of the three-span object, and to dynamically correct the minimum distance based on the uncertainty information to obtain an effective safety distance. Specifically, this module is used to: establish a corresponding three-dimensional safety clearance model based on the type of the three-span object and the spatial boundary combined with prior information; search along the conductor space curve for the point closest to the safety clearance to obtain the geometric minimum distance and record the most dangerous point pair; estimate the conductor's velocity and acceleration based on the conductor space curve at continuous time intervals and predict the conductor's position within a short time range; project various uncertainties onto the direction of the connection between the most dangerous point pairs to calculate the uncertainty of the minimum distance; jointly correct the dynamically predicted minimum distance, distance uncertainty, and conductor sway compensation to obtain the effective safety distance of the conductor; compare the effective safety distance of the conductor with the safety distance of the tower outline, taking the smaller one as the control safety distance, and calculate the safety distance margin in conjunction with the required safety distance.

[0106] The joint judgment and output module 500 is used to jointly determine and output the intrusion risk level based on the effective safety distance and the changing trend of the tower line status. Specifically, this module is used to: correlate the tower status, conductor status, and safety distance status according to the tower line topology diagram; analyze the consistency of abnormal directions of the same tower and its connected conductors to determine associated faults; calculate the joint risk value using the tower risk value, conductor risk value, safety distance risk value, and correlation enhancement value; generate a comprehensive detection uncertainty based on the uncertainty of each information source and determine the comprehensive detection reliability; smooth the joint risk value within the sliding time window and determine the final risk level; and finally output the abnormal location, intrusion status, and risk level.

[0107] This embodiment presents a joint detection system for the status and safe distance of three-span transmission line towers based on multi-source information fusion. This system is used to implement the aforementioned joint detection method for the status and safe distance of three-span transmission line towers based on multi-source information fusion. Therefore, the specific implementation of this system can be found in the embodiments of the aforementioned joint detection method for the status and safe distance of three-span transmission line towers based on multi-source information fusion. For example, the information acquisition module 100, the spatiotemporal unified and adaptive fusion module 200, the joint identification module 300, the safe distance calculation and correction module 400, and the joint judgment and output module 500 are respectively used to implement steps S1, S2, S3, S4, and S5 in the aforementioned joint detection method for the status and safe distance of three-span transmission line towers based on multi-source information fusion. Therefore, the specific implementation can be referred to the descriptions of the corresponding embodiments. To avoid redundancy, further details are omitted here.

[0108] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0109] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0110] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The functions specified in one or more boxes. These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable apparatus for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0111] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A method for joint detection of tower and line status and safe distance of a three-span transmission line based on multi-source information fusion, characterized in that, include: Acquire multi-source detection information and prior information collected by drones; The multi-source detection information is spatiotemporally unified to generate data quality and uncertainty information, and adaptive fusion is performed based on the data quality and uncertainty information to establish a unified spatiotemporal relationship between the tower line and the three-span object; Based on the unified spatiotemporal relationship, the tower line status and spatial morphology are extracted through joint identification. The minimum distance is calculated based on the spatial morphology and the safety clearance of the three-span object, and the minimum distance is dynamically corrected based on the uncertainty information to obtain an effective safety distance; The risk level of intrusion is jointly determined based on the effective safe distance and the changing trend of the tower line status.

2. The method for joint detection of tower and line status and safe distance of a three-span transmission line based on multi-source information fusion as described in claim 1, characterized in that, The process of spatiotemporally unifying the multi-source detection information to generate data quality and uncertainty information, and adaptively fusing the data quality and uncertainty information, includes: Based on a unified detection time, the position, velocity and attitude of the UAV between adjacent sampling times are interpolated, and the visible light image, lidar point cloud and distance measurement information are compensated to the unified detection time. Point-level time correction is performed on the lidar point cloud, and each point in the same frame of point cloud is compensated to the unified detection time according to the scanning time of each point and the UAV pose change. The compensated data is transformed from the coordinate systems of each sensor to a unified world coordinate system, and the image pixels are restored to world coordinates by combining the camera intrinsic parameters and distance measurement results, so that the image target matches the point cloud target space. For dynamic targets such as guide wires, the spatial position changes between continuous detection times are used to estimate the motion components of the guide wires, and the platform's own motion components are eliminated based on the UAV's pose changes. The overall quality scores of each type of data are evaluated, and corresponding uncertainty information is generated based on measurement error, registration residual, identification confidence, and temporal consistency. Based on the comprehensive quality score and uncertainty information, adaptive fusion weights for various types of information are determined, and features from different information sources are combined in a weighted manner. When the overall quality score of any information source is lower than the preset failure threshold, the information source will not participate in the target spatial location calculation at the current moment, but its detection results and quality flags will be retained for subsequent cross-frame verification.

3. The method for joint detection of tower and line status and safe distance of a three-span transmission line based on multi-source information fusion as described in claim 1, characterized in that, The extraction of tower line status and spatial morphology based on the unified spatiotemporal relationship through joint identification includes: Image features, point cloud features, distance features, motion features, prior features, and temporal features obtained after spatiotemporal unification and adaptive fusion are input into a pre-established joint recognition model; The joint identification model synchronously outputs the tower body area, key points of the tower body structure, conductor area, conductor instance, conductor endpoint, conductor suspension point, three-span object type, three-span object spatial boundary, and the position confidence and position uncertainty of each target; When multiple conductors overlap, conductors are obscured, conductors break and their positions change abruptly, or there are linear interference targets in the background, conductor instances are associated by comprehensively considering conductor endpoints, suspension points, extension directions, phase arrangement, crossarm levels, historical topological relationships, and cross-frame motion trajectories. For normal conductors, the phase and span are determined based on the suspension points at both ends, the direction of extension, and the design arrangement sequence; for faulty conductors that have broken off from the suspension points at one or both ends, their original phase, original connecting tower, and original suspension position are restored based on historical topology, break location, spatial continuity, and cross-frame trajectory. Towers, key structural points, crossarms, insulator strings, suspension points, conductors, and three-span objects are represented as target nodes, and support, suspension, connection, adjacency, crossing, and spatial proximity relationships are represented as associated edges to construct a tower-line topology map. The node association score is calculated based on spatial location, extension direction, structural priors, and cross-frame consistency. When the association score exceeds a preset threshold, an associated edge is established.

4. The method for joint detection of tower and line status and safe distance of a three-span transmission line based on multi-source information fusion as described in claim 1, characterized in that, The step of calculating the minimum distance based on the spatial shape and the safety clearance of the three-span object, and dynamically correcting the minimum distance based on the uncertainty information to obtain an effective safety distance, includes: Based on the type of object spanning three sections and its spatial boundaries, and combined with prior information, a corresponding three-dimensional safety clearance model is established: for railway objects, a railway safety clearance is established based on the rail surface, overhead contact line, vehicle operating space, and corresponding safety distances; for highway objects, a highway safety clearance is established based on the road surface, road boundaries, vehicle clearance, and corresponding safety distances; for crossed power transmission lines, a dynamic safety clearance is established based on the spatial curve of the crossed conductor, its normal swing range, voltage level, and electrical safety distance. Search along the spatial curve of the conductor to find the point closest to the safety clearance, and obtain the minimum geometric distance between the conductor and the three-span object; search for the minimum spatial distance between the outer contour of the tower and the safety clearance; record the conductor point or tower point at which the minimum distance is obtained and the corresponding safety clearance point as the most dangerous point pair; Based on the spatial curve of the conductor at continuous time, estimate the velocity and acceleration of the conductor at each position, predict the position of the conductor within a short time range, and take the minimum distance between the conductor curve and the safety clearance within the prediction time as the minimum distance for dynamic prediction. Based on the direction of the connection between the most dangerous points, the uncertainty of traverse reconstruction, the uncertainty of the three-span object boundary, the UAV pose error, the registration error, and the distance measurement error are projected onto this direction to calculate the uncertainty of the minimum distance; The effective safe distance of the conductor is obtained by jointly correcting the dynamic predicted minimum distance, the uncertainty of the minimum distance, and the conductor swing compensation amount. Compare the effective safety distance of the conductor with the safety distance of the tower outline, take the smaller one as the control safety distance, and calculate the safety distance margin in combination with the required safety distance.

5. The method for joint detection of tower and line status and safe distance of a three-span transmission line based on multi-source information fusion as described in claim 1, characterized in that, The method of jointly determining the intrusion risk level based on the effective safety distance and the changing trend of the tower line status includes: Based on the constructed tower-line topology diagram, the tower status, conductor status, and safety distance status are correlated; For the same tower and its connecting conductors, analyze the consistency of the tower tilt direction, suspension point displacement direction, conductor swing direction, most dangerous point direction, and three-span object direction; when the tower continues to tilt towards the three-span object direction and the conductor simultaneously shows displacement, sag increase, or abnormal swing, and the effective safety distance continues to decrease, it is determined to be an associated fault on the same topology chain; Tower risk values ​​are constructed based on tower tilt angle, offset, and degree of bending; conductor risk values ​​are constructed based on conductor breakage, sag, splicing, abnormal sag, and degree of sway; and safety distance risk values ​​are constructed based on effective safety distance, safety margin, and distance change trends. The joint risk value is calculated using the tower risk value, conductor risk value, safety distance risk value, and association enhancement value determined based on topological association and directional consistency. Based on the uncertainties of each information source and the uncertainties of tower, conductor, and safety distance detection, a comprehensive detection uncertainty is generated, and the comprehensive detection reliability is determined accordingly. The joint risk value within the sliding time window is smoothed, and the final risk level is determined by combining the risk change rate, duration, and safety margin: when it is below the first threshold, it is a normal state; when there is an abnormal change but it is not close to the safety limit, it is a state of concern; when there is a broken line, a drop, or the effective safety distance is close to the required value and continues to decrease, it is an alarm state; when the safety margin is less than zero or has violated the safety limit, it is a severe state and an intrusion alarm is immediately output.

6. The method for joint detection of tower and line status and safe distance of a three-span transmission line based on multi-source information fusion as described in claim 1, characterized in that, The multi-source detection information includes visible light image information, lidar point cloud information, distance measurement information, and UAV position and motion status information; the prior information includes prior information of transmission lines and three-span objects, continuous time series and historical reference information; the prior information of transmission lines and three-span objects includes tower type, tower design height, tower structure dimensions, crossarm level, insulator string position, conductor arrangement, conductor phase, span, design sag, suspension point design position, three-span object type, historical spatial range of three-span objects, and safety distance requirements; The continuous time series and historical reference information includes images, point clouds, distances, poses, target recognition and spatial reconstruction results before the current moment, as well as the tower body centerline, tower top position, conductor spatial curve, normal sag range, normal swing range and safe distance range of the same object under the historical normal state.

7. The method for joint detection of tower and line status and safe distance of a three-span transmission line based on multi-source information fusion as described in claim 1, characterized in that, The data quality and uncertainty information includes: the sharpness, contrast, exposure, rain, snow, and fog interference, occlusion ratio, and recognition confidence index of visible light images; the point cloud density, target area integrity, echo stability, outlier ratio, and local registration residual index of lidar point clouds; the continuous fluctuation and abnormal jump ratio index of distance measurements; and the positioning accuracy, attitude accuracy, pose continuity, and motion stability index of UAV positions and motion states. The comprehensive quality score of various types of information at the detection time is the weighted sum of its various quality indicators; the normalization uncertainty of various types of information is determined based on measurement error, spatial registration residual, target recognition confidence, and continuous temporal consistency. When image quality degrades, reduce the weight of image features and increase the weight of point cloud, distance, and temporal features; when point clouds are sparse or missing, reduce the weight of point cloud features and increase the weight of image contour, structural prior, and historical curves; when the UAV pose accuracy decreases, increase the uncertainty of the target's absolute position and use conservative corrections in the safe distance calculation.

8. The method for joint detection of tower and line status and safe distance of a three-span transmission line based on multi-source information fusion according to claim 1, characterized in that, The spatial morphology includes the spatial outline of the tower and the spatial curve of the conductor; the tower-line status includes the tower status and the conductor status. The status of the tower is identified in the following ways: Extract the effective 3D point cloud of the tower based on the tower body area and key structural points; Divide the tower into multiple height levels along the vertical direction of the scene, calculate the center point of the tower in each level, and form a sequence of center points; A robust linear fit was performed on the center point sequence to obtain the main central axis of the tower. The overall tilt angle is determined based on the angle between the main central axis and the vertical direction of the scene, the tilt direction is determined based on the horizontal component of the main central axis, and the horizontal offset of the tower top relative to the tower base is calculated based on the effective detection height. Calculate the lateral deviation of the center point of each level relative to the main central axis; when most levels are offset along the same direction and the tilt angle exceeds the threshold, it is determined to be an overall tilt; when only some continuous levels have increased lateral deviation while other levels are still close to the main central axis, it is determined to be a local bend. The trend of change is determined by the tilt angle, direction, offset, and deviation within the sliding time window; when the apparent tilt direction is consistent with the direction of UAV movement but the key points in the world coordinates are stable, it is determined to be an apparent change caused by the movement of UAV. For the conductor suspension points that are obstructed, their position uncertainty is restored and increased based on the key points of the crossarm structure, the direction of inclination, the designed suspension position, and the historical topology. The actual spatial outline of the tilted tower is constructed based on the tower body point cloud, key structural points, main central axis, and local bending positions.

9. The method for joint detection of tower and line status and safe distance of a three-span transmission line based on multi-source information fusion as described in claim 1, characterized in that, The spatial morphology includes the spatial outline of the tower and the spatial curve of the conductor; the tower-line status includes the tower status and the conductor status. The state of the conductor is identified in the following ways: The image regions, point clouds, endpoints, suspension points, previous time curves, and historical normal curves of traverses belonging to the same span, the same phase, and the same historical suspension relationship are aggregated into a multi-source observation set for a single traverse. The current connection status is determined based on the spatial relationship between the conductor endpoint and the suspension point; For double-ended suspended traverses, the two suspension points are used as endpoint constraints, and the vertical curve is initialized using a catenary shape. For single-ended suspended traverses, the still-connected end is used as the fixed end, and the drooping section is reconstructed based on point cloud, image contour, free end position, and temporal trajectory. For traverses that have detached at both ends or whose towers have collapsed, piecewise curves and local smoothing constraints are used for reconstruction. The reconstruction of traverse curves is constrained by the spatial observation points, the connection of suspension points, the continuity of curves, the phase arrangement relationship, and the temporal consistency. When a section of the conductor lacks observation points due to obstruction, the observation points are supplemented based on the positions on both sides, tangent direction, curvature, suspension relationship, and previous curves; when the same area cannot be restored at multiple consecutive moments and a reasonable connection cannot be formed on both sides, it is identified as a suspected broken line or broken strand location. The maximum sag is calculated based on the spatial chord between the two suspension points of the double-ended suspended conductor; the wind deflection and sag deviation are obtained based on the lateral and vertical deviations of the current curve from historical or design benchmarks; and overlapping and multi-line overlap anomalies are identified based on the minimum distance between adjacent conductors, their intersection positions, and durations. Calculate the dynamic swing amplitude based on the spatial curves of the same conductor at different times within a sliding time window; Anomalies are identified based on the conductor curve, endpoint connection status, suspension point connection status, spatial continuity, and temporal characteristics: a broken line is defined as a continuous interruption of the curve that cannot be reasonably connected; a drooping line is defined as one or both ends detaching from the original suspension point and moving significantly downward; and a single-end suspension anomaly is defined as only one suspension end remaining and moving significantly around that end. When different conductors continuously approach, cross, or overlap, and the minimum phase-to-phase distance is below the threshold, it is considered an overlap or multi-line overlap; when the maximum sag deviates from the historical or design value by more than the threshold, it is considered an abnormal sag; when the swing amplitude, speed, or spatial range exceeds the normal range and shows a continuous trend of approaching the safety limit, it is considered an abnormal swing.

10. A joint detection system for tower and line status and safe distance of a three-span transmission line based on multi-source information fusion, characterized in that, include: The information acquisition module is used to acquire multi-source detection information and prior information collected by the UAV; The spatiotemporal unification and adaptive fusion module is used to perform spatiotemporal unification on the multi-source detection information, generate data quality and uncertainty information, and perform adaptive fusion based on the data quality and uncertainty information to establish a unified spatiotemporal relationship between the tower line and the three-span object. The joint identification module is used to extract the tower line status and spatial morphology based on the unified spatiotemporal relationship through joint identification; The safety distance calculation and correction module is used to calculate the minimum distance based on the spatial shape and the safety clearance of the three-span object, and to dynamically correct the minimum distance based on the uncertainty information to obtain an effective safety distance; The joint determination and output module is used to jointly determine the intrusion risk level and output it based on the effective safe distance and the changing trend of the tower line status.