An overhead power transmission cable inspection robot rapid falling line control method

Abstract

This invention provides a rapid cable-dropping control method for an overhead power transmission cable inspection robot, belonging to the field of intelligent control technology. The method includes: acquiring electromagnetic field strength data monitored by an onboard electromagnetic sensor in real time based on a corrected cable-dropping path; using the electromagnetic field strength data to counteract the pose shift of the drone and work platform caused by electromagnetic interference in real time, thereby obtaining a control signal after offsetting the pose shift; and driving the drone and work platform to perform cable-dropping control actions based on the corrected cable-dropping path and the control signal after offsetting the pose shift, until the work platform achieves stable hovering and positioning at a designated location on the target conductor. This invention improves the automation and intelligence of power transmission cable inspection.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology, and in particular to a method for rapid cable-dropping control of an overhead power transmission cable inspection robot. Background Technology

[0002] In a power transmission cable inspection scenario during cloudy weather in mountainous areas, when a drone carrying an inspection robot approaches a target conductor at an altitude of 800 meters, sudden cloud cover causes sunlight to filter through, creating dappled light and shadow. This results in severe uneven lighting in the conductor texture image captured by the visual sensor. Consequently, the conductor's geometric contour extracted by the edge detection algorithm becomes significantly distorted due to localized overexposure. A circular conductor with a diameter of 20mm appears as an irregular shape in the image, with one side blurred and the other side stretched, with a maximum contour deviation of 5mm.

[0003] Meanwhile, during the scanning process, the lidar encountered partial obstruction from the side branches of the mountain, resulting in approximately 15% missing area in the 3D point cloud data of the target conductor. Under the existing technical framework, the system directly performs a weighted fusion of the deformed visual contour and the incomplete point cloud data: simply stitching together the coordinates of the center point of the visual contour with the average coordinates of the point cloud data. Due to the lack of an effective feature association mechanism, the system cannot identify that the visual contour deformation is caused by illumination interference, nor can it detect the spatial location of the missing point cloud area. The final fusion result shifts the actual spatial pose of the conductor towards the side branches of the mountain by 30mm.

[0004] This fusion deviation directly leads to serious errors in the initial landing path planning: when the drone approaches according to the planned path, the relative position error between the working platform and the ground wire increases from the theoretical value of ±5mm to ±35mm. When the drone attempts to perform the landing action, the anti-collision mechanism is triggered due to inaccurate positioning, forcing it to stop landing and replan the path. This not only prolongs the time of a single inspection operation, but also increases the risk of equipment colliding with obstacles in complex terrain. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a method for rapid cable-dropping control of an overhead power transmission cable inspection robot, which improves the automation and intelligence of power transmission cable inspection.

[0006] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows:

[0007] A method for rapid cable lowering control of an overhead power transmission cable inspection robot, applied to the inspection robot, the method comprising:

[0008] Step 100: Collect three-dimensional point cloud data of the target power transmission cable area using an airborne lidar sensor; based on the three-dimensional point cloud data, identify the three-dimensional spatial coordinates of the conductor and the distribution of nearby obstacles; based on the identified conductor position and obstacle distribution information, calculate and generate an initial landing path to guide the UAV carrying the operation platform to approach the target conductor.

[0009] Step 200: Based on the initial wire drop path, acquire the surface texture feature image of the conductor and ground wire; based on the texture feature image, extract the geometric contour features of the conductor; perform multimodal feature matching and data fusion with the spatial coordinates of the conductor and ground wire obtained from the LiDAR point cloud to obtain the fusion result; based on the fusion result, dynamically calibrate the pose deviation in the initial wire drop path to generate a corrected wire drop path.

[0010] Step 300: Based on the corrected landing path, acquire the electromagnetic field strength data monitored by the airborne electromagnetic sensor in real time; based on the electromagnetic field strength data, cancel the attitude deviation of the UAV and the work platform caused by electromagnetic interference in real time to obtain the control signal after canceling the attitude deviation; based on the corrected landing path and the control signal after canceling the attitude deviation, drive the UAV and the work platform to perform the landing control action until the work platform achieves stable hovering and positioning at the designated position of the target ground wire.

[0011] The above-described solution of the present invention has at least the following beneficial effects:

[0012] By employing multimodal feature matching and data fusion techniques, the 3D spatial coordinates acquired by LiDAR and the geometric contour features extracted by visual sensors are effectively integrated, compensating for the shortcomings of single sensors in complex environments. For example, when the visual contour is deformed due to illumination interference, it can be calibrated using the precise spatial coordinates of LiDAR; and data loss caused by LiDAR occlusion can be supplemented by the continuity of the visual contour, making the fusion result more accurately reflect the true pose of the conductor and significantly reducing the planning error of the initial drop path.

[0013] The adaptive compensation algorithm employed can process electromagnetic field strength data monitored by airborne electromagnetic sensors in real time, specifically offsetting the attitude deviation caused by the strong electromagnetic environment of power transmission cables on the UAV and operating platform. This mechanism solves the problem of positioning drift and inaccurate action that traditional control methods are prone to under electromagnetic interference, ensuring stable control performance in complex scenarios such as mountainous areas, fog, and electromagnetic radiation, and improving the reliability of the line-dropping action.

[0014] A closed-loop control system is formed through a three-stage control logic of "initial path planning - dynamic path correction - real-time interference compensation". The initial landing path is generated based on accurate obstacle distribution information, avoiding invalid path exploration; the dynamic calibration stage quickly corrects pose deviations through multi-modal fusion, reducing path adjustment time; and real-time electromagnetic interference compensation ensures the accuracy of actions during the execution phase. The synergistic effect of these three aspects significantly shortens the operation time from approaching the target to stable hovering, improving the operation efficiency of the inspection robot. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating a rapid cable-dropping control method for an overhead power transmission cable inspection robot provided in an embodiment of the present invention.

[0016] Figure 2 This is an embodiment of the present invention. Figure 1 A flowchart of step 100. Detailed Implementation

[0017] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0018] like Figure 1 As shown, an embodiment of the present invention proposes a rapid cable-dropping control method for an overhead power transmission cable inspection robot, which is applied to the inspection robot. The method includes the following steps:

[0019] Step 100: Collect three-dimensional point cloud data of the target power transmission cable area using an airborne lidar sensor; based on the three-dimensional point cloud data, identify the three-dimensional spatial coordinates of the conductor and the distribution of nearby obstacles; based on the identified conductor position and obstacle distribution information, calculate and generate an initial landing path to guide the UAV carrying the operation platform to approach the target conductor.

[0020] Step 200: Based on the initial wire drop path, acquire the surface texture feature image of the conductor and ground wire; based on the texture feature image, extract the geometric contour features of the conductor; perform multimodal feature matching and data fusion with the spatial coordinates of the conductor and ground wire obtained from the LiDAR point cloud to obtain the fusion result; based on the fusion result, dynamically calibrate the pose deviation in the initial wire drop path to generate a corrected wire drop path.

[0021] Step 300: Based on the corrected landing path, acquire the electromagnetic field strength data monitored by the airborne electromagnetic sensor in real time; based on the electromagnetic field strength data, cancel the attitude deviation of the UAV and the work platform caused by electromagnetic interference in real time to obtain the control signal after canceling the attitude deviation; based on the corrected landing path and the control signal after canceling the attitude deviation, drive the UAV and the work platform to perform the landing control action until the work platform achieves stable hovering and positioning at the designated position of the target ground wire.

[0022] In this embodiment of the invention, the overhead power transmission cable inspection robot (hereinafter referred to as the "inspection robot") is the main working body that directly performs the cable inspection task. Its core function is to move along the power transmission cable and complete detailed inspections of cable surface defects, joint temperature, insulator condition and other details through the detection equipment such as the high-definition camera and infrared thermal imager. However, the inspection robot itself does not have the ability to move long distances and cannot autonomously reach the high-altitude cable from the ground or the initial position. The drone, on the other hand, serves as an "aerial carrier" and undertakes the task of safely and accurately delivering the inspection robot to the target cable.

[0023] The "airborne lidar sensor" and "UAV carrying operation platform" mentioned in step 100 are essentially the work of the UAV carrying the inspection robot (the operation platform is the inspection robot and its auxiliary equipment) to complete the initial approach phase. The UAV provides the inspection robot with the ability to move from a distance to the vicinity of the cable, which is a prerequisite for the inspection robot to "land on the cable" (i.e., transfer from the air to the cable).

[0024] The entire inspection mission is divided into three core phases: "approaching the cable," "dropping the cable," and "inspecting along the cable." Drones and inspection robots play leading roles in different phases, forming a seamless connection.

[0025] Step 100, the "approaching the cable stage," is led by the drone: Utilizing the drone's flight capabilities, the inspection robot departs from its takeoff point (such as a ground base station or inspection vehicle), collects environmental data using lidar, identifies the cable's location and obstacles, plans a safe path, and gradually transports the inspection robot to the target cable's landing area (typically a hovering range of 1-3 meters from the cable). At this point, the inspection robot is in a "waiting to land" state; its own drive system (such as a wheeled walking mechanism) is not yet activated, relying entirely on the drone's positioning and movement control.

[0026] In the subsequent "line-landing phase" (steps 200 to 300), the two work together to transition: the drone guides the inspection robot to the top or side of the cable according to the corrected path, and the inspection robot activates its docking mechanism (such as the cable clamping arm and guide wheel) to complete the mechanical connection with the cable with the assistance of the drone's hovering. After the line is landed, the drone can leave the work area (return to the base station or go to the next work point) or serve as an auxiliary platform to provide support such as communication relay and global monitoring.

[0027] During the "line inspection phase," the inspection robot takes the lead: it moves along the cable entirely on its own walking mechanism to perform detailed inspection tasks, while the drone no longer participates in direct control.

[0028] In this embodiment, the lidar carried by the drone in step 100, the texture feature acquisition device (such as a high-definition camera) mentioned in subsequent steps, and the electromagnetic sensors do not only serve the drone itself, but also provide environmental perception data for the entire "drone-inspection robot" system.

[0029] The 3D point cloud data acquired by the UAV through LiDAR is not only used to plan its own flight path, but more importantly, it provides a precise spatial coordinate reference for the landing position of the inspection robot, ensuring that the inspection robot finally docks with the designated location of the target cable (such as near the defect point or the joint). The flight control system of the UAV is interconnected with the landing control system of the inspection robot. The initial landing path generated in step 100 will be synchronized to the inspection robot, allowing it to adjust its own attitude in advance (such as the opening angle of the clamping arm and the orientation of the guide mechanism) to prepare for the subsequent landing action.

[0030] Specifically, the drone serves as the "mobile carrier and initial navigation platform" for the inspection robot to perform high-altitude operations, while the inspection robot is the "core operating entity" for completing the detailed inspection of cables. Through the process of "transportation-docking-separation (or collaboration)," the two together constitute a complete overhead power transmission cable inspection system. All operations of the drone in step 100 are essentially to ensure that the inspection robot can safely and accurately land on the line and start the inspection task, which is the basic link for the collaborative operation of the entire system.

[0031] In an optional embodiment of the present invention, step 100 involves acquiring three-dimensional point cloud data of the target power transmission cable area using an airborne lidar sensor; and identifying the three-dimensional spatial coordinates of the conductor and the distribution of nearby obstacles based on the three-dimensional point cloud data, including:

[0032] Step 101: Perform voxel mesh filtering downsampling on the original 3D point cloud data acquired by LiDAR to eliminate environmental scattering noise points and output the denoised point cloud dataset.

[0033] Step 102: Based on the denoised point cloud dataset, separate the conductor ground line point cloud clusters; use the RANSAC straight line fitting algorithm to fit the spatial center axis of the conductor ground line point cloud clusters to generate the spatial center axis equation of the conductor ground line and its three-dimensional coordinate parameters.

[0034] Step 103: Based on the separated non-conductor ground wire point cloud clusters, identify obstacle point cloud clusters using the DBSCAN density clustering algorithm; combine the three-dimensional coordinate parameters of the conductor ground wire spatial central axis equation to calculate the minimum Euclidean distance between the outer edge of each obstacle point cloud cluster and the conductor ground wire central axis, and output the spatial position and distance data of the obstacles.

[0035] In this embodiment of the invention, when applied in a specific application, step 101 is implemented as follows:

[0036] First, to achieve effective preprocessing of the raw 3D point cloud data, the spatial resolution of the voxel mesh needs to be determined. Specifically, the acquisition parameters of the airborne LiDAR sensor are first obtained, including but not limited to the sensor's scanning frequency and the generated point cloud density. Based on this, the spatial resolution of the voxel mesh is comprehensively determined by combining the diameter parameters of the target power transmission cable and the inspection distance between the UAV-carried operating platform and the target cable. To avoid the loss of effective point cloud information of the ground wire in subsequent processing, the spatial resolution must be strictly limited, ensuring that the side length of a single voxel mesh is no greater than 1 / 5 of the ground wire diameter. This setting allows for the maximum preservation of the ground wire's characteristic information while removing noise.

[0037] Next, after determining the spatial resolution of the voxel grid, the original 3D point cloud data acquired by the LiDAR is spatially partitioned based on this resolution. Specifically, the 3D space containing the original point cloud data is divided into several cubic regions, each of which is a voxel grid, and all voxel grids use the aforementioned set spatial resolution, thereby achieving the initial spatial discretization processing of the original point cloud data.

[0038] Subsequently, after completing the voxel mesh division, each voxel mesh is traversed one by one, and the number of point clouds contained in each mesh is counted. Based on this, differentiated processing is adopted for voxel meshes with different numbers of point clouds: for voxel meshes containing a certain number of point clouds, the distance from all point clouds in the mesh to the center point of the mesh is calculated, and the point with the smallest distance is selected as the valid point of the mesh and retained; conversely, for voxel meshes with zero point clouds, or voxel meshes with a number of point clouds but only containing isolated points (i.e., the number of point clouds is less than a preset threshold, which is preset to 3-5 points based on the point cloud density generated by the lidar), they are marked as noise regions, and all point clouds in the region are removed. This operation can effectively remove scattering noise points in the environment.

[0039] Finally, after processing all voxel meshes, all valid points retained in the above steps are summarized to form a denoised point cloud dataset. This denoised point cloud dataset, having removed a large amount of noise interference, can provide more reliable basic data support for subsequent point cloud segmentation processing.

[0040] In practical applications, the specific implementation steps of step 102 above are as follows:

[0041] Based on the noise-reduced point cloud dataset obtained in step 101, a preliminary screening operation needs to be performed on the noise-reduced point cloud dataset to accurately separate the conductor and ground wire point clouds. Specifically, this preliminary screening is based on two characteristics: first, the range of the cable erection height. This height range needs to be preset according to the actual erection parameters of the cables in the target inspection area (such as tower height, cable suspension height, etc.). By eliminating point clouds that are lower or higher than this height range, non-conductor and ground wire point clouds such as those on the ground and low-altitude vegetation can be effectively excluded; second, the spatial distribution density of the point clouds. The density is assessed by calculating the number of point clouds contained in a unit space (such as a unit cubic meter). Point cloud clusters with a density higher than a preset threshold are retained, thereby further screening out point clouds that meet the dense distribution characteristics of conductors and ground wires.

[0042] After completing the initial screening, in order to aggregate the discrete point clouds into related point cloud clusters, connected component analysis needs to be performed on the initially screened point clouds. The specific operations are as follows: traverse all the screened point clouds and calculate the Euclidean distance between any two points; set a distance threshold, which is determined based on the actual diameter of the ground wire and the point cloud resolution, usually 2-3 times the diameter of the ground wire; group points whose distance to each other is less than the threshold into the same connected component. Through this operation, several candidate point cloud clusters can be obtained, each candidate point cloud cluster representing a possible continuous object in space.

[0043] Based on the separated candidate point cloud clusters, feature verification is required for each candidate point cloud cluster to accurately identify the point cloud cluster corresponding to the conductor / ground wire. Specifically, the spatial distribution morphology features of each point cloud cluster are calculated, including but not limited to the aspect ratio (i.e., the ratio of the length of the point cloud cluster in the longest extension direction to the maximum width perpendicular to that direction) and orientation consistency (i.e., whether the distribution direction of each point in the point cloud tends to be consistent). These features are compared with the typical linear distribution features of the conductor / ground wire (such as aspect ratio much greater than 1 and consistent orientation height), and point cloud clusters that conform to the linear distribution features are retained and identified as conductor / ground wire point cloud clusters.

[0044] After determining the point cloud cluster of the conductor and ground wire, the RANSAC straight line fitting algorithm is used to process the point cloud cluster in order to accurately obtain the spatial location parameters of the conductor and ground wire. The specific process is as follows: First, three non-collinear points are randomly selected from the ground wire point cloud cluster as initial sample points, and an initial straight line model is constructed based on these three points. Second, the distance from all other points in the ground wire point cloud cluster to the initial straight line model is calculated, and points whose distance is less than a preset error threshold (this threshold is set according to the point cloud accuracy, usually 0.5-1mm) are marked as interior points (i.e., points that conform to the straight line model). Third, the sample points are iteratively updated, sample points are selected again from the point cloud cluster, and a new straight line model is constructed. The process of calculating the number of interior points is repeated until the proportion of interior points to the total number of points in the point cloud cluster exceeds 90%, or the number of iterations reaches a preset value (this preset number is set according to the point cloud size, usually 50-100 times). Finally, based on the finally determined set of interior points, the equation of the central axis straight line that can accurately represent the spatial position of the ground wire is calculated. This equation includes slope parameters and intercept parameters. At the same time, the coordinate parameters of the central axis in the three-dimensional coordinate system are extracted, including the starting coordinates, ending coordinates, and direction vector of the axis, so as to completely describe the spatial extension state of the ground wire.

[0045] The process of calculating the equation of the conductor-ground wire spatial center axis based on the final set of interior points after determining the final set of interior points using the RANSAC straight line fitting algorithm requires determining the spatial parameters of the axis through the following steps to fully describe its structure:

[0046] First, determine the coordinates of the starting and ending points of the axis. Traverse the inner point set of the guide point cloud cluster and calculate the extreme coordinates of the inner points along the extension direction in three-dimensional space. In the main extension direction of the axis (judged by the distribution trend of the inner point set, usually the longest extension direction of the point cloud cluster), select the point with the smallest coordinate value as the starting point of the axis and the point with the largest coordinate value as the ending point of the axis, thereby defining the spatial range of the axis.

[0047] Secondly, the direction vector of the axis is calculated. This direction vector is determined by the coordinate difference between the starting and ending points: subtracting the starting point coordinates from the ending point coordinates yields a three-dimensional vector. The direction of this vector represents the extension direction of the conductor's central axis in three-dimensional space, and its magnitude corresponds to the straight-line distance from the starting point to the ending point. By normalizing this vector (i.e., adjusting the vector length to 1), a unit direction vector is obtained, used to accurately describe the axis's orientation.

[0048] Finally, the core structure of the central axis equation is constructed. This equation uses the starting point coordinates as a reference point and combines them with the direction vector to define the entire straight line: any point on the axis in space can be obtained by extending the starting point coordinates along the direction vector in a certain proportion. Specifically, starting from the starting point, moving any distance along the direction vector will result in a position on the central axis, thus forming a continuous straight line model. Through the combination of the starting point coordinates, ending point coordinates, and direction vector, the central axis equation of the conductor-ground line is completely constructed. This equation can accurately reflect the position, direction, and extension range of the conductor-ground line in three-dimensional space.

[0049] For example, if the starting coordinates of a conductor are P0 = (10, 5, 20) (unit: meters), and the unit direction vector is... Then the equations of the axis are: X = 10 + 0.8t, y = 5 + 0 × t, z = 20 + 0.6t; its physical meaning is that the straight line extending from the point (10, 5, 20) along the direction (0.8, 0, 0.6) completely represents the position, direction, and range of the conductor in three-dimensional space; t is a proportional parameter, representing the direction vector from the starting point P0 along the direction vector. The proportion of the extended distance (t>0 extends in the positive direction of the axis, t<0 extends in the negative direction).

[0050] In practical applications, the specific implementation steps of step 103 above are as follows:

[0051] First, based on the accurate identification of the ground wire point cloud clusters in step 102, a filtering operation needs to be performed on the denoised point cloud dataset obtained in step 101 to further distinguish potential obstacle point clouds. Specifically, the ground wire point cloud clusters identified in step 102 are removed from the denoised point cloud dataset. After this removal operation, the remaining point cloud data is defined as non-ground wire point cloud clusters, which may contain the obstacle point clouds that need to be identified.

[0052] Next, the DBSCAN density clustering algorithm is used to cluster the non-conductive point cloud clusters obtained above to separate potential obstacle point cloud clusters. When performing this clustering process, two key parameters need to be set: the first is the neighborhood radius, which is determined based on the minimum size of obstacles that may exist in the inspection scenario, typically set within the range of 0.3-0.5 meters; the second is the minimum number of points threshold, which is set based on the point cloud density collected by the lidar, generally set to 8-12. After setting the parameters, each point in the non-conductive point cloud is checked one by one: for any point traversed, if the number of points in the neighborhood of the point (i.e., the space formed by the aforementioned neighborhood radius) is not less than the set minimum number of points threshold, then the point is determined as a core point, and the cluster is expanded around the core point, incorporating all points in its neighborhood that meet the conditions into the same cluster, thus forming a cluster; conversely, if the number of points in the neighborhood of the traversed point is less than the minimum number of points threshold, and the point is not contained in the neighborhood of any core point, then the point is marked as a noise point and removed. Through the above operations, several candidate point cloud clusters of obstacles can be obtained in the end.

[0053] After obtaining the candidate point cloud clusters of obstacles, boundary extraction is required for each candidate point cloud cluster to accurately calculate the distance between the obstacle and the ground wire. Specifically, the extraction method involves calculating the extreme coordinates of each candidate point cloud cluster in the x, y, and z axes of three-dimensional space, i.e., the maximum and minimum values ​​in each axis direction. These extreme coordinates determine the boundary range of the point cloud cluster in three-dimensional space, thus obtaining the outer edge point set of the point cloud cluster. These outer edge point sets are key reference points for subsequent distance calculations.

[0054] Finally, combining the equation of the guide wire spatial center axis obtained in step 102, the Euclidean distance from each point in the outer edge point set of each obstacle candidate point cloud cluster to the guide wire spatial center axis is calculated. After obtaining the distances from all outer edge points to the center axis, the minimum distance value is selected, and this minimum value is determined as the minimum distance between the obstacle point cloud cluster and the guide wire. Simultaneously, by calculating the average coordinates of all points within the obstacle point cloud cluster, the center coordinates of the obstacle point cloud cluster are obtained. These center coordinates are used to characterize the spatial position of the obstacle. Furthermore, the obtained obstacle spatial position and minimum distance data are output, providing crucial obstacle information support for subsequent initial drop path planning.

[0055] This invention, through voxel grid filtering downsampling in step 101, effectively eliminates environmental scattering noise in the original point cloud, reducing data volume while preserving key features of the ground wire, providing high-quality basic data for subsequent point cloud segmentation, and avoiding feature recognition deviations caused by noise interference. Secondly, step 102, by separating the ground wire point cloud clusters and combining them with the RANSAC straight-line fitting algorithm, accurately extracts the spatial central axis equation and three-dimensional coordinate parameters of the ground wire, achieving precise characterization of the ground wire's spatial pose, providing a reliable target benchmark for subsequent drop path planning, and ensuring the directional accuracy of the initial path planning. Finally, step 103, based on DBSCAN density clustering to identify obstacle point cloud clusters and combined with the ground wire central axis to calculate the minimum distance, comprehensively and accurately obtains the spatial position and distance information of nearby obstacles, providing detailed obstacle avoidance basis for initial drop path planning, effectively avoiding collision risks, and ensuring the safety and efficiency of the approach process of the work platform. The synergistic effect of these three steps improves the accuracy and stability of subsequent path planning and control.

[0056] In an optional embodiment of the present invention, based on the identified location of the ground wire and the distribution information of obstacles, an initial landing path for guiding the UAV carrying the operating platform to approach the target ground wire is calculated and generated, including:

[0057] Step 104: Based on the spatial location and distance data of obstacles, using the center axis of the grounding wire as the baseline and combining the minimum Euclidean distance of each obstacle, generate a cylindrical collision-free safety corridor with the baseline as the center axis and a radial safety radius ≥ 1.5 meters.

[0058] Step 105: Within the cylindrical collision-free safety corridor space, set up a three-dimensional path node sequence at equal intervals of 2-5 meters along the direction of movement of the UAV approaching the ground wire;

[0059] Step 106: Using the path node sequence as control points, calculate a continuous and smooth spatial trajectory curve using a cubic spline interpolation algorithm to generate an initial landing path containing the UAV's position coordinates, heading angle, and pitch angle control parameters.

[0060] In this embodiment of the invention, when applied in a specific application, step 104 is implemented as follows:

[0061] First, based on the determination of the spatial center axis of the ground wire in step 102 and the acquisition of the spatial positions of each obstacle and their minimum Euclidean distance from the center axis of the ground wire in step 103, in order to construct a channel for the UAV to safely approach the target ground wire with its operating platform, it is necessary to define the safety range based on the above information.

[0062] Specifically, the spatial centerline of the conductor obtained in step 102 is directly used as the baseline, and the spatial range of the cylindrical safety zone is delineated with this baseline as the center. When determining the radial safety radius of this cylindrical collision-free safety corridor, the minimum Euclidean distance between each obstacle and the centerline of the conductor calculated in step 103 must be considered to ensure that the set radial safety radius is not less than 1.5 meters.

[0063] During this process, since the central axis of the cylindrical safety corridor is completely coincident with the central axis of the grounding wire space, and its radial safety radius meets the requirement of not less than 1.5 meters, this setting can ensure that there is sufficient buffer distance between the boundary of the safety corridor and all obstacles, thereby creating a movement range in physical space for the drone and the operation platform that is free from collision risk, and providing a safe and reliable spatial constraint for subsequent path planning.

[0064] In practical applications, the specific implementation steps of step 105 above are as follows:

[0065] Based on the cylindrical collision-free safety corridor generated in step 104, specific path nodes need to be deployed within this safety corridor to transform the abstract safety space into an executable path planning basis. Specifically, firstly, a preset movement direction is determined for the UAV carrying the work platform to approach the target grounding line. This direction is typically set from an initial position away from the grounding line towards the target landing point on the grounding line, and the overall direction maintains a preset relative attitude with the central axis of the grounding line. Next, along the preset movement direction, the spacing of the path nodes is determined at equal intervals of 2-5 meters. This spacing can be adaptively adjusted according to the complexity of the working environment. The more complex the environment, the smaller the spacing can be to improve path accuracy; the simpler the environment, the larger the spacing can be to reduce the amount of computation.

[0066] After determining the intervals, starting from the initial position of the UAV, the spatial positions of each node are sequentially marked along the preset movement direction. This ensures that the three-dimensional coordinates of each marked point fall within the cylindrical collision-free safety corridor generated in step 104, meaning the distance from the point to the center axis of the grounding guide does not exceed the radial safety radius of the safety corridor, thus guaranteeing the safety of each node. Through the above operations, a series of orderly arranged three-dimensional path nodes are formed, which together constitute the path node sequence. By setting equally spaced path nodes in this way, definite control points can be provided for generating continuous and smooth trajectory curves in subsequent steps. This ensures that the final planned path can strictly follow the constraints of the safety corridor while steadily guiding the UAV and the work platform from the initial position to gradually approach the target grounding guide.

[0067] In practical applications, the specific implementation steps of step 106 above are as follows:

[0068] Based on the path node sequence generated in step 105, in order to transform the discrete nodes into a continuous motion trajectory that the UAV and the work platform can actually follow, the path node sequence needs to be used as control points for trajectory fitting, and a cubic spline interpolation algorithm is employed for processing. Specifically, the path node sequence obtained in step 105 is first sorted according to its chronological order in the direction of motion to determine the connection relationship between adjacent nodes. Subsequently, the cubic spline interpolation algorithm is used to interpolate the trajectory between two adjacent path nodes. Through this algorithm processing, the generated trajectory curve can accurately pass through each path node in sequence, while ensuring that the first derivative (i.e., the rate of change of velocity) and the second derivative (i.e., the rate of change of acceleration) of the curve remain continuous at each node. This continuity processing can effectively avoid abrupt changes or angles in the trajectory curve, thereby ensuring the smoothness of the entire trajectory and preventing the UAV and the work platform from experiencing drastic attitude changes or vibrations due to abrupt changes in the trajectory when moving along the trajectory, thus ensuring the stability of equipment operation.

[0069] After obtaining the aforementioned continuous and smooth spatial trajectory curve, key parameters for controlling the UAV's movement are further extracted based on this curve. Specifically, these include: the three-dimensional spatial coordinates corresponding to each position along the trajectory curve, which determine the UAV's target position at different times; and, based on the trajectory curve's direction, calculating the required heading angle (horizontal pointing angle) and pitch angle (vertical tilt angle) for the UAV to move along the curve at each position. By integrating these position coordinates with control parameters such as heading and pitch angles, a complete set of motion commands is formed. This set constitutes the initial landing path used to guide the UAV carrying the work platform from its initial position to gradually approach the target ground wire. This process ensures that the initial landing path conforms to the safety corridor constraints while possessing smooth and controllable motion characteristics.

[0070] Step 104 of this invention constructs a cylindrical collision-free safety corridor, using the center axis of the ground guide as a reference and setting a radial safety radius of not less than 1.5 meters. Combined with minimum obstacle distance data, this defines a definite safe movement boundary for the UAV and operating platform, completely avoiding the risk of collision with obstacles in terms of spatial scope and ensuring basic safety during the approach process. Secondly, step 105 sets a sequence of path nodes at equal intervals along the direction of movement within the safety corridor. This ensures the safety of node positions (all falling within the safety corridor) and provides evenly distributed control points for subsequent trajectory fitting through a reasonable spacing of 2-5 meters. This allows path planning to balance accuracy and efficiency, avoiding trajectory deviations caused by sparse nodes or increased computational load due to overly dense nodes. Finally, step 106 uses a cubic spline interpolation algorithm to generate a continuous and smooth trajectory curve, ensuring that the curve passes through all nodes and that the derivative is continuous at the nodes, effectively avoiding sudden attitude changes during UAV movement. Simultaneously, control parameters such as position coordinates, heading angle, and pitch angle are extracted to form a complete initial landing path, enabling the UAV to stably approach the ground guide along a smooth trajectory.

[0071] In an optional embodiment of the present invention, step 200 involves acquiring a surface texture feature image of the conductor based on the initial grounding path; and extracting the geometric contour features of the conductor based on the texture feature image, including:

[0072] Step 201: Based on the heading angle and pitch angle control parameters of the initial drop line path, adjust the shooting attitude of the airborne high-definition industrial camera, and collect the surface texture feature image of the conductor wire in real time during the flight of the UAV along the initial drop line path.

[0073] Step 202: Perform grayscale conversion and Gaussian filtering preprocessing on the acquired texture feature image to eliminate uneven lighting and image noise interference, and output the enhanced binarized image.

[0074] Step 203: Process the binarized image using the Canny edge detection algorithm to extract the set of continuous geometric contour edge pixels on the surface of the ground wire;

[0075] Step 204: Based on the set of pixel points on the edge of the geometric contour, fit the curvature equation of the center line of the conductor using the least squares method, calculate the boundary direction vector, and output the geometric contour feature parameters containing the radius of curvature and the normal vector.

[0076] In this embodiment of the invention, when applied in a specific application, step 201 is implemented as follows:

[0077] First, to ensure the airborne high-definition industrial camera can accurately capture the texture features of the conductor surface, the camera's shooting attitude needs to be adjusted based on the heading and pitch angle control parameters included in the initial drop path generated in step 106. Specifically, the camera's horizontal shooting direction is adjusted according to the heading angle parameters of the UAV at different positions in the initial drop path, ensuring it is consistent with the UAV's direction of movement; simultaneously, the camera's vertical shooting angle is adjusted according to the pitch angle parameters, ensuring the lens is always pointed at the conductor surface area. Based on this, as the UAV flies along the initial drop path, the camera starts shooting in real time, continuously acquiring texture feature images of the conductor surface, thereby obtaining visual data containing details of the conductor surface and providing raw image material for subsequent geometric contour extraction.

[0078] In practical applications, the specific implementation steps of step 202 above are as follows:

[0079] After acquiring the texture feature image in step 201, preprocessing is required to eliminate interference and improve subsequent processing accuracy. First, the color texture feature image is converted to grayscale, reducing the amount of image data and simplifying subsequent calculations. Next, a Gaussian filter is used to smooth the grayscale image. By weighted averaging of the grayscale values ​​of each pixel and its neighbors, grayscale fluctuations caused by uneven lighting and random image noise are mitigated, resulting in clearer image edge information. After these processes, binarization is performed. By setting an appropriate grayscale threshold, areas with grayscale values ​​above the threshold are marked as foreground (conductor / ground line regions), and areas below the threshold are marked as background. The final output is an enhanced binarized image, providing a high-quality image foundation for edge detection.

[0080] In practical applications, the specific implementation steps of step 203 above are as follows:

[0081] After obtaining the enhanced binarized image in step 202, the Canny edge detection algorithm is used to process the image to extract the geometric contour of the ground wire. Specifically, gradient calculation is first performed on the binarized image to identify regions with drastic changes in grayscale values, which typically correspond to the edges of the ground wire. Then, gradient values ​​are filtered by setting two thresholds, high and low. Edge points with gradient values ​​higher than the high threshold are retained as strong edges, and points with gradient values ​​between the high and low thresholds and connected to strong edges are included in the edge range, thus forming a continuous edge contour. Through this processing, a set of continuous geometric contour edge pixels on the ground wire surface is finally extracted from the image. This set of pixels accurately reflects the spatial distribution of the ground wire in the image.

[0082] In practical applications, the specific implementation steps of step 204 above are as follows:

[0083] After obtaining the set of geometric contour edge pixels in step 203, in order to convert these discrete pixel information into quantifiable conductor geometric feature parameters, the following calculation process is required:

[0084] First, from the set of geometric contour edge pixels obtained in step 203, pixels representing the centerline of the ground guide are selected. Specifically, by traversing the set of edge pixels, the coordinates of corresponding pixels on both sides of the ground guide contour (i.e., the upper and lower edge points at the same horizontal position) are calculated, and the average of the coordinates of the two sides is taken as the pixel coordinate of the centerline of the ground guide at that position. All such coordinates are summarized to form the centerline pixel set, thus reflecting the core direction of the ground guide in the image. Next, the least squares method is used to perform curve fitting on the above centerline pixel set. Based on the morphological characteristics of the ground guide in the image, a quadratic curve is preset as the fitting model, and its formula structure is y = ax 2 +bx+c, where a, b, and c are coefficients to be determined, x is the abscissa of the image, and y is the ordinate of the image; then calculate the centerline pixel (x i y i The longitudinal deviation (i.e., y) from the preset curve model i with ax i 2 +bx i The total error is obtained by summing the squares of all deviations (a, b, and c). Then, the values ​​of a, b, and c are iteratively adjusted to continuously reduce the total error until a set of parameter values ​​is obtained, which makes the total error reach its minimum value (i.e., any small change in any parameter will lead to an increase in the total error). The curve model at this time is the fitted conductor centerline curve. Based on this curve, the curvature equation of the conductor centerline is derived. This equation can be quantified to reflect the degree of curvature of the conductor centerline at each position in the image plane.

[0085] The specific calculation of the curvature equation involves the following steps:

[0086] The first step is to calculate the numerator. Take twice the coefficient 'a' of the quadratic curve, and then take the absolute value of the result. That is, first calculate the product of 2 and 'a' to get 2a, and then take the absolute value of 2a as the numerator of the entire equation. The second step is to calculate the denominator. The denominator is a composite operation result, which needs to be calculated in stages. First, calculate the value of 2ax + b, that is, first multiply the coefficient 'a' by the x-coordinate, then multiply by 2, and then add the coefficient 'b' to get an intermediate result; next, square this intermediate result, that is, calculate the square of (2ax + b); then, add 1 to the result of the square operation to get the sum of 1 and the square of (2ax + b); finally, raise this sum to the power of 3 / 2, that is, first take the square root of the sum, and then cube the result of the square root to get the denominator of the entire equation. The third step is to calculate the curvature value k, by dividing the numerator obtained in the first step (i.e., |2a|) by the denominator obtained in the second step (i.e., [1 + (2ax + b)]). 2 The curvature value k of the conductor-ground line centerline at the x-axis is obtained by using the above steps, based on the coefficients of the fitted quadratic curve and the x-axis of the image. The larger the value, the more significant the curvature of the conductor-ground line at the corresponding position.

[0087] Based on this, the radius of curvature corresponding to the curvature equation is calculated. For the curvature value k at each point in the curvature equation (the larger the curvature value, the more obvious the curvature), its reciprocal (i.e., 1 / k) is the radius of curvature at that point. By calculation, the radius of curvature at each position of the conductor centerline can be obtained, thus intuitively representing the degree of curvature of the conductor. At the same time, the direction vector of the conductor contour boundary is calculated according to the distribution direction of the edge pixel point set obtained in step 203. Specifically, by traversing the set of edge pixels, the coordinates of two adjacent pixels are subtracted (i.e., the coordinates of the latter point minus the coordinates of the former point) to obtain a vector (dx, dy) representing the direction of the line connecting the two points, which is the direction vector of the contour boundary. Then, the normal vector is calculated based on the direction vector, and a vector perpendicular to the direction vector (such as (-dy, dx) or (dy, -dx)) is taken to represent the vertical direction of the contour boundary at that position. Finally, the curvature radii and corresponding normal vectors obtained from the above calculations are summarized and integrated to form a complete set of geometric contour feature parameters and output. These parameters can comprehensively quantify the geometric shape of the ground wire in the visual image, providing accurate visual feature basis for multimodal feature matching and data fusion of the geometric contour features and the ground wire spatial coordinates obtained from the LiDAR point cloud in subsequent steps.

[0088] In this embodiment of the invention, step 201 adjusts the camera's shooting posture by combining the attitude parameters of the initial ground wire path, ensuring that a clear surface texture image of the ground wire can be acquired in real time during the UAV's approach, providing a high-quality visual data source for subsequent feature extraction and avoiding feature loss due to shooting angle deviation. Secondly, the grayscale conversion and Gaussian filtering preprocessing in step 202 effectively eliminates the interference of uneven lighting and image noise on texture features, highlighting the contrast between the ground wire and the background by outputting an enhanced binarized image. Thirdly, step 203 uses the Canny edge detection algorithm to extract continuous geometric contour edges, accurately capturing the boundary features of the ground wire, ensuring the integrity and continuity of the contour pixel set, and avoiding the impact of edge breaks or false edges on subsequent feature calculations. Finally, step 204 uses the least squares method to fit the curvature equation and calculate the normal vector, transforming discrete contour pixels into quantized geometric parameters (radius of curvature, normal vector). This not only achieves accurate characterization of the ground wire's curvature but also provides a reliable visual feature basis for subsequent multimodal feature matching and data fusion, improving the accuracy and robustness of overall environmental perception.

[0089] In an optional embodiment of the present invention, the extracted geometric contour features are subjected to multimodal feature matching and data fusion with the conductor spatial coordinates obtained from the LiDAR point cloud to obtain a fusion result, including:

[0090] Step 205: Based on the normal vector and radius of curvature in the geometric contour feature parameters, establish a local coordinate system for the surface texture of the conductor ground wire; construct a global spatial coordinate system based on the three-dimensional coordinate parameters of the equation of the spatial central axis of the conductor ground wire.

[0091] Step 206: The geometric contour feature parameters are mapped from the local coordinate system to the global spatial coordinate system through a spatial projection transformation matrix to generate a standardized contour feature aligned with the spatial coordinate scale of the LiDAR conductor.

[0092] Step 207: Calculate the difference in the radius of curvature and the angle between the normal vector of the standardized contour feature and the spatial center axis of the LiDAR ground wire at the corresponding position point. When the difference exceeds the threshold, it is marked as a pose deviation area.

[0093] Step 208: Aggregate the curvature deviation and orientation offset of all pose deviation regions and output the fusion result containing the spatial pose error vector.

[0094] In this embodiment of the invention, when applied in a specific application, step 205 is implemented as follows:

[0095] Based on the geometric contour feature parameters (including normal vector and radius of curvature) obtained in step 204 and the three-dimensional coordinate parameters of the spatial center axis equation of the conductor ground line obtained in step 102, the coordinate system is first constructed. Specifically, a local coordinate system for the surface texture of the ground wire is established based on the normal vector and radius of curvature in the geometric contour feature parameters: taking a feature point on the ground wire surface as the origin, the direction of the normal vector of that point is set as one coordinate axis of the local coordinate system (to characterize the vertical direction of the contour), and the direction along the radius of curvature pointing outward of the contour is set as another coordinate axis (to characterize the radial direction of curvature). Then, a third coordinate axis (to characterize the direction along the length of the ground wire) is determined by the right-hand rule, thus forming a local coordinate system that fits the surface texture of the ground wire, realizing the local spatial positioning of visual features; at the same time, a global spatial coordinate system is constructed based on the three-dimensional coordinate parameters (including the starting point coordinates, the ending point coordinates, and the axis direction vector) of the equation of the central axis of the ground wire space obtained in step 102: taking the starting point of the central axis of the ground wire space as the global origin, the direction vector of the axis is set as one coordinate axis of the global coordinate system (to characterize the extension direction of the ground wire), and the direction perpendicular to the axis and pointing to the horizontal ground is set as the second coordinate axis. A third coordinate axis is determined by the right-hand rule, forming a global spatial coordinate system covering the entire space range of the ground wire, providing a unified spatial reference for LiDAR point cloud coordinates.

[0096] In practical applications, the specific implementation steps of step 206 above are as follows:

[0097] First, a spatial projection transformation matrix is ​​constructed. Specifically, the translation parameter between the origin of the local coordinate system and the origin of the global spatial coordinate system is determined. This parameter is obtained by calculating the coordinate difference between the origins of the two coordinate systems in three-dimensional space and is used to characterize the positional offset of the local coordinate system relative to the global spatial coordinate system. Next, when calculating the rotation components, the angles formed between the X, Y, and Z axes of the local coordinate system and the X, Y, and Z axes of the global spatial coordinate system are determined. These angles are obtained by measuring the actual deflection angles of the corresponding coordinate axes in space. For each pair of corresponding coordinate axes (such as the local X-axis and the global X-axis, the local X-axis and the global Y-axis, etc.), the sine and cosine values ​​of the angle are calculated. The cosine value is used to characterize the degree of coincidence of the two coordinate axes in the same direction, and the sine value is used to characterize the degree of deviation of the two coordinate axes in the vertical direction. These sine and cosine values ​​are arranged in order according to the correspondence of the coordinate axes to form rotation components that can reflect the rotation state of the local coordinate system relative to the global spatial coordinate system in the X, Y, and Z dimensions. The magnitude of each component directly corresponds to the rotation amplitude in the corresponding direction.

[0098] When determining the scaling parameter, first select at least three pairs of known corresponding points in the local coordinate system and the global spatial coordinate system. These corresponding points are characteristic points on the surface of the conductor (such as the connection point of the conductor, surface protrusions, etc.), and their coordinates can be accurately obtained in both coordinate systems. Calculate the spatial straight-line distance of each pair of corresponding points in the local coordinate system and the spatial straight-line distance of the pair of corresponding points in the global spatial coordinate system. For each pair of corresponding point distance data, divide the distance value in the local coordinate system by the distance value in the global spatial coordinate system to obtain a single scale value. Take the arithmetic mean of the scale values ​​calculated for all pairs of corresponding points. This average value is the scaling parameter, and its magnitude reflects the scale ratio relationship between the local coordinate system and the global spatial coordinate system.

[0099] When integrating to form a spatial projection transformation matrix, the main framework of the matrix is ​​first constructed based on the rotation components. The rotation components are then filled into the corresponding positions of the matrix according to the coordinate transformation rules of three-dimensional spatial rotation, forming a rotation submatrix. The scaling parameters are multiplied by each element in the rotation submatrix to enable the rotation submatrix to have scale adjustment capabilities. Subsequently, the translation parameters (i.e., the coordinate difference between the origins of the two coordinate systems) are added to the matrix as independent components to form an extended matrix containing translation information. Finally, through matrix concatenation operations, the above-mentioned scaled rotation submatrix and translation components are integrated into a complete four-dimensional spatial projection transformation matrix. Each element of this matrix corresponds to the quantized parameters of translation, rotation, and scaling, and can be directly used to realize coordinate transformation from a local coordinate system to a global spatial coordinate system.

[0100] For the 3D coordinate transformation of each contour point, firstly, the scaling parameter in the spatial projection transformation matrix is ​​used to adjust the scale of the 3D coordinate values ​​of the contour points in the local coordinate system, so that the magnitude of the coordinate values ​​is initially consistent with the global spatial coordinate system; then, the rotation component in the matrix is ​​used to correct the orientation of the scaled coordinate values, so that the orientation of the contour points in 3D space matches the orientation reference of the global coordinate system; finally, the translation parameter in the matrix is ​​used to migrate the position of the rotated coordinate values, so that the contour points are transformed from the origin reference of the local coordinate system to the origin reference of the global coordinate system, and finally the corresponding coordinates of each contour point in the global spatial coordinate system are obtained.

[0101] For the direction transformation of the normal vector, the rotation component in the spatial projection transformation matrix is ​​extracted, and a spatial rotation transformation is performed on the normal vector in the local coordinate system. Specifically, the direction of the normal vector in three-dimensional space is adjusted by the angular relationship corresponding to the rotation component, so that the direction of the transformed normal vector is consistent with the normal direction reference of the guide surface in the global spatial coordinate system, ensuring the uniqueness of the vector direction in the global coordinate system. For the numerical adjustment of the radius of curvature, the value of the radius of curvature in the local coordinate system is proportionally converted according to the scaling parameter in the spatial projection transformation matrix. That is, the value of the radius of curvature in the local coordinate system is enlarged or reduced according to the scale ratio represented by the scaling parameter, so that the adjusted radius of curvature value matches the actual curvature scale of the guide reflected by the LiDAR point cloud in the global spatial coordinate system, ensuring the consistency of the physical meaning of the curvature parameter.

[0102] The specific process for scale alignment verification and correction is as follows:

[0103] First, select characteristic points of the ground wire, including but not limited to extreme bending points, abrupt changes in surface texture, and connection points with other objects. These points have identifiable features in both the local and global coordinate systems. Second, for each characteristic point, obtain the spatial coordinates of the mapped contour point and its corresponding position in the LiDAR point cloud, and calculate the three-dimensional straight-line distance between them as the coordinate deviation value. Obtain the mapped normal vector and the axis normal vector of the corresponding position in the LiDAR point cloud, and measure the spatial angle between them as the direction deviation value. Obtain the mapped radius of curvature and the curvature parameter of the corresponding position in the LiDAR point cloud, and calculate the numerical difference between them as the curvature deviation value. Then, the above... The coordinate deviation, direction deviation, and curvature deviation values ​​are compared with the preset allowable deviation range (which is set according to the accuracy requirements of conductor inspection). If all deviation values ​​are within the allowable range, the scale alignment is deemed qualified. If any deviation value exceeds the allowable range, the scaling parameters in the spatial projection transformation matrix are readjusted (for cases where the coordinate deviation or curvature deviation exceeds the limit) or the rotation component (for cases where the direction deviation exceeds the limit). After adjusting the parameters, the above coordinate mapping process is repeated, and the deviation values ​​of each feature point are calculated and compared again until the coordinate deviation, direction deviation, and curvature deviation values ​​of all feature points are within the preset allowable range. The generated contour feature at this time is the standardized contour feature.

[0104] In practical applications, the specific implementation steps of step 207 above are as follows:

[0105] After obtaining the standardized contour features in step 206 and knowing the coordinates of each position of the LiDAR conductor's spatial center axis, it is necessary to verify the consistency of the features between the two. First, the corresponding points of the standardized contour features and the central axis of the LiDAR ground plane are determined by spatial coordinate matching. That is, in the global spatial coordinate system, the point on the contour feature and the point on the central axis that are closest to each other are found and regarded as a pair of corresponding points. For each pair of corresponding points, the difference in the radius of curvature and the angle between the normal vectors are calculated: the difference in the radius of curvature is the radius of curvature of the standardized contour feature minus the radius of curvature of the LiDAR ground plane at that position (the radius of curvature of the LiDAR is derived from the equation of the central axis); the angle between the normal vectors is the angle between the normal vector of the standardized contour feature and the normal vector of the LiDAR ground plane at that position (derived from the direction of the central axis). Preset thresholds for the difference in the radius of curvature (e.g., a 5% baseline value for the radius of curvature) and the angle between the normal vectors (e.g., 5 degrees) are set. When the difference in the radius of curvature of a corresponding point exceeds the threshold, or the angle between the normal vectors exceeds the angle threshold, the local area where the point is located is marked as a pose deviation area, indicating that the visual features of this area are inconsistent with the LiDAR point cloud features.

[0106] In practical applications, the specific implementation steps of step 208 above are as follows:

[0107] After marking all pose deviation regions in step 207, the deviation data needs to be aggregated to generate a fusion result. First, each pose deviation region is traversed, and the curvature deviation (i.e., the difference in curvature radius calculated in step 207) and direction offset (i.e., the direction offset corresponding to the angle between the normal vectors) of all corresponding positions within the region are extracted. Statistical analysis is performed on the curvature deviation within the same deviation region, and the average or maximum value is taken as the representative curvature deviation of the region. The direction offset is vector synthesized, and combined with the direction parameters of the global coordinate system, the angular offset is converted into an offset vector in the spatial direction. Finally, the representative curvature deviations and direction offset vectors of all pose deviation regions are summarized and integrated to form an error vector that can reflect the overall spatial pose deviation of the conductor and ground wire. This vector contains the magnitude, direction, and spatial distribution information of the deviation.

[0108] In this embodiment of the invention, by constructing a local coordinate system and a global spatial coordinate system, a precise correlation benchmark between visual geometric contour features and LiDAR point cloud spatial coordinates is established, solving the adaptation problem of different sensor data in terms of spatial scale and direction. Coordinate mapping and scale alignment are achieved using a spatial projection transformation matrix, ensuring that visually extracted local detail features (such as normal vectors and radii of curvature) are consistent with LiDAR global spatial coordinates in terms of measurement standards, eliminating feature deviations caused by differences in sensor characteristics. By calculating the angle between the radius of curvature difference and the normal vector and setting a threshold for detection, the limitations of point cloud resolution, visual deformation, or local distortions in the initial path can be quantitatively identified. The millimeter-level pose deviation region caused by occlusion provides a definite "calibration target" for path correction; the curvature deviation and directional offset of the aggregated deviation region form a spatial pose error vector, which intuitively reflects the position, magnitude and direction of the deviation, providing accurate quantitative data support for dynamic calibration of the initial landing path, ensuring that the corrected path has both global spatial consistency and local detail accuracy; through complementary verification of visual and LiDAR data (such as using global coordinates to constrain local features and using local details to correct global deviations), the limitations of a single sensor in complex environments (such as illumination interference and point cloud missing) are effectively overcome, improving the stability and reliability of conductor feature recognition.

[0109] In an optional embodiment of the present invention, based on the fusion result, the pose deviation existing in the initial falling line path is dynamically calibrated to generate a corrected falling line path, including:

[0110] Step 209: Based on the spatial pose error vector, the real-time deviations of the UAV in heading angle, pitch angle and radial distance are decomposed.

[0111] Step 210: Input the real-time deviation into a preset PID controller to generate position correction parameters for the three-dimensional path node sequence in the initial falling path;

[0112] Step 211: Based on the position correction parameters, adjust the spatial coordinates of the three-dimensional path node sequence in real time and output the corrected path node sequence.

[0113] Step 212: Using the corrected path node sequence as new control points, the spatial trajectory curve is recalculated using a cubic spline interpolation algorithm to generate a corrected landing path that includes the calibrated position coordinates, heading angle, and pitch angle control parameters.

[0114] In this embodiment of the invention, step 209 is specifically implemented according to the following process:

[0115] First, the spatial pose error vector output in step 208 is analyzed. This vector contains the position and attitude deviation information of the UAV in three-dimensional space. According to the preset definition of the UAV motion coordinate system (with the UAV's center of gravity as the origin, the forward direction as the vertical axis, the horizontal direction as the horizontal axis, and the vertical direction as the vertical axis), the spatial pose error vector is decomposed into deviations in three dimensions:

[0116] The heading angle deviation is obtained by calculating the angle between the projection of the error vector onto the horizontal plane and the UAV's preset heading, reflecting the UAV's pointing deviation in the horizontal direction; the pitch angle deviation is obtained by calculating the angle between the projection of the error vector onto the vertical plane (the vertical plane along the forward direction) and the UAV's preset pitch angle, reflecting the UAV's tilt deviation in the vertical direction; the radial distance deviation is obtained by extracting the magnitude of the component of the error vector along the radial direction of the ground guide (the direction perpendicular to the center axis of the ground guide), reflecting the distance deviation between the UAV and the ground guide. The deviations in the above three dimensions are recorded separately as real-time deviations.

[0117] When implementing step 210 above, the following process shall be followed:

[0118] First, the preset PID controller is invoked (the proportional coefficient, integral time, and derivative time parameters of the controller have been pre-tuned according to the dynamic characteristics of the UAV and the inspection accuracy requirements); the real-time deviation of heading angle, real-time deviation of pitch angle, and real-time deviation of radial distance obtained in step 209 are respectively input into the three control channels of the PID controller (corresponding to the heading, pitch, and radial dimensions).

[0119] The controller calculates the real-time deviation for each channel:

[0120] The proportional component is proportional to the current value of the real-time deviation and is used to quickly respond to deviations. The integral component calculates the cumulative value of the real-time deviation over a preset time period and is used to eliminate persistent steady-state deviations. The differential component is calculated based on the rate of change of the real-time deviation (i.e., the change in deviation per unit time) and is used to suppress rapid fluctuations in deviations. The three components are superimposed according to preset weights to obtain the position correction components for each path node in the X, Y, and Z spatial axes. After integration, the position correction parameters are formed.

[0121] When implementing step 211 above, the following process shall be followed:

[0122] First, the position correction parameters generated in step 210 are associated with the initial 3D path node sequence obtained in step 105 to determine the X-axis, Y-axis, and Z-axis correction components corresponding to each initial path node. For each initial path node, its original X-axis coordinates are superimposed with the X-axis correction components to obtain the adjusted X-axis coordinates. Similarly, the Y-axis and Z-axis coordinates are superimposed to obtain the adjusted Y-axis and Z-axis coordinates. After the coordinates of all nodes are adjusted, the safety of each path node coordinate is checked to confirm that it is still within the cylindrical collision-free safety corridor generated in step 104 (i.e., the distance from the node to the center axis of the conductor does not exceed the safety radius). If a node exceeds the safety corridor range, the corresponding position correction components are truncated (the correction magnitude is reduced) until the node coordinates return to the safety corridor range. According to the arrangement order of the path nodes, all the adjusted node coordinates are integrated in sequence to form the corrected path node sequence.

[0123] When implementing step 212 above, the following process shall be followed:

[0124] First, the corrected path node sequence is reordered according to the order in which the nodes approach the ground wire of the UAV, ensuring that the logical order between the nodes is consistent with the direction of movement. Then, using the reordered corrected path nodes as new control points, a cubic spline interpolation algorithm is applied, namely:

[0125] The trajectory between two adjacent correction nodes is fitted, and the specific process is as follows:

[0126] First, two adjacent modified path nodes are selected, and their three-dimensional coordinates are used as the start and end points for trajectory fitting. For the trajectory segment between these two nodes, a curve form conforming to the characteristics of a cubic polynomial is set so that the curve can accurately pass through the coordinate positions of these two nodes. Second, to ensure the continuity of the tangent direction at the connection point of adjacent nodes, the tangent direction of the previous trajectory segment at the end point (i.e., the trend of curve change at that point) is calculated, and the tangent direction of the current trajectory segment at the start point is made consistent with it. By adjusting the curve parameters, the smooth transition of the two curve segments at the connection point is ensured. At the same time, to achieve a smooth transition of curvature, the rate of change of curvature at the end point of the previous trajectory segment (i.e., the rate of change of the curvature) is calculated, and the rate of change of curvature at the start point of the current trajectory segment is matched with it. By further fine-tuning the curve parameters, the continuous change of curvature at the connection point of the two curve segments is ensured, avoiding abrupt changes in curvature.

[0127] When extracting calibrated 3D spatial coordinates based on a smooth trajectory curve, the specific process is as follows:

[0128] On the fitted smooth trajectory curve, sampling points are selected according to a preset sampling interval (this interval is set according to the control accuracy requirements of the UAV; the smaller the interval, the denser the coordinate data). For each sampling point, its X-axis, Y-axis, and Z-axis coordinate values ​​in the three-dimensional coordinate system are read. These coordinate values ​​are the calibrated three-dimensional spatial coordinates of the UAV at that location. The coordinates of all sampling points are arranged in the order of the trajectory to form a complete coordinate sequence.

[0129] The specific process for calculating the calibrated heading and pitch angles based on the tangent direction of the trajectory curve is as follows:

[0130] For each sampling point on the trajectory curve, determine the tangent direction of the curve at that point (i.e., the direction in which the curve extends at that point); when calculating the heading angle, project the tangent direction onto the horizontal plane (ignore the vertical component), and measure the angle between the horizontal projection direction and the preset horizontal reference direction (such as true north or the initial forward direction of the UAV). This angle is the calibrated heading angle at that sampling point; when calculating the pitch angle, measure the angle between the tangent direction and the horizontal plane. This angle is the calibrated pitch angle at that sampling point, where an upward tilt of the tangent direction is a positive angle and a downward tilt is a negative angle.

[0131] When integrating control parameters to form a corrected landing path, the specific process is as follows:

[0132] The calibrated three-dimensional spatial coordinates of each sampling point, along with the corresponding calibrated heading angle and pitch angle, are correlated to ensure that each position coordinate corresponds to a unique heading angle and pitch angle parameter. These correlated data are then arranged sequentially according to the order of the sampling points on the trajectory to form a complete sequence containing position and attitude control information. This sequence is the corrected landing path, which can be directly used to control the movement of the UAV, compensate for the attitude deviation in the initial path, and ensure that the UAV approaches the target ground wire along a precise trajectory.

[0133] In this embodiment, step 209 decomposes the spatial pose error vector into heading angle, pitch angle, and radial distance deviations, achieving a refined breakdown of the UAV's pose deviations. This provides a definite target for targeted correction and avoids blind corrections caused by deviation coupling. Step 210 calculates the real-time deviations using a PID controller, generating position correction parameters that can quickly respond to deviation changes. Combined with the real-time adjustment of path nodes in step 211, this enables dynamic adaptability to path correction, allowing real-time compensation for pose shifts caused by environmental interference or sensor errors. Step 212 uses cubic spline interpolation to regenerate the trajectory curve, ensuring continuous tangents and smooth curvature between the corrected path nodes, avoiding sudden attitude changes during UAV movement. Simultaneously, safety corridor constraints are retained during node adjustment to ensure the corrected path remains within a collision-free space, balancing accuracy and safety. Integrating the calibrated position coordinates and attitude parameters, the generated corrected landing path accurately compensates for millimeter-level pose deviations in the initial path, enabling the UAV to maintain sub-centimeter-level relative positioning accuracy during critical phases approaching the ground wire (e.g., within a 10-meter to 1-meter range), providing reliable guidance for the final stable hovering and landing maneuvers. Through the closed-loop control logic of "deviation decomposition - parameter correction - node adjustment - trajectory reconstruction", the system effectively resists the influence of complex environment (such as electromagnetic interference and partial obstruction) on path accuracy, and ensures that even when there is noise in the sensor data, it can still output stable and accurate line landing control commands.

[0134] In an optional embodiment of the present invention, step 300 involves acquiring electromagnetic field strength data monitored by an airborne electromagnetic sensor in real time based on the corrected fall path; and using the electromagnetic field strength data to counteract the attitude shift of the UAV and the work platform caused by electromagnetic interference in real time, so as to obtain a control signal after the attitude shift is counteracted, including:

[0135] Step 301: Based on the heading angle and pitch angle control parameters of the corrected landing path, during the flight of the UAV along the corrected path, the raw data of electromagnetic field intensity along the X-axis, Y-axis or Z-axis monitored by the airborne three-axis electromagnetic sensor are acquired synchronously at a preset sampling period.

[0136] Step 302: Perform sliding window weighted average filtering on the raw electromagnetic field strength data to output smoothed electromagnetic field strength time series data;

[0137] Step 303: Extract the high-frequency oscillation component from the smoothed electromagnetic field intensity time series data, and identify the periodic pose shift characteristics caused by the alternating magnetic field of the transmission line through spectrum analysis.

[0138] Step 304: Based on the pose offset characteristics, establish a mapping relationship model between electromagnetic interference intensity and UAV heading angle deviation, pitch angle deviation and radial displacement deviation;

[0139] Step 305: Based on the mapping relationship model, use the least mean square adaptive filter to dynamically calculate the real-time offset of electromagnetic interference that needs to be compensated in terms of heading angle, pitch angle and radial distance.

[0140] Step 306: The real-time offset is superimposed on the heading angle, pitch angle and position coordinate control parameters of the correct landing path to generate an anti-interference control signal after offsetting the attitude deviation.

[0141] In this embodiment of the invention, step 301 is specifically implemented according to the following process:

[0142] First, based on the heading and pitch control parameters included in the corrected landing path generated in step 212, adjust the measurement attitude of the airborne three-axis electromagnetic sensor so that the X, Y, and Z axes of the sensor are aligned with the forward, lateral, and vertical directions of the UAV, respectively, ensuring that the measurement direction corresponds to the direction of the UAV's attitude change. Second, set a preset sampling period (this period is determined based on the frequency of change of the electromagnetic field of the power transmission line and the control response speed of the UAV, typically 50-100 milliseconds). During the flight of the UAV along the corrected landing path, sensor data acquisition is synchronously started according to this period to obtain the raw electromagnetic field strength data in the X, Y, and Z axes in real time. The raw data in each direction includes the magnetic field strength value at the corresponding moment and the acquisition timestamp, ensuring the temporal correlation between the data and the UAV's flight attitude.

[0143] When implementing step 302 above, the following process shall be followed:

[0144] First, the size of the sliding window is set (the number of sampling points in the window is determined based on the fluctuation characteristics of electromagnetic interference, typically 5-15 consecutive sampling points), and a weight is assigned to each sampling point within the window, with higher weights for sampling points closer to the current time (e.g., using linearly increasing weights, the newest sampling point has the highest weight within the window, and the oldest sampling point has the lowest weight), to highlight the influence of recent data on the filtering result. Second, the raw electromagnetic field strength data obtained in step 301 is sequentially input into the sliding window in chronological order. For all sampling points within the window, their respective weights are multiplied by the corresponding magnetic field strength values ​​to obtain weighted magnetic field strength values. All weighted values ​​are summed and then divided by the sum of all weights within the window to obtain the filtering result for that window. Finally, as time progresses, the sliding window moves point by point along the time axis (removing the oldest sampling point within the window for each new sampling point), repeating the above weighted summation operation to generate continuous, smoothed electromagnetic field strength time-series data, eliminating instantaneous impulse noise in the raw data.

[0145] When implementing step 303 above, the following process shall be followed:

[0146] First, the smoothed electromagnetic field intensity time-series data obtained in step 302 is processed using a high-pass filter to remove low-frequency trend components (such as slow changes in the ambient magnetic field) and retain high-frequency oscillation components (which mainly reflect rapid fluctuations caused by the alternating magnetic field of the transmission line). Second, spectral analysis is performed on the extracted high-frequency oscillation components: the oscillation data in the time domain is converted to the frequency domain, the energy proportion of different frequency components is calculated, and the dominant frequency of energy concentration is identified (this frequency is usually consistent with the power frequency or its harmonic frequency of the transmission line, such as 50Hz or 100Hz). Finally, the periodic pose shift characteristics are determined based on the dominant frequency: the oscillation period (the period is the reciprocal of the frequency) is calculated according to this frequency, and the amplitude (reflecting the shift intensity) and phase (reflecting the time correlation between the shift and the magnetic field change) of the oscillation components are analyzed to fully characterize the periodic pose shift law of the UAV caused by electromagnetic interference.

[0147] When implementing step 304 above, the following process shall be followed:

[0148] First, a simulation experimental platform was built, which included an adjustable-intensity power frequency electromagnetic field generator (simulating the electromagnetic field of a power transmission line), a drone mounting bracket (to maintain the drone's attitude stability), an airborne three-axis electromagnetic sensor (consistent with the sensor model used in actual inspections), and a high-precision optical positioning system (for attitude measurement, with a positioning accuracy of no less than 0.1 mm). Second, the electromagnetic field intensity adjustment range was set: starting from a threshold intensity where the drone is not significantly disturbed (e.g., 50 μT), it was gradually increased to the maximum intensity near the power transmission line (e.g., 500 μT). A test point was set at each preset intensity value (e.g., 50 μT), forming 10-20 intensity gradients. For each intensity gradient, the electromagnetic field generator was activated to output a stable electromagnetic field, and the drone's power system was simultaneously activated (simulating flight). After the electromagnetic field intensity stabilized, the data was recorded synchronously.

[0149] The airborne electromagnetic sensors measure real-time values ​​in the X, Y, and Z axes (recorded every 10 milliseconds for 5 seconds, with the average value taken as the electromagnetic interference intensity at that intensity); the high-precision optical positioning system collects the UAV's real-time pose parameters (including heading angle, pitch angle, and 3D coordinates), and compares them with preset interference-free ideal pose parameters to calculate the heading angle deviation (the difference between the actual heading angle and the ideal heading angle), pitch angle deviation (the difference between the actual pitch angle and the ideal pitch angle), and radial displacement deviation (the difference in radial distance between the actual 3D coordinates and the ideal coordinates); the electromagnetic interference intensity at each intensity gradient and the corresponding three deviations are used as a set of sample data to form a sample dataset.

[0150] The specific process of statistical analysis of sample data is as follows:

[0151] First, each data set in the sample dataset is paired and organized to establish three subsets: "Electromagnetic Interference Intensity - Heading Angle Deviation," "Electromagnetic Interference Intensity - Pitch Angle Deviation," and "Electromagnetic Interference Intensity - Radial Displacement Deviation." Second, each subset is grouped and statistically analyzed: the electromagnetic interference intensity is divided into several intervals according to a preset interval (e.g., 20 μT per interval), and the average electromagnetic interference intensity and corresponding average deviation of all samples within each interval are calculated, resulting in several sets of statistical data points. Next, correlation analysis is performed on the statistical data points: the correlation coefficient (a statistic used to characterize the degree of linear correlation between two variables) is calculated for each subset. When the absolute value of the correlation coefficient is close to 1, it is determined to be a strong linear correlation. When the absolute value of the correlation coefficient is small (e.g., less than 0.3), further observe the distribution trend of the data points and identify possible nonlinear correlations (such as quadratic curve correlation, exponential correlation, etc.) by drawing scatter plots. Finally, determine the correlation trend: for example, if the statistical data points of "electromagnetic interference intensity - heading angle deviation" are approximately linearly distributed and the deviation increases uniformly with the increase of interference intensity, then the two are determined to be positively linearly correlated; if the data points of "electromagnetic interference intensity - radial displacement deviation" show a curved shape and the deviation increases slowly at low intensity and rapidly at high intensity, then it is determined to be a nonlinear correlation.

[0152] The specific process of constructing the mapping relationship model is as follows:

[0153] First, select the model form based on the correlation trend: for linearly correlated variable pairs (such as electromagnetic interference intensity and heading angle deviation), use linear equations as the model basis; for nonlinearly correlated variable pairs (such as electromagnetic interference intensity and radial displacement deviation), use polynomial equations or piecewise functions as the model basis. Second, determine the model parameters through fitting: take the electromagnetic interference intensity in the sample dataset as the independent variable and the corresponding deviation as the dependent variable, substitute them into the selected model equation, and use the least squares method (by adjusting the coefficients to minimize the overall deviation between the calculated value of the equation and the measured value of the sample) to solve for the coefficients in the equation (such as the slope and intercept in linear equations, and the coefficients of each order in polynomial equations).

[0154] For example, for the linearly correlated "electromagnetic interference intensity - heading angle deviation", the fitted equation is: heading angle deviation = k1 × electromagnetic interference intensity + b1 (where k1 is a proportionality coefficient, reflecting the heading deviation caused by unit intensity interference; b1 is a constant term, reflecting the initial deviation when there is zero interference); for the nonlinearly correlated variable pair, the fitted equation contains quadratic terms or higher-order terms.

[0155] Finally, the effectiveness of the model is verified: some sample data that were not involved in the fitting are substituted into the model, and the difference between the deviation predicted by the model and the actual deviation is calculated. If the average difference is less than the preset error threshold (such as 0.1° or 0.5 mm), the model is deemed effective; otherwise, the model form is readjusted or the amount of sample data is increased and fitted again until the model meets the accuracy requirements. The final mapping relationship model can directly calculate the corresponding pose deviation based on the input electromagnetic interference intensity.

[0156] When implementing step 305 above, the following process shall be followed:

[0157] First, determine the initial filter coefficients. Extract the correlation coefficients (such as the proportional coefficients of the linear model and the coefficients of each order of the nonlinear model) from the mapping relationship model constructed in step 304. Assign these coefficients to the initial coefficient matrix of the filter according to the three dimensions of heading angle, pitch angle, and radial distance, so that the initial state of the filter is consistent with the known electromagnetic interference-pose deviation correlation law. Second, set the convergence factor. Based on the fluctuation frequency of the electromagnetic field of the transmission line (such as 50Hz power frequency) and the response speed of the UAV pose control system (such as 100Hz control frequency), select a value in the range of 0.01-0.1 as the convergence factor. If the electromagnetic interference fluctuation is severe (such as high frequency component ratio), select a smaller convergence factor (such as 0.02) to avoid filter output oscillation; if the interference is stable (such as power frequency is the main component), select a larger convergence factor (such as 0.08) to speed up the response speed. Finally, determine the error threshold. Based on the accuracy requirements of the UAV landing line control (such as a heading angle deviation of no more than 0.5° and a radial displacement deviation of no more than 5 mm), set the threshold to 1 / 5 to 1 / 3 of the corresponding accuracy requirements (such as a heading angle error threshold of 0.1° and a radial displacement error threshold of 1 mm) to ensure that the filtering effect meets the actual control requirements.

[0158] The specific process of signal input is as follows:

[0159] First, the input signal is processed by converting the periodic pose offset features (including oscillation frequency, amplitude, and phase) identified in step 303 into a continuous electrical signal in the time domain. The amplitude corresponds to the signal strength, the frequency corresponds to the signal period, and the phase corresponds to the offset at the start of the signal, so that the signal can be directly recognized by the filter. Second, the error signal is acquired: the current actual pose (heading angle, pitch angle, and three-dimensional coordinates) is collected in real time by the high-precision inertial navigation equipment on the UAV and compared with the theoretical pose parameters for correcting the landing path (extracted from the control parameters in step 212). The differences between the two in heading angle, pitch angle, and radial distance are calculated, and these differences are converted into electrical signals and used as error signals input to the filter. Finally, the signal timing is synchronized: the input signal and the error signal are aligned by timestamps to ensure that the electromagnetic interference features and the corresponding pose deviations at the same time are matched in the time dimension, avoiding a decrease in filtering accuracy due to timing misalignment.

[0160] The specific process for adjusting the filter coefficients is as follows:

[0161] First, the difference between the current output and the error is calculated. The filter performs calculations on the input signal based on the current filter coefficients to generate the predicted pose deviation (i.e., the filtered output). This predicted deviation is compared with the actual error signal (measured pose deviation), and the difference between the two (i.e., the residual) is calculated. A positive residual indicates that the predicted deviation is less than the actual deviation, and a negative residual indicates that the predicted deviation is greater than the actual deviation. Second, the coefficients are updated according to the least mean square criterion: the filter coefficients are adjusted according to the magnitude and direction of the residual. If the residual is positive, the filter coefficient in the corresponding dimension is increased (to enhance the prediction of the deviation in that direction). (Measurement); if the residual is negative, reduce the filter coefficient of the corresponding dimension (weaken the prediction of deviation in that direction); the adjustment range is determined by the residual size and the convergence factor. The larger the residual and the larger the convergence factor, the larger the coefficient adjustment range; finally, iterate until the error reaches the target: repeat the above process of "calculating residuals - updating coefficients". After each iteration, calculate the overall deviation between the current filter output and the actual error (such as the sum of squares of the residuals in the three dimensions); when the overall deviation is less than the preset error threshold, stop the iteration. The filter coefficient at this time is the optimal coefficient to adapt to the current electromagnetic interference state.

[0162] The specific process of real-time offset calculation is as follows:

[0163] First, the real-time electromagnetic interference intensity is obtained. From the smoothed electromagnetic field intensity time series data output in step 302, the electromagnetic interference intensity values ​​of the X, Y, and Z axes at the current moment are extracted as the real-time input of the filter. Second, the three-dimensional compensation amount is calculated. The real-time electromagnetic interference intensity is input into the adjusted filter, and the prediction deviation (i.e., the pose shift caused by electromagnetic interference) in the heading angle, pitch angle, and radial distance dimensions is calculated respectively. The prediction deviation in each dimension is reversed (e.g., if the predicted heading angle deviation is +2°, then the compensation offset is -2°) to obtain the real-time compensation offset to be applied, ensuring that the offset can completely cancel the influence of electromagnetic interference. Finally, the rationality of the compensation is verified. It is checked whether the calculated real-time compensation offset is within the adjustment range of the UAV actuator (e.g., the heading angle compensation does not exceed ±5°, and the radial distance compensation does not exceed ±20 mm). If it exceeds the range, it is truncated (taking the limit value within the maximum adjustment range) to ensure that the compensation command can be physically executed.

[0164] When implementing step 306 above, the following process shall be followed:

[0165] First, the control parameters in the corrected landing path generated in step 212 are extracted, including the heading angle, pitch angle, and three-dimensional position coordinates (X-axis, Y-axis, Z-axis coordinates) of each position point. Second, the real-time offset calculated in step 305 is superimposed with the corresponding control parameters. For the heading angle, the heading angle compensation offset is added to the heading angle of the corrected path (a negative offset is equivalent to subtraction); for the pitch angle, the pitch angle compensation offset is added to the pitch angle of the corrected path; for the position coordinates, the radial distance compensation offset is decomposed into components in the X-axis, Y-axis, and Z-axis directions, and added to the corresponding axis coordinates of the corrected path. Finally, the superimposed control parameters are validated for rationality to ensure that the adjusted heading angle and pitch angle are within the physical motion range of the UAV (e.g., the heading angle does not exceed ±180°, and the pitch angle does not exceed ±30°), and the position coordinates remain within the cylindrical safety corridor. After successful validation, these parameters are integrated into an anti-interference control signal to drive the actuators of the UAV and the operating platform in real time, counteracting attitude deviations caused by electromagnetic interference.

[0166] In this embodiment, step 301 synchronously acquires triaxial electromagnetic field data according to the corrected path attitude, and combines this with the sliding window filtering in step 302 to eliminate noise. Step 303 accurately identifies the periodic offset characteristics of the alternating magnetic field of the transmission line (such as 50Hz power frequency oscillation) through spectrum analysis, providing clear characteristic basis for interference compensation and avoiding interference from stray magnetic fields in the environment. Step 304 fits a mapping relationship model using laboratory sample data to associate the abstract electromagnetic interference intensity with specific heading angle, pitch angle, and radial displacement deviation, realizing a quantitative characterization of the interference effect and solving the problem of difficulty in directly mapping electromagnetic interference to attitude deviation. Step 305 uses a minimum mean square adaptive filter based on real-time interference data. Calculating the compensation offset allows for rapid response to changes in electromagnetic field strength (such as increased interference when the drone approaches a conductor), ensuring that the compensation amount is equal in magnitude and opposite in direction to the actual offset, thus canceling out the interference in real time. Step 306 superimposes the compensation offset onto the correction path parameters, and the generated anti-interference control signal can directly correct the pose drift caused by electromagnetic interference (such as centimeter-level offset), avoiding collision risks or misalignment issues caused by magnetic compass interference or abnormal motor control. Through the entire process of electromagnetic interference sensing, modeling, and compensation, the pose control accuracy of the drone in strong electromagnetic field environments is significantly enhanced, ensuring that the operating platform maintains sub-centimeter-level positioning stability throughout the entire process from approaching the ground wire to stable hovering.

[0167] In an optional embodiment of the present invention, based on the control signal after correcting the drop path and offsetting the pose deviation, the UAV and the work platform are driven to perform drop control actions until the work platform achieves stable hovering and positioning at the designated position of the target conductor wire, including:

[0168] Step 307: Analyze the heading angle, pitch angle, and position coordinate parameters in the anti-interference control signal to generate independent drive commands for the UAV flight control and the leveling mechanism of the work platform. This includes: Step 3071: Calculate the displacement control quantities of the UAV in three-dimensional space along the X, Y, and Z axes based on the position coordinate parameters in the anti-interference control signal; Step 3072: Generate the yaw angle control quantity and pitch angle control quantity of the UAV respectively through an attitude decoupling algorithm based on the heading angle and pitch angle parameters; Step 3073: Fuse the displacement control quantity, yaw angle control quantity, and pitch angle control quantity to generate a UAV flight control command containing the reference speed of the quadcopter motor; Step 3074: Extract the pitch angle parameter from the anti-interference control signal, calculate the extension and retraction of the hydraulic outriggers or the rotation angle of the servo motor of the leveling mechanism of the work platform, and generate independent drive commands to maintain the horizontal attitude of the platform.

[0169] Step 308: Based on the UAV flight control commands, calculate the quadcopter motor speed adjustment in real time, drive the UAV to move along the radial target ground line of the corrected landing line, and output the UAV's real-time attitude data, including: Step 3081: Based on the quadcopter motor reference speed in the UAV flight control commands, and combined with the UAV's current attitude angular velocity and acceleration data, dynamically calculate the speed compensation of each motor; Step 3082: Superimpose the reference speed and the speed compensation to generate a real-time drive signal for the quadcopter motor, controlling the UAV to move along the spatial trajectory of the corrected landing line path; Step 3083: Collect the UAV's three-dimensional position, heading angle, and pitch angle data in real time, fuse GPS positioning data and LiDAR ranging data, and output the UAV's real-time attitude data including position coordinates and attitude angles;

[0170] Step 309: Based on the drive command, dynamically adjust the joint angle of the robotic arm to keep the working platform in a horizontal attitude during the movement of the drone, and output the real-time attitude data of the platform.

[0171] Step 310: Respond to the real-time pose data of the UAV. When the distance between the UAV and the target ground wire is less than a preset threshold, a progressive deceleration control strategy is initiated based on the difference between the real-time position coordinates and the specified position of the ground wire to dynamically reduce the flight speed of the UAV and the amplitude of the platform leveling action.

[0172] Step 311: Integrate the real-time pose data of the UAV with the real-time attitude data of the platform, and combine the airborne inertial measurement data and visual feedback to calculate the pose deviation of the platform relative to the ground wire; when the pose deviation continuously meets the preset hovering conditions, output the stable hovering judgment result.

[0173] In this embodiment of the invention, step 307 is specifically implemented according to the following process:

[0174] First, the three-dimensional position coordinates of the target (X target, Y target, Z target) are extracted from the anti-interference control signal. The current three-dimensional position coordinates of the UAV (X current, Y current, Z current) are obtained through the airborne positioning module. The displacement difference of each axis is calculated: X-axis displacement control amount = X target - X current, Y-axis displacement control amount = Y target - Y current, Z-axis displacement control amount = Z target - Z current. Positive values ​​indicate that the UAV needs to move in the positive direction of that axis, and negative values ​​indicate that it needs to move in the negative direction. The absolute value of the displacement control amount directly corresponds to the distance to be moved.

[0175] When implementing step 3072 above, the following process shall be followed:

[0176] First, the target heading angle θ and target pitch angle φ are extracted from the anti-interference control signal. The current heading angle θ and current pitch angle φ are obtained through the attitude sensor. The heading angle deviation Δθ = θ target - θ current and the pitch angle deviation Δφ = φ target - φ current are calculated. The turning direction is determined according to the sign of Δθ (Δθ is positive for clockwise turning and negative for counterclockwise turning). The amplitude of the yaw angle control is calculated according to the absolute value of Δθ and a preset ratio (e.g., 1° deviation corresponds to a turning rate of 0.5° / s). The larger Δθ is, the higher the turning rate. The tilt direction is determined according to the sign of Δφ (Δφ is positive for pitching up and negative for pitching down). The amplitude of the pitch angle control is calculated according to the absolute value of Δφ and a preset ratio (e.g., 1° deviation corresponds to a tilt rate of 0.5° / s). The larger Δφ is, the higher the tilt rate. The two control quantities are processed through an independent channel to ensure that the heading and pitch adjustments do not interfere with each other.

[0177] When implementing step 3073 above, the following process shall be followed:

[0178] First, the X-axis and Y-axis displacement control quantities are mapped to the horizontal thrust distribution of the quadcopter: X-axis displacement corresponds to the thrust difference between the front and rear motors (e.g., positive X-axis displacement requires the front motor thrust to be greater than the rear motor thrust), and Y-axis displacement corresponds to the thrust difference between the left and right motors (e.g., positive Y-axis displacement requires the right motor thrust to be greater than the left motor thrust). The Z-axis displacement control quantity is mapped to the total thrust of the four motors (positive Z-axis displacement requires the total thrust to increase, and negative Z-axis displacement requires it to decrease). Second, the yaw angle control quantity is converted into the speed difference between the left and right motors (clockwise rotation requires the left motor speed to be higher than the right motor speed, and the difference is proportional to the control quantity amplitude), and the pitch angle control quantity is converted into the speed difference between the front and rear motors (tilt-up requires the rear motor speed to be higher than the front motor speed, and the difference is proportional to the control quantity amplitude). Finally, based on the quadcopter motor dynamic characteristics, the thrust distribution, total thrust, and speed difference are integrated into the reference speeds of the four motors (the initial operating speeds of the front left, front right, rear left, and rear right motors), forming the UAV flight control commands.

[0179] When implementing step 3074 above, the following process shall be followed:

[0180] First, the current platform pitch angle φplatform is obtained through the platform tilt sensor and compared with the target horizontal pitch angle (usually 0°) in the anti-interference control signal. The horizontal deviation Δφplatform = φplatform - 0° is calculated. Based on the installation distance (lever length) between the platform and the outriggers, the required outrigger length is calculated according to the ratio that "for every 1° increase in deviation, the corresponding outrigger extension / retraction amount increases by a preset length (e.g., 5mm)" (Δφplatform indicates proper extension of the lower outrigger and negative extension / retraction of the higher outrigger). The servo motor rotation angle is calculated according to the ratio that "for every 1° increase in deviation, the servo motor rotation angle increases by a preset value (e.g., 2°)" (Δφplatform indicates clockwise rotation and negative rotation of counterclockwise rotation). This ensures that the platform approaches horizontality and generates a leveling drive command.

[0181] When implementing step 3081 above, the following process shall be followed:

[0182] First, obtain the reference speed of the quadcopter motor generated in step 3073. Collect the current attitude angular velocity (rotational speed around the X, Y, and Z axes) using a gyroscope and the current acceleration using an accelerometer. When the angular velocity exceeds a preset stable range (e.g., yaw angular velocity > 5° / s), calculate the corresponding motor speed compensation: reduce the motor speed in the overshoot direction (e.g., reduce the right motor speed if the right yaw is too fast), and the compensation amount is proportional to the angular velocity overshoot (1° / s overshoot corresponds to 5% compensation of the reference speed). When the acceleration deviates from the target value (e.g., climb acceleration < 0.5 m / s²),... 2 ), calculate the vertical motor compensation amount, and increase the speed of all motors proportionally (if the acceleration is less than 10%, the speed will be increased by 5%) to ensure that the power output matches the displacement requirements.

[0183] When implementing step 3082 above, the following process shall be followed:

[0184] The motor reference speed in step 3073 and the speed compensation amount in step 3081 are superimposed according to the motor correspondence (e.g., the real-time speed of the front left motor = the reference speed of the front left + the compensation amount of the front left), generating real-time drive signals (voltage signals or pulse width signals) for four motors. The drive signals are then output to the quadcopter motors through the motor controller, controlling the motors to run at the real-time speed, driving the UAV to gradually approach the target ground wire along the spatial trajectory of the corrected landing path.

[0185] When implementing step 3083 above, the following process shall be followed:

[0186] The system acquires real-time latitude and longitude using a GPS module and converts it into three-dimensional spatial coordinates. It measures the straight-line and horizontal distance between the UAV and the grounding wire using a LiDAR sensor to correct GPS coordinate errors (the closer to the grounding wire, the higher the weight of LiDAR data; for example, when the distance is <5 meters, the LiDAR weight accounts for 80%). It collects heading and pitch angle data using an attitude sensor, aligns them with the position coordinates by timestamp, and integrates them into real-time UAV attitude data containing three-dimensional position (X, Y, Z) and attitude angles (heading and pitch angles), and outputs it.

[0187] When implementing step 309 above, the following process shall be followed:

[0188] First, the current platform pitch and roll angles are collected in real time by the platform's attitude sensors and compared with the target horizontal attitude (0° pitch and 0° roll) to calculate the attitude deviation (Δpitch and Δroll). Based on the connection structure between the robotic arm joints and the platform, the influence weight of each joint on the attitude is determined (e.g., the weight of the root joint is higher than that of the end joints). According to the ratio of "for every 1° increase in deviation, the corresponding joint adjustment angle increases by a preset value (e.g., 1.5°)," Δpitch is allocated to the joints controlling pitch, and Δroll is allocated to the joints controlling roll. The angle to be adjusted for each joint is calculated (positive deviation means positive rotation, negative deviation means negative rotation). The angle adjustment is executed by the joint drive motors, and the adjusted attitude data is fed back in real time until the platform attitude deviation is < ±0.5°, at which point the real-time platform attitude data is output.

[0189] When implementing step 310 above, the following process shall be followed:

[0190] A preset distance threshold (e.g., 1 meter) is used to measure the straight-line distance D between the UAV and the target ground wire in real time via a LiDAR sensor. When D > the threshold, the UAV moves at a normal speed (e.g., 1 m / s), and the platform's leveling actions are performed at the normal amplitude. When D ≤ the threshold, gradual deceleration is initiated.

[0191] Flight speed adjustment: Reduce speed by the ratio of "(threshold - D) / threshold" (e.g., when D = 0.5 meters, reduce speed to 50% of normal speed).

[0192] Adjusting the leveling motion amplitude: Simultaneously reduce the extension and retraction of the hydraulic outriggers or the rotation angle of the servo motor by the same proportion (e.g., when D=0.5 meters, reduce the motion amplitude to 50% of the normal range) to avoid excessive motion amplitude when approaching the target, which could cause swaying.

[0193] When implementing step 311 above, the following process shall be followed:

[0194] By integrating the real-time UAV pose data from step 308 with the real-time platform attitude data from step 309, and combining the acceleration and angular velocity data from the airborne inertial measurement unit (IMU) with the position deviation of the ground wire in the image collected by the visual sensor, the three-dimensional position deviation (≤5mm) and attitude deviation (≤0.1°) of the platform relative to the ground wire are calculated.

[0195] Preset hovering conditions: position deviation < 3mm and attitude deviation < 0.05°. The deviation value is monitored in real time. When the hovering conditions are met for 10 consecutive sampling cycles (50ms per cycle), the platform is determined to have achieved stable hovering, and the stable hovering determination result is output.

[0196] This embodiment separates displacement control quantities from attitude control quantities, and independently converts the three-dimensional spatial displacement, yaw angle, and pitch angle requirements into motor drive parameters, thereby avoiding attitude fluctuations caused by multi-dimensional control coupling. The independent generation of leveling commands for the work platform and flight commands for the UAV ensures that the platform remains level during UAV movement, solving the problem of "body tilt causing platform imbalance" in traditional linkage control. Based on the current attitude angular velocity and acceleration, the motor speed is adjusted in real time. By superimposing the reference speed and compensation amount, trajectory deviations caused by airflow disturbances, load changes, and other factors are dynamically offset. Multi-source data fusion (GPS+LiDAR) further corrects position errors, ensuring that the UAV always moves along the corrected path as it approaches the ground wire, ensuring sub-centimeter-level trajectory tracking accuracy and avoiding collision risks caused by path deviations. By dynamically adjusting the joint angles of the robotic arm, the platform tilt caused by UAV movement is compensated in real time, ensuring that the platform remains level (attitude deviation ≤0.5°) during UAV pitch and turn. This adaptive leveling mechanism solves the measurement errors or misalignment of work tools caused by changes in the work platform's body attitude, ensuring the effective working posture of the inspection equipment. Distance threshold-triggered deceleration control gradually reduces the flight speed and leveling amplitude as the UAV approaches the ground wire, avoiding vibrations or overshoot caused by high-speed or large movements near the target. This "faster at a distance, slower at a distance" control logic balances landing efficiency and end-point accuracy, ensuring smooth movement during the critical phase within 1 meter of the ground wire. A mechanism that continuously meets hovering conditions (e.g., deviation ≤ 3mm for 10 consecutive cycles) ensures the reliability of stable hovering results. This redundant verification method effectively filters instantaneous measurement noise, avoiding hovering failures due to misjudgments, ultimately achieving sub-millimeter-level stable positioning of the work platform at the designated location on the ground wire.

[0197] The above description represents the preferred embodiments of the present invention. It should be noted that, for those skilled in the art, various improvements and modifications can be made without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for rapid cable lowering control in an overhead power transmission cable inspection robot, applied to the inspection robot, characterized in that, The method includes: The airborne lidar sensor collects three-dimensional point cloud data of the target power transmission cable area; based on the three-dimensional point cloud data, the three-dimensional spatial coordinates of the conductor and the distribution of nearby obstacles are identified; based on the identified conductor position and obstacle distribution information, the initial landing path for guiding the UAV carrying the operation platform to approach the target conductor is calculated and generated. Based on the initial wire drop path, surface texture feature images of the conductor and ground wire are acquired; based on the texture feature images, geometric contour features of the conductor are extracted; the extracted geometric contour features are then subjected to multimodal feature matching and data fusion with the spatial coordinates of the conductor and ground wire obtained from LiDAR point clouds to obtain the fusion result; based on the fusion result, pose deviations existing in the initial wire drop path are dynamically calibrated to generate a corrected wire drop path, including: Based on the heading and pitch angle control parameters of the initial drop path, the shooting attitude of the airborne high-definition industrial camera is adjusted, and the surface texture feature image of the conductor ground wire is acquired in real time during the UAV's flight along the initial drop path. The acquired texture feature image is preprocessed with grayscale conversion and Gaussian filtering to eliminate uneven lighting and image noise interference, outputting an enhanced binarized image. The binarized image is processed using the Canny edge detection algorithm to extract the continuous geometric contour edge pixel set of the conductor ground wire surface. Based on the geometric contour edge pixel set, the curvature equation of the conductor centerline is fitted using the least squares method, and the boundary direction vector is calculated, outputting geometric contour feature parameters including the radius of curvature and normal vector. The local coordinate system of the surface texture of the ground plane is established using the normal vector and radius of curvature in the contour feature parameters. A global spatial coordinate system is constructed based on the three-dimensional coordinate parameters of the equation of the spatial central axis of the ground plane. The geometric contour feature parameters are mapped from the local coordinate system to the global spatial coordinate system through a spatial projection transformation matrix to generate a standardized contour feature aligned with the spatial coordinate scale of the LiDAR ground plane. The difference in radius of curvature and the angle between the normal vector and the spatial central axis of the LiDAR ground plane at corresponding positions are calculated. When the difference exceeds a threshold, it is marked as a pose deviation region. The curvature deviation and direction offset of all pose deviation regions are aggregated to output a fusion result containing the spatial pose error vector. Based on the corrected landing path, the system acquires electromagnetic field strength data monitored by the airborne electromagnetic sensors in real time. Based on the electromagnetic field strength data, it cancels the attitude deviation of the UAV and the work platform caused by electromagnetic interference in real time to obtain the control signal after canceling the attitude deviation. Based on the corrected landing path and the control signal after canceling the attitude deviation, it drives the UAV and the work platform to perform landing control actions until the work platform achieves stable hovering and positioning at the designated position of the target ground wire.

2. The rapid cable-dropping control method for an overhead power transmission cable inspection robot according to claim 1, characterized in that, Three-dimensional point cloud data of the target power transmission cable area were collected using an airborne lidar sensor. Based on 3D point cloud data, identify the 3D spatial coordinates of the ground wire and the distribution of nearby obstacles, including: Voxel mesh filtering downsampling is performed on the raw 3D point cloud data acquired by LiDAR to eliminate environmental scattering noise points, and the noise-reduced point cloud dataset is output. Based on the denoised point cloud dataset, the ground wire point cloud clusters are separated; the spatial center axis of the ground wire point cloud clusters is fitted by the RANSAC straight line fitting algorithm to generate the spatial center axis equation of the ground wire and its three-dimensional coordinate parameters. Based on the separated non-conductor ground wire point cloud clusters, obstacle point cloud clusters are identified using the DBSCAN density clustering algorithm. Combining the three-dimensional coordinate parameters of the conductor ground wire spatial central axis equation, the minimum Euclidean distance between the outer edge of each obstacle point cloud cluster and the conductor ground wire central axis is calculated, and the spatial position and distance data of the obstacles are output.

3. The rapid cable-dropping control method for an overhead power transmission cable inspection robot according to claim 1, characterized in that, Based on the identified location of the ground wire and the distribution of obstacles, an initial landing path is calculated and generated to guide the UAV carrying the work platform to approach the target ground wire, including: Based on the spatial location and distance data of obstacles, and taking the center axis of the grounding wire as the baseline, combined with the minimum Euclidean distance of each obstacle, a cylindrical collision-free safety corridor with the baseline as the center axis and a radial safety radius of ≥1.5 meters is generated. Within the cylindrical collision-free safety corridor space, a three-dimensional path node sequence is set at equal intervals of 2-5 meters along the movement direction of the UAV approaching the ground wire. Using the path node sequence as control points, a continuous and smooth spatial trajectory curve is calculated using a cubic spline interpolation algorithm to generate an initial landing path containing the UAV's position coordinates, heading angle, and pitch angle control parameters.

4. The rapid cable-dropping control method for an overhead power transmission cable inspection robot according to claim 3, characterized in that, Based on the fusion results, the pose deviations in the initial line-falling path are dynamically calibrated to generate a corrected line-falling path, including: Based on the spatial pose error vector, the real-time deviations of the UAV in heading angle, pitch angle and radial distance are obtained by decomposition. The real-time deviation is input into a preset PID controller to generate position correction parameters for the three-dimensional path node sequence in the initial falling path; Based on the position correction parameters, the spatial coordinates of the three-dimensional path node sequence are adjusted in real time, and the corrected path node sequence is output. Using the corrected path node sequence as new control points, the spatial trajectory curve is recalculated using a cubic spline interpolation algorithm to generate a corrected landing path that includes the calibrated position coordinates, heading angle, and pitch angle control parameters.

5. The rapid cable-dropping control method for an overhead power transmission cable inspection robot according to claim 4, characterized in that, Based on the corrected trajectory, the electromagnetic field strength data monitored by the airborne electromagnetic sensors is acquired in real time. Based on electromagnetic field strength data, the attitude shift of the UAV and operating platform caused by electromagnetic interference is counteracted in real time to obtain the control signal after attitude shift is counteracted, including: Based on the heading and pitch angle control parameters of the corrected landing path, the raw data of electromagnetic field intensity along the X-axis, Y-axis or Z-axis monitored by the airborne triaxial electromagnetic sensor are acquired synchronously at a preset sampling period during the UAV's flight along the corrected path. Perform a sliding window weighted average filter on the raw electromagnetic field intensity data to output smoothed electromagnetic field intensity time series data; High-frequency oscillation components were extracted from the smoothed electromagnetic field intensity time series data, and periodic pose shift characteristics caused by the alternating magnetic field of the transmission line were identified through spectrum analysis. Based on the pose offset characteristics, a mapping model is established between electromagnetic interference intensity and UAV heading angle deviation, pitch angle deviation and radial displacement deviation. Based on the mapping relationship model, the minimum mean square adaptive filter is used to dynamically calculate the real-time offset that electromagnetic interference needs to compensate for in terms of heading angle, pitch angle and radial distance. The real-time offset is superimposed on the heading angle, pitch angle, and position coordinate control parameters of the correct landing path to generate an anti-interference control signal that compensates for the attitude offset.

6. The rapid cable-dropping control method for an overhead power transmission cable inspection robot according to claim 5, characterized in that, Based on the corrected landing path and the control signals after offsetting the pose deviation, the UAV and the work platform are driven to perform landing control actions until the work platform achieves stable hovering and positioning at the designated position of the target ground wire, including: The heading angle, pitch angle and position coordinate parameters in the anti-interference control signal are analyzed to generate UAV flight control commands and independent drive commands for the leveling mechanism of the work platform; Based on the UAV flight control commands, the speed adjustment of the quadcopter motor is calculated in real time, driving the UAV to move along the radial target ground line of the corrected landing line, and outputting the UAV's real-time attitude data. Based on drive commands, the joint angles of the robotic arm are dynamically adjusted to keep the work platform in a horizontal position during the movement of the drone, and the platform's real-time attitude data is output. In response to the real-time pose data of the UAV, when the distance between the UAV and the target ground wire is less than a preset threshold, a progressive deceleration control strategy is initiated based on the difference between the real-time position coordinates and the specified position of the ground wire, dynamically reducing the flight speed of the UAV and the amplitude of the platform leveling action. By integrating real-time UAV pose data with real-time platform attitude data, and combining airborne inertial measurement data and visual feedback, the pose deviation of the platform relative to the ground wire is calculated; when the pose deviation continuously meets the preset hovering conditions, a stable hovering judgment result is output.

7. The rapid cable-dropping control method for an overhead power transmission cable inspection robot according to claim 6, characterized in that, The heading angle, pitch angle, and position coordinate parameters in the anti-interference control signal are analyzed to generate independent drive commands for the UAV flight control command and the leveling mechanism of the work platform, including: Based on the position coordinate parameters in the anti-interference control signal, calculate the displacement control quantities of the UAV in three-dimensional space along the X, Y, and Z axes. Based on the heading and pitch angle parameters, the yaw and pitch angle control quantities of the UAV are generated respectively through the attitude decoupling algorithm. By integrating displacement control, yaw angle control, and pitch angle control, a UAV flight control command containing the reference speed of the quadcopter motor is generated. Extract the pitch angle parameter from the anti-interference control signal, calculate the extension and retraction of the hydraulic outriggers or the rotation angle of the servo motor of the leveling mechanism of the work platform, and generate independent drive commands to maintain the horizontal attitude of the platform.

8. The rapid cable-dropping control method for an overhead power transmission cable inspection robot according to claim 7, characterized in that, Based on the UAV flight control commands, the quadcopter motor speed adjustment is calculated in real time to drive the UAV to move along the radial target ground guide line of the corrected landing path, and the real-time UAV attitude data is output, including: Based on the reference speed of the quadcopter motor in the UAV flight control command, and combined with the current attitude angular velocity and acceleration data of the UAV, the speed compensation of each motor is dynamically calculated. The base speed and the speed compensation amount are superimposed to generate a real-time drive signal for the quadcopter motor, which controls the UAV to move along the spatial trajectory of the corrected landing line path. The system collects real-time three-dimensional position, heading angle, and pitch angle data of the UAV, integrates GPS positioning data and LiDAR ranging data, and outputs real-time UAV attitude data including position coordinates and attitude angles.

Images