Unmanned aerial vehicle power corridor disaster body identification method based on POS auxiliary geometric correction

Abstract

This invention discloses a method for identifying power corridor hazards using unmanned aerial vehicles (UAVs) based on POS-assisted geometric correction. Relating to the field of UAV technology, the method includes: generating a 3D point cloud set of the power corridor based on the original point cloud data of the power corridor in the UAV sensor coordinate system and the UAV's POS data, wherein the POS data includes a position, attitude, and rotation matrix and a position vector; determining the neighborhood point cloud of the guy wire anchor piles in the power corridor based on the 3D coordinates of the guy wire anchor piles and the 3D point cloud set of the corridor; and generating a slope point cloud set and multiple unstable rock block point cloud clusters based on the neighborhood point cloud of the guy wire anchor piles. This invention improves the reliability of hazard identification.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) technology, and in particular to a method for identifying power corridor hazards using UAVs based on POS-assisted geometric correction. Background Technology

[0002] In the construction of power corridors in mountainous areas, to improve the stability of transmission tower structures in undulating terrain, guyed tower structures with multiple inclined guy wires are commonly used. This transfers the tower's stress to anchor piles located on steep slopes or rocky ridges via the guy wires, thereby achieving high pull-out resistance and anchoring stability. The slopes surrounding the guyed anchor piles are often characterized by complex terrain. Due to the combined effects of long-term natural weathering, rainfall erosion, freeze-thaw cycles, and human disturbance, large isolated boulders often form on the slope surface. Even without obvious deformation or crack expansion, these boulders can still roll down under extreme rainfall, strong earthquakes, or soil disturbance. If a rockfall impacts the guy wire or guyed anchor pile, it can easily induce major structural disasters such as tower tilting or guy wire detachment. However, actual slopes often have small-scale micro-topographical structures such as steps, bumps, and rock ridges, which directly affect the stability and roll direction of unstable rock blocks. Existing image reconstruction and terrain interpolation methods often fail to properly handle these structures due to sparse point clouds or resampling operations, resulting in weakened details or even loss of structures, thus misleading the subsequent disaster identification process based on geometric quantities.

[0003] Existing methods for identifying unstable rock masses primarily rely on the geometric relationship between the centroid of the unstable rock mass and the slope position in point clouds, such as elevation difference, slope difference, or projection distance. However, when narrow steps are smoothed out during terrain reconstruction, the actual support position of the unstable rock mass is weakened or distorted. Its position in the point cloud may appear overhanging, tilted, or offset along the slope, thus contradicting the actual support relationship. In this case, relying solely on elevation judgment or slope projection relationship is insufficient to accurately reconstruct the true state of the unstable rock mass, easily leading to misjudgments of its stability or roll-off path, resulting in a serious underestimation or overestimation of the threat level to guyed anchor piles. Especially in high-risk areas near guyed towers, failure to accurately identify blocks in a temporarily stable state but with a tendency to roll off may cause inspection results to deviate from the actual risk situation. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies, such as difficulty in accurately reconstructing the true state of unstable rock blocks and easy misjudgment of their stability or rolling path, and to propose a UAV-based power corridor disaster identification method based on POS-assisted geometric correction.

[0005] To address the problems existing in the prior art, the present invention adopts the following technical solution: A method for identifying power corridor hazards using unmanned aerial vehicles (UAVs) based on POS-assisted geometric correction includes: S1. Based on the original point cloud data of the power corridor in the UAV sensor coordinate system and the UAV's POS data, generate a 3D point cloud set of the power corridor, where the POS data includes the position, attitude, rotation matrix and position vector. S2. Based on the three-dimensional coordinates of the guy wire anchor piles in the power corridor and the three-dimensional point cloud set of the corridor, determine the neighborhood point cloud of the guy wire anchor piles. S3. Based on the neighborhood point cloud of the guy wire anchor pile, generate a set of slope point clouds and multiple dangerous rock block point cloud clusters; S4. Calculate the equivalent radius of the dangerous rock block based on the centroid coordinates of the point cloud cluster of dangerous rock blocks, and calculate the suspension index of the dangerous rock block based on the equivalent radius of the dangerous rock block. S5. Based on the three-dimensional coordinates of the guy wire anchor pile and the centroid coordinates of the point cloud cluster of the unstable rock block, calculate the relative displacement of the unstable rock block along the slope direction. S6. Calculate the rolling risk index of the dangerous rock block based on the slope point cloud set, the suspension index of the dangerous rock block, and the relative displacement of the dangerous rock block along the slope direction. S7. Based on the centroid coordinates and roll-off risk index of the unstable rock blocks, generate a disaster identification record for the power corridor.

[0006] Preferably, based on the original point cloud data of the power corridor in the UAV sensor coordinate system and the UAV's POS data, a 3D point cloud set of the power corridor is generated, including: Acquire raw point cloud data of the power corridor in the UAV sensor coordinate system; Obtain the position, attitude, and rotation matrix of the UAV during flight; Obtain the position vector of the drone during its flight; Perform coordinate rotation transformation on the position and attitude rotation matrix and the original point cloud data to obtain the point cloud data after coordinate rotation transformation; The point cloud data and position vector after coordinate rotation transformation are transformed by coordinate translation to obtain three-dimensional point cloud data points in the world coordinate system. By collecting the 3D point cloud data points in the world coordinate system at all times, a 3D point cloud set of the power corridor is obtained.

[0007] Preferably, based on the three-dimensional coordinates of the guy wire anchor piles in the power corridor and the three-dimensional point cloud set of the corridor, the neighborhood point cloud of the guy wire anchor piles is determined, including: Obtain the three-dimensional coordinates of guy wire anchor piles in the power corridor in the world coordinate system; The Euclidean distance between each 3D point cloud data point in the corridor 3D point cloud set and the 3D coordinates of the guy wire anchor pile is calculated to obtain the first Euclidean distance; If the first Euclidean distance is less than or equal to the preset spatial distance threshold, then the three-dimensional point cloud data point corresponding to the first Euclidean distance is used as the neighborhood point cloud of the guy wire anchor pile.

[0008] Preferably, based on the neighborhood point cloud of the guy wire anchor piles, a slope point cloud set and multiple unstable rock block point cloud clusters are generated, including: Based on the known geometric range of the artificial components, the point cloud of the neighboring area of ​​the guy wire anchor pile is processed to remove the artificial component point cloud, resulting in a natural slope-rock mass point cloud set. In the natural slope-rock mass point cloud set, slope fitting sample points are extracted from the three-dimensional point cloud data points that are close to the guy wire anchor piles and have low height, to obtain a slope fitting point cloud subset. The least squares fit is performed on the subset of the slope fitting point cloud to obtain the slope fitting plane equation; Based on the slope fitting plane equation, the elevation difference of each three-dimensional point cloud data point in the natural slope-rock mass point cloud set is calculated to obtain the elevation deviation. If the elevation deviation is greater than the elevation resolution of the sensor, the three-dimensional point cloud data point corresponding to the elevation deviation is marked as a protrusion point; otherwise, the three-dimensional point cloud data point corresponding to the elevation deviation is marked as a slope point. Spatial adjacency clustering of the protruding points yields multiple clusters of unstable rock blocks. The slope points are processed into a set to obtain a slope point cloud set.

[0009] Preferably, the equivalent radius of the unstable rock mass is calculated based on the centroid coordinates of the unstable rock mass point cloud cluster, including: Calculate the centroid coordinates of the point cloud cluster of unstable rock blocks; Calculate the second Euclidean distance between each 3D point cloud data point and the centroid coordinates in the point cloud cluster of unstable rock blocks; Summing the squares of the second Euclidean distance yields the sum of squares of the Euclidean distances. The average of the sum of squares of the Euclidean distances is obtained by taking the average of the squares of the Euclidean distances. The equivalent radius of the dangerous rock block is obtained by taking the square root of the squared mean of the Euclidean distance.

[0010] Preferably, the suspension index of the unstable rock block is calculated based on its equivalent radius, including: Based on the equation parameters of the plane equation fitted to the slope, the slope normal vector is generated; Determine the slope fitting plane based on the slope fitting plane equation; Based on the slope normal vector, calculate the normal distance between the centroid coordinates of the unstable rock mass point cloud cluster and the slope fitting plane; The suspension index of the unstable rock block is calculated based on the equivalent radius and normal distance of the unstable rock block.

[0011] Preferably, based on the three-dimensional coordinates of the guy wire anchor pile and the centroid coordinates of the point cloud cluster of unstable rock blocks, the relative displacement of the unstable rock blocks along the slope direction is calculated, including: Determine the slope gradient vector based on the equation parameters of the fitted plane equation of the slope. Calculate the unit vector of the slope gradient vector; The vector subtraction operation is performed on the three-dimensional coordinates of the guy wire anchor pile and the centroid coordinates of the unstable rock mass cluster to obtain the pointing vector of the guy wire anchor pile relative to the centroid coordinates. The relative displacement of the unstable rock block along the slope direction is obtained by performing a dot product operation on the unit vector and the pointing vector of the slope gradient vector.

[0012] Preferably, the roll-off risk index of the unstable rock block is calculated based on the slope point cloud set, the suspension degree index of the unstable rock block, and the relative displacement of the unstable rock block along the slope direction, including: Perform a dot product operation between each 3D point cloud data point in the slope point cloud set and the unit vector of the slope gradient vector to obtain the slope aspect projection coordinates of each 3D point cloud data point in the slope aspect direction. The slope extension length is obtained by calculating the difference between the maximum and minimum values ​​of all slope aspect projected coordinates. The risk index of rockfall is calculated based on the suspension index of the rock block, the relative displacement of the rock block along the slope direction, and the slope extension length.

[0013] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention introduces a comprehensive calculation process based on slope fitting plane, suspension degree index, and relative displacement along the slope direction. It can effectively reconstruct the geometric relationship between unstable rock blocks and the real slope when narrow steps are smoothly hidden and unstable rock blocks show a suspended offset in three-dimensional data. This accurately characterizes whether unstable rock blocks are on the edge of instability. The invention uses point cloud to construct slope fitting plane and calculates the normal distance of unstable rock blocks to identify the true degree of dispersion between unstable rock blocks and potential support surfaces. Combined with the equivalent radius of unstable rock blocks, a quantitative measure of stability is formed, enabling the point cloud acquired by UAVs to accurately reflect the stability state of unstable rock blocks and improve the reliability of disaster identification.

[0014] 2. In this invention, by constructing a vector pointing from the centroid of the unstable rock block to the guy wire anchor pile, and combining the projection along the slope gradient direction, the relative displacement of the unstable rock block along the slope direction is obtained. This allows the three-dimensional position error of the unstable rock block caused by the disappearance of the narrow step to be corrected. The relative displacement reflects the potential movement trend of the unstable rock block when it rolls down the slope direction under actual terrain conditions. This effectively compensates for the misjudgment of the rolling path caused by the conventional practice of using a smooth slope surface to replace the real slope surface. The threat level of the unstable rock block relative to the guy wire anchor pile can be accurately recovered through an interpretable geometric method.

[0015] 3. In this invention, by normalizing and integrating the suspension index of unstable rock blocks with the relative displacement along the slope according to the slope extension length, a roll-off risk index is constructed. This enables a quantitative characterization of the risk of unstable rock blocks impacting guy wire anchor piles. The roll-off risk index can simultaneously reflect the stability of unstable rock blocks, the potential roll-off direction, and the geometric relationship with anchor piles. It does not rely on the complete preservation of micro-topography such as narrow steps. Even if narrow steps are hidden in point cloud reconstruction, it can still accurately identify unstable rock blocks that truly pose an impact threat. This improves the ability of UAV inspection to identify disaster bodies in complex mountainous environments and provides a more reliable risk assessment basis for the safe operation and maintenance of power corridors. Attached Figure Description

[0016] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings: Figure 1 This is a flowchart illustrating a method for identifying power corridor hazards using unmanned aerial vehicles (UAVs) based on POS-assisted geometric correction, according to an embodiment of the present invention. Detailed Implementation

[0017] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0018] Example: This example provides a method for identifying power corridor hazards using unmanned aerial vehicles (UAVs) based on POS-assisted geometric correction. See [link / reference]. Figure 1 Specifically, including: S1. Based on the original point cloud data of the power corridor in the UAV sensor coordinate system and the UAV's POS data, generate a 3D point cloud set of the power corridor, where the POS data includes the position, attitude, rotation matrix and position vector. In an embodiment of the present invention, a three-dimensional point cloud set of the power corridor is generated based on the original point cloud data of the power corridor in the UAV sensor coordinate system and the UAV's POS data, including: Acquire raw point cloud data of the power corridor in the UAV sensor coordinate system; Specifically, the drone is equipped with a 3D imaging device such as a lidar sensor or a depth imaging sensor. It flies along a pre-planned power corridor inspection route. The device uses the drone's own sensor coordinate system as a reference and continuously emits laser beams or depth imaging beams to vegetation, slopes, rocks, artificial structures and other ground features in the power corridor area. At the same time, it receives the return signals reflected from the ground features. Based on the flight time, phase difference and other information of the signals, it calculates the X, Y and Z 3D coordinates of each ground feature point relative to the sensor coordinate system. The 3D coordinate information of all ground feature points is recorded and stored in real time, and finally a set of these spatially scattered measurement points is formed. This is used to obtain the original point cloud data of the power corridor in the drone's sensor coordinate system.

[0019] Specifically, a power corridor refers to a continuous operating channel for power transmission lines in the surface space. This channel is mainly composed of transmission towers, conductors, guy wire anchoring foundations, and the necessary safety clearance around them. It is defined by the spatial orientation of the transmission line and the terrain below the line, and is used to ensure the safety distance requirements of high-voltage transmission lines during operation, maintenance, and inspection. Power corridors typically cover the surface area under and to both sides of the conductors, including various complex environmental elements such as vegetation, slopes, rock masses, and man-made structures. Its spatial morphology directly affects the safe operation of the transmission lines; therefore, this area needs to be a key monitoring target during drone inspections and hazard identification.

[0020] Obtain the position, attitude, and rotation matrix of the UAV during flight; Specifically, the UAV is equipped with a combined navigation system consisting of a Global Navigation Satellite System (GNSS) receiver and an inertial measurement unit (IMU). During its flight along the power corridor inspection route, the GNSS receiver receives positioning signals from multiple satellites in real time, calculating the UAV's real-time position and velocity in the world coordinate system. Simultaneously, the IMU collects the UAV's angular velocity and linear acceleration data in three-dimensional space. The position and velocity information output from the GNSS receiver and the angular velocity and acceleration data output from the IMU are input into a pre-defined Kalman filter data fusion algorithm. This algorithm then processes the two types of data... Time synchronization calibration and error compensation are performed to eliminate random and cumulative errors generated during single-device measurement. Then, the roll, pitch, and yaw angles of the UAV relative to the world coordinate system at any flight moment are calculated. Based on these three attitude parameters, rotation sub-matrices around the X, Y, and Z axes of the world coordinate system are constructed respectively. The three rotation sub-matrices are operated according to a preset matrix multiplication order to obtain a matrix that can completely describe the rotation state of the UAV in three-dimensional space. This matrix is ​​the position and attitude rotation matrix of the UAV during flight, which is used for subsequent orientation correction of point cloud data from the sensor coordinate system to the world coordinate system.

[0021] Obtain the position vector of the drone during its flight; Specifically, before the UAV performs its power corridor inspection mission, a Global Navigation Satellite System (GNSS) reference station is deployed in an area with a wide field of vision and no obstructions around the power corridor. This reference station, together with the GNSS mobile receiver on the UAV, forms a real-time dynamic positioning system. The reference station continuously receives signals from multiple satellite constellations and transmits its precise world coordinate system position information and satellite observation data to the UAV in real time. During the UAV's flight along the preset inspection route, its onboard GNSS mobile receiver simultaneously receives satellite signals and differential data transmitted from the reference station. These two types of data are input into the receiver's built-in carrier phase differential positioning algorithm. This algorithm calculates errors such as ionospheric delay, tropospheric delay, satellite clock bias, and receiver clock bias during satellite signal propagation, while also combining the data with the UAV's flight data. The real-time motion status is used to dynamically correct the positioning results. In addition, the inertial measurement unit on the UAV collects the motion acceleration and angular velocity data of the body in real time. This data is synchronized and fused with the positioning data output by the mobile receiver of the Global Navigation Satellite System. This further eliminates the instantaneous errors and positioning interruption problems caused by signal loss that may occur in a single positioning method. Finally, the three-dimensional coordinates of the UAV body in the world coordinate system are calculated at each flight moment. These three-dimensional coordinates correspond to the displacement components in the X, Y, and Z directions of the world coordinate system. The vector formed by these three displacement components is the position vector of the UAV during flight. This position vector is used to perform translation correction on the point cloud data after rotation transformation, so that the point clouds collected at different times can be unified in the world coordinate system.

[0022] Specifically, the raw point cloud data of the power corridor in the UAV sensor coordinate system represents a set of scattered spatial measurement points directly collected by the 3D imaging sensor on the UAV under its own coordinate reference. Each point in this set corresponds to the 3D position information returned after the corridor target surface is illuminated by a laser or depth imaging beam, which is used to reflect the original spatial shape of the ground object. The position and attitude rotation matrix of the UAV during flight represents the instantaneous attitude state of the UAV body relative to the world reference direction. This matrix is ​​formed by combining the rotation angles in three directions and is used to describe the rotation relationship of the body in three-dimensional space and provide a basis for orientation correction for the point cloud to be transformed from the sensor coordinate system to the world coordinate system. The position vector of the UAV during flight represents the spatial position of the UAV body in the world coordinate system. It is composed of displacement components in three directions and is used to provide a unified translation reference for the rotated and corrected point cloud, so that the point cloud at all times can be accurately projected onto a unified world coordinate framework, thereby forming a real 3D structure of the corridor.

[0023] Perform coordinate rotation transformation on the position and attitude rotation matrix and the original point cloud data to obtain the point cloud data after coordinate rotation transformation; The point cloud data and position vector after coordinate rotation transformation are transformed by coordinate translation to obtain three-dimensional point cloud data points in the world coordinate system. Specifically, the raw point cloud data of the power corridor in the UAV sensor coordinate system is first timestamped with the position and attitude rotation matrix recorded during the UAV's flight. This ensures that each set of raw point cloud data acquisition times corresponds to the same position and attitude rotation matrix at that time. Each 3D point in the raw point cloud data is represented by a 3D column vector composed of X, Y, and Z axis coordinate components in the sensor coordinate system. Then, for each 3D point in the raw point cloud data, its corresponding 3D column vector is used as the multiplicand, and the matched position and attitude rotation matrix is ​​used as the left multiplication matrix. Matrix multiplication is performed point by point to eliminate the directional deviation caused by the UAV's body rotation on the raw point cloud coordinates, thus obtaining the 3D coordinate vector of each 3D point in the transition coordinate system, completing the coordinate rotation transformation. The system obtains point cloud data after coordinate rotation transformation. Then, it timestamps the point cloud data after coordinate rotation transformation with the position vector recorded during the UAV flight to ensure that each 3D point in the transition coordinate system corresponds to a position vector at the same moment. This position vector is composed of 3D column vectors with X, Y, and Z axis displacement components in the world coordinate system. Afterward, for each 3D point in the point cloud data after coordinate rotation transformation, it performs vector addition on its 3D coordinate vector in the transition coordinate system and the matched position vector. This operation translates the point cloud coordinates in the transition coordinate system to the world coordinate system. After completing this operation point by point, each 3D point obtains the X, Y, and Z axis coordinate values ​​in the world coordinate system, thus obtaining the 3D point cloud data points in the world coordinate system.

[0024] By collecting the 3D point cloud data points in the world coordinate system at all times, a 3D point cloud set of the power corridor is obtained.

[0025] Specifically, the aggregation of 3D point cloud data points in the world coordinate system at all times includes: During the execution of the UAV inspection mission, after each coordinate rotation and translation transformation of the point cloud, the airborne processing module stores the 3D point cloud data points obtained at the current moment into the buffer data area in chronological order; As the UAV continues to fly along the power corridor, batches of point clouds from different flight positions and observation angles are continuously written into the same data area. The data management program records the time label, point range, and spatial distribution of each batch of point clouds to avoid data overwriting and omission; After the data acquisition is completed, the point cloud aggregation module reads all the stored point cloud batches sequentially and merges each batch of point clouds into a unified point set under a unified coordinate system through spatial stitching. At the same time, duplicate points and abnormal offset points are removed or compressed so that the merged point cloud can completely cover the spatial information of the power corridor area, such as towers, conductors, slopes, and ground surfaces. Finally, a 3D point cloud set with a continuous spatial structure is generated for the corridor, providing a unified input for subsequent disaster body identification calculations.

[0026] S2. Based on the three-dimensional coordinates of the guy wire anchor piles in the power corridor and the three-dimensional point cloud set of the corridor, determine the neighborhood point cloud of the guy wire anchor piles. In an embodiment of the present invention, the neighborhood point cloud of the guy wire anchor piles is determined based on the three-dimensional coordinates of the guy wire anchor piles in the power corridor and the three-dimensional point cloud set of the corridor, including: Obtain the three-dimensional coordinates of guy wire anchor piles in the power corridor in the world coordinate system; Specifically, guy wire anchor piles are ground-based load-bearing foundation components used to fix the guy wires of transmission towers. They are made of concrete or steel and are buried below or above the ground surface to withstand the continuous tension generated by the guy wires under wind loads, conductor tension, and tower attitude changes, so that the towers remain stable and do not tilt within the power corridor. The position and shape of the guy wire anchor piles directly determine the direction of force on the guy wires in space and the overall stability of the tower structure. Therefore, during UAV inspections, they should be regarded as key spatial reference points. By identifying changes in the surrounding terrain slopes and protrusions, it can be used to help determine whether there are any hazards that may affect the safe operation of transmission lines.

[0027] The Euclidean distance between each 3D point cloud data point in the corridor 3D point cloud set and the 3D coordinates of the guy wire anchor pile is calculated to obtain the first Euclidean distance; Specifically, the three-dimensional coordinates of the guy wire anchor piles in the world coordinate system are first obtained through preliminary on-site surveys or power corridor design drawings. These three-dimensional coordinates include coordinate values ​​in the X, Y, and Z axes. Simultaneously, the consistency of these three-dimensional coordinates with the world coordinate system used in the corridor's three-dimensional point cloud set is verified to ensure that the coordinate references are identical, avoiding distance calculation errors due to coordinate system differences. Then, the corridor's three-dimensional point cloud set is loaded into the point cloud data processing platform. The data format of all three-dimensional point cloud data points in the set is parsed, and the X, Y, and Z axis coordinate values ​​corresponding to each three-dimensional point cloud data point are extracted. The extracted coordinate values ​​are then effectively processed. A screening process is performed to remove 3D point cloud data points with abnormal coordinate values ​​due to data storage errors or transmission interference, ensuring the accuracy and reliability of the coordinates of the point cloud data points used in distance calculations. Next, an Euclidean distance calculation module is built into the data processing platform. This module uses the 3D coordinates of the guy wire anchor piles as fixed reference points and the coordinates of each valid 3D point cloud data point in the corridor's 3D point cloud set as the target point coordinates. It performs calculations on the target point coordinates against the reference point coordinates one by one, first calculating the difference between the X-axis coordinates of the target point and the X-axis coordinates of the reference point, the difference between the Y-axis coordinates of the target point and the Y-axis coordinates of the reference point, and the difference between the Z-axis coordinates of the target point and the Z-axis coordinates of the reference point. The differences in the three directions are then squared separately, and the results are summed to obtain the sum of squared differences. Finally, the square root of the sum of squared differences is taken to obtain the straight-line distance between a single 3D point cloud data point and the 3D coordinates of the guy wire anchor pile. To improve processing efficiency during point-by-point calculation, a batch parallel calculation method is adopted. The corridor 3D point cloud set is divided into multiple data batches according to a preset point cloud quantity threshold. Each data batch is calculated independently using Euclidean distance. After the calculation is completed, the results of each batch are summarized. Simultaneously, to ensure the accuracy of the calculation results, after each batch is completed, a certain number of data points are randomly selected from that batch. For the 3D point cloud data points, manually recalculate the Euclidean distance between them and the 3D coordinates of the guy wire anchor piles, and compare the recalculated results with the module calculation results. If the comparison deviation is less than the preset error allowable threshold, the calculation results of this batch are deemed valid. If the deviation exceeds the threshold, the batch of data is recalculated. After all the 3D point cloud data points have completed the Euclidean distance calculation, the calculation results corresponding to each 3D point cloud data point are marked. The distance value after marking is the first Euclidean distance. At the same time, a correlation data table between the coordinates of the 3D point cloud data points and the first Euclidean distance is established so that the neighboring point cloud of the guy wire anchor pile can be selected based on the first Euclidean distance.

[0028] If the first Euclidean distance is less than or equal to the preset spatial distance threshold, then the three-dimensional point cloud data point corresponding to the first Euclidean distance is used as the neighborhood point cloud of the guy wire anchor pile.

[0029] Specifically, the three-dimensional coordinates of the guyed anchor piles are used to represent the actual fixed position of the guyed foundation components in the ground space. The coordinates are calculated by the positioning system carried by the UAV and can serve as spatial reference points for identifying slopes and unstable rock blocks. The three-dimensional point cloud data points in the corridor's three-dimensional point cloud set represent discrete spatial measurement points formed by reflected echoes from ground objects collected by airborne sensors. Each point contains positional information in a unified world coordinate system, used to reflect the actual shape of the terrain and structures in the corridor area. The first Euclidean distance is used to measure the straight-line spatial interval of the measurement point relative to the guyed anchor pile. By calculating this interval, it can be determined whether the measurement point is near the anchoring area. The preset spatial distance threshold is used to limit the scale of the neighborhood range. This threshold is determined based on the ranging capability of the sensor and the actual size of the guyed foundation. It is used to filter out point clouds that are close to the anchor pile, thereby constructing a neighborhood point cloud set distributed around the anchor pile.

[0030] Specifically, when the first Euclidean distance is less than or equal to the preset spatial distance threshold, the corresponding 3D point cloud data point is used as the neighboring point cloud of the guy wire anchor pile. This is because the guy wire anchor pile is a local structure in the power corridor, and its surrounding surface morphology directly reflects the slope conditions, rock mass distribution, and the location of potential hazards. Therefore, it is necessary to select measurement points that are spatially adjacent to the anchor pile from the entire corridor's 3D point cloud set to construct a local topographic point cloud. The preset spatial distance threshold is set based on the actual size of the anchor pile, the range of construction disturbance, and the ranging accuracy of the sensor. It is used to limit the actual stress zone around the anchor pile and the effective spatial range where slope changes may occur. When the distance between the point cloud data point corresponding to the first Euclidean distance and the anchor pile does not exceed the threshold, it indicates that the measurement point is located on the surface or in an adjacent structural area near the anchor pile. Its spatial information directly contributes to the subsequent identification of slope fitting sample points, removal of artificial component point clouds, and extraction of dangerous protrusions. Therefore, such points need to be included in the neighboring point cloud set to ensure the integrity and identification accuracy of the local geometry of the anchoring area.

[0031] Specifically, the generation of the preset spatial distance threshold includes: First, during the task deployment phase, the actual structural dimensions of the guy wire anchor piles and the stress diffusion range formed on the ground surface are obtained based on the power line design data. The scale of the potentially affected surface area around the anchor piles is determined by analyzing the anchor pile burial depth, guy wire length, and construction disturbance zone width. Then, the minimum coverage radius for ensuring the integrity of the neighboring point cloud is determined by combining the ranging accuracy, scanning angle range, and point cloud density requirements of the UAV onboard sensor. Next, the data processing module compares the anchor pile stress zone scale with the sensor coverage radius and selects the larger value that can simultaneously satisfy structural area coverage and point cloud continuity as the base radius. Finally, by conducting test flight point cloud analysis on typical corridor scenarios, the base radius is verified to effectively distinguish between anchor-related terrain and irrelevant surface points far from the anchor piles. Based on the verification results, the base radius is fine-tuned so that it is neither too large, causing noise points to enter the neighborhood set, nor too small, causing slope sample points to be missing, thus forming a spatial distance threshold suitable for this method.

[0032] Specifically, the neighborhood point cloud of the guy wire anchor piles is determined based on the 3D coordinates of the guy wire anchor piles in the power corridor and the 3D point cloud set of the corridor. This is because, in the mountainous guy wire tower scenario, whether a rockfall might collide with the guy wire anchor pile depends on the local slope morphology, support structure, and spatial geometric relationship along the slope of the rockfall relative to the anchor pile. These key geometric clues are concentrated in the local area around the anchor pile. Therefore, it is necessary to first extract the nearest terrain points around the anchor pile from the large-scale 3D point cloud of the corridor to construct a realistic geometric model of the slope where the anchor pile is located. Narrow steps are reconstructed due to the sparse point cloud and smooth terrain. The data is obscured, causing the unstable rock blocks to appear suspended and offset. Therefore, it is necessary to extract the point cloud of the neighborhood based on the location of the anchor pile in order to capture the micro-scale elevation undulations of the area and distinguish the geomorphic details hidden by the smooth slope. By establishing a local point cloud subset centered on the anchor pile, the subsequent slope fitting, protrusion clustering, and relative geometric calculation of the centroid of the unstable rock block and the slope can be limited to the spatial area directly related to the safety of the guy wire stress. This avoids interference from irrelevant terrain far from the anchor pile, thereby ensuring that the support relationship of the unstable rock block in the actual terrain and its roll-off risk relative to the anchor pile can be accurately restored during the identification of the disaster body.

[0033] S3. Based on the neighborhood point cloud of the guy wire anchor pile, generate a set of slope point clouds and multiple dangerous rock block point cloud clusters; In an embodiment of the present invention, based on the neighborhood point cloud of the guy wire anchor pile, a slope point cloud set and multiple unstable rock block point cloud clusters are generated, including: Based on the known geometric range of the artificial components, the point cloud of the neighboring area of ​​the guy wire anchor pile is processed to remove the artificial component point cloud, resulting in a natural slope-rock mass point cloud set. Specifically, the point cloud removal process for the neighboring point cloud of the guy wire anchor pile, based on the known geometric range of the artificial components, includes: First, the structural information management module loads the known geometric parameters such as the external dimensions, embedment depth, and foundation range of the guy wire anchor pile and its ancillary facilities, and establishes a spatial envelope region consistent with this geometric range in the world coordinate system; then, the point cloud traversal module sequentially reads each three-dimensional point cloud data point in the neighboring point cloud, extracts the spatial coordinates of the point, and determines its inclusion relationship with the spatial envelope region. When the coordinates of the point cloud data point are within the envelope region, the point cloud data point is removed from the envelope region. When a point falls within the error buffer zone of the envelope region or the outer edge of the envelope region, it indicates that the point is a reflection caused by artificial components. This point should be marked as an artificial component point cloud and removed from the neighboring point cloud. When a point cloud data point does not meet the above inclusion conditions, it is retained as a naturally formed measurement point on the ground surface. After completing the removal and retention operations of all point clouds, the data processing module merges and sorts the retained point clouds to form a complete set of natural slope-rock mass point clouds that reflects the slope morphology and rock mass distribution, providing real and reliable basic data for subsequent slope fitting and dangerous rock block identification.

[0034] Specifically, the known geometric range of the artificial component is used to represent the fixed shape and size of the guy wire anchor pile and its surrounding ancillary facilities in space. This range is obtained from design data or on-site measurements and can be used to identify measurement points in the point cloud that belong to the artificial structure. The neighborhood point cloud of the guy wire anchor pile refers to the local point cloud set distributed around the anchor pile location. It contains mixed measurement points of the slope, rock mass, and artificial structures and is used to reflect the real surface structure of the anchoring area. The artificial component point cloud removal process is used to remove the point cloud belonging to the anchor pile body or ancillary facilities from the neighborhood point cloud according to the geometric range of the artificial component, so as to avoid the artificial structure from interfering with the slope fitting and the identification of unstable rock blocks.

[0035] In the natural slope-rock mass point cloud set, slope fitting sample points are extracted from the three-dimensional point cloud data points that are close to the guy wire anchor piles and have low height, to obtain a slope fitting point cloud subset. Specifically, the extraction of slope fitting sample points from the natural slope rock mass point cloud set for 3D point cloud data points that are close to the guy wire anchor pile and have a low height includes: First, the point cloud filtering module reads the spatial coordinates of each 3D point cloud data point in the natural slope rock mass point cloud set and calculates the horizontal distance between the point and the guy wire anchor pile. When the horizontal distance is less than the preset neighborhood radius, the point is considered as a candidate point near the anchor pile. Then, the elevation components of all candidate points are statistically analyzed to determine the elevation range of the bottom edge of the slope, and the point cloud in the bottom height range is selected as the initial low point based on the elevation distribution. Elevation samples are filtered out, removing convex points that are significantly higher than the main slope body to ensure that the fitting input points reflect the true slope baseline shape. Next, the sample verification module checks the sparsity and spatial continuity of the initial sample points, eliminating isolated and noise points, and retaining point cloud data that are continuously distributed along the slope direction, close to the slope foot, and conform to the slope trend. After the filtering is completed, all sample points that meet the distance and height conditions are sorted and summarized in spatial order to generate a subset of slope fitting point clouds for fitting the slope geometric plane, providing stable input for the subsequent least squares solution of the slope plane equation.

[0036] Specifically, the natural slope-rock mass point cloud set refers to the original surface measurement points retained after removing artificial components. This set truly reflects the natural morphology of the mountain slope and is the basis for slope modeling and unstable rock detection. The three-dimensional point cloud data points close to the anchor piles and with low height refer to the measurement points located near the anchor piles and at the lower edge of the slope. These points usually best reflect the natural extension of the bottom of the slope and are key to constructing the slope reference surface. The slope fitting sample point extraction selects representative low elevation points from the natural slope rock mass point cloud set as input for geometric fitting, thereby obtaining a subset of the slope fitting point cloud, which provides a reliable original data basis for subsequent slope plane equation fitting.

[0037] The least squares fit is performed on the subset of the slope fitting point cloud to obtain the slope fitting plane equation; Specifically, the least-squares fitting of the slope fitting point cloud subset includes: First, the data reading module sequentially acquires the spatial coordinates of each 3D point cloud data point in the slope fitting point cloud subset, and stores these coordinates into numerical operation arrays according to their lateral, longitudinal, and elevation components. Then, the plane fitting module establishes a linear parametric equation describing the slope plane, expressing the elevation of any point on the plane as a linear combination of the lateral and longitudinal components. The difference between the elevation component in the slope fitting point cloud subset and this linear combination is then used as the residual. Next, the least-squares solution unit performs a summation of squares on all residuals, and establishes a normal equation system by taking the partial derivative of the sum of squares with respect to the plane parameters. The optimal solution of the plane parameters is obtained by matrix inversion or iterative numerical methods, minimizing the sum of squares of the residuals. After the solution is completed, the plane parameters that minimize the fitting error are combined into a complete slope fitting plane equation according to a predetermined format. This equation represents the overall spatial position and tilt direction of the slope in the world coordinate system and serves as a reference datum for subsequent elevation deviation calculation and identification of protruding rock blocks.

[0038] Specifically, the slope fitting point cloud subset represents continuous low-elevation measurement points selected from the natural slope rock mass point cloud that can represent the true orientation of the slope. These points are used to reflect the basic shape of the terrain near the toe of the slope. Least square fitting means finding a plane that best approximates the distribution trend of these points in an overall sense by mathematically solving the three-dimensional coordinates of these representative points. This plane can describe the large-scale tilt direction and undulation trend of the slope, thus forming the slope fitting plane equation. The slope fitting plane equation represents the analytical expression of the fitted plane in three-dimensional space, which is used to describe the geometric position and tilt direction of the slope surface.

[0039] Based on the slope fitting plane equation, the elevation difference of each three-dimensional point cloud data point in the natural slope-rock mass point cloud set is calculated to obtain the elevation deviation. Specifically, the elevation difference calculation for each 3D point cloud data point in the natural slope rock mass point cloud set based on the slope fitting plane equation includes: First, the data reading module sequentially obtains the horizontal coordinates, vertical coordinates, and elevation coordinates of each 3D point cloud data point in the natural slope rock mass point cloud set, and calls the slope fitting plane equation obtained in the previous step, where the plane equation has given the theoretical slope elevation corresponding to any horizontal and vertical coordinates in the form of parameters a, b, and c; then, for the current 3D point cloud data point, its horizontal and vertical coordinates are substituted into the slope fitting plane equation to calculate the elevation difference. The theoretical elevation value of the point on the smooth slope is obtained and used as the slope reference elevation. Then, the difference calculation module subtracts the corresponding slope reference elevation from the actual elevation coordinates of the 3D point cloud data point to obtain the elevation deviation of the point relative to the slope fitting plane. The sign of the deviation determines whether the point is above, below, or close to the plane. After calculating the elevation deviation of one point, the above operation is repeated until all 3D point cloud data points in the natural slope rock mass point cloud set are processed. All elevation deviation information is stored in the original point cloud order to form a complete elevation deviation sequence.

[0040] Specifically, the elevation difference calculation refers to substituting each measurement point in the point cloud set of the natural slope rock mass into the fitted plane expression to calculate the vertical distance between the measurement point and the plane, which is used to reflect the undulation of the point relative to the fitted slope. The elevation deviation represents the degree of height difference of each measurement point relative to the fitted slope, which can be used to distinguish between points that are close to the slope and points that are obviously raised, thus providing a reliable basis for subsequent identification of dangerous rock blocks.

[0041] If the elevation deviation is greater than the elevation resolution of the sensor, the three-dimensional point cloud data point corresponding to the elevation deviation is marked as a protrusion point; otherwise, the three-dimensional point cloud data point corresponding to the elevation deviation is marked as a slope point. Specifically, 3D point cloud data points with elevation deviations greater than the sensor's elevation resolution are marked as protrusion points, while those with elevation deviations no greater than the sensor's elevation resolution are marked as slope points. This is because elevation resolution reflects the minimum height change that the UAV sensor can reliably distinguish in the vertical direction. When the elevation deviation is below this resolution, the deviation is likely due to measurement noise, attitude calculation errors, or minor terrain roughness, and is insufficient to represent a true geometric protrusion, thus it should not be considered an independent hazard candidate. Only when the elevation deviation exceeds the sensor's elevation resolution does it indicate that the point has a significant height increase relative to the fitted slope, possesses an independent geometric profile and volume, and can correspond to protruding structures such as rocks, steps, or deposits on the actual slope. Classifying such points as protrusion points and distinguishing them from slope points can highlight spatially anomalous areas that are meaningful for hazard identification while suppressing measurement noise interference.

[0042] Spatial adjacency clustering of the protruding points yields multiple clusters of unstable rock blocks. Specifically, the spatial adjacency clustering processing for protrusion points includes: First, the data reading module extracts all 3D point cloud data of points identified as protrusions by elevation deviation from the natural slope rock mass point cloud set, and sets a spatial adjacency distance threshold based on the power corridor point cloud resolution and the minimum size of the unstable rock block to describe the maximum allowable spacing between points within the same unstable rock block; then, the clustering processing module sequentially selects protrusion points that have not yet been assigned a category in 3D space as initial clustering seeds, searches for other protrusion points within the spatial adjacency distance threshold range of the seed point, and groups points that meet the distance condition into the same temporary point set. The process involves repeating the adjacent search process for each newly added point until no new protrusion points appear within the adjacent range, thus obtaining a spatially connected and concentrated set of protrusion point clouds. Next, this set of protrusion point clouds is marked as a candidate cluster of dangerous rock blocks. Points that have already been classified are removed from the set of protrusion points to be processed. The above steps are repeated to perform cyclic clustering on the remaining protrusion points until all protrusion points are assigned to the corresponding candidate clusters. Finally, the result processing module removes candidate clusters that are too small in scale or do not have enough points to constitute actual rock blocks, and determines the candidate clusters that meet the point number threshold and spatial scale requirements as multiple dangerous rock block point cloud clusters.

[0043] The slope points are processed into a set to obtain a slope point cloud set.

[0044] Specifically, the set construction process for slope points includes: first, traversing all 3D point cloud data points in the natural slope rock mass point cloud set; identifying the 3D point cloud data points marked as slope points based on the elevation deviation determination results from the previous step; and storing these slope points separately from the 3D point cloud data points marked as protrusion points; then, the set construction module sequentially writes all slope points into the same data structure, simultaneously recording the spatial coordinate information of the slope points in the world coordinate system and the auxiliary attribute information related to the slope fitting plane equation, ensuring subsequent projection along the slope direction. The calculation can be directly called; then the written data is sorted and organized according to the projection position or height along the slope direction, and isolated outliers that may be caused by measurement noise are removed, so that the remaining slope points are spatially continuous and completely cover the slope area around the guy wire anchor pile; after the above sorting is completed, the result output module uniformly names the sorted slope point data as a slope point cloud set, and uses the slope point cloud set as a special data object to describe the slope geometry for subsequent slope extension length calculation and roll-off risk index calculation steps.

[0045] Specifically, the sensor's elevation resolution represents the smallest reliably distinguishable change in altitude in the vertical direction by the measurement equipment on board the UAV, serving as the benchmark for determining whether elevation deviation has practical terrain significance. Protrusion points refer to point cloud measurement points whose elevation deviation exceeds the sensor's elevation resolution and is significantly higher than the fitted slope surface; these points typically correspond to rock blocks, deposits, or other protruding objects in the terrain. Slope points refer to point cloud measurement points whose elevation deviation does not exceed the sensor's elevation resolution and is close to the fitted slope surface height; these points reflect the continuous reference surface of the slope. Spatial adjacency relationships... Clustering refers to classifying point clouds that are close to each other and belong to the same rock mass or the same protrusion structure into the same category based on the spatial distance and distribution continuity between protrusion points, so as to form a point cloud cluster that completely represents the morphology of a single dangerous rock block. The dangerous rock block point cloud cluster is a collection of point clouds composed of protrusion points obtained after clustering. They represent discrete large-volume rock blocks that may roll down the slope. The slope point cloud collection is a continuous point cloud composed of all measurement points that are identified as slope points. It is used to accurately reflect the overall geometric shape of the slope and the reference slope structure required for subsequent geometric calculations along the slope direction.

[0046] Specifically, the generation of slope point cloud sets and multiple unstable rock mass point cloud clusters based on the neighborhood point cloud of the guyed anchor pile is because in the mountainous guyed tower scenario targeted by this invention, after the narrow steps are hidden in the data processing, the unstable rock masses resting on the narrow steps will appear as suspended offset states in the 3D data. If geometric analysis is performed directly on the global point cloud of the entire power corridor, it will be impossible to distinguish which points belong to the continuous smooth slope and which points belong to discrete unstable rock masses that pose a potential impact risk to the guyed anchor pile. Therefore, it is necessary to first use the neighborhood point cloud of the guyed anchor pile as the data starting point, and separate the local areas directly related to the safe stress of the guyed anchor pile from the global data, and then use elevation deviation determination and spatial clustering. The class unifies the point cloud closely attached to the fitted slope into the slope point cloud set, and divides the point cloud of protrusions that are significantly higher than the slope into multiple dangerous rock block point cloud clusters according to their spatial adjacency. Thus, under the same coordinate frame, the geometric reference of the continuous slope and the volume and position expression of discrete dangerous rock blocks are obtained respectively. This allows the subsequent calculation of the suspension index of dangerous rock blocks and the derivation of the relative displacement and roll-off risk index along the slope direction to be carried out on the real terrain around the guy wire anchor pile. It accurately reflects the actual threat relationship of dangerous rock blocks to guy wire anchor piles under the condition of narrow step concealment. At the same time, it ensures that the naming and referencing of data objects such as the slope point cloud set, dangerous rock block point cloud clusters and neighboring point clouds of guy wire anchor piles are consistent throughout the entire processing flow.

[0047] S4. Calculate the equivalent radius of the dangerous rock block based on the centroid coordinates of the point cloud cluster of dangerous rock blocks, and calculate the suspension index of the dangerous rock block based on the equivalent radius of the dangerous rock block. In an embodiment of the present invention, calculating the equivalent radius of a dangerous rock block based on the centroid coordinates of the point cloud cluster of dangerous rock blocks includes: Calculate the centroid coordinates of the point cloud cluster of unstable rock blocks; Specifically, calculating the centroid coordinates of the unstable rock mass point cloud cluster includes: First, the data reading module counts the total number of 3D point cloud data points in the unstable rock mass point cloud cluster, and uses this number as the count value for subsequent averaging; then, iterates through each 3D point cloud data point in the unstable rock mass point cloud cluster, reading the horizontal, vertical, and elevation coordinates of each point in the world coordinate system, and summing them in the three coordinate directions to form the sum of the horizontal, vertical, and elevation coordinates; after completing the traversal of all point cloud data points, the centroid solving module divides the sum of the horizontal coordinates by the total number of points to obtain the horizontal component of the centroid coordinates, divides the sum of the vertical coordinates by the total number of points to obtain the vertical component of the centroid coordinates, and divides the sum of the elevation coordinates by the total number of points to obtain the elevation component of the centroid coordinates. These three components together constitute the centroid coordinates of the unstable rock mass point cloud cluster, which represent the overall geometric center position of the unstable rock mass in three-dimensional space.

[0048] Calculate the second Euclidean distance between each 3D point cloud data point and the centroid coordinates in the point cloud cluster of unstable rock blocks; Specifically, calculating the second Euclidean distance between each 3D point cloud data point in the unstable rock mass cluster and its centroid coordinates includes: first, the data reading module acquires the centroid coordinates of the unstable rock mass cluster, and stores the horizontal, vertical, and elevation coordinate components of the centroid in the world coordinate system in memory; then, iterates through each 3D point cloud data point in the unstable rock mass cluster, reads the horizontal, vertical, and elevation coordinate components of the current 3D point cloud data point in the world coordinate system, and the distance calculation module calculates the distance between the horizontal coordinate component of the current 3D point cloud data point and the horizontal coordinate component of the centroid. The differences between the longitudinal coordinate component and the centroid's longitudinal coordinate component, and the differences between the elevation coordinate component and the centroid's elevation coordinate component, are squared and summed to obtain the squared distance of the 3D point cloud data point relative to the centroid. The square root of this squared distance is then performed, and the result is defined as the second Euclidean distance between the 3D point cloud data point and the centroid coordinates. This second Euclidean distance is then stored in a one-to-one correspondence with the corresponding 3D point cloud data point index. The above steps are repeated until all 3D point cloud data points in the unstable rock block point cloud cluster have been traversed, thus forming a complete second Euclidean distance sequence.

[0049] Specifically, the centroid coordinates of the point cloud cluster of the unstable rock block are used to represent the overall position of the unstable rock block in three-dimensional space. They are composed of the average positions of all the point clouds that make up the unstable rock block in three coordinate directions and are used to describe the spatial center position of the rock block. The second Euclidean distance between each three-dimensional point cloud data point in the unstable rock block point cloud cluster and the centroid coordinates is used to reflect the degree of dispersion of the point relative to the center of the rock block. Its magnitude reflects the degree of spatial diffusion of the point cloud inside the rock block.

[0050] Summing the squares of the second Euclidean distance yields the sum of squares of the Euclidean distances. The average of the sum of squares of the Euclidean distances is obtained by taking the average of the squares of the Euclidean distances. The equivalent radius of the unstable rock block is obtained by taking the square root of the squared mean of the Euclidean distance.

[0051] Specifically, summing the squares of the second Euclidean distance and taking the average and square root of each sum, and using the result as the equivalent radius of the unstable rock block, is based on the unified law between the spatial distribution of the point set and the geometric inertia of a rigid body: In the point cloud cluster of the unstable rock block, each three-dimensional point cloud data point can be regarded as a unit mass particle, and its second Euclidean distance to the center of mass represents the radial position of the particle relative to the geometric center. The sum of the squares of all points corresponds to the second geometric moment of the discrete particle system relative to the center of mass, which is the same as the rotational inertia determined by the radius in classical rigid body mechanics. Therefore, when averaging the sum of the squares of the distances by the number of points to obtain the average square distance, we are actually calculating the average radial expansion scale of the entire unstable rock block point cloud relative to the center of mass. Then, we take the square root of this average value to obtain a spatial scale quantity consistent with the root mean square radius of a uniform sphere. This scale quantity can characterize the overall size of the irregular unstable rock block in three-dimensional space with a radius parameter, so that the subsequent suspension index and roll-off risk index can be compared and calculated under a unified length scale, thereby making the complex rock block shape equivalent to a spherical unstable rock block with the same average radial energy distribution.

[0052] Specifically, the square of the second Euclidean distance is used to quantify the energy level of the offset of a single point relative to the centroid, enabling the offsets of multiple points to be accumulated using a unified dimension. The sum of squares of the Euclidean distances represents the total offset of all points within the unstable rock block relative to the centroid, used to measure the overall volume distribution of the rock block. The average of the squares of the Euclidean distances represents the average distribution of this offset among all points, serving as a comprehensive measure of the sparsity of the point cloud within the rock block. The square root of the average of the squares of the Euclidean distances is used to convert these offsets into scale values ​​with length meaning, thereby obtaining the equivalent radius of the unstable rock block. This equivalent radius is used to describe the overall size of the rock block, replacing the true geometric outline of the irregular rock block, and providing a unified spatial scale parameter for subsequent calculations of the suspension index and roll-off risk index.

[0053] In an embodiment of the present invention, the suspension index of the unstable rock block is calculated based on the equivalent radius of the unstable rock block, including: Based on the equation parameters of the plane equation fitted to the slope, the slope normal vector is generated; Specifically, the process of generating the slope normal vector based on the equation parameters of the slope fitting plane equation is as follows: First, read the parameters a, b, and c of the slope fitting plane equation obtained by least squares fitting, where parameters a and b represent the two inclination coefficients of the slope in the horizontal direction, and parameter c represents the height offset of the slope in the vertical direction; then, rearrange the slope fitting plane equation in the form of ax + by - z + c = zero, so that the coefficients in the equation can directly correspond to the direction information of the plane in three-dimensional space; based on this, take the coefficients corresponding to x, y, and z in the equation as the three components of the slope normal vector, namely parameters a, b, and negative one, and construct the slope normal vector according to the method of vector n equal to a, b, and negative one, which is used to represent the direction of the slope perpendicular to the slope surface in three-dimensional space; through the above processing, a slope normal vector that is strictly consistent with the slope fitting plane can be obtained, which provides a directional reference for the subsequent calculation of the normal distance between the centroid coordinates of the unstable rock mass cloud cluster and the slope fitting plane.

[0054] Specifically, the equation parameters of the slope fitting plane equation are coefficients and constants used to characterize the slope's tilt direction and height position in space. These values ​​reflect the amplitude of the slope's undulation in the horizontal direction and the degree of overall uplift or subsidence. The slope normal vector is a direction quantity perpendicular to the slope derived from these equation parameters. It is used to represent the slope's orientation in three-dimensional space and is the reference direction used when calculating the vertical distance from a point to the slope.

[0055] Determine the slope fitting plane based on the slope fitting plane equation; Specifically, the process of determining the slope fitting plane based on the slope fitting plane equation is as follows: First, the parameters a, b, and c of the slope fitting plane equation are read from the results of the slope fitting calculation. Parameters a and b are dimensionless coefficients describing the degree of slope inclination along the x and y directions, and parameter c is a constant term describing the overall elevation offset of the slope. Then, in a unified world coordinate system, the three-dimensional coordinates x, y, and z of any spatial point are substituted into the equation z equals a multiplied by x plus b multiplied by y plus c. This equation treats all three-dimensional points satisfying the equation as the same geometric plane, thus using the parameter set a, b, c, and the corresponding coordinate range as the unique identifier of the slope fitting plane in the data structure. Based on this, the slope fitting plane is used as the reference plane for subsequent geometric calculations to constrain the relative positional relationship between the centroid coordinates of the unstable rock mass cluster and the slope, as well as to determine the relevant normal distances, ensuring that the slope fitting plane remains consistent with the parameters of the slope fitting plane equation throughout the entire processing flow.

[0056] Based on the slope normal vector, calculate the normal distance between the centroid coordinates of the unstable rock mass point cloud cluster and the slope fitting plane; Specifically, firstly, based on the slope normal vector corresponding to the slope fitting plane equation determined earlier using the least squares fitting method, the three component values ​​of this normal vector correspond to the direction cosine values ​​of the plane normal on the three orthogonal coordinate axes in space. Next, based on the spatial coordinates of each point in the rock mass cluster, the centroid coordinates of the cluster are obtained; that is, the average value of all points along each coordinate axis is taken to form the three-dimensional centroid. Then, the obtained centroid coordinates are substituted into the aforementioned slope fitting plane equation, and the perpendicular distance from the centroid to the fitting plane is calculated using the point-to-plane distance formula. In this calculation, the numerator is the algebraic value obtained by substituting the centroid coordinates into the plane equation plus the absolute value of the intercept constant, and the denominator is the square root of the sum of the squares of the normal vector components plus one. The final distance obtained is the normal distance between the centroid of the rock mass and the slope fitting plane, which can quantitatively represent the geometric height position of the rock mass centroid relative to the smooth slope.

[0057] Specifically, the slope fitting plane is an idealized slope geometry model used to approximate the real slope surface, allowing subsequent distance and position calculations to be performed on a continuous, smooth plane; the unstable rock mass point cloud cluster is a collection of multiple three-dimensional point cloud data belonging to the same unstable rock mass, representing the three-dimensional shape and distribution of the unstable rock mass in space; the normal distance between the centroid coordinates of the unstable rock mass point cloud cluster and the slope fitting plane is the shortest distance from the centroid of the unstable rock mass to the slope fitting plane along the slope normal vector direction, used to characterize the degree of protrusion or suspension of the unstable rock mass relative to the slope, thereby reflecting the support relationship and potential instability trend between the unstable rock mass and the slope.

[0058] Based on the equivalent radius and normal distance of the unstable rock block, the suspension index of the unstable rock block is calculated. The calculation formula for the suspension index is as follows: In the formula, It is the equivalent radius of the unstable rock block. It is the normal distance. It is an indicator of the suspension degree of dangerous rock blocks.

[0059] Specifically, the suspension index is typically calculated based on the normal distance between the centroid of the unstable rock mass point cloud cluster and the slope fitting plane. This distance reflects the geometric state of whether the unstable rock mass is actually supported. When the centroid of the unstable rock mass is significantly higher than the slope fitting plane, especially when there are small-scale step structures that are not explicitly preserved in the data, it is easy to appear as if the bottom is overhanging or suspended in mid-air in the point cloud or image data. In this case, its suspension index is high, indicating that the rock mass has poor stability and a high risk of rolling down. With the assistance of UAV 3D data, the suspension index can be used to determine whether the unstable rock mass is on the actual contact surface. It is particularly suitable for assessing scenarios where support information is lost due to terrain smoothing, and for accurately identifying potential threats to facilities such as power corridors.

[0060] Specifically, the reason why the suspension degree index of a unstable rock block can be calculated through the proportional relationship between the equivalent radius and the normal distance is based on the fundamental laws of solid contact geometry and support stability in classical mechanics. When a unstable rock block is in a stable supported state, the vertical distance from its center of mass to the slope surface should generally be less than or close to the lower half of its own dimensions, that is, the center of mass falls within the area that can form an effective supporting reaction force. When the center of mass gradually rises above the support range of the slope fitting plane, the contact zone that can provide supporting reaction force will significantly decrease, causing the unstable rock block to be partially suspended, thus reducing its anti-sliding and anti-toppling ability. Therefore, comparing the normal distance from the center of mass to the slope surface with the equivalent radius of the unstable rock block can reflect the degree to which the bottom of the unstable rock block is supported. When the normal distance is close to the equivalent radius, the center of mass is close to the geometric center of the unstable rock block, and the supporting surface can hardly provide effective constraint on it, and the unstable rock block is in a highly suspended state; when the normal distance is much smaller than the equivalent radius, a large area of ​​the unstable rock block falls within the slope support zone, and the whole is in a relatively stable state. Based on the aforementioned geometric and stability principles, by subtracting the normal distance from the equivalent radius and then normalizing the result, a suspension index reflecting the degree of support of the unstable rock block can be obtained. This index is used to quantify the suspension status and potential instability risk of the unstable rock block.

[0061] S5. Based on the three-dimensional coordinates of the guy wire anchor pile and the centroid coordinates of the point cloud cluster of the unstable rock block, calculate the relative displacement of the unstable rock block along the slope direction. In an embodiment of the present invention, based on the three-dimensional coordinates of the guy wire anchor pile and the centroid coordinates of the point cloud cluster of unstable rock blocks, the relative displacement of the unstable rock blocks along the slope direction is calculated, including: Determine the slope gradient vector based on the equation parameters of the fitted plane equation of the slope. Specifically, the equation parameters of the slope fitting plane equation are coefficients used to describe the slope's horizontal inclination and overall elevation change. Two inclination coefficients reflect the slope's upward or downward trend along the east-west and north-south directions, respectively, while a constant term describes the overall vertical uplift of the slope. The slope gradient vector is a horizontal component vector constructed based on these inclination coefficients, representing the slope aspect along the direction of maximum change, i.e., the horizontal projection direction formed along the steepest direction on the slope.

[0062] Specifically, the steps for determining the slope gradient vector based on the equation parameters of the slope fitting plane equation are as follows: First, extract the coefficients from the general form equation representing the slope fitting plane. This equation is in the form that one variable equals a linear combination of other variables plus a constant term, where the slope parameter describes the rate of change of the plane along the two horizontal axes. Next, take the slope coefficients corresponding to the two horizontal directions in the equation, forming two components of a two-dimensional vector, and set the vertical component to zero to represent the projection characteristics of the vector in the horizontal direction. Then, perform a magnitude normalization operation on this two-dimensional vector to convert it into a unit vector, thereby obtaining the direction information of the slope gradient vector, ensuring that the vector only expresses direction and does not contain length information. Finally, this unit vector is the unit vector of the slope gradient direction, which can be used to describe the direction of the maximum steepness of the slope surface in the horizontal direction, serving as the basic parameter for subsequent projection calculations and distance determination.

[0063] The unit vector of the slope gradient vector is calculated using the following formula: In the formula, It is the unit vector of the slope gradient vector. It is the slope gradient vector. , These are the equation parameters of the plane equation fitted to the slope; Specifically, the unit vector of the slope gradient vector is obtained by normalizing the slope gradient vector, so that it retains only directional information and does not contain length dimensions. It is used as the standard vector for the projection direction in subsequent calculations. In the expression, the symbols 'a' and 'b' represent the slope inclination coefficients in two horizontal directions, and their values ​​together determine the overall aspect of the slope. The vector formed by combining symbols 'a' and 'b' represents the direction of slope change in the horizontal plane. The denominator is the square root of the length of this slope vector, used to achieve normalization, so that the resulting unit vector can represent both the most important downward or upward direction on the slope and serve as the directional reference for subsequent calculations of the displacement of the unstable rock block along the slope aspect.

[0064] The vector subtraction operation is performed on the three-dimensional coordinates of the guy wire anchor pile and the centroid coordinates of the unstable rock mass cluster to obtain the pointing vector of the guy wire anchor pile relative to the centroid coordinates. The relative displacement of the unstable rock block along the slope direction is obtained by performing a dot product operation on the unit vector and the pointing vector of the slope gradient vector.

[0065] Specifically, firstly, the three-dimensional spatial coordinate data of the guy wire anchor pile and the centroid coordinate data of the point cloud cluster of unstable rock blocks are extracted. Then, a dimension-by-dimensional vector subtraction operation is performed using the centroid coordinates as the subtrahend and the guy wire anchor pile coordinates as the minuend to obtain the pointing vector representing the guy wire anchor pile pointing towards the centroid of the unstable rock block. Subsequently, the slope gradient vector calculated based on the plane equation parameters in the aforementioned fitted slope is extracted and normalized to serve as the unit vector of the slope gradient direction. This unit vector is then used to perform a vector dot product operation with the pointing vector to obtain the projected distance of the unstable rock block's centroid relative to the anchor pile in the slope gradient direction. This distance serves as the relative displacement of the unstable rock block along the slope direction, characterizing the spatial proximity of the unstable rock block relative to the anchor point in the direction of gravity sliding tendency, thus providing input data for hazard level analysis.

[0066] Specifically, the three-dimensional coordinates of the guy wire anchor pile are used to determine its absolute position in space, and the centroid coordinates of the point cloud cluster of unstable rock blocks represent the average position of the block in the overall point cloud. The vector subtraction between the two is the pointing vector of the guy wire anchor pile relative to the centroid of the unstable rock block. This vector reflects the directional relationship from the anchor pile to the centroid of the unstable rock block. The unit vector of the slope gradient vector represents the direction of the maximum slope in the horizontal direction, characterizing the standardized directional information of the slope towards the sliding direction. By performing a dot product operation between this unit vector and the pointing vector, the projection component of the unstable rock block along the slope gradient direction can be calculated. This value is the relative displacement of the unstable rock block along the slope direction, which is used to assess the directionality and quantification of its sliding trend and potential movement path.

[0067] Specifically, the relative displacement of the unstable rock block along the slope direction can be obtained by performing a dot product operation between the pointing vector and the unit vector of the slope gradient direction. This is based on the combined principle of the classical vector projection formula and the law of slope motion dominated by gravity. In a slope environment, the main tendency direction of object sliding is determined by the maximum downward direction on the slope, which is represented by the slope gradient vector. After normalization, this vector can be used as the standard direction vector characterizing the potential sliding direction under gravity. The vector obtained by subtracting the centroid coordinates of the unstable rock block from the three-dimensional coordinates of the guy wire anchor pile can represent the most direct spatial pointing relationship between the two. Projecting this pointing vector along the slope gradient direction is equivalent to calculating the component distance of the unstable rock block relative to the anchor pile in the direction of maximum slope gradient. This component is the key geometric quantity for measuring the potential sliding of the unstable rock block towards the anchor pile along the slope direction. The dot product operation essentially gives the degree of consistency of the directions of two vectors and the magnitude of the projection of the pointing vector in the unit slope direction. Therefore, the dot product result can accurately reflect the tendency of the unstable rock block to approach or move away from the anchor pile along the slope direction. The resulting quantitative result is the relative displacement of the unstable rock block along the slope direction, which provides a reliable basis for further assessment of the potential rolling path and impact risk of the unstable rock block.

[0068] Specifically, calculating the relative displacement of the unstable rock mass along the slope direction based on the three-dimensional coordinates of the guy wire anchor piles and the centroid coordinates of the unstable rock mass point cloud cluster is to assess the potential movement trend and threat level of the unstable rock mass relative to the pre-set interception or support structure under gravity. Since unstable rock masses on slopes often slide or roll along the slope gradient direction due to gravity, the initial state of their possible paths can be quantified by the relative geometric relationship between their centroid position and the anchor piles. The pointing vector obtained by vector subtraction represents the spatial direction of the unstable rock mass relative to the anchor piles. Combining this with the dot product of the unit vectors along the slope gradient direction, the component displacement of the unstable rock mass in the main sliding direction on the slope can be obtained. This effectively reflects whether the unstable rock mass is sliding towards the anchor piles, providing a key reference for hazard classification, evaluation of the rationality of anchor layout, and optimization of support measures.

[0069] S6. Calculate the rolling risk index of the dangerous rock block based on the slope point cloud set, the suspension index of the dangerous rock block, and the relative displacement of the dangerous rock block along the slope direction. In embodiments of the present invention, the roll-off risk index of the unstable rock block is calculated based on the slope point cloud set, the suspension degree index of the unstable rock block, and the relative displacement of the unstable rock block along the slope direction, including: Perform a dot product operation between each 3D point cloud data point in the slope point cloud set and the unit vector of the slope gradient vector to obtain the slope aspect projection coordinates of each 3D point cloud data point in the slope aspect direction. The slope extension length is obtained by calculating the difference between the maximum and minimum values ​​of all slope aspect projected coordinates. Specifically, each 3D point cloud data point in the slope point cloud set represents an actual sampling location on the landslide slope, and this data set records the spatial distribution of the landslide slope. The unit vector of the slope gradient vector represents the direction of the maximum steepness of the slope in the horizontal direction, and its direction is consistent with the maximum horizontal gradient of the slope, with a magnitude of one. By performing a dot product operation between the 3D coordinates of each slope point cloud and the unit vector of the slope gradient, the projected coordinates of that point in the slope aspect direction can be obtained, representing the relative position of that point in the slope aspect direction. The difference between the maximum and minimum values ​​of the slope aspect projected coordinates of all points is the spatial expansion range of the entire slope in the slope aspect direction. This range is defined as the slope aspect extension length, which is used to characterize the geometric dimensions of the landslide slope in the main sliding direction and reflects the degree of extension of the slope along the main deformation direction.

[0070] Specifically, the 3D point cloud data points in the slope point cloud set represent spatially distributed discrete points on the slope, and the unit vector of the slope gradient vector represents the normalized directional information of the slope along the direction of maximum steepness. By performing a dot product operation on each 3D point cloud data point and the unit vector of the slope gradient vector, the projected coordinates of that point along the slope direction can be calculated, i.e., the spatial projected position of that point in the slope direction. Subsequently, the difference between the maximum and minimum values ​​of all projected coordinates is calculated, and the resulting difference is the extension distance of the entire slope point cloud set in the slope direction. This extension distance reflects the spatial length of the slope along the direction of maximum steepness, and can thus be used to characterize the slope aspect development scale. This method, based on the linear projection principle of dot product operation, projects 3D spatial data onto a scale analysis in a specific direction, possessing clear geometric meaning and computational feasibility.

[0071] Based on the suspension index of the unstable rock block, the relative displacement of the unstable rock block along the slope direction, and the slope extension length, the roll-off risk index of the unstable rock block is calculated. The formula for calculating the roll-off risk index of the unstable rock block is as follows: In the formula, It is the risk index of falling loose rocks. It is an indicator of the suspension degree of unstable rock blocks. It is the relative displacement of the unstable rock block along the slope direction. It is the slope extension length.

[0072] Specifically, the rockfall risk index is used to comprehensively measure the overall danger of a rockfall sliding or rolling downhill under current terrain conditions and potentially colliding with critical structures. This index is determined by both the suspension state of the rockfall and its relative displacement trend along the slope. The suspension state reflects the adequacy of support between the bottom of the rockfall and the slope surface. When only a small portion of the bottom of the rockfall is supported by the slope surface, its overall stability decreases significantly, making it more susceptible to instability under external disturbances. The relative displacement trend along the slope describes the proximity of the rockfall to important facilities in the potential sliding direction. The danger increases as the rockfall is located in the maximum downward direction of the natural slope and the spatial distance between its center of mass and the facility continuously decreases.

[0073] Specifically, the reason why the risk index of a rockfall block can be calculated by combining the suspension degree index with the normalized ratio of the relative displacement along the slope direction over the slope extension length is based on the fundamental mechanical law that the weakening of support capacity and the proximity of the sliding path jointly determine the probability of instability in typical slope stability analysis. The suspension degree index reflects the degree to which the bottom of the rockfall block is adequately supported by the slope surface. When the suspension degree is high, the effective contact area at the bottom of the rockfall block shrinks, its anti-sliding and anti-toppling capacity is significantly weakened, and it is more likely to displace under the action of gravity. The relative displacement along the slope direction represents the geometric proximity of the rockfall block to an important target in the direction of maximum descent of the slope surface. The ratio of this displacement to the slope extension length can measure the position of the rockfall block in the entire potential sliding path. When this ratio is close to one, it indicates that the rockfall block is close to the lower edge of the slope surface and is more likely to slide outward. By subtracting one from this ratio, a relationship can be formed in which the risk increases monotonically with the proximity. Multiplying the suspension index by the normalized displacement means that the rollover risk index only increases when the unstable rock block simultaneously exhibits a high degree of suspension and a significant downward trend. This aligns with the actual mechanism where instability is often triggered by both insufficient support and increased sliding force. Therefore, this combined formula comprehensively reflects the degree of support weakening and the strength of the sliding trend of the unstable rock block, and the result is the rollover risk index of the unstable rock block.

[0074] Specifically, the slope point cloud is a digital description of the slope's topographic geometry. Its three-dimensional data points, through spatial distribution, represent the slope's morphological boundaries and aspect extension range. The suspension degree index of the unstable rock block is an empirical parameter characterizing the stability of the unstable rock block, reflecting its contact degree and attachment status with the supporting structure. The relative displacement of the unstable rock block along the slope direction is a physical quantity that measures the degree of its center of gravity shift relative to the anchor piles in the slope direction, reflecting whether it has a downward trend. By fusing and analyzing these three types of information, a roll-off risk assessment model that better conforms to the gravity-driven slide risk mechanism can be constructed, based on quantifying the slope morphology, assessing the degree of unstable rock block suspension, and perceiving its downward displacement trend along the slope direction. Therefore, calculating the roll-off risk index of the unstable rock block based on the slope point cloud, the suspension degree index, and the relative displacement of the unstable rock block along the slope direction can comprehensively consider topographic constraints, support conditions, and movement trends, significantly improving the accuracy and relevance of risk assessment.

[0075] S7. Based on the centroid coordinates and roll-off risk index of the unstable rock blocks, generate a disaster identification record for the power corridor.

[0076] Specifically, the core data corresponding to all unstable rock mass clusters is first extracted from the preliminary data processing results. This includes the unique identifier of each unstable rock mass cluster, the X, Y, and Z axis components of the centroid coordinates in the world coordinate system, the calculated roll-off risk index, and the identifier of the guy wire anchor pile associated with the unstable rock mass cluster. Simultaneously, basic scene information such as the power corridor section number, inspection date, and time recorded during drone inspections is retrieved. This data is then aggregated into the same data processing module for initial integration. Subsequently, the integrated data undergoes validity verification, checking whether the centroid coordinates of each unstable rock mass are within a reasonable spatial range for the power corridor section and whether the roll-off risk index meets the requirements. The pre-set calculation logic is based on the relative displacement of the unstable rock block along the slope direction and the slope extension length, calculated by formula. If data anomalies are found, the corresponding calculation step is traced back to obtain accurate data again, ensuring that all data included in the record is accurate and reliable. Then, according to the disaster risk classification standard formulated by the power industry and combined with the actual application scenario of this method, the classification threshold of the roll-off risk index is set. For example, when the roll-off risk index is less than 0.3, it is judged as low risk; when the index is between 0.3 and 0.7, it is judged as medium risk; and when the index is greater than 0.7, it is judged as high risk. The roll-off risk index of each unstable rock block is compared with the classification threshold to determine the corresponding risk level. Next, a structured framework for disaster identification and recording was constructed. This framework includes a section for identifying unstable rock blocks, where a unique identifier and associated guy wire anchor pile identifier are entered; a spatial location information section, where the X, Y, and Z axis values ​​of the centroid coordinates are entered; a risk assessment information section, where the roll-off risk index and corresponding risk level are entered; and a scenario supplementary information section, where the power corridor section number, inspection date, and time are entered. Additionally, a section for basic parameters of unstable rock blocks is added, where the equivalent radius of the unstable rock block is entered to reflect its size. Then, all verified data is entered one by one according to the sections of the structured framework, ensuring that the information in each section is complete and corresponds one-to-one with the unstable rock block. For low-risk unstable rock blocks… The system supplements the record with recommendations for regular monitoring cycles. For medium-risk unstable rock blocks, it recommends increasing the frequency of inspections. For high-risk unstable rock blocks, it provides warnings for immediate on-site verification and handling. Finally, all the completed unstable rock block hazard identification information is summarized and formatted into a unified power corridor hazard identification record document. This document can be exported to an electronic format that conforms to power industry data management standards and can be linked to the power corridor spatial information management system. This allows maintenance personnel to quickly locate unstable rock blocks based on the centroid coordinates in the record and formulate targeted safety control measures according to the risk level, thereby completing the generation of the power corridor hazard identification record.

[0077] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for identifying power corridor hazards using unmanned aerial vehicles (UAVs) based on POS-assisted geometric correction, characterized in that, Includes the following steps: S1. Based on the original point cloud data of the power corridor in the UAV sensor coordinate system and the UAV's POS data, generate a 3D point cloud set of the power corridor, where the POS data includes the position, attitude, rotation matrix and position vector. S2. Based on the three-dimensional coordinates of the guy wire anchor piles in the power corridor and the three-dimensional point cloud set of the corridor, determine the neighborhood point cloud of the guy wire anchor piles. S3. Based on the neighborhood point cloud of the guy wire anchor pile, generate a set of slope point clouds and multiple dangerous rock block point cloud clusters; S4. Calculate the equivalent radius of the dangerous rock block based on the centroid coordinates of the point cloud cluster of dangerous rock blocks, and calculate the suspension index of the dangerous rock block based on the equivalent radius of the dangerous rock block. S5. Based on the three-dimensional coordinates of the guy wire anchor pile and the centroid coordinates of the point cloud cluster of the unstable rock block, calculate the relative displacement of the unstable rock block along the slope direction. S6. Calculate the rolling risk index of the dangerous rock block based on the slope point cloud set, the suspension index of the dangerous rock block, and the relative displacement of the dangerous rock block along the slope direction. S7. Based on the centroid coordinates and roll-off risk index of the unstable rock blocks, generate a disaster identification record for the power corridor.

2. The method for identifying power corridor hazards using unmanned aerial vehicles (UAVs) based on POS-assisted geometric correction according to claim 1, characterized in that, Based on the raw point cloud data of the power corridor in the UAV sensor coordinate system and the UAV's POS data, a 3D point cloud set of the power corridor is generated, including: Acquire raw point cloud data of the power corridor in the UAV sensor coordinate system; Obtain the position, attitude, and rotation matrix of the UAV during flight; Obtain the position vector of the drone during its flight; Perform coordinate rotation transformation on the position and attitude rotation matrix and the original point cloud data to obtain the point cloud data after coordinate rotation transformation; The point cloud data and position vector after coordinate rotation transformation are transformed by coordinate translation to obtain three-dimensional point cloud data points in the world coordinate system. By collecting the 3D point cloud data points in the world coordinate system at all times, a 3D point cloud set of the power corridor is obtained.

3. The method for identifying power corridor hazards using unmanned aerial vehicles based on POS-assisted geometric correction according to claim 1, characterized in that, Based on the three-dimensional coordinates of the guy wire anchor piles in the power corridor and the three-dimensional point cloud set of the corridor, the neighborhood point cloud of the guy wire anchor piles is determined, including: Obtain the three-dimensional coordinates of guy wire anchor piles in the power corridor in the world coordinate system; The Euclidean distance between each 3D point cloud data point in the corridor 3D point cloud set and the 3D coordinates of the guy wire anchor pile is calculated to obtain the first Euclidean distance; If the first Euclidean distance is less than or equal to the preset spatial distance threshold, then the three-dimensional point cloud data point corresponding to the first Euclidean distance is used as the neighborhood point cloud of the guy wire anchor pile.

4. The method for identifying power corridor hazards by unmanned aerial vehicles based on POS-assisted geometric correction according to claim 1, characterized in that, Based on the neighborhood point cloud of the guy wire anchor piles, a slope point cloud set and multiple unstable rock block point cloud clusters are generated, including: Based on the known geometric range of the artificial components, the point cloud of the neighboring area of ​​the guy wire anchor pile is processed to remove the artificial component point cloud, resulting in a natural slope-rock mass point cloud set. In the natural slope-rock mass point cloud set, slope fitting sample points are extracted from the three-dimensional point cloud data points that are close to the guy wire anchor piles and have low height, to obtain a slope fitting point cloud subset. The least squares fit is performed on the subset of the slope fitting point cloud to obtain the slope fitting plane equation; Based on the slope fitting plane equation, the elevation difference of each three-dimensional point cloud data point in the natural slope-rock mass point cloud set is calculated to obtain the elevation deviation. If the elevation deviation is greater than the elevation resolution of the sensor, the three-dimensional point cloud data point corresponding to the elevation deviation is marked as a protrusion point; otherwise, the three-dimensional point cloud data point corresponding to the elevation deviation is marked as a slope point. Spatial adjacency clustering of the protruding points yields multiple clusters of unstable rock blocks. The slope points are processed into a set to obtain a slope point cloud set.

5. The method for identifying power corridor hazards using unmanned aerial vehicles based on POS-assisted geometric correction according to claim 1, characterized in that, The equivalent radius of the unstable rock mass is calculated based on the centroid coordinates of the point cloud cluster, including: Calculate the centroid coordinates of the point cloud cluster of unstable rock blocks; Calculate the second Euclidean distance between each 3D point cloud data point and the centroid coordinates in the point cloud cluster of unstable rock blocks; Summing the squares of the second Euclidean distance yields the sum of squares of the Euclidean distances. The average of the sum of squares of the Euclidean distances is obtained by taking the average of the squares of the Euclidean distances. The equivalent radius of the dangerous rock block is obtained by taking the square root of the squared mean of the Euclidean distance.

6. The method for identifying power corridor hazards by unmanned aerial vehicles based on POS-assisted geometric correction according to claim 4, characterized in that, The suspension index of the unstable rock block is calculated based on its equivalent radius, including: Based on the equation parameters of the plane equation fitted to the slope, the slope normal vector is generated; Determine the slope fitting plane based on the slope fitting plane equation; Based on the slope normal vector, calculate the normal distance between the centroid coordinates of the unstable rock mass point cloud cluster and the slope fitting plane; The suspension index of the unstable rock block is calculated based on the equivalent radius and normal distance of the unstable rock block.

7. The method for identifying power corridor hazards by unmanned aerial vehicles based on POS-assisted geometric correction according to claim 4, characterized in that, Based on the three-dimensional coordinates of the guy wire anchor piles and the centroid coordinates of the point cloud cluster of unstable rock blocks, the relative displacement of the unstable rock blocks along the slope direction is calculated, including: Determine the slope gradient vector based on the equation parameters of the fitted plane equation of the slope. Calculate the unit vector of the slope gradient vector; The vector subtraction operation is performed on the three-dimensional coordinates of the guy wire anchor pile and the centroid coordinates of the unstable rock mass cluster to obtain the pointing vector of the guy wire anchor pile relative to the centroid coordinates. The relative displacement of the unstable rock block along the slope direction is obtained by performing a dot product operation on the unit vector and the pointing vector of the slope gradient vector.

8. The method for identifying power corridor hazards by unmanned aerial vehicles based on POS-assisted geometric correction according to claim 7, characterized in that, The roll-off risk index of the unstable rock block is calculated based on the slope point cloud set, the suspension degree index of the unstable rock block, and the relative displacement of the unstable rock block along the slope direction, including: Perform a dot product operation between each 3D point cloud data point in the slope point cloud set and the unit vector of the slope gradient vector to obtain the slope aspect projection coordinates of each 3D point cloud data point in the slope aspect direction. The slope extension length is obtained by calculating the difference between the maximum and minimum values ​​of all slope aspect projected coordinates. The risk index of rockfall is calculated based on the suspension index of the rock block, the relative displacement of the rock block along the slope direction, and the slope extension length.

Images