An unmanned aerial vehicle-based construction period pumped storage power station inspection processing method

Abstract

The application discloses a kind of construction period pumped storage power station inspection processing methods based on unmanned aerial vehicle, it is related to unmanned aerial vehicle inspection technical field, including, acquisition field control data and unmanned aerial vehicle real-time state, field control data is preprocessed, generates working condition single, determines candidate space based on working condition single and filters available candidate section, and obtains control corridor by connected analysis, according to control corridor, generate control corridor normal subzone in corridor, according to continuous flyable length, determine reference area;According to the transverse axis direction of control corridor main direction, determine the first inspection advancing direction and generate the first inspection track;The reference area matching of first inspection section is carried out, and the first inspection callback index is calculated;According to the set first inspection trigger threshold and determine normal first inspection section or suspected first inspection section, suspected first inspection section is merged to form a doubtful single.The application realizes the fine identification and reliability confirmation of construction surface anomaly.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) inspection technology, and in particular to a method for inspecting and processing pumped storage power stations during the construction period based on UAVs. Background Technology

[0002] With the expansion of pumped-storage power station construction and the increasing complexity of construction environments, traditional inspection methods are insufficient to meet the dynamic monitoring needs during the construction period. In the engineering field, routine inspections mainly rely on a combination of manual ground patrols and fixed sensor networks. Manual patrols record the appearance of the construction surface through visual inspection and combine this with local measurements using equipment such as total stations; fixed monitoring systems deploy sensors for temperature, humidity, displacement, etc., at pre-set locations to achieve continuous data collection in key areas. In addition, recent research has introduced unmanned aerial vehicles (UAVs) into engineering inspections, using planned flight paths to perform visible light and thermal imaging scans of the construction area, acquiring surface images and temperature distribution data, and using point cloud modeling to reconstruct 3D scenes. These methods provide fundamental data support for quality and safety management during the construction period and have become an important development direction for technological applications in this field.

[0003] However, the aforementioned conventional drone inspection methods still have certain limitations. On the one hand, in dynamic construction environments, the alternation of operating machinery, temporary facilities, and work processes leads to limited and highly time-varying flight space. Existing flight path planning largely relies on static geofencing, making it difficult to integrate dynamic constraints such as construction machinery envelopes, restricted areas, and work windows in real time, which can easily result in flight path interruptions or blind spots. On the other hand, the identification of suspected abnormal areas usually relies on comparing single-collection data with fixed thresholds, lacking a cross-validation mechanism for the spatial consistency and multi-source feature consistency of suspicious points at multiple data levels. Under complex lighting, temperature differences, and temporary obstruction conditions, false alarms or missed alarms may occur, affecting the reliability of inspection results. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a method for inspecting pumped storage power stations during the construction period based on unmanned aerial vehicles (UAVs) to solve the problems of insufficient continuity of UAV inspection trajectories and low reliability of anomaly detection under dynamic construction constraints in the prior art.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: This invention provides a method for inspection and processing of pumped-storage power stations during construction based on unmanned aerial vehicles (UAVs). The method includes: collecting on-site control data and real-time UAV status; preprocessing the on-site control data to generate a work status sheet; determining candidate spaces and filtering available candidate segments based on the work status sheet; obtaining a control corridor through connectivity analysis; generating normal sub-regions within the control corridor based on the control corridor; determining a reference zone based on the continuous flyable length; determining the initial inspection advancement direction based on the transverse axis of the main direction of the control corridor and generating an initial inspection trajectory; matching the initial inspection segment with the reference zone and calculating the initial inspection callback index; and determining the normal initial inspection based on a set initial inspection trigger threshold. The system detects fragments or suspected initial inspection fragments, groups suspected initial inspection fragments into a suspicious point list, generates a re-inspection list based on the suspicious point list, determines the re-inspection trajectory and speed, and performs local re-inspection, matches, unifies coordinates, and extracts regions from the re-inspection sampling data to obtain the re-inspection point set, re-inspection outline, re-inspection visible area, and re-inspection hot area, and summarizes them to generate a re-inspection local information set, determines the re-inspection suspicious area and reference area based on the re-inspection datum, calculates spatial consistency and symbol consistency, calculates the re-inspection confirmation value based on spatial consistency and symbol consistency and generates a confirmation threshold, determines whether the suspicious area passes the re-inspection and generates an inspection processing order.

[0007] As a preferred embodiment of the UAV-based inspection and processing method for pumped storage power stations during construction as described in this invention, the on-site control data includes the construction area boundary, mechanical motion envelope, temporary control area, start and end times of permitted flight in the work area, design datum or previous qualified datum, and areas that have been inspected and found to be normal; the real-time status of the UAV includes cruise speed, braking capacity parameters, and external dimensions.

[0008] As a preferred embodiment of the UAV-based inspection and processing method for pumped storage power stations during construction, the preprocessing of on-site control data to generate a work status sheet comprises the following steps: The construction area boundary, mechanical motion envelope, temporary control area, design datum or previous qualified datum, and verified normal areas are uniformly converted to the on-site three-dimensional coordinate system; using the UAV takeoff time as the reference time, the start and end times of permitted flights in the work area are time-converted to obtain the work time window, and invalid time windows that do not overlap with the UAV mission time period are deleted; the converted construction area boundary, mechanical motion envelope, temporary control area, design datum or previous qualified datum, verified normal areas, and valid work time windows are summarized to generate a work status sheet.

[0009] As a preferred embodiment of the UAV-based inspection and processing method for pumped storage power stations during construction, the following steps are taken to determine the reference area based on the continuous flyable length: Candidate spaces are determined based on the construction area boundary, mechanical motion envelope, and temporary control area in the work condition sheet; the candidate spaces are discretized using voxels to obtain candidate segments, and usable candidate segments are selected; all usable candidate segments are merged according to voxel adjacency relationships, retaining usable connected components that are connected to the voxel where the UAV takeoff point is located, and the usable connected component with the largest number of voxels is determined as the control corridor; the verified normal areas are discretized and overlapped with the control corridor to obtain a set of normal voxels within the control corridor, and connectivity is divided to obtain normal sub-regions within the control corridor; the main direction of the control corridor is determined, and based on the continuous flyable length of each normal sub-region within the control corridor along the main direction of the control corridor, the normal sub-region within the control corridor with the largest continuous flyable length is selected as the reference area.

[0010] As a preferred embodiment of the UAV-based inspection method for pumped storage power stations during construction as described in this invention, the steps for determining the initial inspection direction and generating the initial inspection trajectory are as follows: Extract all voxel center points in the control corridor; using the voxel center point of the UAV takeoff point as the starting point, determine the voxel center point with the furthest horizontal axis distance and use this to determine the initial inspection direction; Perform vertical slicing along the initial inspection direction, using the voxel side length as the step size, to obtain the initial inspection slice, and extract the voxel center points falling within the slice; Calculate the average values ​​of the horizontal, vertical, and height coordinates of all voxel center points in the initial inspection slice to obtain the path center point of the initial inspection slice; Connect all path center points sequentially according to the initial inspection direction to form the initial inspection trajectory, and collect the initial inspection sampling data.

[0011] As a preferred embodiment of the UAV-based inspection and processing method for pumped storage power stations during the construction period described in this invention, the initial inspection sampling data includes visible light images, thermal imaging images, and distance points on the main surface of the pumped storage power station, as well as UAV pose data.

[0012] As a preferred embodiment of the UAV-based inspection and processing method for pumped storage power stations during construction as described in this invention, the steps for grouping suspected first-inspection segments into a list of suspected segments are as follows: The distance point of the first-inspection segment is converted to the on-site three-dimensional coordinate system and projected vertically onto a reference surface; the surface contour of the segment and its thermal and visible light images are extracted; it is determined whether the center point of the first-inspection segment falls within the reference area; the signed normal height difference, signed temperature difference, and gap value are calculated, and the first-inspection callback index is calculated; a first-inspection trigger threshold is set, and the first-inspection callback index is compared with the first-inspection trigger threshold to determine whether the first-inspection segment is a normal first-inspection segment or a suspected first-inspection segment; adjacent suspected first-inspection segments are merged along the first-inspection advancement direction, and the boundary and long side dimensions of the suspected area are extracted to form a list of suspected segments.

[0013] As a preferred embodiment of the UAV-based inspection and processing method for pumped storage power stations during construction as described in this invention, the steps for generating a re-inspection form based on the suspicious area list are as follows: The projection of the suspicious area is obtained based on the boundary of the suspicious area, and the horizontal long side direction and offset direction are determined based on the long side direction of the suspicious area; the re-inspection offset distance is calculated based on the long side dimension and gap value of the suspicious area, and a central re-inspection line and lateral re-inspection lines located on both sides are generated using the center point of the suspicious area projection as the centerline reference; the central re-inspection line and lateral re-inspection lines are intersected with the control corridor, retaining the line segments located inside the control corridor and covering the projection range of the suspicious area, and connecting them through transition sections to form a re-inspection trajectory; the re-inspection speed is calculated based on the UAV's cruising speed and gap value, and the suspicious area projection, the center point of the suspicious area, the re-inspection offset distance, the re-inspection trajectory, and the re-inspection speed are summarized into a re-inspection form.

[0014] As a preferred embodiment of the UAV-based inspection and processing method for pumped storage power stations during construction as described in this invention, the steps for generating a local re-inspection information set are as follows: Perform a local re-inspection along the re-inspection trajectory at the re-inspection speed, and re-collect visible light images, thermal images, distance points, and pose data; using the timestamp of the re-collected pose data as a reference, match the re-collected distance points, visible light images, and thermal images to form a re-inspection sampling data set, and uniformly convert the distance points to the on-site three-dimensional coordinate system to obtain a re-inspection point set; extract the re-inspection comparison base surface, re-inspection contour, re-inspection visible area, and re-inspection thermal area from the re-inspection point set; summarize the re-inspection point set, re-inspection comparison base surface, re-inspection contour, re-inspection visible area, re-inspection thermal area, re-inspection sampling data set, and corresponding re-inspection trajectory to generate a local re-inspection information set.

[0015] As a preferred embodiment of the UAV-based inspection and processing method for pumped storage power stations during construction as described in this invention, the steps for determining whether a suspicious area passes re-inspection and generating an inspection processing order are as follows: Based on the boundary of the suspicious area, determine the suspicious area and the re-inspection reference area on the re-inspection comparison datum; overlap and compare the re-inspection outline with the suspicious area to obtain spatial consistency; calculate the marked re-inspection normal elevation difference and the marked re-inspection temperature difference, and obtain the sign consistency with the marked normal elevation difference and the marked temperature difference; calculate the re-inspection confirmation value based on the spatial consistency and the sign consistency; obtain the reference confirmation value using a sliding window method within the re-inspection reference area, and set a confirmation threshold based on the reference confirmation value; compare the re-inspection confirmation value with the confirmation threshold to determine whether the suspicious area passes re-inspection, and summarize the relevant information of the suspicious areas that fail re-inspection to generate an inspection processing order.

[0016] The beneficial effects of this invention are as follows: control corridors are obtained through connectivity analysis, and normal sub-regions within the control corridors are generated based on the control corridors. The reference area is determined according to the continuous flyable length, ensuring the continuity and integrity of the flight path. By determining whether suspicious areas have passed the re-inspection, the refined identification and reliability confirmation of anomalies on the construction surface are achieved. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of a method for inspecting and handling pumped storage power stations during the construction period based on drones.

[0019] Figure 2 A flowchart for determining the reference area.

[0020] Figure 3 A flowchart for creating a list of suspicious points.

[0021] Figure 4 A flowchart for generating inspection processing orders.

[0022] Figure 5 This is a graph showing the relationship between the gap value and the initial inspection pullback index.

[0023] Figure 6 This is a graph showing the influence of the gap value on the re-inspection confirmation value. Detailed Implementation

[0024] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0025] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0026] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0027] Reference Figures 1-6This is one embodiment of the present invention, which provides a method for inspection and processing of pumped storage power stations during the construction period based on unmanned aerial vehicles (UAVs), including the following steps: S1. Collect on-site control data and real-time status of UAVs, preprocess the on-site control data, generate a working condition sheet, determine candidate spaces and screen available candidate segments based on the working condition sheet, obtain the control corridor through connectivity analysis, generate normal sub-regions within the control corridor based on the control corridor, and determine the reference area according to the continuous flyable length.

[0028] Collect on-site control data, perform data preprocessing on the on-site control data, and obtain the operating condition sheet.

[0029] It should be noted that the on-site control data includes the construction area boundary, the mechanical motion envelope, the temporary control area, the start and end times of permitted flights in the work area, the design datum or the previous qualified datum, and the areas that have been inspected and found to be normal.

[0030] The construction area boundary, machinery movement envelope, and temporary control zone are all transformed into the same three-dimensional coordinate system. Using the vertical projection of the UAV takeoff point on the ground as the origin, and east as the horizontal axis, north as the vertical axis, and upward as the height axis, a unified on-site three-dimensional coordinate system is established. Boundary point data of the construction area boundary are read and uniformly converted to the aforementioned on-site three-dimensional coordinate system. If the original boundary point data has engineering coordinates, its specific position in the on-site three-dimensional coordinate system is directly read. If the original boundary point data contains engineering coordinates, the specific position of the origin of the three-dimensional coordinate system in the engineering coordinates is identified to obtain the origin's engineering coordinates. The difference between the engineering coordinates of the original boundary point data and the origin's engineering coordinates is calculated along each axis of the engineering coordinate system to obtain the relative translation of the original boundary point data's engineering coordinates relative to the origin's engineering coordinates. The rotation angle of the engineering coordinate system relative to the on-site three-dimensional coordinate system is calculated. Based on the rotation formula, the specific coordinates of the original boundary point data converted to the on-site three-dimensional coordinate system are calculated. The expression is: ; ; ; in, The x-axis coordinates of the original boundary point data in the on-site three-dimensional coordinate system. This represents the horizontal translation of the engineering coordinates of the original boundary point data relative to the origin engineering coordinates. For rotation angle, This represents the relative translation of the engineering coordinates of the original boundary point data with respect to the vertical axis of the origin engineering coordinates. This represents the ordinate of the original boundary point data in the on-site three-dimensional coordinate system. These are the high-axis coordinates of the original boundary point data in the on-site three-dimensional coordinate system. The relative translation of the engineering coordinates of the original boundary point data with respect to the high axis of the origin engineering coordinates.

[0031] Based on the on-site three-dimensional coordinate system, a unified coordinate expression is used to represent the spatial range and trajectory of the mechanical motion envelope, as well as the boundary points and elevation range of the temporary control area.

[0032] Using the drone's takeoff time as the baseline, the start and end times of permitted flights within the work area are calculated by subtracting them from the baseline to obtain the permitted start and end times. These permitted start and end times are then divided into work area time windows, with all time measured in seconds. The drone's time coverage is determined based on its set mission duration, and the end time of this coverage is taken as the end of the time period. The work area time window is checked for overlap with the coverage. If the permitted end time is earlier than the baseline, the work area time window has already ended before the drone's takeoff and is no longer valid within the drone's executable time period; in this case, the work area time window is deleted. If the permitted start time is later than the end of the time period, the entire work area time window is located after the coverage and does not coincide with the drone's inspection execution time; in this case, the work area time window is deleted. Only work area time windows that overlap with the coverage are retained. For work area time windows where only a portion overlaps with the coverage, only the overlapping segment is extracted.

[0033] Following the method of converting the original boundary point data with engineering coordinates to the on-site three-dimensional coordinate system, the design datum or the previous qualified datum and the verified normal area are converted to the on-site three-dimensional coordinate system.

[0034] The construction area boundaries, mechanical motion envelopes, temporary control areas, design datum or previous qualified datum, inspected normal areas, and operation time windows with unified time references, which are converted to a unified on-site three-dimensional coordinate system, are summarized into a work status sheet.

[0035] Read the drone's real-time status, which includes cruise speed, braking capability parameters, and external dimensions.

[0036] It should be noted that the braking capability parameter refers to the maximum deceleration that the UAV can stably achieve along the current heading under the current sortie, current load, current battery level, and current wind disturbance conditions.

[0037] Extract the construction area boundary from the work status sheet, deduct the mechanical motion envelope and the temporary control area within the construction area boundary to obtain the candidate space, discretize the candidate space using voxelization, and treat each voxel as a candidate segment to be judged.

[0038] It should be noted that the voxel side length is set to 1 times the length of the UAV fuselage outline.

[0039] Specifically, the voxel side length is set to 1 times the length of the UAV's outer contour because if the voxel is too large, it will obscure the small flyable gaps near the mechanical envelope boundary; if the voxel is too small, it will increase the computational load. Using the length of the UAV's outer contour as the voxel side length ensures that the voxel-level flyability judgment is consistent with the aircraft's passability, and the voxel side length can be adjusted within a range of 1 to 2 times the length of the UAV's outer contour.

[0040] Calculate the shortest distance from the UAV's takeoff point to the center of each candidate segment, and use the ratio of the shortest distance to the cruise speed to obtain the earliest arrival time of the candidate segment. For each candidate segment, calculate the minimum distance between the outer surface of the candidate segment and the nearest mechanical motion envelope or the nearest temporary containment zone boundary to obtain the minimum clearance distance of the candidate segment.

[0041] Based on the cruise speed and braking capability parameters, the response time of the UAV is statistically analyzed, and the braking distance is calculated using the following expression: ; in, The distance required for braking. For cruising speed, For response time, For maximum deceleration, This refers to the distance the drone travels at its cruising speed during the response time phase.

[0042] It should be noted that the response time for stable deceleration was measured through historical tests.

[0043] For the work area time window, earliest arrival time, minimum clearance distance, and braking distance, a multiplicative ratio is used to construct the available values ​​of candidate segments, expressed as follows: ; in, Available values ​​for candidate segments To allow for the end time of the flight, The earliest arrival time, Minimum clearance distance, The permitted start time for flight.

[0044] when At that time, the drone can arrive at the candidate segment before the end of the corresponding time window; when At that time, the drone can complete stable braking without touching the surrounding restricted boundaries.

[0045] only and Only when both conditions are met can it be said that the candidate segment can be actually reached by the drone before the end of the permitted flight time window, and has sufficient clearance to complete stable braking. Therefore, when the available value of the candidate segment is greater than 0, the candidate segment is retained as an available candidate segment, and when the available value of the candidate segment is not greater than 0, the candidate segment is directly eliminated.

[0046] All available candidate segments are merged according to voxel adjacency relationships to form available connected components. If an available connected component is not connected to the voxel where the UAV's takeoff point is located, the UAV cannot enter the available connected component and is directly eliminated. Therefore, only available connected components connected to the voxel where the UAV's takeoff point is located are retained. The number of voxels in each available connected component is counted, and the available connected component with the largest number of voxels is selected as the control corridor.

[0047] It should be noted that adjacency means that two voxels share at least one face in space, thus determining that the two voxels are geometrically connected.

[0048] The verified normal regions in the work condition sheet are discretized using voxel side lengths identical to those in the candidate space, resulting in a set of normal voxels corresponding to the verified normal regions. The set of normal voxels is then compared with the set of voxels in the control corridor. If the absolute values ​​of the differences between the voxel center coordinates of each voxel in the normal voxel set and each voxel in the control corridor along the horizontal, vertical, and height axes are all less than half the voxel side length, the two voxels are considered to overlap, and the overlapping voxels are retained, thus obtaining the set of normal voxels within the control corridor. The set of normal voxels within the control corridor is then partitioned based on voxel adjacency relationships to obtain normal sub-regions within the control corridor.

[0049] Extract all voxels from the control corridor, calculate the horizontal distance between the center points of any two voxels for each pair, and define the pair of voxel center points with the largest horizontal distance as the main direction of the control corridor. Take the voxel center point on the horizontal axis closer to the UAV takeoff point as the starting point of the main direction, and the other end as the ending point.

[0050] Project the center point of each voxel in the normal sub-region of the control corridor onto the main direction of the control corridor, and calculate the distance from each projection point to the starting point of the main direction to obtain the control corridor projection value. Extract the maximum and minimum control corridor projection values ​​and calculate the difference to obtain the continuous flyable length of the normal sub-region of the control corridor. Use the normal sub-region of the control corridor with the largest continuous flyable length as the reference region.

[0051] It should be noted that the continuous flyable length refers to the maximum length that a UAV can continuously pass through in the main direction of the control corridor, provided that there are no interruptions, omissions, or non-adjacent disconnections within the voxel projection interval corresponding to the normal sub-region within the same control corridor.

[0052] It should be noted that if there is no verified normal area in the test condition sheet, the previous round of qualified datum surface shall be used as the basis for determining the reference area; if there is no previous round of qualified datum surface in the test condition sheet, the design datum surface shall be used as the basis for determining the reference area. The reference area shall be determined using the same method as for verified normal areas.

[0053] S2. Determine the first inspection advancement direction and generate the first inspection trajectory based on the horizontal axis direction of the main control corridor; perform benchmark area matching on the first inspection segment and calculate the first inspection callback index; determine the normal first inspection segment or the suspected first inspection segment based on the set first inspection trigger threshold, and classify the suspected first inspection segment to form a suspicious point list.

[0054] The horizontal axis of the control corridor, from its starting point to its ending point, is used as the initial inspection direction. Vertical slices are made along this direction, using the voxel side length as the step size, to obtain initial inspection slices. Within each initial inspection slice, the voxel center points of the control corridor falling within it are extracted, and the horizontal, vertical, and vertical coordinates of these voxel center points are averaged to obtain the path center points of the initial inspection slice. The path center points of the initial inspection slices are then connected sequentially along the initial inspection direction to form the initial inspection trajectory. The UAV performs the initial inspection along this trajectory at its cruising speed, collecting initial inspection sampling data. The area covered by the UAV between two adjacent path center points in the initial inspection trajectory is taken as the initial inspection segment.

[0055] The initial sampling data included visible light images, thermal images, and distance points on the main surface of the pumped storage power station, as well as UAV pose data.

[0056] The visible light image, thermal image, distance point, and UAV pose data in the first inspection segment are aligned according to the timestamp. Using the timestamp of the UAV pose data as a reference, the distance point, visible light image, and thermal image are matched to the sampling time closest to the timestamp to obtain the first inspection sampling data group.

[0057] If a qualified datum surface from the previous round exists in the work condition sheet, it is used as the reference datum surface for the first inspection segment; if no qualified datum surface exists in the work condition sheet, the design datum surface is used as the reference datum surface for the first inspection segment. Following the same method as converting the original boundary point data with engineering coordinates to the on-site three-dimensional coordinate system, all distance points within the first inspection segment are uniformly converted to the on-site three-dimensional coordinate system to obtain the segment surface point set for the first inspection segment. All distance points in the segment surface point set are projected vertically onto the reference datum surface to obtain reference projection points. The maximum and minimum coordinate values ​​of the reference projection points in the horizontal axis direction and the maximum and minimum coordinate values ​​in the vertical axis direction constitute the minimum enclosing range of the reference projection points. The portion of the reference datum surface falling within this minimum enclosing range is used as the comparison datum surface for the first inspection segment. The contour of the reference projection point on the reference base surface is taken as the surface contour of the segment; in the thermal imaging image corresponding to the first inspection segment, all thermal imaging pixels inside the surface contour of the segment are extracted to obtain the thermal region of the first inspection segment; in the visible light image corresponding to the first inspection segment, all visible light pixels inside the surface contour of the segment are extracted to obtain the visible region of the first inspection segment.

[0058] The center point of the first detected segment is compared with the center points of all voxels in the reference area one by one. If the absolute value of the coordinate difference between the center point of the first detected segment and the center point of the voxels in the reference area in the horizontal, vertical and vertical directions is less than half of the side length of the voxel, then the center point of the first detected segment is determined to fall into the reference area, and the first detected segment that falls into the reference area is retained as the reference segment.

[0059] Calculate the initial inspection projection value of the center point of the first inspection segment in the initial inspection advancement direction. Calculate the reference projection values ​​of the center points of each reference segment in the initial inspection advancement direction. Calculate the absolute difference between the initial inspection projection value and all reference projection values, and select the reference segment with the smallest absolute difference as the synchronous comparison segment. If two or more reference segments have the same projection value difference, the reference segment corresponding to the minimum absolute difference between the center elevation of the initial inspection segment and the center elevation of the reference segments is selected as the synchronous comparison segment.

[0060] Calculate the vertical distance between the distance point and the reference projection point within the first inspection segment. If the distance point is located on the positive side of the height axis of the comparison base plane, the vertical distance is recorded as positive; if the distance point is located on the negative side of the height axis of the comparison base plane, the vertical distance is recorded as negative. Calculate the average of all signed vertical distances within the first inspection segment to obtain the signed normal height difference. Calculate the average temperature of all thermal imaging pixels within the thermal region of the first inspection segment to obtain the average temperature of the first inspection segment. Calculate the average temperature of all thermal imaging pixels within the thermal region of the synchronous comparison segment to obtain the average temperature of the synchronous comparison segment. Calculate the difference between the average temperature of the first inspection segment and the average temperature of the synchronous comparison segment to obtain the signed temperature difference of the first inspection segment. Statistically calculate the total boundary length of the segment surface contour on the comparison base plane. Then examine the visible boundary pixel length corresponding to the segment surface contour in the visible area of ​​the first inspection segment. Calculate the ratio of the visible boundary length to the total boundary length of the segment surface contour on the comparison base plane to obtain the visible scale. Based on the visible scale, calculate the gap value, expressed as: ; in, This is the gap value. This is the viewing ratio.

[0061] It should be noted that a larger visible ratio indicates a more complete contour information of the initial inspection segment, meaning the initial inspection has provided sufficient morphological evidence, and the necessity for subsequent local re-inspection is lower. Conversely, a smaller visible ratio indicates that the initial inspection segment has occlusions, missing boundaries, limited viewing angles, or insufficient local sampling. Although abnormal signs have been observed, the evidence of abnormal contours is incomplete, making it more necessary to trigger local re-inspection in subsequent steps. The gap value is defined as... This indicates a gap in evidence and can directly convert "the degree of completeness of evidence" into "the degree of lack of evidence".

[0062] The median of the absolute value of the signed normal elevation difference across all synchronized comparison segments is calculated to obtain the median normal elevation difference. Similarly, the median of the absolute value of the signed temperature difference across all synchronized comparison segments is calculated to obtain the median temperature difference. Based on the median normal elevation difference, median temperature difference, and gap value, the first-inspection pullback index is calculated using the following expression: ; in, The initial check is a pullback index. For the signed normal elevation difference, This is the median value of the normal elevation difference. For temperature differences with signs, This is the median value of the temperature difference. For normalized protection parameters, These are the temperature difference normalization protection parameters.

[0063] It should be noted that the ratio of the absolute value of the signed normal elevation difference to the median value of the normal elevation difference reflects the geometric deviation of the first-inspection segment relative to the baseline state; the larger the value, the more abnormal the geometric shape. The ratio of the absolute value of the signed temperature difference to the median value of the temperature difference reflects the thermal deviation of the first-inspection segment relative to the baseline state; the larger the value, the more abnormal the thermal state. The gap value reflects the evidence gap in the first inspection. The geometric mean is used to ensure that the three factors mutually constrain and work together; the first-inspection pullback index only increases significantly when the combined geometric deviation, thermal deviation, and evidence gap are high. The normalization protection parameter and the temperature difference normalization protection parameter are both minimum values ​​to ensure that the denominator is not zero.

[0064] Following the same method used to calculate the first-inspection callback index, the synchronization callback index of the synchronized comparison segment is calculated. The synchronization callback indices are sorted in ascending order and numbered. If there is only one synchronization callback index, it is used as the first-inspection trigger threshold. If there is more than one synchronization callback index, the 95th percentile position is determined using the following expression: ; in, This corresponds to the 95th percentile. To synchronize the number of pullback indexes.

[0065] If the position corresponding to the 95th percentile is an integer, the synchronization callback index of the corresponding number is taken as the first detection trigger threshold. If the position corresponding to the 95th percentile is not an integer, the synchronization callback index of the two adjacent numbers at the 95th percentile position is linearly interpolated according to the position ratio to obtain the first detection trigger threshold.

[0066] It should be noted that the 95th percentile value is used to determine the first detection trigger threshold in order to ensure that only first detection segments that significantly deviate from the normal distribution are triggered as suspected first detection segments, thereby taking into account both false alarm suppression and abnormal detection sensitivity.

[0067] The first inspection callback index is compared with the first inspection trigger threshold. If the first inspection callback index is less than the first inspection trigger threshold, the corresponding first inspection segment is regarded as a normal first inspection segment; if the first inspection callback index is not less than the first inspection trigger threshold, the corresponding first inspection segment is judged as a suspected first inspection segment.

[0068] All suspected first-inspection segments are arranged sequentially according to the first-inspection progress direction. If the projection contours of adjacent suspected first-inspection segments on the comparison base overlap, or if the projection distance of the center point of adjacent suspected first-inspection segments in the first-inspection progress direction is not greater than the side length of one voxel, then adjacent suspected first-inspection segments are merged into the same suspicious area.

[0069] For each suspicious region, extract the projected contour of the suspicious region on the comparison base plane as the boundary of the suspicious region; connect any two points on the boundary of the suspicious region with a line and calculate the length of each line, select the longest line as the long side of the suspicious region, and use the length of the long side of the suspicious region as the long side dimension of the suspicious region; summarize all suspicious regions into a suspicious list.

[0070] To verify the invention's ability to accurately identify surface anomalies during the initial inspection stage, Figure 5 This illustrates the correlation between the gap value and the first-inspection pullback index. The blue curve represents the change in the first-inspection pullback index for normal first-inspection fragments under different gap value conditions, while the yellow curve represents the change in the first-inspection pullback index for suspected first-inspection fragments under different gap value conditions. Figure 5 As can be seen, with the change of the gap value, the first inspection pullback index corresponding to the suspected first inspection segment is generally higher than that of the normal first inspection segment. This indicates that by integrating the signed normal elevation difference, the signed temperature difference and the gap value, the present invention can better distinguish between the normal first inspection segment and the suspected first inspection segment in the first inspection stage, thus demonstrating the refined identification effect of the construction surface anomaly.

[0071] S3. Generate a re-inspection order based on the suspicious point list, determine the re-inspection trajectory and re-inspection speed, and perform local re-inspection; match, unify coordinates, and extract regions from the re-inspection sampling data to obtain the re-inspection point set, re-inspection outline, re-inspection visible area, and re-inspection hot area, and summarize to generate a re-inspection local information set.

[0072] Projecting the suspicious area onto a horizontal plane yields the projection of the suspicious area. Project the long side of the suspicious area onto the horizontal plane to obtain the horizontal long side direction; within the horizontal plane, take a direction perpendicular to the horizontal long side direction, define it as the offset direction, and calculate the re-inspection offset distance, expressed as: ; in, For the re-inspection offset distance, This represents the length of the long side of the suspicious area.

[0073] It should be noted that the re-inspection offset distance is related to both the spatial dimensions of the suspicious area itself and the degree of missing evidence in the initial inspection. Using half the length of the longer side as the base length allows the re-inspection trajectory to naturally unfold around the suspicious area. Combined with the radical amplification corresponding to the gap value, the offset distance can be kept to change gradually when the gap is small, and the re-inspection range on both sides can be gradually expanded when the gap increases, avoiding the offset distance from increasing too rapidly.

[0074] Using the center point of the projection of the suspicious area as the centerline reference, extend along the long side to generate the center re-inspection line. At the positions on both sides of the center re-inspection line, which differ by the width of the re-inspection offset distance, generate two lateral re-inspection lines parallel to the center re-inspection line.

[0075] Intersecting the three re-inspection lines with the control corridor, only the segments located within the control corridor and covering the projected area of ​​the suspected area are retained, resulting in three re-inspection line segments. These three re-inspection line segments are designated as the first lateral re-inspection line segment, the intermediate re-inspection line segment, and the second lateral re-inspection line segment, respectively, and are parallel to each other. Along the length direction, the first starting point and first ending point of the first lateral re-inspection line segment, the intermediate starting point and intermediate ending point of the intermediate re-inspection line segment, and the second starting point and second ending point of the second lateral re-inspection line segment are determined. The shortest connecting path within the control corridor between the first ending point and the intermediate ending point is taken as the first transition segment, and the shortest connecting path within the control corridor between the intermediate starting point and the second starting point is taken as the second transition segment. These segments are connected in the order of the first lateral re-inspection line segment, the first transition segment, the intermediate re-inspection line segment, the second transition segment, and the second lateral re-inspection line segment to form the re-inspection trajectory.

[0076] Read the drone's cruising speed, combine it with the gap value, and calculate the re-inspection speed. The expression is: ; in, To speed up the re-inspection.

[0077] It should be noted that the re-inspection is a local supplementary flight in the UAV inspection process. The re-inspection speed should be kept within the stable speed range that the UAV can execute under the current flight state. Therefore, the current cruising speed is used as the base speed. The greater the degree of evidence missing in the suspicious area, the less sufficient the contour information of the area was obtained in the first round of observation. During the re-inspection, it is necessary to increase the sampling density per unit length and the continuous frame coverage density by reducing the flight speed, so as to obtain more complete local observation results. Using the radical form in the denominator can make the re-inspection speed decrease monotonically as the degree of evidence missing increases, while keeping the deceleration process smooth and continuous, and preventing sudden speed changes due to slight fluctuations in the degree of evidence missing. Furthermore, when the degree of evidence missing is small, the re-inspection speed is still close to the current cruising speed, while when the degree of evidence missing is large, the re-inspection speed gradually decreases, thus balancing the efficiency of re-inspection and the precision of local observation.

[0078] The region projection outline, region center point, re-inspection offset distance, re-inspection trajectory, and re-inspection speed are summarized into a re-inspection form.

[0079] The drone follows the continuous re-inspection trajectory and performs local re-inspection according to the re-inspection speed. During the local re-inspection process, the drone continuously collects visible light images, thermal images and distance points on the main surface of the pumped storage power station, and continuously records the drone's pose data.

[0080] Using the timestamp of the pose data as the reference timestamp, the distance points, visible light images, and thermal imaging images are matched for the most recent time to form a re-inspection sampling data group. This re-inspection sampling data group is arranged according to the chronological order of the re-inspection trajectory to form a continuous re-inspection sequence. Following the same method as transforming the original boundary point data with engineering coordinates to the on-site 3D coordinate system, all distance points in the re-inspection sampling data group of the continuous re-inspection sequence are uniformly transformed to the on-site 3D coordinate system. If a qualified datum surface from the previous round exists in the work condition sheet, it is used as the re-inspection reference datum surface; otherwise, the design datum surface is used. The spatial range covered by the re-inspection trajectory is used as the re-inspection coverage area. Distance points falling within the re-inspection coverage area are extracted from all distance points already transformed to the on-site 3D coordinate system to form a re-inspection point set.

[0081] Project all distance points in the re-inspection point set onto the re-inspection reference datum plane along the normal direction to obtain the re-inspection projection points. Extract the maximum and minimum values ​​of the re-inspection projection points on the horizontal coordinate and the maximum and minimum values ​​on the vertical coordinate, and use these four boundary values ​​to form the minimum enclosing range; extract the portion of the re-inspection reference datum plane that falls within this minimum enclosing range, and use it as the re-inspection comparison datum plane.

[0082] Extract the outer contour of all re-inspection projection points on the re-inspection comparison base surface as the re-inspection contour. Determine the boundary position of the re-inspection contour in the corresponding visible light image and the boundary position in the corresponding thermal imaging image point by point. Using the boundary position as the boundary, extract all visible light pixels inside the boundary position in the visible light image to form the re-inspection visible area, and extract all thermal imaging pixels inside the boundary position in the thermal imaging image to form the re-inspection thermal area.

[0083] The re-inspection point set, re-inspection comparison datum, re-inspection outline, re-inspection visible area, re-inspection hot area, re-inspection continuous sequence, and corresponding re-inspection trajectory are uniformly summarized into a re-inspection local information set.

[0084] S4. Based on the re-inspection datum, determine the re-inspection suspicious area and reference area, calculate the spatial consistency and symbolic consistency, calculate the re-inspection confirmation value and generate the confirmation threshold according to the spatial consistency and symbolic consistency, determine whether the suspicious area passes the re-inspection and generate the inspection processing form.

[0085] The boundary of the suspicious area in the suspicious area list is orthogonally projected onto the re-inspection comparison base plane along the normal direction to obtain the re-inspection suspicious area. Outside the re-inspection suspicious area, the length of the suspicious area is extended outward along the re-inspection comparison base plane by half to obtain the annular zone. The re-inspection suspicious area is then subtracted from the extended zone to obtain the re-inspection reference area.

[0086] It should be noted that extending the suspicious area outward by half the length of its longer side along the re-examination comparison baseline serves two purposes. First, if the extension distance is too small, the resulting ring area will be too close to the boundary of the suspicious area, making it susceptible to the influence of the suspicious area's edge and weakening the control effect of the re-examination reference area. Second, if the extension distance is too large, it may introduce irrelevant areas far from the suspicious area, reducing the relevance of the re-examination reference area to the current suspicious area. Using half the length of the suspicious area as the extension distance ensures that the resulting ring area is both adjacent and continuous to the suspicious area and has sufficient area to construct the re-examination reference area, thus balancing local comparability and the stability of the re-examination confirmation.

[0087] The re-inspection contour and the re-inspection suspicious area are compared on the re-inspection comparison base plane. The overlapping area and the union area are calculated, and the ratio of the overlapping area and the union area is calculated to obtain the spatial consistency.

[0088] Read the marked normal elevation difference and marked temperature difference from the suspicious point list; according to the order of the continuous re-inspection sequence, extract the portion of the re-inspection point set that falls within the projection of the suspicious point area onto the re-inspection comparison base, thus obtaining a subset of re-inspection points. Project all distance points of the re-inspection point subset onto the re-inspection comparison base along the normal direction of the re-inspection comparison base, calculate the distance between each distance point and the projection point, and add a sign to obtain the marked re-inspection normal distance. Calculate the average value of all marked re-inspection normal distances to obtain the marked re-inspection normal elevation difference. According to the order of the continuous re-inspection sequence, extract the portion falling into the re-inspection suspicious point area from the re-inspection hot area, and calculate the average temperature to obtain the average temperature of the suspicious point; extract the portion falling into the re-inspection reference area from the re-inspection hot area, and calculate the average temperature to obtain the reference average temperature; calculate the difference between the average temperature of the suspicious point and the reference average temperature to obtain the marked re-inspection temperature difference.

[0089] It should be noted that if the distance point of the re-examination point subset is located on the positive side of the high axis of the re-examination comparison base plane, the sign is positive; if the distance point of the re-examination point subset is located on the negative side of the high axis of the re-examination comparison base plane, the sign is negative.

[0090] The number of items in the continuous re-inspection sequence is counted. The sign of the signed normal elevation difference is judged against the sign of the signed re-inspection normal elevation difference. At the same time, the sign of the signed temperature difference is judged against the sign of the signed re-inspection temperature difference. The number of items that simultaneously satisfy the conditions of "the sign of the signed re-inspection normal elevation difference is consistent with the sign of the signed normal elevation difference in the suspicious list" and "the sign of the signed re-inspection temperature difference is consistent with the sign of the signed temperature difference in the suspicious list" is recorded. The number of items with consistent signs is obtained. The ratio of the number of items with consistent signs to the number of items in the continuous re-inspection sequence is calculated to obtain the degree of consistency of signs.

[0091] The re-verification value is calculated using spatial consistency and symbolic consistency, expressed as follows: ; in, This is a value confirmed by re-inspection. For spatial consistency, This represents the degree of symbol consistency.

[0092] It should be noted that the re-inspection confirmation value is calculated using the geometric mean of spatial consistency and symbol consistency. This is because only when the re-inspection contour and the suspected area in the first inspection are spatially aligned, and the signed normal height difference and signed temperature difference obtained from the re-inspection are consistent with the results of the first inspection, can it be determined that the suspected area has been effectively re-inspected and confirmed.

[0093] Within the re-inspection reference area, a sliding window of the same area as the re-inspection suspicious area is used, sliding window by window along the extension direction of the re-inspection trajectory with a step size of half the side length of the voxel. For each sliding window, the same calculation process as for the re-inspection suspicious area is repeated to obtain the corresponding re-inspection confirmation value, forming a set of reference confirmation values. All reference confirmation values ​​are arranged in ascending order, and the confirmation threshold is calculated using the same method as for calculating the initial inspection trigger threshold.

[0094] It should be noted that sliding window by window along the extension direction of the re-inspection trajectory with a step size of half the voxel's side length has two advantages. On the one hand, if the step size is too large, the intervals between adjacent sliding windows may become too wide, potentially missing local changes within the re-inspection reference area and affecting the stability of the reference confirmation value. On the other hand, if the step size is too small, although it can increase the sampling density, it will cause excessive repetitive calculations between sliding windows, increasing the processing burden. Using half the voxel's side length as the step size allows adjacent sliding windows to maintain a moderate overlap, which is beneficial for continuously reflecting local changes in the re-inspection reference area along the extension direction of the re-inspection trajectory, while also balancing computational complexity and the stability of the confirmation threshold calculation.

[0095] The re-inspection confirmation value is compared with the confirmation threshold. If the re-inspection confirmation value is less than the confirmation threshold, the suspicious area is marked as having no abnormality. If the re-inspection confirmation value is not less than the confirmation threshold, the suspicious area is marked as having an abnormality.

[0096] The following information is collected and summarized to obtain the inspection record: the suspicious areas marked as having anomalies, their corresponding boundaries, center coordinates, long side dimensions, signed normal elevation differences, signed temperature differences, gap values, re-inspection contours, continuous re-inspection sequences, corresponding re-inspection trajectories, re-inspection confirmation values, and confirmation thresholds. The time of failure to pass re-inspection is then recorded in each inspection record. All inspection records that failed re-inspection are sorted and summarized according to the time of failure to pass re-inspection to obtain the inspection processing sheet.

[0097] To verify the reliability of this invention in confirming suspicious areas during the re-inspection stage Figure 6The diagram illustrates the correspondence between gap values ​​and re-inspection confirmation values. The yellow curve represents the change in re-inspection confirmation values ​​under different gap value conditions with fixed re-inspection, while the green curve represents the change in re-inspection confirmation values ​​under different gap value conditions with the adaptive re-inspection of this invention. When the gap value changes, the re-inspection confirmation values ​​obtained by the adaptive re-inspection of this invention maintain good stability. This demonstrates that this invention can adjust the re-inspection method according to the gap value, enabling the re-inspection stage to re-acquire more sufficient anomaly information, thereby avoiding the influence of insufficient initial inspection evidence on the final judgment. In other words, even if there is a lack of evidence in the initial inspection stage, reliable confirmation of suspicious areas can still be made through re-inspection.

[0098] In summary, this invention achieves the continuity and integrity of flight routes by: obtaining control corridors through connectivity analysis, generating normal sub-regions within the control corridors based on the control corridors, and determining the reference area according to the continuous flyable length; and by determining whether suspicious areas pass the re-inspection, it realizes the refined identification and reliability confirmation of anomalies on construction surfaces.

[0099] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for inspection and handling of pumped-storage power stations during the construction period based on unmanned aerial vehicles (UAVs), characterized in that, include: Collect on-site control data and UAV real-time status, preprocess the on-site control data to generate a working condition sheet, determine the candidate space and screen the available candidate segments based on the working condition sheet, obtain the control corridor through connectivity analysis, generate normal sub-regions within the control corridor based on the control corridor, and determine the reference area according to the continuous flyable length. The first inspection advance direction is determined according to the horizontal axis direction of the main control corridor, and the first inspection trajectory is generated; the first inspection segment is matched with the reference area and the first inspection callback index is calculated; the first inspection trigger threshold is set and the normal first inspection segment or the suspected first inspection segment is determined; the suspected first inspection segment is grouped and a suspicious point list is formed. A re-inspection order is generated based on the suspicious point list, the re-inspection trajectory and speed are determined, and local re-inspection is performed; the re-inspection sampling data is matched, coordinates are unified, and regions are extracted to obtain the re-inspection point set, re-inspection outline, re-inspection visible area, and re-inspection hot area, and the re-inspection local information set is generated by summarizing them. Based on the re-inspection datum, the suspicious areas and reference areas are determined, spatial consistency and symbolic consistency are calculated, and the re-inspection confirmation value is calculated and a confirmation threshold is generated based on the spatial consistency and symbolic consistency. The questioned areas are then judged to pass the re-inspection and an inspection processing order is generated.

2. The method for inspection and handling of pumped storage power stations during the construction period based on unmanned aerial vehicles (UAVs) as described in claim 1, characterized in that, The on-site control data includes the construction area boundary, mechanical motion envelope, temporary control area, start and end times of permitted flight in the work area, design datum or previous qualified datum, and areas that have been inspected and found to be normal. The real-time status of the UAV includes cruise speed, braking capability parameters, and external dimensions.

3. The method for inspection and handling of pumped storage power stations during the construction period based on unmanned aerial vehicles (UAVs) as described in claim 2, characterized in that, The steps for preprocessing the on-site control data to generate a work status sheet are as follows: The construction area boundary, mechanical movement envelope, temporary control area, design datum or previous qualified datum, and inspected normal area are uniformly converted to the on-site three-dimensional coordinate system. Using the drone takeoff time as the base time, the start and end times of permitted flights in the work area are converted to obtain the work time window, and invalid time windows that do not overlap with the drone mission time are deleted. The converted construction area boundaries, mechanical movement envelopes, temporary control areas, design base surfaces or previous qualified base surfaces, inspected normal areas, and effective operation time windows are summarized to generate a work status sheet.

4. The method for inspection and handling of pumped storage power stations during the construction period based on unmanned aerial vehicles (UAVs) as described in claim 3, characterized in that, The steps for determining the reference zone based on the continuous flyable length are as follows: Candidate spaces are determined based on the construction area boundary, mechanical motion envelope, and temporary control area in the work condition sheet. The candidate spaces are then discretized using voxelization to obtain candidate segments, and usable candidate segments are selected. All available candidate segments are merged according to voxel adjacency, and available connected segments that are connected to the voxel where the UAV takeoff point is located are retained. The available connected segment with the largest number of voxels is determined as the control corridor. The verified normal regions are discretized and overlapped with the control corridor to obtain the set of normal voxels in the control corridor. At the same time, connectivity is divided to obtain normal sub-regions in the control corridor. Determine the main direction of the control corridor, and based on the continuous flyable length of the normal sub-regions within each control corridor along the main direction of the control corridor, select the normal sub-region within the control corridor with the largest continuous flyable length as the reference region.

5. The method for inspection and handling of pumped storage power stations during construction based on unmanned aerial vehicles (UAVs) as described in claim 1, characterized in that, The steps for determining the initial inspection direction and generating the initial inspection trajectory are as follows: Extract all voxel center points in the control corridor, take the voxel center point of the UAV take-off point as the starting point, determine the voxel center point with the farthest horizontal axis distance, and use this to determine the first inspection advancement direction; Along the direction of the first inspection, vertical slicing is performed with the voxel side length as the step size to obtain the first inspection slice, and the center point of the voxel falling into the slice is extracted. Calculate the average values ​​of the x-axis, y-axis, and height axis coordinates of the center points of all voxels in the first inspection slice to obtain the path center point of the first inspection slice; Connect all path center points sequentially according to the direction of the first inspection to form the first inspection trajectory, and collect the first inspection sampling data.

6. The method for inspection and handling of pumped storage power stations during construction based on unmanned aerial vehicles (UAVs) as described in claim 5, characterized in that, The initial inspection sampling data includes visible light images, thermal images, and distance points on the main surface of the pumped storage power station, as well as UAV pose data.

7. The method for inspection and handling of pumped storage power stations during the construction period based on unmanned aerial vehicles (UAVs) as described in claim 6, characterized in that, The steps for grouping suspected initial inspection fragments into a list of suspicious fragments are as follows: The distance point of the first inspection segment is transformed into the on-site three-dimensional coordinate system and projected onto the reference datum plane in the vertical direction to extract the surface contour of the segment and its thermal and visible light images. Determine whether the center point of the first inspection segment falls into the reference area, calculate the signed normal elevation difference, the signed temperature difference and the gap value, and calculate the first inspection pullback index; Set a first inspection trigger threshold, compare the first inspection callback index with the first inspection trigger threshold, and determine whether the first inspection segment is a normal first inspection segment or a suspected first inspection segment. Adjacent suspected initial inspection segments are merged along the initial inspection progress direction, and the boundaries and long side dimensions of the suspected area are extracted to form a suspected segment list.

8. The method for inspection and handling of pumped storage power stations during the construction period based on unmanned aerial vehicles (UAVs) as described in claim 1, characterized in that, The steps for generating a re-inspection form based on the suspicious item form are as follows: The projection of the suspicious area is obtained based on the boundary of the suspicious area, and the horizontal long side direction and offset direction are determined based on the long side direction of the suspicious area. The re-inspection offset distance is calculated based on the long side dimension and gap value of the suspicious area, and the center re-inspection line and the lateral re-inspection lines on both sides are generated with the center point of the projection of the suspicious area as the center line reference. Intersect the central re-inspection line and the lateral re-inspection line with the control corridor, retain the line segment located inside the control corridor that covers the projection range of the suspicious area, and connect them through the transition section to form the re-inspection trajectory; The re-inspection speed is calculated based on the drone's cruising speed and the gap value, and the projection of the suspicious area, the center point of the suspicious area, the re-inspection offset distance, the re-inspection trajectory, and the re-inspection speed are summarized into a re-inspection form.

9. The method for inspection and handling of pumped storage power stations during construction based on unmanned aerial vehicles (UAVs) as described in claim 1 or 8, characterized in that, The steps for summarizing and generating the partial information set for re-examination are as follows: Perform local re-inspection along the re-inspection trajectory at the re-inspection speed, and re-acquire visible light images, thermal images, distance points, and pose data; Based on the timestamp of the reacquired pose data, the reacquired distance points, visible light images and thermal imaging images are matched to form a re-inspection sampling data set, and the distance points are uniformly transformed to the on-site three-dimensional coordinate system to obtain the re-inspection point set; Extract the re-inspection comparison base surface, re-inspection outline, re-inspection visible area, and re-inspection hot area from the re-inspection point set; The re-inspection point set, re-inspection comparison datum, re-inspection outline, re-inspection visible area, re-inspection hot area, re-inspection sampling data group and corresponding re-inspection trajectory are summarized to generate a re-inspection local information set.

10. The method for inspection and handling of pumped storage power stations during construction based on unmanned aerial vehicles (UAVs) as described in claim 9, characterized in that, The steps for determining whether a suspicious area passes re-inspection and generating an inspection processing order are as follows: Based on the boundaries of the suspicious area, determine the suspicious area and the reference area for re-inspection on the re-inspection comparison datum. The re-inspection contour is overlapped and compared with the re-inspection suspicious area to obtain the spatial consistency. The signed re-inspection normal elevation difference and the signed re-inspection temperature difference are calculated and compared with the sign of the signed normal elevation difference and the signed temperature difference to obtain the sign consistency. The re-inspection confirmation value is calculated based on spatial consistency and symbol consistency; Within the re-inspection reference area, a sliding window method is used to obtain the reference confirmation value, and a confirmation threshold is set based on the reference confirmation value; The re-inspection confirmation value is compared with the confirmation threshold to determine whether the suspicious area has passed the re-inspection. The relevant information of the suspicious areas that have not passed the re-inspection is summarized and an inspection processing order is generated.

Images