An unmanned aerial vehicle inspection positioning method and system based on an asymmetric tower

Abstract

This invention discloses a UAV inspection and positioning method and system based on an asymmetric tower. It uses 3D point cloud data to correct the verticality of the tower space, and utilizes geometric priors such as the top plane of the tower to eliminate attitude deviations caused by vibration or tilting of the ground platform, thereby establishing a unified vertical coordinate system. This provides a stable spatial reference for subsequent positioning, avoiding systematic errors caused by ground reference drift. Furthermore, iterative matching of the acquired real-time 3D point cloud data with the initial reference map of the tower space continuously suppresses drift caused by micro-motions or environmental interference on the ground platform during long-term operation, solving the problem of positioning error accumulation in long-endurance UAV inspections and ensuring positioning stability across the entire altitude range. Finally, separating the UAV point cloud from the tower environment point cloud yields reliable real-time UAV position information, resulting in more stable UAV positioning and higher inspection reliability.

Description

Technical Field

[0001] This invention belongs to the field of UAV inspection and positioning technology, and relates to a UAV inspection and positioning method and system based on asymmetric tower. Background Technology

[0002] As a crucial infrastructure component of wind power systems, the regular inspection of the internal structure of wind turbine towers is essential for ensuring the safe and stable operation of wind turbines. With the continuous expansion of wind power installations, tower height and internal space dimensions are increasing, placing higher demands on the efficiency, stability, and automation level of inspection tasks. However, the interior of wind turbine towers, being a large-scale, enclosed, and highly structured working environment, still presents numerous technical challenges when conducting drone inspections. Wind turbine towers have large internal spaces with multiple layers, requiring drones to make multiple up-and-down flights during inspection missions, resulting in lengthy overall inspection times. In the absence of GNSS signals, positioning methods relying solely on the drone's own sensing capabilities are prone to accumulating positioning errors over extended periods, causing the pose estimation to gradually deviate from the true position and failing to meet the positioning stability requirements for full-height tower inspections.

[0003] In highly structured scenarios with repetitive geometric features, such as tower structures, autonomous localization methods based on vision or lidar suffer from feature degradation, leading to decreased localization performance or even complete failure, which seriously jeopardizes the reliability and safety of UAV flight. Furthermore, autonomous localization methods require the UAV to possess its own perception capabilities, and the additional onboard load reduces the UAV's maneuverability and endurance, further decreasing inspection efficiency.

[0004] The interior space of wind turbine towers is relatively open, but there are usually suspended structures such as steel cables and power cables distributed in the air. These slender, non-rigid, or partially obstructed structures can interfere with the positioning of drones, making it difficult for traditional methods to reliably distinguish between the drone and the tower environment, further affecting the positioning accuracy of the drone and the safety of inspections.

[0005] In summary, traditional inspection methods that rely on single-drone autonomous positioning or simple geometric processing are difficult to adapt to the complex working conditions inside wind turbine towers, which involve large scale, long duration, and repetitive structural height. There is an urgent need to study an asymmetric cooperative positioning method for tower inspection scenarios, so as to achieve stable and reliable drone positioning and autonomous inspection while reducing the onboard burden of drones. Summary of the Invention

[0006] The purpose of this invention is to solve the problems of unstable UAV positioning and low inspection reliability caused by the lack of GNSS signals and highly repetitive scene structures during the internal inspection of wind turbine towers in the prior art, and to provide a UAV inspection and positioning method and system based on asymmetric towers.

[0007] To achieve the above objectives, the present invention employs the following technical solution: A UAV inspection and positioning method based on asymmetric towers includes the following steps: Acquire real-time 3D point cloud data of the internal environment of the drone and the tower; Verticality correction of the tower space is performed based on 3D point cloud data. The corrected 3D point cloud data is then overlaid and fused to construct an initial reference map of the tower space. The acquired real-time 3D point cloud data is iteratively matched with the initial reference map of the tower space to estimate the attitude change of the real-time 3D point cloud data relative to the initial reference map. Based on the attitude change, the attitude compensation of the current 3D point cloud data is performed to obtain the corrected and compensated 3D point cloud data. Semantic segmentation is performed on the corrected and compensated 3D point cloud data to obtain separate UAV point clouds and tower environment point clouds. The UAV's position information is determined based on the UAV point cloud, and the inspection task is performed based on the UAV's position information and the tower environment point cloud.

[0008] A further improvement of the present invention is that: The acquisition of real-time 3D point cloud data of the internal environment of the drone and the tower includes: A 3D LiDAR is installed on the ground inside the tower to scan the drone and the internal environment of the tower in real time, obtaining real-time 3D point cloud data of the internal environment of the tower, including the drone.

[0009] The verticality correction of the tower space based on 3D point cloud data includes: The collected raw 3D point cloud data is fitted with a planar model, and distance constraint thresholds and interior point number thresholds are set. Based on the distance constraint threshold and the number of interior points threshold, the largest plane with the most interior points in the space and located in the upper region is identified as the top plane of the tower. Calculate the deviation between the normal vector of the top plane of the tower and the theoretical vertical direction. Based on the deviation calculation results, continue to rotate and compensate the current point cloud to complete the verticality correction of the tower space.

[0010] The process of overlaying and fusing the corrected 3D point cloud data to construct an initial reference map of the tower space includes: The point cloud with verticality correction completed is downsampled, and the point cloud of multiple consecutive frames is superimposed and fused based on the downsampling results to construct an initial reference map of the tower space.

[0011] The process of iteratively matching the acquired real-time 3D point cloud data with the initial reference map of the tower space, estimating the attitude change of the real-time 3D point cloud data relative to the initial reference map, and performing attitude compensation on the current 3D point cloud data based on the attitude change to obtain the corrected and compensated 3D point cloud data includes: The pose transformation of the current point cloud relative to the initial reference map is estimated using the generalized iterative nearest point method.

[0012] The process involves semantic segmentation of the corrected and compensated 3D point cloud data to obtain separate UAV point clouds and tower environment point clouds. Based on the UAV point clouds, the UAV's position information is determined. Based on the UAV's position information and the tower environment point cloud, an inspection task is performed, including: The corrected and compensated 3D point cloud data is input into the LENet 3D semantic segmentation network for stationary point classification, resulting in the separated UAV point cloud and tower environment point cloud. Spatial clustering analysis is performed on the separated UAV point cloud, and the cluster center is located as the real-time three-dimensional position of the UAV to complete the UAV localization. The flight path of the UAV is determined based on the point cloud of the tower environment, and the inspection task is performed in combination with the UAV positioning results.

[0013] A UAV inspection and positioning system based on an asymmetric tower includes: The data acquisition module is used to acquire real-time 3D point cloud data of the internal environment of the drone and the tower; The initial reference map construction module is used to perform verticality correction of the tower space based on 3D point cloud data, and to overlay and fuse the corrected 3D point cloud data to construct an initial reference map of the tower space. The point cloud registration module is used to iteratively match the acquired real-time 3D point cloud data with the initial reference map of the tower space, estimate the attitude change of the real-time 3D point cloud data relative to the initial reference map, perform attitude compensation on the current 3D point cloud data based on the attitude change, and obtain the corrected and compensated 3D point cloud data. The positioning module is used to perform semantic segmentation on the corrected and compensated 3D point cloud data to obtain separate UAV point clouds and tower environment point clouds. Based on the UAV point cloud, the position information of the UAV is determined, and the inspection task is performed based on the position information of the UAV and the tower environment point cloud.

[0014] A computer program product includes a computer program that, when executed by a processor, implements any one of the methods described.

[0015] A terminal device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the steps of any of the methods described above.

[0016] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of any of the methods described herein.

[0017] Compared with the prior art, the present invention has the following beneficial effects: This invention discloses a UAV inspection and positioning method based on an asymmetric tower. It corrects the verticality of the tower space using 3D point cloud data, and utilizes geometric priors such as the top plane of the tower to eliminate attitude deviations caused by vibration or tilt of the ground platform, thereby establishing a unified vertical coordinate system. This provides a stable spatial reference for subsequent positioning, avoiding systematic errors caused by ground reference drift. Furthermore, it iteratively matches the acquired real-time 3D point cloud data with the initial reference map of the tower space, estimating the attitude transformation of the real-time 3D point cloud data relative to the initial reference map. Based on this attitude transformation, attitude compensation is performed on the current 3D point cloud data, continuously suppressing drift caused by micro-motions or environmental interference on the ground platform during long-term operation. This solves the problem of accumulated positioning errors in long-endurance UAV inspections, ensuring positioning stability across the entire altitude range. Finally, it separates the UAV point cloud from the tower environment point cloud, overcoming self-localization failures caused by repetitive internal tower structures and feature degradation, and eliminating interference from suspension structures such as steel cables and power cables on target recognition. This provides reliable real-time UAV position information, resulting in more stable UAV positioning and higher inspection reliability. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart disclosed in an embodiment of the present invention; Figure 2a This is a schematic diagram of the top surface point cloud disclosed in an embodiment of the present invention; Figure 2b This is a diagram of the point cloud and plane normal vectors before correction, as disclosed in an embodiment of the present invention. Figure 2c This is the corrected point cloud and plane normal vector diagram disclosed in an embodiment of the present invention; Figure 3This is a schematic diagram of the point cloud registration result disclosed in the embodiments of the present invention (where a represents the original point cloud image; b represents a schematic diagram of the point cloud translated by 0.5 meters; c represents a schematic diagram of the point cloud rotated by 30 degrees around the z-axis; d represents a schematic diagram of the point cloud simultaneously translated and rotated; e represents a schematic diagram of the point cloud after registration). Figure 4 The image shown is a semantic segmentation result of UAV point cloud disclosed in an embodiment of the present invention (where a represents the complete point cloud image; b represents the UAV point cloud image; and c represents the tower point cloud image). Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0021] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0022] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0023] In the description of the embodiments of the present invention, it should be noted that if terms such as "upper," "lower," "horizontal," or "inner" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship commonly used when the product of the invention is in use, they are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present invention. Furthermore, terms such as "first" and "second" are only used to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0024] Furthermore, the use of the term "horizontal" does not imply that the component must be absolutely horizontal, but rather that it can be slightly tilted. For example, "horizontal" simply means that its direction is more horizontal than "vertical," and does not mean that the structure must be completely horizontal, but can be slightly tilted.

[0025] In the description of the embodiments of the present invention, it should also be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," "connect," and "link" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in the present invention according to the specific circumstances.

[0026] The present invention will now be described in further detail with reference to the accompanying drawings: See Figures 1 to 4 This invention discloses a UAV inspection and positioning method based on asymmetric towers. By introducing a collaborative working mechanism between a ground platform and the UAV, it fully utilizes the advantages of the ground platform in terms of perception capabilities and computing resources, reducing the UAV's onboard burden while achieving stable and reliable positioning of the UAV in an environment where GNSS signals are denied inside the wind turbine tower. The method mainly includes: tower point cloud perception and correction technology based on a ground platform, UAV positioning technology based on point cloud semantic segmentation, and a ground-air collaborative positioning mechanism based on asymmetric structures. The main innovation of this embodiment lies in: First, an asymmetric ground-air cooperative positioning method for wind turbine tower interior scenarios is proposed. This method utilizes a ground platform to replace UAVs in performing high-load sensing and computing tasks, effectively avoiding the degradation problem of UAV autonomous positioning methods. Furthermore, while ensuring positioning accuracy and system stability, it significantly reduces the UAV's onboard burden, improving the overall system endurance and engineering feasibility.

[0027] Secondly, based on the prior geometric characteristics inside the tower, verticality correction and drift suppression of point clouds were achieved, providing a stable spatial reference for long-term inspection tasks.

[0028] Third, a 3D point cloud semantic segmentation method is introduced to reliably distinguish between the UAV and the tower environment, avoiding the influence of interfering structures such as suspension cables on the positioning results. It also facilitates the expansion of multi-UAV collaborative inspection scenarios.

[0029] Example 1 This invention discloses a UAV inspection and positioning method based on an asymmetric tower, comprising the following steps: Step 1: Acquire real-time 3D point cloud data of the internal environment of the drone and the tower; Step 2: Perform verticality correction of the tower space based on 3D point cloud data, and overlay and fuse the corrected 3D point cloud data to construct an initial reference map of the tower space; Step 3: Iteratively match the acquired real-time 3D point cloud data with the initial reference map of the tower space, estimate the attitude change of the real-time 3D point cloud data relative to the initial reference map, perform attitude compensation on the current 3D point cloud data based on the attitude change, and obtain the corrected and compensated 3D point cloud data. Step 4: Perform semantic segmentation on the corrected and compensated 3D point cloud data to obtain separate UAV point cloud and tower environment point cloud. Determine the UAV's position information based on the UAV point cloud, and perform the inspection task based on the UAV's position information and the tower environment point cloud.

[0030] Specifically, the present invention will be further explained and illustrated through the following embodiments: Example 2 This invention discloses a UAV inspection and positioning method based on an asymmetric tower, comprising the following steps: Step 1: Tower point cloud sensing and correction technology based on ground platform, specifically including: Using a 3D lidar deployed on a ground platform, the internal environment of the tower is continuously scanned to obtain 3D point cloud data containing the tower structure and UAV targets.

[0031] Furthermore, to address attitude disturbances caused by uneven ground platform installation or operation, the collected point cloud is vertically corrected based on the geometric prior characteristics of the top plane inside the tower, ensuring that the tower's axis is aligned with a unified reference coordinate system. Simultaneously, continuous point clouds are matched using a point cloud registration method, effectively suppressing translational and rotational drift generated by the ground platform during long-term operation, providing a stable spatial reference for subsequent UAV positioning.

[0032] It should be noted that in this step, the ground platform, as the main sensing and computing unit in the collaborative system, completes the mapping of the internal environment of the wind turbine tower, the processing of 3D point cloud data, and the estimation of the UAV's position. The UAV is only equipped with flight and basic sensing sensors, and completes its own motion control and the inspection of the inside of the wind turbine tower based on the positioning results fed back by the ground platform.

[0033] Step 2: UAV localization technology based on point cloud semantic segmentation, specifically including: Based on the point cloud correction and preprocessing, a lightweight semantic segmentation network for 3D LiDAR point clouds is introduced to perform semantic parsing on the point cloud inside the tower, effectively distinguishing the UAV target point cloud from the tower environment point cloud.

[0034] By performing geometric statistics on the segmented UAV point cloud, the spatial position of the UAV in the lidar coordinate system is calculated, enabling the UAV's external perception and localization. Simultaneously, the segmented tower environment point cloud provides necessary environmental information support for subsequent path planning, obstacle avoidance, and mission status assessment.

[0035] This invention discloses an asymmetric ground-air cooperative positioning mechanism. Addressing the issues of missing GNSS signals inside the tower and the tendency for errors to accumulate and degrade in UAV autonomous positioning, this mechanism constructs an asymmetric ground-air cooperative positioning mechanism. This mechanism uses a ground platform as the primary sensing and computing unit, which completes environmental mapping, point cloud processing, and UAV position estimation. The UAV only needs to carry necessary flight and basic sensing sensors, and completes motion control and inspection tasks based on the position information provided by the ground platform. This significantly reduces the UAV's onboard load and extends its endurance while ensuring positioning accuracy and system stability.

[0036] This invention constructs an asymmetric collaborative positioning core framework between a ground platform and a UAV, directly addressing the instability in UAV positioning caused by missing GNSS signals and structural repetition within wind turbine towers. By employing a division of labor where the ground platform leads perception and computation, and the UAV only performs flight inspections, the error accumulation and feature degradation issues inherent in single-UAV autonomous positioning are largely avoided. Simultaneously, the UAV's onboard configuration is simplified, balancing positioning stability and inspection efficiency. This process forms a complete closed loop from point cloud acquisition and processing to positioning feedback and inspection execution. The results of each step provide data support for subsequent steps, ensuring the continuity of positioning and inspection. It is adaptable to the complex, large-scale, long-term inspection conditions inside wind turbine towers, laying the foundation for subsequent detailed technical optimization.

[0037] Example 3 This invention discloses a UAV inspection and positioning method based on an asymmetric tower, comprising the following steps: Step 1: System Initialization: Start the ground platform and UAV collaborative system, complete the power-on operation and driver loading of hardware devices such as lidar, industrial control host, and communication module, and establish a TCP / IP communication link between the UAV and the ground platform.

[0038] Step 2: Raw point cloud data acquisition: The ground platform activates a 3D lidar to continuously scan the internal space of the tower, acquiring raw 3D point cloud data containing the UAV and the tower structure, providing stable and reliable input for subsequent planar extraction and attitude correction.

[0039] Step 3: Perpendicularity correction based on plane fitting: See Figures 2a to 2cThe random sampling consistency method is used to fit the plane model to the collected raw point cloud data. Under the conditions of setting distance constraints and threshold for the number of interior points, the largest plane with the most interior points in the space and located in the upper region is extracted and identified as the ceiling structure of the current tower layer.

[0040] Analyze the deviation between the normal vector direction of the plane and the theoretical vertical direction to determine if the ground platform is tilted. When the deviation exceeds a set threshold, perform corresponding rotation compensation on the entire point cloud frame to restore the ceiling plane to a horizontal state, thereby completing the verticality correction of the tower space and eliminating attitude errors caused by ground vibration or slight equipment tilt.

[0041] This step clarifies the geometric priors for point cloud verticality correction and proposes a precise correction scheme to address point cloud data deviations caused by uneven ground platform installation and operational attitude disturbances. The correction method, based on the geometric characteristics of the tower's internal top plane, can quickly align the tower's axis with a unified reference coordinate system, eliminating the interference of platform attitude errors on subsequent point cloud processing and positioning calculations, resulting in more accurate spatial reference for the tower's point cloud map. Compared to generic correction methods without specific targeting, this approach relies on the tower's own structural features, making the correction process more aligned with actual inspection scenarios. It eliminates the need for additional calibration equipment, reducing operational complexity, and provides standardized point cloud data for subsequent point cloud registration and semantic segmentation, improving the fundamental data quality of the overall positioning process.

[0042] Step 4: Initial map construction: After verticality correction, the point cloud is subjected to voxel downsampling to reduce the number of points while preserving spatial structure features, thus lowering computational complexity. Subsequently, multiple consecutive frames of point clouds are overlaid and fused within a short time window to construct an initial reference map for subsequent continuous point cloud matching and drift suppression.

[0043] This step defines the target scenario for point cloud preprocessing, clarifying that preprocessing aims to address attitude disturbances during the installation and operation of the ground platform, making the point cloud preprocessing more precise and targeted. By correcting point cloud deviations caused by disturbances in advance, invalid and distorted point cloud data can be filtered out, preventing the accumulation and amplification of disturbance errors in subsequent steps such as verticality correction and point cloud registration, thus ensuring the original accuracy of the point cloud data. As a prerequisite for point cloud processing, this preprocessing step makes subsequent correction and registration operations more efficient, eliminating the need for repeated filtering and correction of complex data, reducing the computational load on the ground platform, and making the constructed tower point cloud map more closely resemble the actual environment, providing a more reliable spatial data foundation for UAV positioning.

[0044] Step 5: Drift suppression based on point cloud registration: See Figure 3During the inspection process, the point cloud data collected in real time for each frame is iteratively matched with the constructed initial reference map. The optimal pose transformation relationship of the current point cloud relative to the reference map is estimated by the generalized iterative nearest point method. This transformation is used to perform attitude compensation on the current frame point cloud, continuously correcting the position offset and rotation error caused by ground platform vibration, slight movement, or environmental interference, thereby suppressing attitude drift during the inspection process and ensuring the robustness of the system.

[0045] This step refines the suppression targets for point cloud registration, specifically addressing translational and rotational drift during long-term operation of the ground platform. It fills the gap in detail by only mentioning attitude drift, making the technical objectives of point cloud registration more concrete. By matching continuous point clouds through point cloud registration, the two types of positional deviations generated during long-term platform inspections can be offset in real time, effectively maintaining the spatial stability of the tower point cloud map and preventing positioning reference frame shifts due to platform drift. This ensures that subsequent UAV positioning calculations are always based on a stable spatial reference. This feature guarantees that the ground platform can still provide continuously accurate point cloud data during long-term, uninterrupted tower inspections, adapting to the multi-layered, high-altitude inspection needs within the tower and improving the long-term operational reliability of the method.

[0046] Step 6: Drone and obstacle recognition based on semantic segmentation: See Figure 4 After completing verticality correction and drift suppression, the previously fitted top surface point cloud is removed to ensure the structural consistency of the tower point cloud in the environment. The processed 3D point cloud data is then input into the LENet 3D semantic segmentation network for point-by-point classification, achieving separation between the UAV point cloud and the tower environment point cloud.

[0047] This step clarifies the type and core functions of the semantic segmentation network, adopting a lightweight semantic segmentation network for 3D LiDAR point clouds, balancing segmentation accuracy and computational efficiency. The lightweight design adapts to the computing resources of the ground platform, avoiding computational latency caused by complex networks. Furthermore, the dedicated design for 3D LiDAR point clouds enables accurate parsing of the spatial characteristics of the point cloud inside the tower, effectively distinguishing the UAV from the tower environment. Compared to general semantic segmentation networks, this feature better adapts to scenarios with interfering structures such as steel cables and electrical cables inside the tower, reducing interference from environmental point clouds on the UAV point cloud, improving the accuracy of point cloud segmentation, and providing clean target point cloud data for subsequent UAV spatial positioning calculations.

[0048] This step refines the method for calculating the UAV's spatial position. By performing geometric statistics on the UAV point cloud, positioning is achieved, making the positioning process more operable and accurate. Geometric statistics can extract precise position parameters from the spatial distribution characteristics of the UAV point cloud, directly calculating the UAV's position in the LiDAR coordinate system, avoiding errors caused by complex coordinate transformations and parameter estimations. This calculation method relies on the segmented, clean UAV point cloud, reducing environmental interference and resulting in positioning results that more closely match the UAV's true position. Furthermore, the calculation process is simple and efficient, quickly feeding the positioning results back to the UAV, meeting the real-time requirements of UAV flight control and improving the motion control accuracy of the UAV during inspections.

[0049] Step 7: Drone positioning and obstacle information output: For the UAV point cloud set obtained from semantic segmentation, its spatial cluster center is calculated as the UAV's real-time 3D position. Simultaneously, the spatial distance relationship between the UAV point cloud and the environmental point cloud is analyzed to extract the nearest obstacle distance information. Finally, the UAV position information and obstacle distance information are fed back to the UAV flight control system in real time via a communication link to complete the UAV localization task in GNSS-denied environments and provide obstacle avoidance assistance information.

[0050] This study unlocks the added value of tower environmental point clouds, enabling the segmented point clouds to serve as auxiliary support for UAV inspections. This achieves full utilization of point cloud data and avoids data resource waste. The environmental point clouds provide precise information about the tower's internal environment for UAV path planning and obstacle avoidance, allowing UAVs to proactively avoid suspended structures such as steel cables and power cables, thus improving flight safety. Simultaneously, it provides data for task status assessment, allowing UAVs to accurately determine whether the inspection area has been completed based on environmental characteristics, enhancing the intelligence and accuracy of the inspection. This feature extends the value of point cloud semantic segmentation from simple localization to the entire inspection process, deeply integrating collaborative localization methods with UAV inspection tasks and improving the overall practicality of the inspection system.

[0051] Furthermore, this embodiment clarifies the asymmetric division of labor mechanism for ground-air collaboration, defines the core functional boundaries of the ground platform and the UAV, and further strengthens the core innovation of the collaborative positioning method. The ground platform, as the main sensing and computing unit, undertakes high-load point cloud processing and positioning calculation tasks, completely avoiding the reduced mobility and endurance issues caused by the high-load equipment onboard the UAV; the UAV only carries basic sensors and performs tasks based on ground positioning results, significantly reducing the onboard burden and extending endurance. This division of labor fully leverages the advantages of the ground platform's computing resources and sensing capabilities, while allowing the UAV to focus on the core inspection task, significantly improving inspection efficiency and engineering feasibility while ensuring positioning accuracy and system stability.

[0052] Furthermore, this embodiment clarifies two key characteristics of the method: UAV movement does not affect point cloud registration accuracy, and it can be expanded to multi-UAV collaborative inspection, significantly improving the method's practicality and scalability. UAV movement does not interfere with registration accuracy, ensuring the continuous stability of the positioning reference system during inspection and preventing UAV flight from interfering with point cloud processing on the ground platform, thus improving the continuity and accuracy of positioning. Support for multi-UAV collaborative expansion adapts to the inspection needs of large spaces and multiple levels inside the tower, and can significantly shorten inspection time and improve inspection efficiency through simultaneous operation of multiple UAVs. This characteristic makes the method not only suitable for single-UAV inspection but also meets the needs of large-scale and efficient inspections, aligning with the development trend of expanding installed capacity and increasing tower size in the wind power industry, and possessing stronger engineering application value.

[0053] The method of this invention constructs an asymmetric ground-air cooperative positioning system, which accurately solves the positioning problem inside the tower, ensures the accuracy of point cloud through calibration and registration, avoids environmental interference through semantic segmentation, reduces the load of UAVs and improves their endurance through division of labor, makes full use of data and supports multi-UAV expansion, provides stable positioning and efficient inspection, and is suitable for the actual engineering needs of the wind power industry.

[0054] The present invention also discloses a simulation experiment based on this embodiment. To verify the technical advantages of the asymmetric tower UAV cooperative positioning method described in this invention, a 1:1 simulation experimental environment of the wind turbine tower interior (tower height 80m, with multiple platforms and suspension interference structures such as steel cables and cables) was built, taking into account the actual working conditions of missing GNSS signals, repetitive structural features, and the presence of suspended interference structures inside the wind turbine tower. Traditional UAV single-unit visual positioning methods and traditional lidar autonomous positioning methods were selected as comparison groups. Simulation experiments were conducted to test the four core indicators of the method of this invention: positioning accuracy, positioning stability, endurance, and inspection efficiency. At the same time, the engineering applicability of the method was verified through actual tower field experiments. The experimental and simulation data are as follows, and the technical advantages of this invention are further discussed in conjunction with the data.

[0055] (a) Positioning accuracy test Refer to Table 1. Test scenario: Inspection process of the tower at full height (80m). Five key height measuring points of 10m, 30m, 50m, 70m and 80m were selected to test the positioning error of the three methods respectively. The average value of multiple tests was taken, and the unit is centimeters (cm).

[0056] Table 1 Positioning accuracy test data

[0057] In actual 80m wind turbine tower inspections, the average positioning error of the method of this invention is 1.8cm, and the maximum positioning error does not exceed 2.0cm, which meets the millimeter-level positioning accuracy requirements for tower inspections.

[0058] (ii) Positioning stability test Refer to Table 2. Test scenario: continuous 2-hour round-trip inspection inside the tower, recording the cumulative positioning error and the number of positioning failures for the three methods. The positioning failure criterion is that the positioning error exceeds 50cm.

[0059] Table 2 Positioning Stability Test Data

[0060] During a 3-hour continuous tower inspection, the cumulative error of the method of this invention was only 2.5cm, with no positioning failures throughout the process and a consistently stable positioning status.

[0061] (III) Unmanned Aerial Vehicle Endurance Test Refer to Table 3. Test scenario: With the same drone hardware configuration (same model drone, same capacity battery), three positioning methods were used to test the duration of tower inspection and the height of the inspectable tower when the drone is fully charged. Units: minutes (min) and meters (m).

[0062] Table 3. Test data on drone endurance

[0063] In actual inspections, the UAV using the method of this invention can complete full-height inspections of three 80m towers when fully charged, improving endurance efficiency by more than 120% compared to traditional methods.

[0064] (iv) Inspection efficiency test See Table 4. Test scenario: Inspection of the entire internal area of ​​a single 80m wind turbine tower. Record the total inspection time for the three methods, in minutes (min), including the entire process of drone flight, positioning, and data acquisition.

[0065] Table 4 Inspection Efficiency Test Data

[0066] Two drones using the method of this invention conduct collaborative inspections, reducing the total inspection time for a single tower to 12 minutes, and improving inspection efficiency by more than 80% compared to the traditional single-drone method.

[0067] Simulation tests have verified that the present invention has the following advantages: (i) Positioning accuracy and stability are improved by orders of magnitude, solving the problem of positioning inside the tower. As can be seen from the positioning accuracy data, the positioning error of traditional positioning methods increases significantly with the tower height. The traditional visual positioning method has a positioning error of 38.6cm at a height of 80m, and the lidar autonomous positioning method has an error of 24.7cm. In contrast, the positioning error of the method of this invention is controlled within 1.6cm at all heights, and the error in the field experiment does not exceed 2.0cm, which is more than 10 times higher than that of traditional methods. In terms of positioning stability, traditional methods suffer from significant error accumulation due to missing GNSS signals and feature degradation. The cumulative error exceeds 30cm in 2 hours, and there are multiple positioning failures. In contrast, the method of this invention relies on the point cloud correction and registration technology of the ground platform and the asymmetric cooperative mechanism. The cumulative error is only 1.9cm in 2 hours, with no positioning failures throughout the process. It remains stable even after 3 hours of continuous on-site inspection. The core reason is that this invention achieves point cloud verticality correction and drift suppression based on the prior geometric characteristics of the tower, providing a stable spatial reference for positioning. At the same time, it achieves accurate differentiation between the UAV and the environment through three-dimensional point cloud semantic segmentation, avoiding the influence of interfering structures such as steel cables and cables. From the data level, it is verified that this invention completely solves the technical pain points of unstable positioning, error accumulation, and easy failure of traditional methods.

[0068] (ii) The endurance of drones has been greatly improved, making them suitable for large-scale, long-term inspection needs of towers. Simulation and experimental data show that the UAV's full-charge inspection duration under the method of this invention reaches 85 minutes, more than double that of the traditional visual positioning method (42 minutes) and lidar autonomous positioning method (38 minutes). The fully charged inspection altitude increases from over 100 meters in the traditional method to 280 meters. This data directly confirms the technical advantages of the asymmetric ground-air collaborative division of labor mechanism of this invention: the ground platform undertakes the high-load point cloud perception, processing, and computing tasks, while the UAV only needs to carry basic flight and perception sensors and does not need to bear the high-load autonomous positioning calculations, significantly reducing the airborne load and the UAV's energy consumption. Traditional methods, due to the continuous high-load operation of the UAV's onboard perception and computing equipment, result in rapid battery consumption and limited endurance, making it difficult to meet the long-term inspection requirements of large-scale, high-altitude towers. This invention, through energy consumption optimization, achieves a leapfrog improvement in endurance. In field experiments, it can complete the full-height inspection of three 80-meter towers, significantly reducing the number of UAV charging cycles and improving the continuity of actual engineering inspections.

[0069] (III) Inspection efficiency is significantly improved, taking into account both single-machine efficiency and the feasibility of multi-machine expansion. Single-tower inspection data shows that the method of this invention reduces the total inspection time from over 50 minutes using traditional methods to 22 minutes, more than doubling the inspection efficiency. The core reasons are twofold: First, the positioning process of this invention eliminates the need for the UAV to autonomously extract, match, and calculate features. The ground platform can quickly complete point cloud processing, positioning calculation, and feedback the results, significantly improving the real-time performance of UAV flight control and reducing positioning waiting time. Second, the segmented tower environment point cloud provides precise support for UAV path planning and obstacle avoidance, avoiding the time loss caused by obstacle avoidance errors and path deviations in traditional methods. Simultaneously, in multi-UAV expansion experiments, collaborative inspection by two UAVs reduced the single-tower inspection time to 12 minutes, verifying the advantage of this invention's method in facilitating multi-UAV collaborative expansion. Because the ground platform's point cloud semantic segmentation can accurately distinguish the point cloud targets of multiple UAVs, no additional hardware modifications to the UAVs are required; only multi-target point cloud positioning calculations need to be completed on the ground platform. This adapts to the actual engineering needs of the wind power industry, which is expanding in installed capacity and increasing in the number of towers. Multi-UAV collaboration can further improve inspection efficiency and solve the problem of multi-UAV collaborative positioning that is difficult to achieve with traditional methods.

[0070] (iv) It has strong engineering applicability, and the simulation data is highly consistent with the field experimental data, which can be directly applied in practice. The simulation test results of this invention closely match the actual on-site experimental data of wind turbine towers. In the simulation, the average positioning error of the method was approximately 1.5 cm, while in the on-site experiment it was 1.8 cm. The simulated endurance time was 85 minutes, and the actual endurance efficiency in the on-site experiment was improved by more than 120%. This demonstrates that the technical solution of this invention is not merely a theoretical design, but rather fully integrates the actual inspection conditions of wind turbine towers, possessing strong engineering feasibility. While traditional methods have a certain positioning effect in ideal laboratory scenarios, their positioning performance significantly decreases in the complex environment of actual wind turbine towers due to feature repetition and the presence of interfering structures. This invention addresses the actual problems inside the tower, with the ground platform's 3D lidar adapting to the enclosed, large-scale environment of the tower. Point cloud correction and segmentation technologies specifically solve the problems of platform attitude disturbance and environmental interference. On-site experimental results show that it can be directly applied to the inspection work of actual wind turbine towers, filling the gap in the engineering applicability of traditional methods.

[0071] Therefore, simulation and field experimental data quantitatively verify the technical advantages of the asymmetric tower UAV collaborative positioning method of this invention from multiple dimensions such as positioning accuracy, stability, endurance, inspection efficiency, and engineering applicability. Compared with the traditional UAV single-unit autonomous positioning method, this invention completely solves the problems of GNSS signal loss, structural feature duplication, and unstable positioning and low inspection reliability caused by interference structures inside wind turbine towers through core technologies such as asymmetric collaboration between the ground platform and the UAV, point cloud correction and drift suppression, and 3D point cloud semantic segmentation. At the same time, it achieves a dual improvement in UAV endurance and inspection efficiency, and supports multi-UAV collaborative expansion. The high degree of consistency between simulation and field data also proves that the method of this invention has strong engineering applicability and can adapt to the development trend of increasing tower height and expanding installed capacity in the wind power industry. It provides a stable, reliable, and efficient positioning solution for UAV inspection inside wind turbine towers and has important engineering application value.

[0072] Example 4 This invention discloses a UAV inspection and positioning system based on an asymmetric tower, comprising: The data acquisition module is used to acquire real-time 3D point cloud data of the internal environment of the drone and the tower; The initial reference map construction module is used to perform verticality correction of the tower space based on 3D point cloud data, and to overlay and fuse the corrected 3D point cloud data to construct an initial reference map of the tower space. The point cloud registration module is used to iteratively match the acquired real-time 3D point cloud data with the initial reference map of the tower space, estimate the attitude change of the real-time 3D point cloud data relative to the initial reference map, perform attitude compensation on the current 3D point cloud data based on the attitude change, and obtain the corrected and compensated 3D point cloud data. The positioning module is used to perform semantic segmentation on the corrected and compensated 3D point cloud data to obtain separate UAV point clouds and tower environment point clouds. Based on the UAV point cloud, the position information of the UAV is determined, and the inspection task is performed based on the position information of the UAV and the tower environment point cloud.

[0073] A schematic diagram of a terminal device according to an embodiment of the present invention. The terminal device of this embodiment includes: a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps in the various method embodiments described above. Alternatively, when the processor executes the computer program, it implements the functions of each module / unit in the various device embodiments described above.

[0074] The computer program can be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention.

[0075] The terminal device can be a desktop computer, laptop computer, cloud server, or other device with strong computing power. The terminal device may include, but is not limited to, a processor and memory.

[0076] The optimal choice for the processor is a multi-core high-speed central processing unit (CPU).

[0077] The memory can be used to store the computer program and / or module. The processor implements various functions of the terminal device by running or executing the computer program and / or module stored in the memory and calling the data stored in the memory.

[0078] If the modules / units integrated into the terminal device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0079] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. An unmanned aerial vehicle inspection positioning method based on an asymmetric tower, characterized in that, Includes the following steps: Acquire real-time 3D point cloud data of the internal environment of the drone and the tower; Verticality correction of the tower space is performed based on 3D point cloud data. The corrected 3D point cloud data is then overlaid and fused to construct an initial reference map of the tower space. The acquired real-time 3D point cloud data is iteratively matched with the initial reference map of the tower space to estimate the attitude change of the real-time 3D point cloud data relative to the initial reference map. Based on the attitude change, the attitude compensation of the current 3D point cloud data is performed to obtain the corrected and compensated 3D point cloud data. Semantic segmentation is performed on the corrected and compensated 3D point cloud data to obtain separate UAV point clouds and tower environment point clouds. The UAV's position information is determined based on the UAV point cloud, and the inspection task is performed based on the UAV's position information and the tower environment point cloud.

2. The UAV inspection and positioning method based on an asymmetric tower according to claim 1, characterized in that, The acquisition of real-time 3D point cloud data of the internal environment of the drone and the tower includes: A 3D LiDAR is installed on the ground inside the tower to scan the drone and the internal environment of the tower in real time, obtaining real-time 3D point cloud data of the internal environment of the tower, including the drone.

3. The UAV inspection and positioning method based on an asymmetric tower according to claim 1, characterized in that, The verticality correction of the tower space based on 3D point cloud data includes: The collected raw 3D point cloud data is fitted with a planar model, and distance constraint thresholds and interior point number thresholds are set. Based on the distance constraint threshold and the number of interior points threshold, the largest plane with the most interior points in the space and located in the upper region is identified as the top plane of the tower. Calculate the deviation between the normal vector of the top plane of the tower and the theoretical vertical direction. Based on the deviation calculation results, continue to rotate and compensate the current point cloud to complete the verticality correction of the tower space.

4. The UAV inspection and positioning method based on an asymmetric tower according to claim 3, characterized in that, The process of overlaying and fusing the corrected 3D point cloud data to construct an initial reference map of the tower space includes: The point cloud with verticality correction completed is downsampled, and the point cloud of multiple consecutive frames is superimposed and fused based on the downsampling results to construct an initial reference map of the tower space.

5. The UAV inspection and positioning method based on an asymmetric tower according to claim 1, characterized in that, The process of iteratively matching the acquired real-time 3D point cloud data with the initial reference map of the tower space, estimating the attitude change of the real-time 3D point cloud data relative to the initial reference map, and performing attitude compensation on the current 3D point cloud data based on the attitude change to obtain the corrected and compensated 3D point cloud data includes: The pose transformation of the current point cloud relative to the initial reference map is estimated using the generalized iterative nearest point method.

6. The UAV inspection and positioning method based on an asymmetric tower according to claim 1, characterized in that, The process involves semantic segmentation of the corrected and compensated 3D point cloud data to obtain separate UAV point clouds and tower environment point clouds. Based on the UAV point clouds, the UAV's position information is determined. Based on the UAV's position information and the tower environment point cloud, an inspection task is performed, including: The corrected and compensated 3D point cloud data is input into the LENet 3D semantic segmentation network for stationary point classification, resulting in the separated UAV point cloud and tower environment point cloud. Spatial clustering analysis is performed on the separated UAV point cloud, and the cluster center is located as the real-time three-dimensional position of the UAV to complete the UAV localization. The flight path of the UAV is determined based on the point cloud of the tower environment, and the inspection task is performed in combination with the UAV positioning results.

7. A UAV inspection and positioning system based on an asymmetric tower, characterized in that, include: The data acquisition module is used to acquire real-time 3D point cloud data of the internal environment of the drone and the tower; The initial reference map construction module is used to perform verticality correction of the tower space based on 3D point cloud data, and to overlay and fuse the corrected 3D point cloud data to construct an initial reference map of the tower space. The point cloud registration module is used to iteratively match the acquired real-time 3D point cloud data with the initial reference map of the tower space, estimate the attitude change of the real-time 3D point cloud data relative to the initial reference map, perform attitude compensation on the current 3D point cloud data based on the attitude change, and obtain the corrected and compensated 3D point cloud data. The positioning module is used to perform semantic segmentation on the corrected and compensated 3D point cloud data to obtain separate UAV point clouds and tower environment point clouds. Based on the UAV point cloud, the position information of the UAV is determined, and the inspection task is performed based on the position information of the UAV and the tower environment point cloud.

8. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1-6.

9. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1-6.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1-6.

Images