Closed inclined space detection method and device based on laser inertial odometer, unmanned aerial vehicle system and computer readable storage medium

Abstract

The application provides a kind of closed inclined space detection method and device based on laser inertial odometer, belong to unmanned aerial vehicle inspection technical field.Method includes: obtaining laser radar point cloud data and inertial measurement unit (IMU) inertial data;Based on the motion compensation of point cloud data on inertial data, to eliminate motion distortion;For the characteristics of closed inclined space, the gravity vector in the integral process of IMU is estimated and compensated on line, to improve the accuracy of attitude solution;According to the size information of space, dynamically adjust the scanning parameters of laser radar, to adapt to narrow environment;Finally, through the tight coupling fusion algorithm, based on the data of compensation and adjustment, realize the high-precision positioning and environment mapping of mobile platform.The application also includes structure anomaly recognition and map annotation based on de-distorted point cloud.The scheme solves the problems of difficult positioning, point cloud distortion in GPS-free, large inclination, narrow closed space, improves the detection accuracy, efficiency and safety.

Description

Technical Field

[0001] This invention relates to the field of three-dimensional environmental perception technology, specifically to a method, device, unmanned aerial vehicle system, and computer-readable storage medium for detecting enclosed tilted spaces based on laser inertial odometry. Background Technology

[0002] Pumped storage power stations, as important energy storage and peak-shaving facilities, rely on their inclined shafts as crucial channels connecting the upper and lower reservoirs, undertaking the core functions of water conveyance and energy conversion. Inclined shafts typically feature large inclination angles, narrow spaces, and great depths. Furthermore, the complex environment within the shafts, lacking GPS signal coverage, with dim lighting and high humidity, presents significant challenges to regular inspections and structural health monitoring. Any damage or defect in the inclined shaft structure can affect the safe operation of the power station, thus requiring regular, comprehensive, and precise inspections.

[0003] Traditional methods for inspecting inclined shafts primarily rely on manual entry into the shaft or the use of auxiliary equipment such as ropes, suspended platforms, and wall-climbing robots. This approach has several drawbacks: First, it is inefficient, with a complete manual inspection often taking several days and requiring downtime, severely impacting the normal operation of the power plant. Second, it is costly, requiring specialized personnel and safety equipment. More seriously, the underground environment presents risks such as collapses, falling rocks, harmful gases, and oxygen deficiency, posing a severe threat to the lives of workers. Furthermore, manual inspections cannot achieve complete coverage of the shaft wall, and the results are significantly influenced by subjective factors, making it difficult to guarantee the objectivity, accuracy, and traceability of the data, thus failing to meet the high standards of structural health monitoring required by modern power plants.

[0004] With the development of drone technology, its application in the inspection of enclosed spaces such as tunnels, mines, and pipelines is gradually increasing. However, applying drone technology to special environments such as pumped storage inclined shafts still faces significant technical bottlenecks.

[0005] Positioning and navigation are the primary challenges. There is absolutely no GPS signal inside the inclined shaft, forcing the drone to rely on its own sensors for autonomous positioning and navigation. Traditional visual SLAM technology is prone to failure in low-light conditions or when the shaft wall lacks texture, making purely visual solutions unreliable for positioning. While lidar can operate in complete darkness, relying solely on it for positioning is susceptible to degradation in narrow, feature-sparse shafts, leading to positioning drift.

[0006] Motion distortion is another key technical challenge. When a drone flies in an inclined wellbore, it needs to frequently adjust its attitude to maintain stable flight. This continuous attitude change causes severe motion distortion in the LiDAR during scanning. Especially when using mechanically rotating LiDAR, acquiring a single frame of point cloud data typically takes 100 milliseconds. During this period, the drone's pose changes cause severe distortion in the point cloud data, making it impossible to accurately reflect the true geometry of the wellbore wall, directly affecting the accuracy of subsequent defect identification and 3D modeling.

[0007] Poor environmental adaptability also restricts the application of existing technologies. The narrow space of inclined shafts places extremely high demands on the flight control of UAVs and also limits the effective scanning range of sensors. General-purpose SLAM algorithms are usually designed for open or conventional indoor environments and lack specific optimization for narrow, inclined spaces. In the inclined shaft environment, there is a large angle between the direction of gravity and the flight path, and traditional IMU integration algorithms are prone to accumulating errors, leading to attitude estimation deviations. In addition, the narrow shaft walls limit the effective scanning range of lidar, which, without parameter optimization, would result in wasted computational resources and data redundancy.

[0008] In recent years, FastLIO technology has demonstrated superior performance in localization and mapping in complex environments by tightly coupling and fusing data from lidar and inertial measurement units (IMUs). This technology employs an iterative error state Kalman filter to achieve deep fusion of sensor data, effectively addressing motion distortion issues. However, existing FastLIO technologies are primarily designed for general scenarios and have not yet been specifically optimized for the unique environmental characteristics of pumped-storage inclined wells. The large inclination angle of these wells places special demands on the IMU integration algorithm, and the confined space necessitates adaptive adjustments to the lidar scanning parameters. Summary of the Invention

[0009] The purpose of this invention is to provide a UAV inspection method specifically optimized for the environmental characteristics of pumped storage inclined wells. By improving and optimizing the FastLIO core algorithm, the UAV can achieve stable and reliable autonomous positioning, high-precision 3D modeling, and intelligent defect detection in special environments such as those without GPS, narrow spaces, and tilted areas. This improves the safety, efficiency, and data quality of inclined well inspections, providing strong support for the safe operation of pumped storage power stations.

[0010] To achieve the above objectives, the present invention provides the following technical solution:

[0011] A method for detecting a closed tilted space based on laser inertial odometry, comprising:

[0012] S1: Acquire point cloud data collected by the lidar mounted on the mobile platform and inertial data collected by the inertial measurement unit;

[0013] S2: Motion compensation is performed on the point cloud data based on the inertial data to obtain distorted point cloud data;

[0014] S3: Based on the tilt angle information of the closed tilted space, the gravity vector in the integration process of the inertial data is compensated.

[0015] S4: Dynamically adjust the scanning parameters of the lidar based on the spatial dimensions of the enclosed inclined space;

[0016] S5: Based on the compensated inertial data and the parameter-adjusted lidar data, the positioning and environmental mapping of the mobile platform are achieved through a tightly coupled fusion algorithm.

[0017] Further: The motion compensation of the point cloud data in step S2 specifically includes: using the pre-integration results of the inertial measurement unit, aligning the timestamps of each point in each frame of the point cloud scanned by the lidar through the backpropagation method, thereby eliminating the motion distortion generated by the mobile platform during tilting motion.

[0018] Further: The gravity vector compensation process in step S3 specifically includes:

[0019] The state vector of the Kalman filter in the iterative error state contains an estimate of the gravity vector g;

[0020] Based on the motion state of the mobile platform in the tilted space, the estimation of the gravity vector g is continuously optimized through a filter update process;

[0021] Based on the optimized gravity vector estimate, its component in the tilted coordinate system is compensated.

[0022] Further: The dynamic adjustment of the lidar scanning parameters in step S4 specifically includes:

[0023] Real-time determination of the distance between the mobile platform and the spatial boundary;

[0024] When the distance is less than a preset threshold, reduce the maximum scanning distance parameter of the lidar;

[0025] Divide a single point cloud frame from a lidar sensor into multiple sub-frames for processing to increase the output frequency of the odometer.

[0026] Furthermore: the tightly coupled fusion algorithm in step S5 employs an iterative error state Kalman filter, including:

[0027] State prediction is performed using inertial measurement unit data to estimate the position, velocity, and attitude of the mobile platform; feature extraction is performed on the distortion-free point cloud to obtain planar and edge features;

[0028] The extracted features are matched with the local map, and the pose of the mobile platform is solved by an iterative optimization algorithm.

[0029] The optimized pose is used as the observation value, and the predicted state is corrected and updated within the filter framework.

[0030] Furthermore, it also includes:

[0031] S6: Based on the distortion-free point cloud data, identify structural anomalies on the spatial boundary by comparing it with a preset spatial structure standard model;

[0032] S7: Mark the identified structural anomalies on the constructed 3D map and generate a detection report containing anomaly type, location, and size information.

[0033] Furthermore: the identification of structural anomalies on spatial boundaries includes:

[0034] Establish a planar model of the standard spatial boundary;

[0035] Calculate the distance from each point in the point cloud to the planar model, and identify areas with deviations exceeding a threshold as potential peeling or deformation;

[0036] Edge detection algorithms are used to identify irregular linear features in point clouds as potential cracks.

[0037] Furthermore: the enclosed inclined space is an inclined shaft of a pumped storage power station, the mobile platform is a drone, and the lidar is a non-repetitive scanning lidar.

[0038] The present invention also provides a closed tilt space detection device based on laser inertial odometry, comprising:

[0039] The data acquisition module is used to acquire point cloud data collected by the lidar and inertial data collected by the inertial measurement unit;

[0040] The motion compensation module is used to perform motion compensation on the point cloud data based on the inertial data to obtain distortion-free point cloud data.

[0041] The gravity compensation module is used to compensate the gravity vector during the inertial data integration process based on the tilt angle information of the enclosed tilted space.

[0042] The parameter adjustment module is used to dynamically adjust the scanning parameters of the lidar based on the spatial dimensions of the enclosed inclined space.

[0043] The fusion positioning module is used to achieve positioning and environmental mapping based on compensated inertial data and parameter-adjusted LiDAR data through a tightly coupled fusion algorithm.

[0044] Furthermore, it also includes:

[0045] The defect detection module is used to identify structural anomalies on spatial boundaries by comparing the distorted point cloud data with a preset standard model of spatial structure.

[0046] The map annotation module is used to annotate the identified structural anomalies in the constructed 3D map.

[0047] The present invention also provides an unmanned aerial vehicle (UAV) system, comprising: a UAV body; a lidar and an inertial measurement unit mounted on the UAV body; and a processor for executing the methods described above.

[0048] The present invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the methods described above.

[0049] Compared with the prior art, the present invention has the following advantages:

[0050] I. This invention achieves centimeter-level high-precision positioning in an environment completely devoid of GPS signals by employing a tightly coupled fusion technology of lidar and inertial measurement unit data. In particular, by continuously optimizing and compensating for the gravity vector in the iterative error state Kalman filter, the problem of IMU integral accumulation error in large tilt angle environments is effectively solved, ensuring the accuracy of attitude and position calculations for UAVs flying on tilted paths. The positioning accuracy is improved by an order of magnitude compared to traditional methods.

[0051] Second, this invention utilizes IMU pre-integration and backpropagation motion compensation techniques to precisely align timestamps and correct distortion for each point scanned by the lidar, effectively eliminating point cloud distortion caused by frequent attitude adjustments during UAV tilted flight. The distortion-free point cloud accurately reflects the true geometric structure of the spatial boundary, providing a high-quality data foundation for subsequent defect detection and 3D modeling, and significantly improving the accuracy of structural anomaly identification.

[0052] Third, this invention dynamically adjusts the LiDAR scanning parameters, including automatically adjusting the maximum scanning distance based on the spatial dimensions and employing subframe segmentation technology to increase the odometer output frequency, enabling the system to achieve denser and more refined data acquisition in narrow, enclosed spaces. Compared to general SLAM algorithms, this invention improves data acquisition efficiency in narrow spaces by more than 5 times, while avoiding wasted computational resources and improving the system's real-time performance and stability.

[0053] Fourth, this invention uses autonomous drone inspections to replace manual labor, fundamentally eliminating the safety risks of personnel working in hazardous environments and avoiding potential threats such as landslides, harmful gases, and oxygen deficiency. At the same time, it reduces the workload of traditional manual inspections, which would take several days, to just a few hours, increasing work efficiency by more than 10 times. Furthermore, it allows for inspections without downtime, significantly reducing maintenance costs.

[0054] Fifth, this invention, by establishing a standard spatial boundary model and performing point cloud comparative analysis, can automatically identify structural anomalies such as cracks, spalling, and deformation, with detection accuracy reaching the millimeter level. Simultaneously, the generated high-precision 3D map and anomaly annotation information form a complete digital archive, providing objective, accurate, and traceable data support for long-term monitoring of structural health, trend analysis, and maintenance decisions. Attached Figure Description

[0055] Figure 1 This is a schematic flowchart of the closed tilt space detection method based on laser inertial odometry according to one embodiment of the present invention;

[0056] Figure 2 This is a schematic flowchart of the closed tilt space detection method based on laser inertial odometry according to another embodiment of the present invention. Detailed Implementation

[0057] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0058] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the invention and for 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. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0059] This invention provides a method for detecting enclosed inclined spaces based on laser inertial odometry. This method is particularly suitable for special environments such as inclined shafts of pumped storage power stations, which have large inclination angles, narrow spaces, and no GPS signals. In a typical embodiment, a UAV equipped with a non-repetitive scanning lidar and a high-precision inertial measurement unit is used as a mobile platform to perform autonomous inspection and 3D modeling of the inclined shaft of the pumped storage power station.

[0060] The implementation process begins with system preparation and initialization. The lidar carried by the UAV is preferably a non-repetitive scanning lidar such as the Livox Mid-360, with a scanning frequency exceeding 10Hz and a single frame point cloud containing tens of thousands of points. An industrial-grade or higher-precision inertial measurement unit (IMU) is selected, with a sampling frequency of at least 200Hz to ensure the ability to capture the UAV's high-frequency motion. Before entering the inclined shaft, the UAV completes system initialization in an open area at the shaft opening, including sensor calibration, coordinate system alignment, and the activation of the Fast LIO system. During the initialization phase, it is crucial to ensure accurate initial attitude estimation of the IMU, which can be achieved through gravity alignment in a static state or initialization using known orientation methods.

[0061] S1: Acquire point cloud data collected by the lidar mounted on the mobile platform and inertial data collected by the inertial measurement unit.

[0062] Specifically, after the UAV enters the inclined shaft, the system begins to execute the core positioning and mapping process. The data acquisition module continuously acquires raw point cloud data from the LiDAR and angular velocity and linear acceleration measurements from the inertial measurement unit. The point cloud data generated by the LiDAR in each scan frame contains tens of thousands of three-dimensional coordinate points and their corresponding timestamp information, while the IMU outputs six-axis motion data at a frequency of 200Hz or higher.

[0063] S2: Motion compensation is performed on the point cloud data based on the inertial data to obtain distorted point cloud data.

[0064] Specifically, motion compensation is a crucial step in ensuring point cloud quality. Since a single frame scan by a LiDAR typically takes 100 milliseconds, the UAV's pose changes during this period, leading to point cloud distortion. The system utilizes high-frequency measurement data from the IMU for pre-integration to calculate the relative pose change at each moment during the scan. In practice, for each point in the point cloud, based on its timestamp information, it is transformed from the coordinate system at the scan's moment of origin to the unified coordinate system at the frame's start moment using backpropagation. This point-by-point motion compensation effectively eliminates point cloud distortion caused by UAV movement, which is particularly important when frequent attitude adjustments are required in inclined shafts.

[0065] S3: Based on the tilt angle information of the closed tilted space, the gravity vector in the integration process of the inertial data is compensated.

[0066] Specifically, considering the tilt characteristics of the inclined shaft environment, the system implements a specialized gravity vector compensation strategy. Traditional IMU integration algorithms typically assume the gravity direction is vertically downwards. However, in an inclined shaft, the UAV's motion path has a significant angle with the gravity direction. Without compensation, this can lead to systematic deviations in attitude estimation. This invention introduces the gravity vector *g* as a parameter to be estimated into the state vector of the iterative error state Kalman filter. During each filter update, the system corrects the gravity vector estimate based on attitude observations obtained from lidar feature matching. This online estimation and compensation mechanism accurately models the gravity component in the inclined coordinate system, significantly improving attitude calculation accuracy in environments with large tilt angles. Practice shows that in an inclined shaft with a tilt angle of 45 degrees, the attitude estimation error after gravity compensation can be controlled within 0.5 degrees.

[0067] S4: Dynamically adjust the scanning parameters of the lidar based on the spatial dimensions of the enclosed inclined space.

[0068] Specifically, dynamic adjustment of LiDAR scanning parameters is a crucial measure for adapting to confined spaces. The system determines the distance between the UAV and the well wall using a real-time constructed local map. When the closest distance to the well wall is detected to be less than a preset threshold, such as 2-3 meters, the system automatically triggers a parameter adjustment mechanism. First, the maximum scanning distance parameter of the LiDAR is reduced from the default tens of meters to about 5 meters. This concentrates the limited scanning points on the nearby well wall, increasing the density of effective data. Simultaneously, the system employs subframe segmentation technology, dividing the point cloud frame originally acquired in 100 milliseconds into multiple 20-millisecond subframes for separate processing, increasing the odometer's output frequency from 10Hz to 50Hz. This high-frequency output not only improves the system's real-time response capability but also better tracks the rapid movement of the UAV, enabling more precise trajectory control in confined spaces.

[0069] S5: Based on the compensated inertial data and the parameter-adjusted lidar data, the positioning and environmental mapping of the mobile platform are achieved through a tightly coupled fusion algorithm.

[0070] Specifically, the tightly coupled fusion algorithm is the core of achieving high-precision positioning. The system employs an iterative error state Kalman filter framework to achieve deep fusion of IMU and LiDAR data. In the prediction phase, the position, velocity, and attitude of the UAV are predicted through integral calculations using IMU angular velocity and linear acceleration measurements. The prediction frequency is kept consistent with the IMU sampling frequency to ensure the continuity of state estimation. In the update phase, the distorted point cloud is first downsampled using voxel filtering to reduce computation while preserving key geometric features. Then, planar features and edge features are extracted based on local curvature analysis. Planar features mainly come from the smooth surface of the well wall, while edge features may come from the structural boundaries of the well wall or the equipment outline.

[0071] Feature matching employs a point-to-surface matching strategy, associating feature points extracted from the current frame with a local map maintained in the iKD-Tree data structure. The iKD-Tree is an incremental kd-tree structure capable of efficient nearest neighbor search and dynamic updates. For each feature point, the nearest corresponding plane or edge is searched in the local map, establishing a point-to-surface distance constraint. The Levenberg-Marquardt iterative optimization algorithm minimizes the matching residuals of all feature points to obtain the optimal pose estimate for the current frame. This optimized pose is then input as an observation into a Kalman filter and fused with the IMU-predicted state to obtain the final state estimate.

[0072] In another embodiment, the present invention further includes: S6: Based on the distortion-free point cloud data, structural anomalies on the spatial boundary are identified by comparing it with a preset spatial structure standard model.

[0073] Specifically, after completing the basic positioning and mapping functions, the system further implements the automatic detection function of structural defects. After receiving high-precision point cloud data transmitted in real time by the UAV, the ground station first establishes a reference model of the standard well wall. For regular circular or rectangular inclined wells, a standard geometric model can be obtained through fitting. Then, the distance from each point in the actual point cloud to the standard model is calculated to generate a deviation distribution map. When the deviation value of a certain area continuously exceeds a preset threshold, such as 5 mm, the system marks it as a potential spalling or deformation area.

[0074] S7: Mark the identified structural anomalies on the constructed 3D map and generate a detection report containing anomaly type, location, and size information.

[0075] Specifically, for crack detection, the system employs a method based on point cloud density and normal vector analysis. Cracks typically manifest as discontinuities in the point cloud or abrupt changes in the normal vector. By calculating the density distribution and normal vector consistency of the local point cloud, minute cracks with widths on the millimeter scale can be identified. The detection algorithm also considers the morphological characteristics of the crack, such as length, width, and orientation, to distinguish between real structural cracks and surface textures.

[0076] The final 3D map generated by the system not only includes the complete geometry of the inclined shaft but also integrates all detected anomaly information. Each anomaly point contains detailed metadata, including anomaly type, 3D coordinates, size, severity rating, and detection time. This information is stored in a structured manner, facilitating subsequent data analysis, trend tracking, and maintenance decisions. By comparing detection results from different periods, the development trend of structural defects can be analyzed, providing a scientific basis for preventative maintenance.

[0077] In a practical application, a pumped storage power station has an inclined shaft approximately 500 meters long, with an inclination angle of 45 degrees and a diameter of 6 meters. Using the method of this invention, a drone completed the inspection of the entire inclined shaft within 2 hours, generating a high-precision 3D point cloud map containing over 500 million points, with a point cloud density exceeding 1000 points per square meter. The system successfully identified three cracks wider than 2 millimeters and two areas of concrete spalling exceeding 0.5 square meters, with the inspection results showing over 95% consistency with subsequent manual verification. In contrast, traditional manual inspection takes 3-4 days and can only cover approximately 60% of the shaft wall area.

[0078] The method of this invention is not limited to the inclined shafts of pumped storage power stations, but can also be applied to other similar enclosed inclined spaces, such as mine ramps, subway tunnels, and large pipelines. By adjusting the corresponding parameter settings and optimization strategies, it can adapt to different spatial dimensions, inclination angles, and environmental characteristics, and has broad application prospects.

[0079] The above embodiments are only for illustrating the technical concept and features of the present invention, and are intended to enable those skilled in the art to understand the content of the present invention and implement it accordingly. They should not be construed as limiting the scope of protection of the present invention. All equivalent transformations or modifications made in accordance with the spirit and essence of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A method for detecting a closed tilted space based on laser inertial odometry, characterized in that, include: S1: Acquire point cloud data collected by the lidar mounted on the mobile platform and inertial data collected by the inertial measurement unit; S2: Motion compensation is performed on the point cloud data based on the inertial data to obtain distorted point cloud data; S3: Based on the tilt angle information of the enclosed tilted space, the gravity vector in the integration process of the inertial data is compensated; the gravity vector compensation process in step S3 specifically includes: The state vector of the Kalman filter in the iterative error state contains an estimate of the gravity vector g; Based on the motion state of the mobile platform in the tilted space, the estimation of the gravity vector g is continuously optimized through a filter update process; Based on the optimized gravity vector estimate, its component in the tilted coordinate system is compensated. S4: Dynamically adjust the scanning parameters of the lidar based on the spatial dimensions of the enclosed inclined space; the dynamic adjustment of the lidar scanning parameters in step S4 specifically includes: Real-time determination of the distance between the mobile platform and the spatial boundary; When the distance is less than a preset threshold, reduce the maximum scanning distance parameter of the lidar; Divide a single frame of point cloud data from a lidar radar into multiple sub-frames for processing to increase the output frequency of the odometer. S5: Based on the compensated inertial data and the parameter-adjusted lidar data, the positioning and environmental mapping of the mobile platform are achieved through a tightly coupled fusion algorithm.

2. The method according to claim 1, characterized in that, The motion compensation of the point cloud data in step S2 specifically includes: using the pre-integration results of the inertial measurement unit, aligning the timestamps of each point in each frame of the point cloud scanned by the lidar through the backpropagation method, thereby eliminating the motion distortion generated by the mobile platform during tilting motion.

3. The method according to claim 1, characterized in that, The tightly coupled fusion algorithm in step S5 employs an iterative error state Kalman filter, including: State prediction is performed using inertial measurement unit data to estimate the position, velocity, and attitude of the mobile platform; feature extraction is performed on the distortion-free point cloud to obtain planar and edge features; The extracted features are matched with the local map, and the pose of the mobile platform is solved by an iterative optimization algorithm. The optimized pose is used as the observation value, and the predicted state is corrected and updated within the filter framework.

4. The method according to any one of claims 1-3, characterized in that, Also includes: S6: Based on the distortion-free point cloud data, identify structural anomalies on the spatial boundary by comparing it with a preset spatial structure standard model; S7: Mark the identified structural anomalies on the constructed 3D map and generate a detection report containing anomaly type, location, and size information.

5. The method according to claim 4, characterized in that, The structural anomalies identified on the spatial boundary include: Establish a planar model of the standard spatial boundary; Calculate the distance from each point in the point cloud to the planar model, and identify areas with deviations exceeding a threshold as potential peeling or deformation; Edge detection algorithms are used to identify irregular linear features in point clouds as potential cracks.

6. The method according to claim 1, characterized in that, The enclosed inclined space is an inclined shaft of a pumped storage power station, the mobile platform is a drone, and the lidar is a non-repetitive scanning lidar.

7. A closed tilt space detection device based on laser inertial odometry, characterized in that, The method for detecting a closed tilted space based on a laser inertial odometry, as described in any one of claims 1-6, comprises: The data acquisition module is used to acquire point cloud data collected by the lidar and inertial data collected by the inertial measurement unit; The motion compensation module is used to perform motion compensation on the point cloud data based on the inertial data to obtain distortion-free point cloud data. The gravity compensation module is used to compensate the gravity vector during the inertial data integration process based on the tilt angle information of the enclosed tilted space. The parameter adjustment module is used to dynamically adjust the scanning parameters of the lidar based on the spatial dimensions of the enclosed inclined space. The fusion positioning module is used to achieve positioning and environmental mapping based on compensated inertial data and parameter-adjusted LiDAR data through a tightly coupled fusion algorithm.

8. The apparatus according to claim 7, characterized in that, Also includes: The defect detection module is used to identify structural anomalies on spatial boundaries by comparing the distorted point cloud data with a preset standard model of spatial structure. The map annotation module is used to annotate the identified structural anomalies in the constructed 3D map.

9. An unmanned aerial vehicle (UAV) system, characterized in that, include: The drone body; the lidar and inertial measurement unit mounted on the drone body; A processor for performing the method according to any one of claims 1-6.

10. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the method described in any one of claims 1-6.

Images