SLAM-based obstacle avoidance and navigation method for blade interior inspection robots using lidar.

Abstract

This invention provides a SLAM-based lidar obstacle avoidance and navigation method for a blade interior inspection robot, belonging to the field of radar obstacle avoidance technology. This method acquires the original point cloud of the blade's internal environment and performs intensity compensation correction and sampling density adjustment, adapting to the reflective characteristics of the blade's internal materials and surface geometric changes, thus improving the accuracy of the point cloud data and its environmental characterization capabilities. By mapping the updated point cloud to a polar coordinate system to construct a ring-shaped obstacle density heatmap, it can intuitively reflect the distribution of obstacles around the robot, providing structured perception input for real-time obstacle avoidance. By fusing historical pose sequences with the SLAM map to achieve map correction and dynamically adjusting the obstacle avoidance strategy based on obstacle type, it can achieve more reliable environmental modeling, more accurate obstacle differentiation, and more adaptive navigation and obstacle avoidance control in the complex, unstructured environment inside the blade, thereby improving the autonomous operation safety and operational efficiency of the inspection robot.

Description

Technical Field

[0001] This invention relates to the field of radar obstacle avoidance technology, and in particular to a SLAM-based laser radar obstacle avoidance navigation method for a blade interior inspection robot. Background Technology

[0002] When existing radar obstacle avoidance technology is applied to the complex environment inside wind turbine blades, it generally adopts a fixed sampling rate or voxelization processing method to collect point clouds. This single sampling strategy is prone to insufficient sampling in areas with rapid curvature changes inside the blade (such as blade tips and joints), resulting in the loss of key geometric information and difficulty in accurately perceiving fine structures. In flat areas, it will collect a large amount of redundant data, causing excessive storage and computing burden and affecting real-time performance. At the same time, traditional forward obstacle avoidance strategies are difficult to achieve effective passage in narrow channels inside the blade, and SLAM maps are prone to cumulative drift and are not accurate enough in distinguishing obstacle types, resulting in untimely adjustment of obstacle avoidance strategies or excessive avoidance. This cannot meet the high-precision and high-reliability obstacle avoidance and navigation requirements of blade internal inspection robots. Summary of the Invention

[0003] The purpose of this invention is to provide a SLAM-based lidar obstacle avoidance and navigation method for blade interior inspection robots, which enables more reliable environmental modeling, more accurate obstacle differentiation, and more adaptive navigation and obstacle avoidance control in the complex and unstructured environment inside blades, thereby improving the autonomous operation safety and work efficiency of the inspection robot.

[0004] In a first aspect, the present invention provides a SLAM-based obstacle avoidance and navigation method for a blade interior inspection robot using lidar, comprising: acquiring a SLAM map of the blade interior environment constructed in real time by the inspection robot, the historical pose sequence of the inspection robot, and the original point cloud of the blade interior environment collected by its onboard lidar; performing intensity compensation correction and sampling density adjustment on the original point cloud to obtain an updated point cloud; mapping the updated point cloud to a polar coordinate system centered on the inspection robot to construct a ring-shaped obstacle density heatmap representing the distribution of obstacles in the blade interior environment; determining a new SLAM map based on the historical pose sequence and the SLAM map; identifying the obstacle types in the ring-shaped obstacle density heatmap to dynamically adjust the obstacle avoidance strategy according to the obstacle type, and planning an obstacle avoidance path based on the new SLAM map to control the inspection robot to perform corresponding obstacle avoidance and navigation actions.

[0005] In an optional implementation, the intensity compensation correction includes: acquiring the reflectivity of the blade material and real-time temperature data collected by the temperature sensor inside the lidar; determining the temperature drift compensation coefficient based on the real-time temperature data and a preset temperature drift curve; calculating the laser incident angle corresponding to each three-dimensional coordinate point in the original point cloud; and performing point-by-point intensity compensation correction on the original point cloud based on the laser incident angle, the reflectivity of the blade material, and the temperature drift compensation coefficient to obtain a calibrated point cloud.

[0006] In an optional implementation, the sampling density adjustment includes: calculating the local curvature of the local neighborhood where each point in the calibration point cloud is located; if the local curvature is greater than a first curvature threshold, increasing the point cloud sampling density of the local neighborhood to a first density value; if the local curvature is less than a second curvature threshold, decreasing the point cloud sampling density of the local neighborhood to a second density value.

[0007] In an optional implementation, constructing an annular obstacle density heatmap characterizing the distribution of obstacles within the blade's internal environment includes: dividing the polar coordinate system into a preset number of sectors; counting the number of polar coordinate points in each sector whose distance from the inspection robot is less than a preset threshold to obtain the number of valid points; generating an initial heatmap based on the number of valid points corresponding to each sector and the correspondence between the number of valid points and the heatmap value; and performing morphological expansion processing on the void areas of the initial heatmap to fill the void areas generated by the LiDAR scanning gap to obtain the annular obstacle density heatmap.

[0008] In an optional implementation, a new SLAM map is determined based on the historical pose sequence and the SLAM map, including: extracting consecutive keyframes from the historical pose sequence and calculating the principal curvature direction vector of the point cloud of each keyframe; calculating the standard deviation of the angle between the principal curvature direction vectors of adjacent keyframes; if the standard deviation is less than a preset standard deviation threshold, the SLAM map is optimized based on the point cloud of the consecutive keyframes to obtain a corrected new SLAM map.

[0009] In an optional implementation, identifying the obstacle type in the ring obstacle density heatmap includes: calculating the density gradient of the ring obstacle density heatmap and predicting the moving speed of the obstacle in combination with the motion state of the inspection robot; if the density gradient is less than a preset gradient threshold and the moving speed is greater than a preset speed threshold, the obstacle is determined to be a dynamic obstacle; if the density gradient is greater than the preset gradient threshold and the moving speed is zero, the obstacle is determined to be a static wall obstacle.

[0010] In an optional implementation, the obstacle avoidance strategy is dynamically adjusted according to the type of obstacle, including: if the obstacle is a static wall obstacle, the inspection robot is controlled to maintain a preset safe distance from the static wall obstacle and an obstacle avoidance path is planned along its tangent direction; if the obstacle is a dynamic obstacle, the inspection robot is controlled to stop moving, the motion trajectory of the dynamic obstacle is calculated, and an obstacle avoidance path is planned along the direction away from the motion trajectory.

[0011] Secondly, this invention provides a SLAM-based obstacle avoidance and navigation device for a blade interior inspection robot using a lidar system, comprising: an acquisition module for acquiring a real-time SLAM map of the blade interior environment constructed by the inspection robot, the inspection robot's historical pose sequence, and the original point cloud of the blade interior environment collected by its onboard lidar; a calibration module for performing intensity compensation correction and sampling density adjustment on the original point cloud to obtain an updated point cloud; a construction module for mapping the updated point cloud to a polar coordinate system centered on the inspection robot to construct a ring-shaped obstacle density heatmap representing the distribution of obstacles in the blade interior environment; a determination module for determining a new SLAM map based on the historical pose sequence and the SLAM map; and an obstacle avoidance module for identifying obstacle types in the ring-shaped obstacle density heatmap, dynamically adjusting the obstacle avoidance strategy according to the obstacle type, and planning an obstacle avoidance path based on the new SLAM map to control the inspection robot to perform corresponding obstacle avoidance and navigation actions.

[0012] Thirdly, the present invention provides an electronic device, including a memory and a processor, wherein the memory stores a computer program that can run on the processor, and the processor executes the computer program to implement the SLAM-based laser radar obstacle avoidance navigation method for blade interior inspection robots as described in any of the foregoing embodiments.

[0013] Fourthly, the present invention provides a computer-readable storage medium storing computer instructions, which, when executed by a processor, implement the SLAM-based laser radar obstacle avoidance and navigation method for blade interior inspection robots as described in any of the foregoing embodiments.

[0014] This invention collects the original point cloud of the internal environment of the blade and performs intensity compensation correction and sampling density adjustment, which can adapt to the reflective characteristics of the internal material and the geometric changes of the curved surface, thereby improving the accuracy of the point cloud data and the environmental characterization capability. By mapping the updated point cloud to the polar coordinate system to construct a ring obstacle density heat map, the distribution of obstacles around the robot can be intuitively reflected, providing structured perception input for real-time obstacle avoidance. By fusing historical pose sequences with SLAM maps to achieve map correction and dynamically adjusting the obstacle avoidance strategy based on obstacle type, more reliable environmental modeling, more accurate obstacle differentiation, and more adaptive navigation and obstacle avoidance control can be achieved in the complex and unstructured environment inside the blade, thereby improving the autonomous operation safety and work efficiency of the inspection robot. Attached Figure Description

[0015] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0016] Figure 1 A flowchart illustrating a SLAM-based obstacle avoidance and navigation method for a blade interior inspection robot using lidar, provided as an embodiment of the present invention. Figure 2 This is a flowchart of a strength compensation correction method provided in an embodiment of the present invention; Figure 3 A functional block diagram of a SLAM-based laser radar obstacle avoidance and navigation device for a blade interior inspection robot, provided for an embodiment of the present invention; Figure 4 This is a schematic diagram of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0017] 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.

[0018] 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.

[0019] The following detailed description of some embodiments of the present invention is provided in conjunction with the accompanying drawings. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0020] Example 1 Figure 1 A flowchart illustrating a SLAM-based obstacle avoidance and navigation method for a blade interior inspection robot using lidar is provided in this embodiment of the invention. Figure 1 As shown, the method specifically includes the following steps: Step S102: Obtain the SLAM map of the blade's internal environment constructed in real time by the inspection robot, the historical pose sequence of the inspection robot, and the original point cloud of the blade's internal environment collected by its onboard LiDAR.

[0021] Specifically, at the beginning of the method, the lidar and SLAM system on the inspection robot need to be activated. The lidar continuously collects three-dimensional raw point cloud data of the internal environment of the blade by combining horizontal rotation of the gimbal with vertical scanning of the internal lens. Simultaneously, it records the pose information output by the SLAM front-end odometer during the movement of the inspection robot and forms a historical pose sequence. At the same time, the SLAM system dynamically constructs and updates the SLAM map of the internal environment of the blade based on the real-time collected point cloud data and pose information. Finally, the real-time constructed SLAM map of the internal environment of the blade, the complete historical pose sequence, and the raw point cloud collected by the lidar are obtained.

[0022] Step S104: Perform intensity compensation correction and sampling density adjustment on the original point cloud to obtain the updated point cloud.

[0023] After acquiring the raw point cloud, to improve the accuracy and reliability of the point cloud data, this embodiment of the invention further preprocesses the acquired raw point cloud data. Specifically, firstly, intensity compensation correction is performed on the data based on key factors affecting the point cloud intensity, eliminating perception biases caused by environmental and equipment characteristics, resulting in a more accurate calibration point cloud. Then, the local geometric features of the calibration point cloud are analyzed, and the point cloud sampling density in different regions is dynamically adjusted according to feature changes, reducing redundant data while retaining key information, and generating an updated point cloud.

[0024] Step S106: Map the updated point cloud to a polar coordinate system centered on the inspection robot to construct a ring-shaped obstacle density heat map that characterizes the distribution of obstacles in the internal environment of the blade.

[0025] Next, this embodiment of the invention will update the point cloud to a polar coordinate system centered on the inspection robot. By performing density statistics and visualization processing on the polarized point cloud data, a ring-shaped obstacle density heat map that can intuitively reflect the distribution of obstacles inside the blade will be constructed, providing a clear basis for obstacle identification.

[0026] Specifically, the polar coordinate transformation operation involves updating the 3D point cloud coordinates in the point cloud. Convert to polar coordinates ,in Indicates radial distance. Indicates the horizontal azimuth angle. Indicates the vertical elevation angle, ignoring height. The absolute change of the axis is projected onto a two-dimensional polar coordinate grid.

[0027] Step S108: Determine a new SLAM map based on the historical pose sequence and the SLAM map.

[0028] Subsequently, the consistency of environmental geometric features is analyzed based on the robot's pose sequence. Optionally, the corresponding geometric consistency curve is calculated to determine whether the preset similarity conditions are met. When the conditions are met, a closed-loop detection process is triggered to optimize and correct the real-time constructed SLAM map, so as to eliminate accumulated errors, ensure the accuracy of the map, and obtain a new SLAM map.

[0029] Step S110: Identify the obstacle types in the ring obstacle density heat map, dynamically adjust the obstacle avoidance strategy according to the obstacle type, and plan the obstacle avoidance path based on the new SLAM map to control the inspection robot to perform the corresponding obstacle avoidance navigation actions.

[0030] Finally, based on the ring-shaped obstacle density heatmap, the specific types of obstacles are identified, and the obstacle avoidance strategy is dynamically adjusted according to the characteristics of different obstacle types. These obstacle types include dynamic obstacles and static wall obstacles. Simultaneously, this embodiment of the invention uses a revised, high-precision SLAM map as a basis to plan a safe and efficient obstacle avoidance path for the inspection robot, controlling the robot to execute obstacle avoidance navigation actions according to the planned path, ensuring stable completion of the inspection task in the complex environment inside the blade.

[0031] This invention, through collecting the original point cloud of the blade's internal environment and performing intensity compensation correction and sampling density adjustment, can adapt to the reflective characteristics of the blade's internal materials and the geometric changes of its curved surface, thereby improving the accuracy of the point cloud data and its environmental characterization capabilities. By mapping the updated point cloud to a polar coordinate system to construct a ring-shaped obstacle density heatmap, it can intuitively reflect the distribution of obstacles around the robot, providing structured perception input for real-time obstacle avoidance. By fusing historical pose sequences with SLAM maps to achieve map correction and dynamically adjusting obstacle avoidance strategies based on obstacle types, it can achieve more reliable environmental modeling, more accurate obstacle differentiation, and more adaptive navigation and obstacle avoidance control in the complex, unstructured environment inside the blade, thereby improving the autonomous operation safety and operational efficiency of the inspection robot.

[0032] In one alternative implementation, such as Figure 2 As shown, in step S104 above, the strength compensation correction includes the following steps: Step S201: Obtain the reflectivity of the blade material and the real-time temperature data collected by the temperature sensor inside the lidar.

[0033] Specifically, the reflectivity of the blade material is determined by conducting pre-tests on the reflectivity characteristics of commonly used composite materials (such as glass fiber and carbon fiber) inside the blade. The reflectivity data of different materials at various angles are measured by a photometer, and then selected and set as a fixed reflectivity constant, for example, 0.6 for glass fiber.

[0034] This embodiment of the invention also reads the value measured by the temperature sensor in the IMU inertial measurement unit integrated inside the lidar. This value reflects the current operating temperature of the laser emitter and receiving diode, providing a basis for subsequent temperature drift compensation.

[0035] Step S202: Determine the temperature drift compensation coefficient based on real-time temperature data and preset temperature drift curve.

[0036] The preset temperature drift curve is generated in advance for this model of lidar in a laboratory constant temperature chamber. For example, the curve records the light intensity attenuation ratio corresponding to every 0.5 degrees Celsius within the range of -20 degrees Celsius to 60 degrees Celsius. Next, the real-time temperature data collected in step S201 above is matched with this curve. Optionally, a linear interpolation method is used to calculate the light intensity attenuation ratio corresponding to the current real-time temperature data. This ratio is the temperature drift compensation coefficient, used to offset the influence of temperature changes on the lidar's reflection intensity detection.

[0037] Step S203: Calculate the laser incident angle corresponding to each three-dimensional coordinate point in the original point cloud.

[0038] Specifically, each 3D coordinate point in the original point cloud is traversed, and the KD-Tree neighborhood search algorithm is used to find the 10 to 20 nearest neighbor points (example values) for each point. Then, a local covariance matrix is ​​constructed based on these neighbor points and eigenvalue decomposition is performed to extract the normal vector of the plane where the point is located. Then, the cosine value of the angle between the direction vector of the laser ray emitted by the lidar and the normal vector is calculated through vector operations. The angle corresponding to the cosine value is the laser incident angle corresponding to the 3D coordinate point, and its value ranges from 0 degrees to 90 degrees.

[0039] Step S204: Based on the laser incident angle, blade material reflectivity and temperature drift compensation coefficient, the original point cloud is subjected to point-by-point intensity compensation correction to obtain the calibrated point cloud.

[0040] Finally, based on the blade material reflectivity obtained in step S201, the temperature drift compensation coefficient determined in step S202, and the laser incident angle calculated in step S203, the reflection intensity of the original point cloud is corrected point by point using a preset correction model. The correction model is expressed as: ;in, This represents the reflection intensity value of the original point cloud. Indicates the laser incident angle. Indicates the reflectivity of the blade material. This represents the temperature drift compensation coefficient. This represents the corrected reflection intensity value.

[0041] By substituting the original reflection intensity value, laser incident angle, blade material reflectivity, and temperature drift compensation coefficient of each point in the original point cloud into the above correction model, the corrected reflection intensity of each point can be calculated.

[0042] If any values ​​in the calculation result exceed the range of the intensity value, the reflection intensity of all points is normalized to ensure that the reflection intensity of all points is within a reasonable range. Finally, all the corrected point cloud data is repackaged to obtain a calibration point cloud with higher accuracy and more accurate reflection intensity.

[0043] In an optional implementation, step S104 above, the sampling density adjustment includes the following: Calculate the local curvature of the local neighborhood of each point in the calibration point cloud.

[0044] If the local curvature is greater than the first curvature threshold, the point cloud sampling density of the local neighborhood is increased to the first density value.

[0045] If the local curvature is less than the second curvature threshold, the point cloud sampling density of the local neighborhood is reduced to the second density value.

[0046] Specifically, when calculating the local curvature of the local neighborhood of each point in the calibration point cloud, this embodiment of the invention uses principal component analysis to analyze the local geometric features of each point. Specifically, for each point, a predetermined number of neighboring points are found, and a local covariance matrix is ​​constructed based on these neighboring points. Then, eigenvalue decomposition is performed on this covariance matrix to obtain three eigenvalues. And there are .

[0047] Subsequently, based on the preset curvature calculation formula By substituting the three characteristic values ​​mentioned above, the curvature change of the surface at that point can be calculated. This curvature change is the local curvature corresponding to that point, which reflects the complexity of the geometric structure of the region.

[0048] After obtaining the local curvature of each point, it is compared with the preset first curvature threshold and second curvature threshold respectively, thereby dynamically adjusting the point cloud sampling density of the corresponding local neighborhood: If the local curvature of a point is determined to be greater than the first curvature threshold, it indicates that the local neighborhood is located in a region rich in geometric features and prone to collisions, such as the blade's internal reinforcing ribs, corners, or blade tips. A higher sampling resolution is needed to fully preserve key geometric information. Therefore, the point cloud sampling density of this local neighborhood is increased to the first density value. For example, by setting the voxel grid size of the voxel filter to between 0.01 meters and 0.015 meters, the number of points retained per unit area is increased.

[0049] If the local curvature of a point is determined to be less than the second curvature threshold, it indicates that the local neighborhood is located on a flat blade wall with simple geometric features, and therefore does not require a high sampling density. Thus, the point cloud sampling density of this local neighborhood is reduced to the second density value, reducing redundant data and computational burden while maintaining positioning accuracy. For example, sparsification can be performed by setting the voxel grid size of the voxel filter to between 0.03 meters and 0.05 meters.

[0050] Optionally, the first curvature threshold can be set in the range of 0.15-0.2m. -1 This range is based on the lower limit of curvature in high-risk areas, derived from statistical analysis of a large amount of internal blade scanning test data. The first density value is set to be greater than or equal to 5000 pts / m. 2 This ensures that the contours of minute cracks or protrusions can be clearly constructed; the second curvature threshold ranges from 0.05 to 0.1 m. -1 This range marks the boundary between the smooth wall surface and the transition zone, and the second density value is set to be less than or equal to 1000 pts / m². 2 This is done to minimize the computational load while ensuring the accuracy of wall positioning.

[0051] In an optional embodiment, step S106 above, which involves constructing an annular obstacle density thermogram characterizing the distribution of obstacles in the internal environment of the blade, specifically includes the following steps: Step S401: Divide the polar coordinate system into a preset number of sectors on an average basis.

[0052] Specifically, after establishing a polar coordinate system with the center of the inspection robot as the origin, the 360-degree horizontal field of view of the polar coordinate system is divided into a preset number of sectors according to a preset angular resolution. The number of sectors is determined based on the scanning accuracy of the LiDAR and the environmental perception requirements, ensuring that each sector can accurately cover the corresponding angular range. This provides a uniform angular division basis for subsequent point cloud quantity statistics and obstacle distribution representation, enabling obstacle information from different directions to be clearly captured. Optionally, this grid divides the 360-degree horizontal field of view into 360 angular sectors, with each sector corresponding to an angular resolution of 1 degree.

[0053] Step S402: Count the number of polar coordinate points in each sector whose distance from the inspection robot is less than a preset threshold, and obtain the number of valid points.

[0054] For the point cloud data converted to polar coordinates, all polar coordinate points within each sector are traversed, and the radial distance between each polar coordinate point and the inspection robot is calculated one by one. This radial distance is compared with a preset threshold (i.e., the robot collision warning distance), and polar coordinate points with distances less than the preset threshold are selected. The total number of polar coordinate points in each sector that meet this condition is counted, and this total number is the number of valid points for the corresponding sector. The number of valid points directly reflects the density of obstacles in that sector.

[0055] Step S403: Based on the number of valid points corresponding to each sector, and combining the correspondence between the number of valid points and the thermal value, an initial heat map is generated.

[0056] In this embodiment of the invention, a pre-defined correspondence between the number of valid points and thermal values ​​is established. For example, when the number of valid points is 0, a cool color tone is used (indicating no obstacles or low risk), and when the number of valid points exceeds a preset density threshold (e.g., 50), a warm color tone is used (indicating high obstacle density or high risk). Based on this correspondence, the number of valid points in each sector is converted into a corresponding thermal value. Then, according to the angular position of each sector and its corresponding thermal value, a visual rendering is performed on a two-dimensional polar coordinate grid to generate an initial thermal map that can intuitively reflect the distribution of obstacles inside the blade.

[0057] Step S404: Morphological expansion processing is performed on the void areas of the initial heat map to fill the void areas generated by the LiDAR scanning gap, thereby obtaining the density heat map of the annular obstacle.

[0058] Because of the scanning gaps in lidar, discontinuous voids may appear in the initial heatmap, affecting the accuracy of obstacle distribution assessment. Therefore, a morphological dilation algorithm is used to process the initial heatmap. By setting a structure element matrix of appropriate dimensions (e.g., 3×3) and performing convolution operations on the heatmap matrix, high-thermal-value regions expand to their surrounding neighborhoods, thereby filling the voids created by the lidar scanning gaps. This makes the heatmap continuous and complete, ultimately yielding a ring-shaped obstacle density heatmap that accurately characterizes the distribution of obstacles within the blade's internal environment.

[0059] In an optional implementation, step S108, which determines a new SLAM map based on the historical pose sequence and the SLAM map, specifically includes the following steps: Step S501: Extract consecutive keyframes from the historical pose sequence and calculate the principal curvature direction vector of the point cloud for each keyframe.

[0060] Specifically, from the historical pose sequence of the inspection robot, continuous keyframes reflecting key positions of the robot's trajectory are first selected according to preset frame intervals or feature extraction rules. For each keyframe, its corresponding local point cloud map is obtained, and surface normals are estimated for each frame's point cloud. By statistically analyzing the distribution patterns of local surface normals, the principal curvature direction vector representing the overall geometric structure of the environment in that frame is extracted. This vector corresponds to the eigenvector corresponding to the largest eigenvalue of the point cloud covariance matrix and is the core parameter characterizing the environmental structural features.

[0061] Step S502: Calculate the standard deviation of the angle between the principal curvature direction vectors between adjacent keyframes.

[0062] For the extracted consecutive keyframes, the angle between the principal curvature direction vectors corresponding to two adjacent keyframes is calculated sequentially. Specifically, the cosine value of the angle between adjacent principal curvature direction vectors is first obtained through vector dot product operation, and then the specific angle is calculated through inverse cosine operation. The average angle between all adjacent keyframes is then calculated to obtain the mean angle.

[0063] Subsequently, the standard deviation of the included angle was calculated using statistical methods, with the following formula: ,in, Indicates the standard deviation of the included angle. Indicates the total number of keyframes. Indicates the first The angle between the principal curvature direction vectors between adjacent keyframes in a group. This represents the mean of the included angle. The standard deviation reflects the stability of the environmental structural features corresponding to consecutive keyframes during continuous motion; the smaller the standard deviation, the more stable the environmental structural features.

[0064] Step S503: If the standard deviation is less than the preset standard deviation threshold, the SLAM map is optimized based on the continuous keyframe point cloud to obtain a corrected new SLAM map.

[0065] Finally, the calculated standard deviation is compared with a preset standard deviation threshold, which is determined based on a large amount of internal navigation test data of the blades and is used to determine whether the conditions for map optimization are met. If the standard deviation is less than the preset standard deviation threshold, it is determined that the structural features of the continuous keyframe point cloud are highly consistent. That is, it means that the environmental geometric features observed by the robot on a continuous path are highly consistent. In this case, the robot may have returned to the area it has previously passed through or is in a structurally symmetrical area. At this time, the closed-loop detection process is triggered: based on the matching results of the continuous keyframe point cloud and the historical frame point cloud, the cumulative error in the original SLAM map is corrected, the position coordinates and structural relationships of the point cloud in the map are adjusted, the map drift caused by long-term navigation is eliminated, and finally a new SLAM map with higher accuracy and stronger stability is obtained.

[0066] In an optional implementation, step S110 above, identifying the obstacle type in the annular obstacle density heatmap, specifically includes the following: Calculate the density gradient of the density heatmap of the ring-shaped obstacle, and predict the moving speed of the obstacle by combining the motion state of the inspection robot.

[0067] If the density gradient is less than the preset gradient threshold and the moving speed is greater than the preset speed threshold, the obstacle is determined to be a dynamic obstacle; if the density gradient is greater than the preset gradient threshold and the moving speed is zero, the obstacle is determined to be a static wall obstacle.

[0068] Specifically, the matrix data corresponding to the constructed annular obstacle density heatmap is first subjected to numerical differentiation to calculate the difference in thermal values ​​between adjacent sectors, thereby obtaining the density gradient in the annular obstacle density heatmap.

[0069] Simultaneously, the instantaneous velocity vector and angular velocity vector of the inspection robot output by the SLAM system are used as a reference. The Kalman filter algorithm is used to continuously track the centroid position of the highlighted area (i.e., the area where the obstacle is located) in the density heatmap of the ring obstacle, and the position coordinates of the centroid in the heatmap of consecutive frames are recorded. By calculating the ratio of the displacement of the centroid position between adjacent frames to the time interval, the moving speed of the obstacle can be obtained.

[0070] After obtaining the density gradient and obstacle movement speed, they are compared with preset gradient thresholds and preset speed thresholds respectively to determine the obstacle type: if the calculated density gradient is less than the preset gradient threshold and the predicted movement speed is greater than the preset speed threshold, it indicates that the obstacle has blurred edges and is in the process of displacement, which is consistent with the characteristics of dynamic objects such as people and moving tools, so the obstacle is determined to be a dynamic obstacle; if the density gradient is greater than the preset gradient threshold and the predicted movement speed is zero, it indicates that the obstacle has clear and sharp edges and remains stationary relative to the global map coordinates, which is consistent with the characteristics of fixed structures such as blade inner walls and partitions, so the obstacle is determined to be a static wall obstacle.

[0071] In an optional implementation, step S110 above, which dynamically adjusts the obstacle avoidance strategy according to the obstacle type, specifically includes the following: If the obstacle is a static wall obstacle, the inspection robot is controlled to maintain a preset safe distance from the static wall obstacle and to plan an obstacle avoidance path along its tangent direction.

[0072] The preset safety distance is a fixed value determined based on the width of the internal channel of the blade, the robot's own size, and its mobility, ensuring that the robot will not collide with the static wall during its movement. The path planning uses an artificial potential field method, treating the static wall obstacle as a repulsive force source, generating an outward repulsive force in the normal direction of the obstacle to prevent the robot from approaching. Simultaneously, the target point of the inspection task is treated as a gravitational field source, applying a guiding force in the tangential direction to guide the robot to move parallel to the wall, achieving wall-following inspection while avoiding obstacles, ensuring the continuity and integrity of the inspection path.

[0073] If the obstacle is a dynamic obstacle, the inspection robot is controlled to stop moving, the trajectory of the dynamic obstacle is calculated, and an obstacle avoidance path is planned in the direction that deviates from the trajectory.

[0074] Specifically, if the obstacle is determined to be dynamic, the inspection robot must be immediately stopped to avoid the risk of collision due to continued movement. Then, based on the dynamic obstacle's displacement data over a recent period, its future trajectory is calculated using linear extrapolation to determine the obstacle's direction of movement, speed, and potential reachable area. Combining this predicted trajectory, a danger zone is constructed with the dynamic obstacle as its center and a dynamically expanding radius over time. The size of this zone is determined by the obstacle's speed and the robot's avoidance response time. Finally, within the inspection robot's current reachable space, an escape path is searched and planned that does not intersect with the danger zone and deviates from the predicted trajectory of the dynamic obstacle. This ensures that the robot will not come into contact with the dynamic obstacle during the avoidance process. Normal inspection movement resumes only after the obstacle has moved away or the avoidance action is completed.

[0075] In an optional embodiment, the method provided in this invention is configured to run on an embedded computing platform mounted on an inspection robot. The embedded computing platform uses a single-threaded processing flow to perform intensity compensation correction and closed-loop detection, and the computation latency of the single-threaded processing flow is configured to be less than 50 milliseconds.

[0076] Specifically, the method is configured to run on an embedded computing platform mounted on an inspection robot. This platform is built on an ARM architecture processor with a main frequency of no less than 1.5 GHz and a memory of no less than 4 GB. To adapt to the characteristics of the non-networked environment inside the blades and the extremely high real-time requirements, the operating system adopts a simplified Linux real-time kernel, eliminating the context switching overhead and resource contention risks brought by multi-threaded concurrency. The embedded computing platform adopts a single-threaded processing flow execution intensity compensation correction and closed-loop detection, that is, all point cloud preprocessing, feature extraction, matching and optimization steps are executed sequentially in the same execution thread. A fixed memory pool is pre-allocated to avoid jitter caused by dynamic memory allocation, and the code of each step is optimized, for example, using SIMD instructions to accelerate matrix operations, controlling the execution time of each link, and ensuring that the total time from receiving the LiDAR data packet to outputting the control command does not exceed the set time limit. The calculation latency of the single-threaded processing flow is configured to be less than 50 milliseconds.

[0077] In summary, this invention overcomes the limitations of traditional radar obstacle avoidance technology in complex environments such as wind turbine blades by acquiring and correcting lidar point cloud data within the blade's interior. By compensating for and correcting point cloud intensity based on the laser incident angle and blade material reflectivity, it suppresses reflection attenuation and distortion caused by material properties, ensuring the accuracy and reliability of the original sensing information. Furthermore, by dynamically adjusting the point cloud sampling density according to local curvature, it improves sensing resolution in high-curvature regions and reduces redundancy in flat regions, avoiding information loss or excessive computational burden caused by fixed sampling strategies. The system addresses several key issues. Furthermore, by acquiring the historical pose sequence of the inspection robot and calculating the standard deviation of the angle between the principal curvature direction vectors of adjacent keyframes, closed-loop detection is triggered when the standard deviation is less than a preset standard deviation threshold. This corrects the cumulative drift of the SLAM map, ensuring the accuracy of long-term autonomous navigation. Additionally, a circular obstacle density heatmap is constructed in polar coordinates to identify obstacle types within the heatmap and dynamically adjust obstacle avoidance strategies accordingly. This distinguishes between static walls and dynamic obstacles, preventing excessive avoidance of non-dangerous areas or slow response to real hazards, thus improving operational efficiency.

[0078] Example 2 This invention also provides a SLAM-based lidar obstacle avoidance and navigation device for a blade interior inspection robot. This device is mainly used to execute the SLAM-based lidar obstacle avoidance and navigation method for a blade interior inspection robot provided in Embodiment 1 above. The device provided in this invention will be described in detail below.

[0079] Figure 3 A functional block diagram of a SLAM-based laser radar obstacle avoidance and navigation device for a blade interior inspection robot is provided in an embodiment of the present invention, as shown below. Figure 3 As shown, the device mainly includes: an acquisition module 10, a calibration module 20, a construction module 30, a determination module 40, and an obstacle avoidance module 50, wherein: The acquisition module 10 is used to acquire the SLAM map of the blade's internal environment built in real time by the inspection robot, the historical pose sequence of the inspection robot, and the original point cloud of the blade's internal environment collected by its onboard lidar.

[0080] The calibration module 20 is used to perform intensity compensation correction and sampling density adjustment on the original point cloud to obtain an updated point cloud.

[0081] Module 30 is used to map the updated point cloud to a polar coordinate system centered on the inspection robot, and to construct a ring-shaped obstacle density heat map that characterizes the distribution of obstacles in the internal environment of the blade.

[0082] The determination module 40 is used to determine a new SLAM map based on the historical pose sequence and the SLAM map.

[0083] The obstacle avoidance module 50 is used to identify the types of obstacles in the ring-shaped obstacle density heat map, so as to dynamically adjust the obstacle avoidance strategy according to the obstacle type and plan the obstacle avoidance path based on the new SLAM map, so as to control the inspection robot to perform corresponding obstacle avoidance navigation actions.

[0084] This invention, through collecting the original point cloud of the blade's internal environment and performing intensity compensation correction and sampling density adjustment, can adapt to the reflective characteristics of the blade's internal materials and the geometric changes of its curved surface, thereby improving the accuracy of the point cloud data and its environmental characterization capabilities. By mapping the updated point cloud to a polar coordinate system to construct a ring-shaped obstacle density heatmap, it can intuitively reflect the distribution of obstacles around the robot, providing structured perception input for real-time obstacle avoidance. By fusing historical pose sequences with SLAM maps to achieve map correction and dynamically adjusting obstacle avoidance strategies based on obstacle types, it can achieve more reliable environmental modeling, more accurate obstacle differentiation, and more adaptive navigation and obstacle avoidance control in the complex, unstructured environment inside the blade, thereby improving the autonomous operation safety and operational efficiency of the inspection robot.

[0085] Optionally, the calibration module 20 is specifically used for: Acquire the reflectivity of the blade material and real-time temperature data collected by the temperature sensor inside the lidar.

[0086] Based on real-time temperature data and a preset temperature drift curve, the temperature drift compensation coefficient is determined.

[0087] Calculate the laser incident angle corresponding to each 3D coordinate point in the original point cloud.

[0088] Based on the laser incident angle, blade material reflectivity, and temperature drift compensation coefficient, the original point cloud is subjected to point-by-point intensity compensation correction to obtain a calibrated point cloud.

[0089] Optionally, the calibration module 20 is also used for: Calculate the local curvature of the local neighborhood of each point in the calibration point cloud.

[0090] If the local curvature is greater than the first curvature threshold, the point cloud sampling density of the local neighborhood is increased to the first density value.

[0091] If the local curvature is less than the second curvature threshold, the point cloud sampling density of the local neighborhood is reduced to the second density value.

[0092] Optionally, module 30 is specifically used for: The polar coordinate system is divided into a predetermined number of sectors.

[0093] The number of polar coordinate points in each sector whose distance from the inspection robot is less than a preset threshold is counted to obtain the number of valid points.

[0094] An initial heatmap is generated based on the number of valid points corresponding to each sector and the correspondence between the number of valid points and the heat value.

[0095] Morphological expansion processing is performed on the void areas of the initial heatmap to fill the void areas generated by the LiDAR scanning gap, resulting in a density heatmap of the annular obstacle.

[0096] Optionally, module 40 is specifically used for: Extract consecutive keyframes from the historical pose sequence and calculate the principal curvature direction vector of the point cloud for each keyframe.

[0097] Calculate the standard deviation of the angle between the principal curvature direction vectors between adjacent keyframes.

[0098] If the standard deviation is less than the preset standard deviation threshold, the SLAM map is optimized based on the continuous keyframe point cloud to obtain a corrected new SLAM map.

[0099] Optionally, the obstacle avoidance module 50 is specifically used for: Calculate the density gradient of the density heatmap of the ring-shaped obstacle, and predict the moving speed of the obstacle by combining the motion state of the inspection robot.

[0100] If the density gradient is less than a preset gradient threshold and the moving speed is greater than a preset speed threshold, then the obstacle is determined to be a dynamic obstacle.

[0101] If the density gradient is greater than the preset gradient threshold and the moving speed is zero, the obstacle is determined to be a static wall obstacle.

[0102] Optionally, the obstacle avoidance module 50 is also used for: If the obstacle is a static wall obstacle, the inspection robot is controlled to maintain a preset safe distance from the static wall obstacle and to plan an obstacle avoidance path along its tangent direction.

[0103] If the obstacle is a dynamic obstacle, the inspection robot is controlled to stop moving, the trajectory of the dynamic obstacle is calculated, and an obstacle avoidance path is planned in the direction that deviates from the trajectory.

[0104] Example 3 See Figure 4 This invention provides an electronic device, which includes a processor 60, a memory 61, a bus 62, and a communication interface 63. The processor 60, the communication interface 63, and the memory 61 are connected via the bus 62. The processor 60 is used to execute executable modules, such as computer programs, stored in the memory 61.

[0105] The memory 61 may include high-speed random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Communication between this system network element and at least one other network element is achieved through at least one communication interface 63 (which can be wired or wireless), such as the Internet, wide area network, local area network, metropolitan area network, etc.

[0106] Bus 62 can be an ISA bus, PCI bus, or EISA bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 4 The symbol is represented by a single double-headed arrow, but this does not mean that there is only one bus or one type of bus.

[0107] The memory 61 is used to store programs. After receiving an execution instruction, the processor 60 executes the program. The method executed by the apparatus defined by the process disclosed in any of the foregoing embodiments of the present invention can be applied to the processor 60 or implemented by the processor 60.

[0108] Processor 60 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of processor 60 or by instructions in software form. Processor 60 can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this invention. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this invention can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in memory 61. Processor 60 reads the information in memory 61 and, in conjunction with its hardware, completes the steps of the above method.

[0109] The computer program product of the SLAM-based blade internal inspection robot lidar obstacle avoidance and navigation method provided in this embodiment of the invention includes a computer-readable storage medium storing non-volatile program code executable by a processor. The instructions included in the program code can be used to execute the methods described in the preceding method embodiments. For specific implementation, please refer to the method embodiments, which will not be repeated here.

[0110] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0111] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0112] 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.

[0113] 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, or the orientation or positional relationship commonly used when the product of this invention is in use. They are only for the convenience of describing this 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 this invention. In addition, the terms "first," "second," "third," etc., are only used to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0114] Furthermore, terms such as "horizontal," "vertical," and "sag" do not imply that components must be absolutely horizontal or suspended, but rather that they can be slightly tilted. For example, "horizontal" simply means that its direction is more horizontal relative to "vertical," and does not mean that the structure must be completely horizontal, but can be slightly tilted.

[0115] In the description of this 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 this invention based on the specific circumstances.

[0116] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A SLAM-based laser radar obstacle avoidance and navigation method for a blade interior inspection robot, characterized in that, include: The system acquires a real-time SLAM map of the blade's internal environment constructed by the inspection robot, the inspection robot's historical pose sequence, and the raw point cloud of the blade's internal environment collected by its onboard LiDAR. The original point cloud is subjected to intensity compensation correction and sampling density adjustment to obtain an updated point cloud; The updated point cloud is mapped to a polar coordinate system centered on the inspection robot to construct an annular obstacle density heat map characterizing the distribution of obstacles in the internal environment of the blade. Based on the historical pose sequence and the SLAM map, a new SLAM map is determined; The obstacle types in the ring-shaped obstacle density heatmap are identified, and the obstacle avoidance strategy is dynamically adjusted according to the obstacle types. An obstacle avoidance path is planned based on the new SLAM map to control the inspection robot to perform corresponding obstacle avoidance and navigation actions.

2. The SLAM-based laser radar obstacle avoidance and navigation method for blade interior inspection robots according to claim 1, characterized in that, Strength compensation correction includes: The reflectivity of the blade material and real-time temperature data collected by the temperature sensor inside the lidar are obtained. Based on the real-time temperature data and the preset temperature drift curve, the temperature drift compensation coefficient is determined. Calculate the laser incident angle corresponding to each three-dimensional coordinate point in the original point cloud; Based on the laser incident angle, the reflectivity of the blade material, and the temperature drift compensation coefficient, the original point cloud is subjected to point-by-point intensity compensation correction to obtain a calibrated point cloud.

3. The SLAM-based laser radar obstacle avoidance and navigation method for blade interior inspection robots according to claim 2, characterized in that, Sampling density adjustment includes: Calculate the local curvature of the local neighborhood of each point in the calibration point cloud; If the local curvature is greater than the first curvature threshold, then the point cloud sampling density of the local neighborhood is increased to the first density value; If the local curvature is less than the second curvature threshold, then the point cloud sampling density of the local neighborhood is reduced to the second density value.

4. The SLAM-based laser radar obstacle avoidance and navigation method for blade interior inspection robots according to claim 1, characterized in that, Constructing an annular obstacle density heatmap characterizing the distribution of obstacles within the blade's internal environment, including: The polar coordinate system is divided into a predetermined number of sectors on an average basis; The number of polar coordinate points in each sector whose distance from the inspection robot is less than a preset threshold is counted to obtain the number of valid points. Based on the number of valid points corresponding to each sector, and combined with the correspondence between the number of valid points and the heat value, an initial heat map is generated; The void regions in the initial heatmap are subjected to morphological expansion processing to fill the void regions generated by the LiDAR scanning gap, thereby obtaining the density heatmap of the annular obstacle.

5. The SLAM-based laser radar obstacle avoidance and navigation method for blade interior inspection robots according to claim 1, characterized in that, Based on the historical pose sequence and the SLAM map, a new SLAM map is determined, including: Extract consecutive keyframes from the historical pose sequence and calculate the principal curvature direction vector of the point cloud for each keyframe. Calculate the standard deviation of the angle between the principal curvature direction vectors between adjacent keyframes; If the standard deviation is less than a preset standard deviation threshold, the SLAM map is optimized based on the continuous keyframe point cloud to obtain a corrected new SLAM map.

6. The SLAM-based laser radar obstacle avoidance and navigation method for blade interior inspection robots according to claim 1, characterized in that, Identifying the obstacle types in the ring-shaped obstacle density heatmap includes: Calculate the density gradient of the density heatmap of the ring obstacle, and predict the moving speed of the obstacle by combining the motion state of the inspection robot; If the density gradient is less than a preset gradient threshold and the moving speed is greater than a preset speed threshold, then the obstacle is determined to be a dynamic obstacle. If the density gradient is greater than the preset gradient threshold and the moving speed is zero, then the obstacle is determined to be a static wall obstacle.

7. The SLAM-based laser radar obstacle avoidance and navigation method for blade interior inspection robots according to claim 1, characterized in that, Dynamically adjust the obstacle avoidance strategy according to the type of obstacle, including: If the obstacle is a static wall obstacle, the inspection robot is controlled to maintain a preset safe distance from the static wall obstacle and to plan an obstacle avoidance path along its tangent direction; If the obstacle is a dynamic obstacle, the inspection robot is controlled to stop moving, the trajectory of the dynamic obstacle is calculated, and an obstacle avoidance path is planned in a direction that deviates from the trajectory.

8. A SLAM-based lidar obstacle avoidance and navigation device for a blade interior inspection robot, characterized in that, include: The acquisition module is used to acquire the SLAM map of the blade's internal environment built in real time by the inspection robot, the historical pose sequence of the inspection robot, and the original point cloud of the blade's internal environment collected by its onboard LiDAR. The calibration module is used to perform intensity compensation correction and sampling density adjustment on the original point cloud to obtain an updated point cloud; The construction module is used to map the updated point cloud to a polar coordinate system centered on the inspection robot, and construct an annular obstacle density heat map characterizing the distribution of obstacles in the internal environment of the blade. The determination module is used to determine a new SLAM map based on the historical pose sequence and the SLAM map; The obstacle avoidance module is used to identify the types of obstacles in the ring-shaped obstacle density heat map, dynamically adjust the obstacle avoidance strategy according to the obstacle type, and plan the obstacle avoidance path based on the new SLAM map, so as to control the inspection robot to perform corresponding obstacle avoidance navigation actions.

9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program executable on the processor, characterized in that, When the processor executes the computer program, it implements the SLAM-based laser radar obstacle avoidance and navigation method for blade interior inspection robots as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions, which, when executed by a processor, implement the SLAM-based laser radar obstacle avoidance and navigation method for blade interior inspection robots as described in any one of claims 1 to 7.

Images