Airplane wheel identification method and system based on three-dimensional point cloud

Abstract

According to the three-dimensional point cloud-based aircraft wheel identification method and system provided by the invention, the aircraft wheel is identified by integrating a clustering algorithm, a least square method circle fitting method, a position feature discrimination method and the like, the consumption of computing resources is further reduced, and the method can be applied by modifying parameters according to the data of a special aircraft; the method does not need learning training, can be applied to a specific aircraft more quickly, has relatively high robustness, and can realize unmanned and full-automatic target recognition stage of a follow-up task.

Description

Technical Field

[0001] The present invention belongs to the technical field related to aircraft wheel identification, and particularly relates to a method and system for aircraft wheel identification based on three-dimensional point cloud. Background Technique

[0002] The statements in this part only provide background technical information related to the present invention, and do not necessarily constitute prior art.

[0003] In the task of adding and hanging inside and outside the belly of the aircraft, identifying the spatial position and body attitude of the target hanging rack to be added and hung is the basis and prerequisite for realizing the entire automated operation process. Since the hanging rack is located at a specific position under the belly of the aircraft, it is difficult to directly position it outside the fuselage. According to field research, the ideal starting position for the adding and hanging operation is 5 - 10 meters away from the aircraft. Within this distance, the fuselage size of the entire aircraft is too large, and the complete identification effect is not good. Other parts of the fuselage are also much higher than the body. Within the height range of the adding and hanging body, the aircraft wheels are relatively independent of the fuselage and have obvious features.

[0004] In recent years, the target recognition algorithm based on three-dimensional point cloud has developed rapidly, and has been widely applied in fields such as autonomous driving, robot navigation, environmental modeling, and unmanned aerial vehicles. Early target recognition algorithms mainly relied on traditional geometric feature extraction and matching methods, including techniques such as geometric model-based matching, point cloud segmentation, and feature description. Target recognition was completed by matching the objects in the point cloud with predefined geometric models (such as spheres, cubes, cylinders, etc.). These methods had good effects when the object shapes were simple and regular.

[0005] With the rapid development of deep learning technology, especially the successful application of convolutional neural network (CNN) in image processing, more and more research has begun to introduce deep learning methods into three-dimensional point cloud target recognition. The most representative algorithm is PointNet, which first proposed a method for directly processing unordered point clouds and overcame the limitation of point cloud disorder through global feature pooling, achieving good classification and segmentation effects. However, it lacks in-depth mining of local geometric information, so there are still certain limitations in dealing with objects with complex shapes. PointNet++ introduced a hierarchical feature learning and local feature aggregation mechanism, which can effectively extract local features at different scales and gradually construct global features, thereby improving the accuracy and robustness of target recognition. With the continuous development of these deep learning algorithms, more point cloud-based target recognition methods have emerged, such as DGCNN (Dynamic Graph CNN), VoxelNet, PointCNN, etc. These methods further improve the performance of three-dimensional point cloud target recognition by introducing innovative technologies such as graph convolution, voxelization processing, or dynamic graph structures.

[0006] In the target recognition task based on 3D point cloud, especially in specific scenarios such as the recognition of special-shaped aircraft wheels, although deep learning algorithms can provide powerful feature learning capabilities, in this practical application, the requirement of pre-training the model may be difficult to achieve. This is because this method usually requires a large amount of labeled data for training, while the data of special-shaped aircraft is not open, and it is often difficult to obtain the 3D point cloud dataset of aircraft wheels. In this case, the applicability of deep learning methods is limited.

[0007] In summary, how to meet the real-time recognition of aircraft wheels, especially special-shaped aircraft wheels, through the extraction and analysis of point cloud geometric features in the absence of a large amount of labeled data, so as to realize the automation of the addition and hanging inside and outside the belly cabin, is the problem that needs to be solved at present. Summary of the Invention

[0008] To overcome the deficiencies of the above-mentioned prior art, the present invention provides a method and system for identifying aircraft wheels based on 3D point cloud, which combines methods such as clustering algorithm, least squares fitting circle, and position feature discrimination to identify aircraft wheels, further reducing the consumption of computing resources and making the algorithm more applicable.

[0009] To achieve the above object, the present invention adopts the following technical solutions:

[0010] In the first aspect, the present invention provides a method for identifying aircraft wheels based on 3D point cloud, including:

[0011] Obtain the lidar point cloud data of the aircraft wheel and perform preprocessing;

[0012] Use the clustering neighborhood radius and minimum number of points to cluster the preprocessed lidar point cloud data to obtain multiple point cloud clusters, and determine the position relationship of each point cloud cluster according to the actual position relationship of the aircraft wheel;

[0013] For the point cloud clusters that meet the position relationship, use the least squares method to fit the distribution of points to find the best fitting circle, and determine whether the best fitting circle is valid;

[0014] Calculate the coverage of the valid best fitting circle, sort the corresponding point cloud clusters according to the calculated coverage, and determine the point cloud cluster of the wheel target.

[0015] In the second aspect, the present invention provides a system for identifying aircraft wheels based on 3D point cloud, including:

[0016] An acquisition module, which is configured to: obtain the lidar point cloud data of the aircraft wheel and perform preprocessing;

[0017] A position determination module, which is configured to: perform clustering on the preprocessed lidar point cloud data by using a clustering neighborhood radius and a minimum number of points to obtain a plurality of point cloud clusters, and determine the position relationship of each point cloud cluster according to the actual position relationship of the aircraft wheels;

[0018] A fitting module, which is configured to: for the point cloud clusters that meet the position relationship, use the least squares method to fit the distribution of points to find the best fitting circle, and determine whether the best fitting circle is valid;

[0019] An identification module, which is configured to: calculate the coverage of the valid best fitting circle, sort the corresponding point cloud clusters according to the calculated coverage, and determine the wheel target point cloud cluster.

[0020] In a third aspect, the present invention provides an electronic device, including a memory, a processor, and computer instructions stored on the memory and running on the processor. When the computer instructions are run by the processor, the method described in the first aspect is completed.

[0021] In a fourth aspect, the present invention provides a computer-readable storage medium for storing computer instructions. When the computer instructions are executed by a processor, the method described in the first aspect is completed.

[0022] In a fifth aspect, the present invention provides a computer program product, including a computer program. When the computer program is executed by a processor, the method described in the first aspect is implemented.

[0023] The above one or more technical solutions have the following beneficial effects:

[0024] In the present invention, methods such as a clustering algorithm, least squares method for fitting a circle, and position feature discrimination are integrated to identify aircraft wheels, further reducing the consumption of computing resources. By modifying the parameters according to the data of the special aircraft itself, it can be applied without the need for learning and training, and can be applied to specific aircraft faster, with strong robustness, and can realize the unmanned and fully automatic identification stage of subsequent hanging tasks.

[0025] The advantages of the additional aspects of the present invention will be partially given in the following description, partially become obvious from the following description, or be understood through the practice of the present invention. Description of the Drawings

[0026] The specification drawings forming a part of the present invention are used to provide a further understanding of the present invention. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention.

[0027] Figure 1 It is a flowchart of a method for identifying aircraft wheels based on three-dimensional point clouds in Embodiment 1 of the present invention;

[0028] Figure 2 This is the schematic diagram of the region of interest division in the first embodiment of the present invention;

[0029] Figure 3 This is the schematic diagram of the radius filtering in the first embodiment of the present invention;

[0030] Figure 4 This is the schematic diagram of the DBSCAN clustering algorithm in the first embodiment of the present invention;

[0031] Figure 5(a) is the effect diagram of the wheel point cloud fitting circle in the first embodiment of the present invention;

[0032] Figure 5(b) is the effect diagram of the debris point cloud fitting circle in the first embodiment of the present invention;

[0033] Figure 6 This is the schematic diagram of the calculation of the final target position and pose in the first embodiment of the present invention. Detailed implementation manners

[0034] It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention belongs.

[0035] It should be noted that the terms used herein are only for describing specific implementation manners and are not intended to limit the exemplary embodiments according to the present invention.

[0036] In the case of no conflict, the embodiments in the present invention and the features in the embodiments can be combined with each other.

[0037] Embodiment 1

[0038] This embodiment discloses a method for identifying aircraft wheels based on three-dimensional point clouds, including:

[0039] Obtaining the lidar point cloud data of the aircraft wheels and performing preprocessing;

[0040] Using the clustering neighborhood radius and the minimum number of points to cluster the preprocessed lidar point cloud data to obtain multiple point cloud clusters, and determining the position relationship for each point cloud cluster according to the actual position relationship of the aircraft wheels;

[0041] For the point cloud clusters that meet the position relationship, using the least squares method to fit the distribution of points to find the best fitting circle, and determining whether the best fitting circle is valid;

[0042] Calculating the coverage of the valid best fitting circles, sorting the corresponding point cloud clusters according to the calculated coverage, and determining the wheel target point cloud clusters;

[0043] According to the determined target point cloud clusters of the wheels, based on the positional relationship between the belly of the aircraft and the wheels of the aircraft, calculate the midpoint coordinates of the rear two wheels of the aircraft to determine the target position of the belly of the aircraft; and determine the target pose of the belly of the aircraft according to the positional relationship between the installation orientation of the lidar and the orientation of the aircraft.

[0044] In this embodiment, the lidar is set on the trailer. By integrating algorithms such as clustering, least squares circle fitting, and position feature discrimination, the consumption of computing resources is further reduced, and the applicability of the algorithm is stronger. It can be applied by modifying the parameters according to the data of the special aircraft itself without learning and training, and can be applied to specific aircraft faster, with strong robustness.

[0045] The following combines Figure 1 to give a detailed description of a method for identifying aircraft wheels based on three-dimensional point clouds proposed in this embodiment, specifically including:

[0046] Step 1: Obtain the lidar point cloud data of the aircraft and perform preprocessing.

[0047] Specifically, after capturing the lidar point cloud data, first perform filtering processing. By setting the boundaries of the coordinate values of the point cloud data, divide the region of interest, and use boolean indexing to screen out the points located within the region of interest from the original point cloud; send these screened points into the adaptive radius outlier removal (AROR) algorithm for denoising processing, and automatically adjust the radius of the neighborhood according to the local density of each point, so as to provide a more accurate filtering effect in different density regions, thereby cleaning up noise and irrelevant points, and sending the remaining point cloud data into subsequent clustering and positional relationship determination.

[0048] Region of Interest (ROI) partitioning is an important technique in three-dimensional point cloud processing, aiming to screen out the points within a specific region from a large amount of point cloud data, so as to focus on the target objects related to the task, reduce the amount of calculation, and improve the processing efficiency. As Figure 2 shown, set a set of boundary values (such as the minimum and maximum values on the X, Y, and Z axes) according to prior knowledge or task requirements to limit the spatial region of interest. In the task of aircraft wheel recognition, the space under the belly of the aircraft is preset as the region of interest: the forward direction of the trailer coincides with the positive direction of the X axis of the lidar, and the initial position of the trailer is generally between 5 - 15m from the target. Considering the positional relationship between the three wheels and the wheel height, the x-axis range (depth) can be limited within 18m, the y-axis range (lateral) can be limited within 5m, and the z-axis range (height) can be limited within 1m. In summary, the region of interest of the lidar is limited to:

[0049] 1 ≤ x ≤ 18

[0050] -5 ≤ y ≤ 5

[0051] 0 < z ≤ 1

[0052] According to the minimum and maximum value ranges of each set coordinate axis, all points that meet the conditions in the coordinates are screened out, and the points outside the range are removed. This method is both simple and efficient, and can flexibly adapt to different application scenarios. Subsequent processing processes such as clustering and feature determination can be concentrated in these key areas, significantly improving the processing accuracy and speed.

[0053] Adaptive Radius Outlier Removal (AROR) is a dynamic radius filtering method based on local density. Compared with Radius Outlier Removal (ROR), it can adaptively adjust the filtering radius according to the point cloud density, ensuring that a sufficient clustering neighborhood radius is used in sparse areas, while avoiding using too large a neighborhood range in dense areas. This can ensure that irrelevant points in sparse areas can be better removed and details in local dense areas can be better retained. As Figure 3 shown, first calculate the local density and set a fixed radius range r 0 (in this embodiment, take 0.3), select a point p i , and the number of points within its clustering neighborhood radius r 0 is N(p i ):

[0054] N(p i ) = {p j |d(p i , p j (≤ r 0}

[0055] where d(p i , p j ) is the Euclidean distance between point p i and point p j . The neighborhood is spherical and its volume is:

[0056]

[0057] On this basis, the local density v i is defined as the ratio of the number of neighborhood points to the neighborhood volume, so the local density is:

[0058]

[0059] Next, the process of dynamically adjusting the radius is carried out according to the local density. In areas with higher local density, there are more points in the neighborhood, so a smaller radius can be selected to avoid the inability to remove edge-irrelevant points, which may affect the filtering effect. On the contrary, in sparse areas, there are fewer neighborhood points, and the radius needs to be increased to ensure that enough neighborhood points are obtained to participate in the filtering. Specifically, the radius can be dynamically adjusted through the following formula:

[0060]

[0061] where C is a constant that controls the maximum value of the radius. This formula indicates that the radius is smaller in areas with larger local density and larger in areas with lower density. To avoid the radius being too small, a minimum radius value r min :

[0062] r i ≥r min

[0063] In this way, the radius in dense areas will automatically become smaller, while in sparse areas, the radius will increase to ensure the filtering effect.

[0064] Then, when performing filtering, set the neighborhood point number threshold to τ (usually τ = 5). For each point p i , the dynamically adjusted radius r i can be used to find its neighborhood points. If the number of neighborhood points is less than the preset threshold, then this point is considered an outlier and can be deleted or adjusted.

[0065] The advantage of adaptive radius filtering is that the algorithm is simple and efficient, especially suitable for sparse point clouds or scenarios with large local density changes, such as scanning in open experimental sites or real-time processing tasks.

[0066] Step 2: Use the clustering neighborhood radius and minimum number of points to cluster the preprocessed lidar point cloud data to obtain multiple point cloud clusters, and determine the positional relationship of each point cloud cluster according to the actual positional relationship of the aircraft wheels.

[0067] Specifically, in order to initially segment the point cloud data clearly, first perform DBSCAN clustering on the point cloud data retained after Step 1. By optimizing the clustering neighborhood radius and minimum number of points, adjacent points are aggregated into a cluster. Then, count the number of clusters and determine the positional relationship of each cluster of point clouds.

[0068] The improved DBSCAN (Density-Based Spatial Clustering of Applications with Noise) is a density-based clustering algorithm that aims to divide regions with high density into the same cluster and regard low-density regions as noise. Different from traditional partitioning-based clustering methods, DBSCAN does not require the number of clusters to be specified in advance, but automatically identifies the cluster structure through the density relationship in the point cloud data.

[0069] The improvement in this embodiment mainly focuses on the optimization of two parameters, namely, the method for selecting the clustering neighborhood radius based on the k-distance graph and the method for selecting the minimum number of points in the neighborhood based on Gaussian kernel density estimation.

[0070] In the traditional DBSCAN algorithm, the parameters ∈ (clustering neighborhood radius) and MinPts (minimum number of points) are the key parameters that determine the clustering result. The selection of these parameters has a great impact on the clustering result, and often needs to be manually adjusted on different data sets, increasing the difficulty of using the algorithm. The purpose of optimizing the parameter selection is to automatically adjust these parameters according to the data distribution and local density changes, so as to improve the clustering effect and adaptability of DBSCAN in different scenarios.

[0071] The method for selecting the clustering neighborhood radius based on the k-distance graph calculates the local change through the k-distance graph and automatically adjusts the clustering neighborhood radius of each point. The k-distance graph evaluates the local change of a point by calculating the distance from each point to its k-th nearest neighbor. Commonly used k values are 4 or 5 (usually corresponding to MinPts, but not necessarily the same).

[0072] The specific steps are as follows: First, for each data point p i , calculate its distance d i,k to its k nearest neighbors, and sort the k-nearest neighbor distances of all points in ascending order; then, plot the distance between each point and its k-nearest neighbors in a graph to construct a k-distance graph. The horizontal axis in the k-distance graph represents the index of the point, and the vertical axis represents the distance from the point to its k nearest neighbors. In the k-distance graph, there is usually an obvious inflection point or elbow (i.e., the position where the distance changes greatly), indicating that after this distance, the local density of the data points begins to drop sharply. Calculate the index number of the point corresponding to the elbow point in the k-distance graph, and find the distance from the corresponding point to p i according to the index number. This distance is used as the new local clustering neighborhood radius ∈'. To detect the elbow position of the k-distance graph, the "inflection point" of the curve, that is, the elbow, can be found through the second-order difference (i.e., curvature):

[0073] Δ 2 d i =(d i+2,k - d i+1,k )-(di+1,k -d i,k )

[0074] Where, Δd i Representative point p j The distance to the kth nearest neighbor point and point p i+1 The difference in distance to the kth nearest neighbor. The elbow usually corresponds to the point with the largest second-order difference, that is:

[0075] i * = argmax(Δ 2 d i )

[0076] By finding the index i corresponding to the maximum second-order difference * , the elbow position in the k-distance graph can be found, and the ∈′ value corresponding to this position is the optimal clustering neighborhood radius.

[0077] The method of selecting the minimum number of points in the neighborhood based on Gaussian kernel density estimation optimizes the traditional method of selecting MinPts in DBSCAN. The fixed value may not be able to effectively process data for areas with different densities. Therefore, the method of selecting the minimum number of points in the neighborhood based on Gaussian kernel density estimation can dynamically adjust the point count threshold MinPts of its neighborhood according to the Gaussian kernel density of the area where each point is located, thereby improving the clustering quality. Gaussian kernel density estimation is to calculate the number of points p at each point. i The local density of a point is estimated by taking the weighted average of the points in its neighborhood. The formula is as follows:

[0078]

[0079] Among them, ρ(p i ) is point p i Gaussian kernel density of ||p i -p j || is the Euclidean distance between two points; σ is the standard deviation of the Gaussian kernel, which controls the rate of distance decay; |N(p i )| represents the number of neighborhood points. This formula weights the points in the neighborhood by the Gaussian kernel function, close to p i Points closer to the center have a larger weight, while points farther away have a smaller weight. By taking the weighted sum of all points in the neighborhood, we can calculate the value of point p. i The local density of point p i The "crowding degree" in its local area. For each point p i , you can use the following formula to adaptively adjust MinPts:

[0080]

[0081] Among them, MinPts i It's point pi The corresponding optimized MinPts′, where α is a constant, usually chosen as 2 or 3, is used to adjust the value of MinPts′ to better conform to the density characteristics of the data. According to the formula, in regions with higher local density, the value of MinPts′ is smaller because the region is already dense enough and does not require too many neighboring points to be determined as core points. In low-density regions, the value of MinPts′ is larger because the region is sparse and may require more neighboring points to be judged as core points.

[0082] According to the optimized selection method, the appropriate clustering neighborhood radius ∈′ and the minimum number of points MinPts′ are obtained. As Figure 4 shown, in this embodiment, the algorithm processes each point as follows: First, the algorithm checks whether there are at least MinPts′ points within the ∈′ neighborhood of a certain point; if this condition is met, the point is regarded as a core point, and the surrounding points will be grouped into the same cluster. The core point and the points within its neighborhood form a cluster, and other points in this cluster can also continue to expand until all points are clustered or marked as noise; through clustering, objects without adjacent ones within the ∈′ distance are considered a cluster of point clouds, and the point cloud clusters of the landing gear are initially aggregated into a whole, which is beneficial for subsequent determination.

[0083] Step 3: For the point cloud clusters that meet the positional relationship, use the least squares method to fit the distribution of the points to find the best fitting circle, and determine whether the best fitting circle is valid.

[0084] Specifically, according to the actual positional relationship of the aircraft landing gear, the target landing gears should present an isosceles triangle relationship. Extract the centroid coordinates of each point cloud cluster, set the side length difference threshold, calculate the Euclidean distance between the centroid points of each cluster of point clouds, and within the allowable threshold range, retain the point cloud clusters that meet the relative position conditions, and add the center points of the clusters to the candidate set for subsequent geometric feature determination.

[0085] Specifically, the aircraft landing gear has obvious cylindrical characteristics and fixed geometric dimensions, and the determination direction of geometric features is relatively clear. First, according to the point coordinate boundaries of the point cloud cluster, check the height consistency of the points within each cluster, extract the z coordinates of all points in the point cloud cluster, take the maximum value, minimum value and compare them with the set range to ensure that the upper and lower surface heights of the target object are reasonable. For those that do not meet the range requirements, delete the point cloud cluster from the target candidate set. For those that meet the requirements, use the least squares method to fit the distribution of the points within the cluster to find the best fitting circle, estimate the center and radius of each cluster accordingly, and judge whether the fitting circle is valid according to the set size of the fitting circle.

[0086] Least squares circle fitting is a common method often used in 3D point cloud data processing to determine the optimal circular fit by minimizing the error. Its basic principle is to obtain the optimal parameters of the circle (center coordinates and radius) by minimizing the sum of the squares of the distances between the data points and the fitted circle. The goal of the least squares method is to find a circle such that the distance from each point to the circle is as small as possible, so that the fitted result is as close as possible to the actual point cloud distribution.

[0087] In a two-dimensional plane, the equation of a circle is usually expressed as:

[0088] (x - a) 2 +(y - b) 2 =r 2

[0089] where (a, b) are the coordinates of the center of the circle and r is the radius of the circle.

[0090] The goal of least squares circle fitting is to determine the center (a, b) and radius r such that the sum of the squares of the distances from each point in the point set to the circle is minimized. That is, to minimize the following error function:

[0091]

[0092] Here, n is the number of data points, and xi and yi are the coordinates of the i-th point. By minimizing this error function, the optimal center coordinates and radius can be obtained.

[0093] To facilitate the calculation, the equation can be linearized. By expanding the equation and introducing auxiliary variables, the least squares method usually solves it through matrix operations.

[0094] First, transform the equation of the circle into:

[0095] x 2 +y 2 +Dx + Ey + F=0

[0096] where D=-2a, E=-2b, and F=a 2 +b 2 -r 2 are the parameters to be determined. By performing least squares fitting on all points, the parameters D, E, and F can be obtained, and then the values of the center and radius can be calculated. This process can be achieved by solving a system of linear equations.

[0097] The effect of the fitted circle in the experiment is as shown in Figure 5(a) - Figure 5(b)As shown in the figure, on the left is the effect of fitting a circle to the point cloud cluster of the wheel, and on the right is the effect of fitting a circle to the point cloud cluster of the pallet. In the actual scenario, when the wheel point cloud is projected onto the xz plane along the Y axis, an obvious arc structure can be seen. This arc structure is also the key feature that differentiates the wheel from other objects. Therefore, it is chosen to first project all the point cloud clusters in the scene onto the xz plane along the Y axis, and then fit a circle. The closer the shape of the fitted circle is to the distribution of the point cloud, the higher the value of the subsequent calculated coverage, which is beneficial to the final discrimination of the target.

[0098] Step 4: Calculate the coverage of the effective best-fitting circle, sort the corresponding point cloud clusters according to the calculated coverage, and determine the point cloud cluster of the wheel target.

[0099] Specifically, calculate the coverage of the best-fitting circle to measure the matching degree between the distribution of all points in each cluster of point clouds and the corresponding best-fitting circle, and judge whether the coverage exceeds a predetermined threshold. Sort the point cloud clusters that meet the conditions, and select the top three clusters in descending order of coverage as the final point cloud clusters of the wheel target.

[0100] The calculation of the coverage of the best-fitting circle of the point cloud is an index to measure the matching degree between the fitting circle and the distribution of the actual point cloud data. The higher the coverage, the better the best-fitting circle can represent the distribution of the point cloud data.

[0101] First, for each point (x i , y i ) in the point cloud cluster, calculate its distance to the best-fitting circle. Assume that the center of the best-fitting circle is (a, b) and the radius is r. Then the distance d i from this point to the best-fitting circle can be calculated by the following formula:

[0102]

[0103] where d i represents the distance from the i-th point to the best-fitting circle.

[0104] The calculation method of the coverage can be expressed as:

[0105]

[0106] Among them, the threshold is used to judge whether the point is within the effective range of the fitting circle. Only when the difference between the distance from the point to the center of the circle and the radius is less than or equal to this threshold, the point is considered to have an impact on the coverage. The value of the coverage usually ranges between 0 and 1. The larger the value, the better the coverage effect of the fitting circle. In the project, the coverage of the wheel target is between 0.85 and 0.93, and the coverage of other objects is between 0.1 and 0.8. According to this determination condition, the target can be well identified.

[0107] Step 5: According to the determined target point cloud clusters of the landing gears, calculate the midpoint coordinates of the rear two landing gears of the aircraft based on the positional relationship between the belly of the aircraft and the landing gears of the aircraft, and determine the target position of the belly of the aircraft; and determine the target pose of the belly of the aircraft according to the positional relationship between the installation orientation of the lidar and the orientation of the aircraft.

[0108] Specifically, after obtaining the target point cloud clusters of the landing gears, distinguish the front and rear wheels according to the coordinate attributes of the points in each point cloud cluster, and further estimate the target position and target pose from the isosceles triangle positional relationship of the three landing gears.

[0109] According to the actual layout of the aircraft, the horizontal position of the midpoint of the belly coincides with the horizontal position of the midpoint of the two rear wheels. The in-cabin hangers are located on both sides inside the cabin, and their relative positions are fixed. The calculated midpoint coordinates of the two rear wheels are the final target points, and the horizontal coordinates of the hangers can be obtained through appropriate data compensation.

[0110] Since the lidar is installed on the trailer, the positive direction of the X-axis of its radar coordinate system coincides with the forward direction of the vehicle body. At the same time, the front wheel is on the vertical line of the midpoint of the connection line of the two rear wheels. Therefore, by calculating the angle between the connection line of the final target point and the midpoint of the front wheel in the lidar coordinate system and the radar X-axis, the target pose can be determined.

[0111] As Figure 6 shown, XOY is the lidar coordinate system. In the figure, a, b, and c are the midpoints of the point cloud clusters of the front wheel and the two rear wheels respectively. z is the midpoint of the bc line segment and also the required final target position. Subsequently, through the z-point coordinates and the aircraft's own dimension data, it is easy to obtain the coordinates of the in-cabin hanger in the belly. The angle θ between az and the X-axis is the angle between the forward direction of the aircraft in the radar coordinate system and the positive direction of the X-axis of the radar coordinate system. When the radar is installed, the positive direction of the X-axis is the same as the forward direction of the trailer. At the same time, it is required that the forward direction of the trailer is the same as the forward direction of the aircraft during the hanging operation. Therefore, θ is the pose angle that the trailer needs to deflect, that is, the target pose.

[0112] Through experimental verification, the size of the algorithm model in this embodiment is only 18 kb, the recognition accuracy rate of the landing gear target is 93.7%, and the recognition time is 0.15 s. From the data, it meets the actual requirements of the hanging task, that is, high accuracy, lightweight model, and fast recognition speed. At the same time, it can be used for another special-shaped aircraft only by modifying the landing gear and compensation parameters, and has high universality.

[0113] Embodiment 2

[0114] The purpose of this embodiment is to provide an aircraft landing gear recognition system based on three-dimensional point cloud, including:

[0115] An acquisition module, which is configured to: acquire the lidar point cloud data of the aircraft landing gear and perform preprocessing;

[0116] A position determination module, configured to: cluster the preprocessed lidar point cloud data by using a clustering domain radius and a minimum number of points to obtain a plurality of point cloud clusters, and determine the positional relationship of each point cloud cluster according to the actual positional relationship of the aircraft wheels;

[0117] A fitting module, configured to: for the point cloud clusters that conform to the positional relationship, fit the distribution of points by using the least squares method to find the best fitting circle, and determine whether the best fitting circle is valid;

[0118] An identification module, configured to: calculate the coverage of the valid best fitting circle, sort the corresponding point cloud clusters according to the calculated coverage, and determine the wheel target point cloud cluster.

[0119] In more embodiments, there is also provided:

[0120] An electronic device, including a memory, a processor, and computer instructions stored on the memory and running on the processor. When the computer instructions are run by the processor, the method described in Embodiment 1 is completed. For the sake of brevity, it will not be elaborated here.

[0121] It should be understood that in this embodiment, the processor may be a central processing unit CPU, and the processor may also be other general-purpose processors, digital signal processors DSP, application-specific integrated circuits ASIC, off-the-shelf programmable gate arrays FPGA, or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0122] The memory may include a read-only memory and a random access memory, and provide instructions and data to the processor. A part of the memory may also include a non-volatile random access memory. For example, the memory may also store information about the device type.

[0123] A computer-readable storage medium, for storing computer instructions, which when executed by the processor, complete the method described in Embodiment 1.

[0124] The method in Embodiment 1 can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules in the processor. The software module may be located in a mature storage medium in the art such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. This storage medium is located in the memory, and the processor reads the information in the memory and combines its hardware to complete the steps of the above method. To avoid repetition, it will not be described in detail here.

[0125] A computer program product, including a computer program, which when executed by the processor, implements the method described in Embodiment 1.

[0126] The present invention also provides at least one computer program product tangibly stored on a non-transitory computer-readable storage medium. The computer program product includes computer-executable instructions, such as instructions included in program modules, which are executed in a device on a target real or virtual processor to perform the processes / methods as described above. Generally, program modules include routines, programs, libraries, objects, classes, components, data structures, etc. that perform specific tasks or implement specific abstract data types. In various embodiments, the functions of program modules can be combined or divided as needed among program modules. The machine-executable instructions for program modules can be executed within local or distributed devices. In a distributed device, program modules can be located in local and remote storage media.

[0127] The computer program code for implementing the method of the present invention can be written in one or more programming languages. This computer program code can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when the program code is executed by the computer or other programmable data processing device, the functions / operations specified in the flowchart and / or block diagram are implemented. The program code can be executed entirely on the computer, partially on the computer, as a stand-alone software package, partially on the computer and partially on a remote computer, or entirely on a remote computer or server.

[0128] In the context of the present invention, the computer program code or related data can be carried by any suitable carrier such that a device, apparatus, or processor can perform the various processes and operations described above. Examples of carriers include signals, computer-readable media, and the like. Examples of signals can include electrical, optical, radio, acoustic, or other forms of propagated signals, such as carrier waves, infrared signals, etc.

[0129] Those of ordinary skill in the art can realize that the units and algorithm steps of the examples described in conjunction with this embodiment can be implemented by electronic hardware or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. A professional technician can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0130] Although the specific implementation manners of the present invention have been described above in conjunction with the accompanying drawings, it is not a limitation to the protection scope of the present invention. Those skilled in the art should understand that, based on the technical solution of the present invention, various modifications or deformations that can be made by those skilled in the art without creative efforts are still within the protection scope of the present invention.

Claims

1. A method for aircraft wheel recognition based on three-dimensional point cloud, characterized in that: include: Obtain the LiDAR point cloud data of the aircraft wheels and perform preprocessing; The pre-processed LiDAR point cloud data is clustered using the clustering neighborhood radius and the minimum number of points to obtain multiple point cloud clusters. The position relationship of each point cloud cluster is determined based on the actual position relationship of the aircraft wheels. The point cloud clusters that meet the positional relationship are fitted with the distribution of points using the least squares method to find the best fitting circle, and determine whether the best fitting circle is valid; The coverage of the effective best-fit circle is calculated, and the corresponding point cloud clusters are sorted according to the calculated coverage to determine the wheel target point cloud cluster.

2. The method for aircraft wheel recognition based on three-dimensional point cloud according to claim 1, characterized in that: The obtained LiDAR point cloud data of the aircraft wheels is preprocessed, including: By setting the boundary of the point cloud data coordinate value, the region of interest is divided, and the point cloud data located in the region of interest is filtered out from the laser radar point cloud data using Boolean indexing; The selected point cloud data is denoised using an adaptive radius filtering algorithm, and outliers are filtered out based on the set radius and neighborhood point count threshold.

3. The method for identifying aircraft wheels based on three-dimensional point clouds according to claim 1, characterized in that: The preprocessed LiDAR point cloud data is clustered using the clustering neighborhood radius and the minimum number of points to obtain multiple point cloud clusters. The position relationship of each point cloud cluster is determined based on the actual position relationship of the aircraft wheels. Specifically: Based on the clustering neighborhood radius and the minimum number of points, adjacent points are clustered into a point cloud cluster; According to the geometric position relationship of the aircraft wheels, the centroid coordinates of each point cloud cluster are extracted, and the Euclidean distance between the centroid points of each cluster point cloud is calculated; The positional relationship of the corresponding point cloud clusters is determined according to the calculated Euclidean distance.

4. The method for identifying aircraft wheels based on three-dimensional point clouds according to claim 1, characterized in that: The point cloud clusters that meet the positional relationship are fitted with the distribution of points using the least squares method to find the best fitting circle and determine whether the best fitting circle is valid; specifically: Based on the geometric features of the aircraft wheels, the consistency of the point heights within each point cloud cluster is checked according to the point coordinate boundaries of the point cloud cluster; For point cloud clusters that meet high consistency, the least squares method is used to fit the distribution of the point cloud clusters to find the best fitting circle; Based on the set fitting circle size, determine whether the best fitting circle corresponding to the point cloud cluster is valid.

5. The method for identifying aircraft wheels based on three-dimensional point cloud according to claim 1, characterized in that: Calculate the coverage of the effective best-fit circle, sort the corresponding point cloud clusters according to the calculated coverage, and determine the wheel target point cloud cluster; specifically: Calculate the distance between each point cloud in the point cloud cluster and the center of the corresponding best fitting circle; The number of point clouds within the effective range of the best fitting circle is determined by the set distance threshold, and the coverage corresponding to the point cloud cluster is calculated based on the determined point cloud data; The corresponding point cloud clusters are sorted according to the calculated coverage to determine the wheel target point cloud cluster.

6. The method for identifying aircraft wheels based on three-dimensional point cloud according to any one of claims 1 to 5, characterized in that: Also includes: According to the determined wheel target point cloud cluster and the positional relationship between the aircraft belly and the aircraft wheels, the midpoint coordinates of the two rear wheels of the aircraft are calculated to determine the target position of the aircraft belly; and according to the positional relationship between the laser radar installation orientation and the aircraft orientation, the target posture of the aircraft belly is determined.

7. The method for identifying aircraft wheels based on three-dimensional point clouds according to claim 1, characterized in that: The cluster neighborhood radius is determined based on the k-distance graph, and the minimum number of points is determined based on the Gaussian kernel density estimation.

8. An aircraft wheel recognition system based on three-dimensional point cloud, characterized in that: include: An acquisition module is configured to: acquire laser radar point cloud data of aircraft wheels and perform preprocessing; The position determination module is configured to: cluster the pre-processed laser radar point cloud data using a clustering neighborhood radius and a minimum number of points to obtain a plurality of point cloud clusters, and determine the position relationship of each point cloud cluster according to the actual position relationship of the aircraft wheels; A fitting module is configured to: fit the distribution of the point cloud clusters that meet the positional relationship using the least squares method to find the best fitting circle, and determine whether the best fitting circle is valid; The identification module is configured to: calculate the coverage of the effective best-fit circle, sort the corresponding point cloud clusters according to the calculated coverage, and determine the wheel target point cloud cluster.

9. An electronic device, characterized in that: The method comprises a memory and a processor and computer instructions stored in the memory and executed on the processor, wherein when the computer instructions are executed by the processor, the method according to any one of claims 1 to 7 is completed.

10. A computer-readable storage medium, characterized in that: Used to store computer instructions, which, when executed by a processor, complete the method described in any one of claims 1 to 7.

Images