A method for generating the tip trajectory of an aviation blade based on structured light vision

Abstract

The present invention discloses a method for generating the tip trajectory of an aviation blade based on structured light vision, comprising the following steps: capturing the tip point cloud of the aviation blade by a 3D surface structured light camera and filtering and denoising it to obtain a homogenized point cloud; using the principal component analysis method to determine the local normal and the tangent plane, and determining the edge point cloud in the homogenized point cloud from the angle of the projection vector in the tangent plane; using the principal component analysis method to determine the first principal direction of the tip represented by the edge points in the homogenized point cloud, constructing a new space coordinate system, and solving and fitting the intermediate line by the least square method with polynomials with endpoint constraints in the XY and XZ dimensions, and then converting the fitted intermediate line back to the original coordinates. The present invention can adapt to aviation tips of different sizes, extract the intermediate line thereof, and has improvements in terms of adaptability, computational efficiency, and noise resistance.

Description

Technical Field

[0001] The present invention relates to a method for generating the tip trajectory of a blade, belonging to the technical field of irregular contour measurement. Background Art

[0002] Aeroengine blades are the core components of an aeroengine. As a typical representative of complex curved surface workpieces, their surfaces have non-uniform curvature and complex three-dimensional geometric structures, and need to meet strict hydrodynamic requirements. This characteristic leads to drastic changes in the tip surface and frequent small-radius bending, posing high requirements for the accuracy of centerline extraction, and at the same time, strict control over the smoothness and stability of the centerline is required.

[0003] The existing technical solutions mainly have problems of poor adaptability to asymmetric regular blades or regions with high curvature changes on the blades, resulting in poor extraction accuracy and high computational complexity. Summary of the Invention

[0004] Aiming at the above-mentioned defects of the existing technology, the task of the present invention is to provide a method for generating the tip trajectory of an aeroengine blade based on structured light vision, which solves the problems of poor adaptability to the curvature change region of the blade, low accuracy and high computational complexity.

[0005] The technical solution of the present invention is as follows: A method for generating the tip trajectory of an aeroengine blade based on structured light vision, comprising the steps of:

[0006] Capturing the tip point cloud of the aeroengine blade by a 3D surface structured light camera and filtering and denoising to obtain a homogenized point cloud;

[0007] For each point in the homogenized point cloud, using the principal component analysis method to determine the local normal and construct a tangent plane, calculating the projection of the neighboring points of the point on the tangent plane, and screening the points with the angular difference of the projection polar angle exceeding a set angle threshold as edge points;

[0008] Using the principal component analysis method to determine the first principal direction of the tip represented by the edge points in the homogenized point cloud, and constructing a new coordinate system Make Align the first principal direction, transform the edge point cloud from the original coordinate system to the new coordinate system, solve and fit the centerline by the least squares method with a polynomial with endpoint constraints in the XY and XZ dimensions, and then transform the fitted centerline back to the original coordinates.

[0009] Further, after the centerline is transformed back to the original coordinates, the centerline is smoothed and uniformly segmented.

[0010] Further, the steps of the uniform segmentation include first calculating the cumulative moment height of the entire centerline, then determining the positions of each sampling point according to the total length and the target number of segments, and obtaining the coordinates of the corresponding points by linear interpolation.

[0011] Further, the filtering and denoising includes first performing height filtering by setting a Z-axis numerical range, and then performing voxel filtering.

[0012] Further, determining the local normal and constructing a tangent plane for each point in the homogenized point cloud by using the principal component analysis method includes selecting points from the point cloud of k nearest neighbor points, denoted as , where , where , is the abscissa of point , is the ordinate of point , is the height coordinate of point ;

[0013] Zero-mean the point cloud data to , where:

[0014] ,

[0015] ,

[0016] ,

[0017] where: ;

[0018] Calculate the eigenvalues and eigenvectors of the covariance matrix G to obtain three eigenvalues and the corresponding eigenvectors , and the eigenvector corresponding to the minimum eigenvalue is the normal vector of point , and the plane orthogonal to the normal vector is the tangent plane, , where is the transpose of .

[0019] Further, k is not less than 50.

[0020] Further, when screening the angular difference of the projection polar angle, the head and tail angular difference is , represents the first projection polar angle, represents the last projection polar angle.

[0021] Further, determining the first principal direction of the edge points in the homogenized point cloud by using the principal component analysis method includes:

[0022] Set the three-dimensional edge point cloud data in the original coordinate system , where:

[0023] ,

[0024] ,

[0025] ,

[0026] Zero-mean the point cloud data to , where:

[0027] ,

[0028] ,

[0029] ,

[0030] Among them, N is the total number of edge point clouds, ,

[0031] Calculate the eigenvalues and eigenvectors of the covariance matrix C, and the eigenvector corresponding to the largest eigenvalue represents the first principal direction of the data,

[0032] , where is transpose of.

[0033] Furthermore, when constructing a new coordinate system, select a reference vector that is not parallel to the axis in space, axis is , is unit vector of the axis, and then determine the axis through the right-hand screw rule, axis is .

[0034] Furthermore, the fitting of the intermediate line by the least squares method with polynomials with endpoint constraints in the XY and XZ dimensions includes:

[0035] Construct a linear interpolation function passing through the endpoints with the extreme points in the first principal direction of the leaf tip as the endpoints,

[0036] ,

[0037] ,

[0038] The coordinates of the extreme points are ;

[0039] Define the residual between the point cloud data and the linear basis difference through polynomial correction:

[0040] ,

[0041] ,

[0042] where, ,

[0043] is the weighting term, k is the polynomial order,

[0044] Convert the problem into a system of linear equations and solve for the polynomial coefficients a and b using the least squares method.

[0045] The advantages of the present invention compared with the prior art are as follows:

[0046] The present invention can adapt to aviation blade tips of different sizes and stably extract the center line under different postures and positions. This method has improved in terms of adaptability, computational efficiency, and noise resistance, and optimizes the trajectory consistency through a uniform segmentation strategy, making it have more extensive application potential in the task of extracting aviation blade tip features under complex working conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 FIG. is the extraction result diagram after smoothing the method for generating the aviation blade tip trajectory based on structured light vision in the embodiment.

[0048] Figure 2 FIG. is the extraction result diagram after uniform segmentation of the method for generating the aviation blade tip trajectory based on structured light vision in the embodiment.

[0049] Figure 3 FIG. is the extraction result diagram of the method for generating the aviation blade tip trajectory based on structured light vision in the embodiment on five types of data. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0050] The following further describes the present invention in conjunction with embodiments, but does not limit the present invention.

[0051] The method for generating the aviation blade tip trajectory based on structured light vision in this embodiment includes the following steps:

[0052] Capture the blade tip point cloud through a 3D surface structured light camera.

[0053] The captured point cloud data contains a large number of outlier noises. The noise points may come from environmental interference, sensor errors, etc. These noises will have an adverse impact on the subsequent feature extraction process. Therefore, it is necessary to further clean and filter the point cloud data to reduce noise and unnecessary information.

[0054] The process of denoising the point cloud data includes height filtering and voxel filtering. In this embodiment, a PassThrough filter based on height is adopted. According to the tip heights of blades of different models, specific Z-axis numerical intervals are manually set to screen the target point cloud. Height filtering helps with data dimensionality reduction and noise suppression, and can extract effective information within a specific height region, thereby improving the accuracy and efficiency of subsequent analysis and processing algorithms.

[0055] After height filtering, there are still a large number of points in the point cloud. It is also necessary to perform downsampling through voxel filtering to homogenize the point cloud. The voxel grid filter divides the space into cubic voxels, averages the points within each voxel, and generates a new point cloud. By voxel filtering, the resolution of the point cloud is reduced, redundant points in dense regions are eliminated, the number of data points is reduced, the point cloud becomes more uniform, and at the same time, the key feature information of the original point cloud is retained, thereby improving the processing speed and reducing the impact of noise.

[0056] Extract the edge point cloud for the homogenized point cloud. Specifically,

[0057] Define the point cloud coordinate system as , let the three-dimensional point cloud data in the original coordinate system be , and the total number of points in the point cloud after voxel filtering downsampling is , where:

[0058] ,

[0059] ,

[0060] ,

[0061] Select points 's k (in this embodiment, select k=60 ) nearest neighbor points, denoted as , where . Among them , is the abscissa of point , is the ordinate of point , is the height coordinate of point .

[0062] Zero-mean the point cloud data to , where:

[0063] ,

[0064] ,

[0065] ,

[0066] where: .

[0067] Calculate the covariance matrix:

[0068] , where is the transpose of.

[0069] From the formula calculate the eigenvalues and eigenvectors of the covariance matrix G, where is the eigenvalue, representing the data variance magnitude in that direction, is the eigenvector, representing the main distribution trend of the data in that direction, obtaining three eigenvalues and the corresponding eigenvectors , it is agreed that , then the eigenvector corresponding to the smallest eigenvalue is the normal vector of the point . .

[0070] Define the tangent plane of the point (i.e., the plane orthogonal to ). For each neighborhood point of the point , project the vector onto the tangent plane to obtain the projection vector , where " " represents the dot product operation. Convert all the projection vectors to polar coordinate representation (on the tangent plane), calculate their polar angles, and let these angles be . After sorting all the angles, calculate the difference between two adjacent angles and consider the head and tail connected.

[0071] .

[0072] If there exists an angle difference greater than a preset angle threshold (the threshold set in this embodiment is ), then it can be considered that the point is located on the edge, thereby finding all the edge points in the homogenized point cloud.

[0073] Aero tip point cloud data usually has problems such as surface mutation regions, local data loss, or uneven sampling. Relying solely on normal estimation cannot fully distinguish these boundary features. The above method can sensitively identify and locate the boundary regions where the point cloud density changes significantly, geometric features are discontinuous, or there are data gaps on the surface by analyzing the projection distribution and angular interval of neighboring points on the local tangent plane, thereby effectively determining the positions of the true edge points in the point cloud.

[0074] The aero tip has an obvious streamlined structure. No matter what its position or attitude is, the first principal direction can always be adaptively found through PCA, that is, the first principal direction of the tip represented by the edge points in the homogenized point cloud is calculated by the principal component analysis method, which is the direction with the largest data variance in three-dimensional space. Specifically,

[0075] Let the total number of edge point clouds be N, and the three-dimensional edge point cloud data in the original coordinate system , where:

[0076] ,

[0077] ,

[0078] ,

[0079] Zero-mean the point cloud data to , where:

[0080] ,

[0081] ,

[0082] ,

[0083] where .

[0084] Calculate the covariance matrix:

[0085] , where is transpose of.

[0086] From the formula Calculate the eigenvalues and eigenvectors of the covariance matrix C, and the eigenvector corresponding to the largest eigenvalue represents the first principal direction of the data, that is, the direction with the largest data variance.

[0087] Establish a new local space coordinate system so that the axis in the new coordinate system is aligned with the first principal direction calculated above, facilitating the subsequent extraction of the middle line.

[0088] Specifically, the first principal direction is normalized,

[0089] ,

[0090] and is used as the axis in the new coordinate system.

[0091] Next, a reference vector that is not parallel to is selected in space, and is calculated, where " " represents the cross product. Further normalization is performed on it:

[0092] ,

[0093] and is used as the axis in the new coordinate system.

[0094] is determined by the right-hand screw rule . is normalized , and is used as the axis in the new coordinate system. axis.

[0095] Finally, the new coordinate system is .

[0096] A rotation matrix R is constructed to transform the edge point cloud from the original coordinate system to the new coordinate system.

[0097] .

[0098] The matrix R is used to transform the point cloud from the original coordinate system to the new coordinate system. The column vectors of the rotation matrix R are three normalized unit direction vectors, that is, among them

[0099] .

[0100] The points in the new coordinate system are obtained through the transformation formula , and its expansion is

[0101] ,

[0102] where is the three-dimensional point cloud data in the new coordinates, is the centroid of the point cloud: .

[0103] The middle line of the aviation blade tip is a curve, and the subsequent blade tip repair work requires that the fitted middle line passes through the front and rear endpoints of the blade tip. Therefore, polynomial fitting of the middle line is carried out by endpoint constraints.

[0104] The point cloud in the new coordinate system is unfolded along the axis direction (i.e., the first principal direction). The endpoint constraints are determined based on the extreme points in the first principal direction ( axis), that is, the starting point and the ending point of the middle line correspond to the two points at the frontmost and rearmost ends of the first principal direction respectively. That is, the two endpoints are respectively:

[0105] .

[0106] First, construct a linear interpolation function passing through the endpoints as the reference curve, whose function is to ensure that the fitted curve passes through the endpoints.

[0107] ,

[0108] ,

[0109] Define the residual between the point cloud data and the linear basis difference as

[0110] ,

[0111] where .

[0112] Correct the residual by polynomial while ensuring that the residual at the endpoints is 0,

[0113] ,

[0114] ,

[0115] where

[0116] ,

[0117] is the weighting term, k is the polynomial order.

[0118] The role of the weighting term is:

[0119] When or , , that is, it is 0 at the endpoints, ensuring that the fitted curve strictly passes through the endpoints.

[0120] And when , , that is, in the middle region, the polynomial is allowed to freely fit the residual.

[0121] Then solve it by the least squares method.

[0122] Convert the problem into a system of linear equations:

[0123] ,

[0124] in ,Right now:

[0125] ,

[0126] in: .

[0127] ,

[0128] .

[0129] Expanding A gives:

[0130] ,

[0131] Where a and b are the coefficients to be solved:

[0132] .

[0133] Minimize the residual sum of squares using the least squares method and , calculate the optimal solution of polynomial coefficients a and b respectively, , , thus obtaining a higher precision fitting effect of the middle line of the point cloud under the endpoint constraints.

[0134] The point on the middle line Back to the original coordinate system, the conversion formula is: . Expanding it yields the following result:

[0135] .

[0136] Since the middle line extracted by polynomial fitting is usually accompanied by certain noise and local fluctuations, it is necessary to smooth it. This process helps to effectively suppress noise and eliminate irregular fluctuations, so that the middle line is more in line with the requirements of the actual trajectory and improves the stability and availability of the trajectory. The above structure is smoothed. In this embodiment, a sliding window averaging method is specifically used, that is, for each point, the average value of the points within a certain range before and after it is taken as the coordinates of the new point.

[0137] Indicates the half-width of the window: ,parameter Specifies the size of the sliding window.

[0138] Let the total number of point clouds on the fitted intermediate line be H. For the i-th point on the intermediate line, where , it is necessary to extract the neighborhood points within the window. If the i-th point , then its neighborhood point set is:

[0139] ,

[0140] Among them, represents the neighborhood set of the i-th point, which contains all points within the legal index range.

[0141] Calculate the new smoothed point , where

[0142] ,

[0143] ,

[0144] ,

[0145] n is the number of valid points in the neighborhood: . After such processing, the coordinates of each point are smoothed by the average value of its surrounding neighborhood. The extraction result of the intermediate line after smoothing is as Figure 1 shown.

[0146] Figure 1 In (a) of Figure 1 is the intermediate line in the original coordinate system, Figure 1 In (b) of

[0147] is the position map of the extracted intermediate line and the edge point cloud, indicating that the extracted intermediate line is located at the center of the edge point cloud in the original coordinate system. Figure 1 In (c) of

[0147] is the position map of the extracted intermediate line and the original point cloud, indicating that the finally extracted intermediate line is successfully located at the center position of the original point cloud. The above comparison shows that the extraction process of the intermediate line is highly consistent with the structure of the point cloud.

[0147] After the smoothing process, the intermediate line contains a large number of trajectory points. If all the trajectory points are directly used for subsequent trajectory guidance, it will significantly increase the system calculation burden and affect the operation efficiency. To effectively control the computational complexity while ensuring the trajectory accuracy, it is necessary to downsample the intermediate line using a length-based uniform segmentation strategy, so that the distance between adjacent trajectory points remains consistent, thereby generating uniformly distributed and representative sampling points as the final guidance trajectory points. In terms of the selection of the number of segments, a reasonable balance between accuracy and computational load is also required: if the number of segments is too large, although the fitting accuracy of the trajectory can be improved, it will also significantly increase the system calculation amount; conversely, if the number of segments is too small, it may lead to too sparse distribution of trajectory points, thereby affecting the smoothness and stability of the guidance process. Therefore, reasonable setting of the number of segments is crucial for achieving efficient and accurate trajectory guidance.

[0148] The steps of downsampling using the uniform segmentation strategy include:

[0149] For every two points on the smoothed intermediate line and , calculate the Euclidean distance between them:

[0150] .

[0151] Then construct an array of cumulative lengths , where . The calculation formula is , where when , , when , is denoted as the total length of the intermediate line I .

[0152] Divide the intermediate line evenly into segments, then the length of each segment is . For each segmented sampling point j ( j from 0 to ), the target cumulative length on its entire line is . If the cumulative length of the target sampling point is equal to the cumulative length of the existing point on the intermediate line, then the existing point on the intermediate line is the th target sampling point.

[0153] For each target cumulative length , it is necessary to find the interval where it is located in the cumulative length array. However, the cumulative length of the target sampling point is not necessarily equal to the cumulative length of the existing point on the intermediate line. If , is the cumulative length of the point before the target sampling point, is the cumulative length of the point after the target sampling point. idx is the index found in k . The scale factor for this segment is: , t ranges from (0, 1), and use t to perform linear interpolation to calculate the new target sampling point.

[0154] ,

[0155] ,

[0156] ,

[0157] where As the th new target sampling point on the center line, represents the abscissa of the i th point among the existing points on the center line, represents the ordinate of the i th point among the existing points on the center line, represents the height coordinate of the i th point among the existing points on the center line. represents the abscissa of the th point among the existing points on the center line, and so on.

[0158] When the target length is exactly equal to 0, directly take the first point on the center line. When the target length is longer than the total length, directly take the last point. Through the above process, finally, a new point set can be resampled at equal intervals according to equal lengths on the center line after smoothing processing, realizing uniform segmentation. According to the characteristics of the aviation blade, the selected in this embodiment is 4, and finally 5 key feature points are obtained. The final result after uniform segmentation is as Figure 2 shown.

[0159] In Figure 2 , the 5 points in (a) are the new target sampling points, which evenly divide the center line into four segments. Figure 2 The 5 points in (b) are the Figure 1 5 new target sampling points in (a). Take these 5 points as the trajectory points to represent the generated trajectory. As can be seen from Figure 2 (c), the generated trajectory is located at the center position of the original point cloud, indicating that the generation of the trajectory is highly consistent with the geometric distribution of the original point cloud. Through the above comparison, it shows that the extracted trajectory points have high accuracy and consistency, thus ensuring the smoothness of the trajectory.

[0160] Verify the effect of the method of the present invention through experiments. The experimental data collection is divided into five categories. The first category of data is collected by a 3D structured light camera for a 1:1 model of a real aviation small blade tip. A total of 14 groups of data are collected. Among the 14 groups of data, there are two types: one is that the position remains unchanged and the attitude of the aviation blade tip model is different, and one group is collected every 45 degrees, for a total of 8 groups (A1 - A8). The other is collected at any position and any attitude, for a total of 6 groups (A9 - A14).

[0161] The second type of data is the real aviation large leaf tips collected by a 3D structured light camera, the third type of data is the real aviation small leaf tips collected by a 3D structured light camera, and the fourth type of data comes from the real aviation large and small leaf tip data collected by the second sensor. To verify the anti-noise performance of the algorithm, the fifth type of data is also the real aviation large and small leaf tip data collected by the second sensor, but it is obtained under overexposure conditions, and its exposure rate is 1.37 times higher than that of the fourth type of data.

[0162] Figure 3 Rows a, b, c, d, and e respectively correspond to the first, second, third, fourth, and fifth types of data. Figure 3 In it, the first column is the original point cloud, the five trajectory points in the second column evenly segment the middle line and have the same structural height as the original point cloud, and the five trajectory points in the third column have the same structural height as the original point cloud. Although some point clouds are missing in the data of row c, the middle line can still be extracted. Due to the missing point cloud data, the data distribution may be uneven during polynomial fitting, resulting in insufficient fitting constraints in local areas, thus reducing the smoothness of the middle line. The data of row e is collected under overexposure conditions, but the method of the present invention can still extract the middle line well, further indicating that the method has good robustness.

[0163] Further verification of the accuracy of the extraction results is carried out. For the same type of data, due to different acquisition positions and postures, it may affect the consistency of the middle line extraction. Therefore, a point cloud is selected from the same type of data, and the middle line extracted from it is used as the reference point cloud, while the middle lines extracted from the remaining point clouds are used as the registration point clouds. The ICP algorithm is used to perform rigid registration on the registration point clouds and the reference point cloud, and its registration error is calculated. For the same type of data, through the ICP technique, multiple repeated tests are carried out to analyze the change of the registration error of different point clouds to evaluate the stability and consistency of the middle line extraction. Finally, by calculating the Final RMS error, the repeatability and reliability of the algorithm are quantified, and the results of the first type of data are shown in Table 1.

[0164] Table 1 Final RMS error

[0165]

[0166] The same operations are performed on the remaining four types of data, the Final RMS errors of each type of data are calculated respectively, and finally the Final RMS errors of the five types of data are summarized to calculate their Final RMS mean value to evaluate the stability and consistency of the algorithm. The results are shown in Table 2.

[0167] Table 2 Final RMS error mean value

[0168]

[0169] There are 37 groups of data in total for the experiment. Excluding the case of self-registration with itself, among the remaining 30 groups of data, the proportion of those with a Final RMS error greater than 0.1 is 6.7%. The reason is that when extracting the middle line, the endpoints are not accurately found, resulting in the entire extracted middle line deviating slightly from the center trajectory. The proportion of those less than 0.1 is 93.3%. Among them, the proportion of those with an error between 0.05 and 0.1 is 73.3%, and the proportion of those less than 0.05 is 26.7%. As can be seen from Table 2, even under the influence of overexposure, the average value of the Final RMS error of the data is still below 0.1, indicating that the method of the present invention has good robustness. Repeated experiments verified by means of ICP technology also illustrate the stability and feasibility of the method of the present invention.

Claims

1. A method for generating the tip trajectory of an aviation blade based on structured light vision, characterized in that, Including the following steps: Capturing the tip point cloud of the aviation blade by a 3D structured light camera and filtering and denoising it to obtain a homogenized point cloud; For each point in the homogenized point cloud, using the principal component analysis method to determine the local normal and construct a tangent plane, calculating the projection of the neighboring points of the point on the tangent plane, and screening out the points with the angular difference of the projection polar angle exceeding the set angle threshold as edge points; The principal component analysis method is used to determine the first principal direction of the leaf tip represented by the edge points in the homogenized point cloud, and a new coordinate system is constructed. Make Align with the first principal direction, transform the edge point cloud from the original coordinate system to the new coordinate system, solve the fitting intermediate line by the least squares method with polynomials with endpoint constraints in the XY and XZ dimensions, and then transform the obtained intermediate line back to the original coordinates. Using the principal component analysis method to determine the first principal direction of the edge points in the homogenized point cloud, including: Set the three-dimensional edge point cloud data in the original coordinate system , where: ， ， ， Zero-mean the point cloud data into , where: ， ， ， where N is the total number of edge point clouds, , Calculate the eigenvalues and eigenvectors of the covariance matrix C, and the eigenvector corresponding to the largest eigenvalue represents the first principal direction of the data , where is the transpose of; When constructing a new coordinate system, select a reference vector in space that is not parallel to the axis , axis is , is the unit vector of the axis, and then determine the axis through the right-hand screw rule. The axis is ; The step of solving the fitting intermediate line by the least squares method with a polynomial with endpoint constraints in the XY and XZ dimensions includes: Taking the extreme points in the first principal direction of the leaf tip as endpoints and constructing a linear interpolation function passing through the endpoints; ， ， The coordinates of the extreme point are ; Correcting the residual between the point cloud data and the linear basis by a polynomial; ， ， Among them, , is a weighted term, k is the polynomial order, Converting the problem into a system of linear equations: ， Among them , Wherein: ; ， ， a and b are coefficients to be solved; , the least squares method is used to minimize the sum of squared residuals and , the optimal solutions of the polynomial coefficients a and b are calculated respectively, , .

2. The method for generating the tip trajectory of an aviation blade based on structured light vision according to claim 1, wherein After the intermediate line is transformed back to the original coordinates, smoothing the intermediate line and performing uniform segmentation.

3. The method for generating the tip trajectory of an aviation blade based on structured light vision according to claim 2, wherein The step of uniform segmentation includes first calculating the cumulative length of the whole intermediate line, then determining the positions of each sampling point according to the total length and the target number of segments, and obtaining the coordinates of the corresponding points by linear interpolation.

4. The method for generating the tip trajectory of an aviation blade based on structured light vision according to claim 1, characterized in that The filtering and denoising includes first performing height filtering by setting the Z-axis numerical interval, and then performing voxel filtering.

5. The method for generating the tip trajectory of an aviation blade based on structured light vision according to claim 1, wherein For each point in the homogenized point cloud, the principal component analysis method is used to determine the local normal and construct the tangent plane, including selecting a point from the point cloud of k nearest neighbor points, denoted as , where , where , is the abscissa of point , is the ordinate of point , is the height coordinate of point ; Zero-mean the point cloud data into , where: ， ， ， Wherein: ; Calculate the eigenvalues and eigenvectors of the covariance matrix G to obtain three eigenvalues and the corresponding eigenvectors , and the eigenvector corresponding to the smallest eigenvalue is the normal vector of the point . The plane orthogonal to the normal vector is the tangent plane , where is the transpose of​ 6. The method for generating the tip trajectory of an aviation blade based on structured light vision according to claim 5, characterized in that, k The value of is not less than 50.

7. The method for generating the tip trajectory of an aviation blade based on structured light vision according to claim 1, wherein The angular difference between the head and tail when screening the angular difference of the projection polar angle is , denotes the first projection polar angle, denotes the last projection polar angle.

Images