Irregular skin point cloud contour extraction method for micro-plane intersection projection

Abstract

The invention belongs to the technical field of three-dimensional digital measurement and aeronautical manufacturing, and discloses an irregular skin point cloud contour extraction method for micro-plane intersection projection. Firstly, skin point cloud is simplified through a voxel downsampling method, and outlier noise points are removed at the same time. Secondly, based on local gradient threshold segmentation of the global point cloud, an end face part in the skin point cloud is separated out, and a surface part in the skin point cloud is separated out with a curvature difference as a growth limiting condition through a region growing method; and finally, traversing all the end surface points, dividing and fitting tiny planes of the end surface and the nearest surface, solving an intersection line of the two tiny planes, projecting the end surface points to the intersection line, projecting all the end surface points to the corresponding intersection line to obtain complete skin contour points, and performing spline curve fitting to obtain a smooth and continuous skin boundary contour line. And an accurate edge contour can be provided for trimming and trimming or shape calibration of the skin.

Description

Technical Field

[0001] This invention belongs to the field of three-dimensional digital measurement and aerospace manufacturing technology, and in particular relates to the extraction of irregular aircraft skin contours based on laser scanning point clouds. Background Technology

[0002] Aircraft skins come in a wide variety of types and shapes, and are mostly irregular thin-walled parts. Because the skins deform during the forming process, it is crucial to quickly and accurately acquire their actual contours during assembly to guide subsequent edge trimming or shape calibration, resolving forced assembly issues caused by deformation, and thus achieving low-stress assembly of the skins. Contact-based contour measurement methods such as laser trackers and coordinate measuring machines can only measure discrete points, making it difficult to characterize continuous skin contours, or their application in manufacturing is limited by measurement range. Laser scanning can quickly acquire complete point clouds of the skin parts, and then extract the complete skin contour through point cloud processing, offering advantages in both efficiency and ease of operation.

[0003] Currently, commonly used point cloud contour extraction methods are mostly based on geometric features such as curvature and normal vectors for segmentation. In the patent "A Contour Extraction Method Based on Point Cloud Data" (Publication No. CN118982552A), points are projected onto a plane, and the probability that a point is an edge point is calculated to determine boundary points and internal points to obtain the contour. However, the rounded corners in the transition area between the surface and the end face affect the accuracy of boundary recognition, and the position of the end face is not considered. The extracted contour does not provide effective guidance for the assembly of skin with end face fitting. In "A Contour Extraction Method Based on Additive Remanufacturing Point Cloud Model" (Patent No. CN108765434A), the idea of ​​obtaining an ordered contour line from contour feature points is used to obtain the contour. However, this method has certain limitations and is not applicable to skins with multiple arcs or other irregular shapes. The contour extraction method by finding the intersection of the end face and the surface can accurately extract straight contours, but it is not applicable to skins with multiple arcs or other irregular shapes. Summary of the Invention

[0004] To address the problem of boundary misidentification in irregular skin contour extraction, and considering the fitting relationship of end faces during skin assembly, this invention proposes a method for extracting irregular skin point cloud contours using microplane intersection projection. Through steps such as "point cloud preprocessing - end face separation - upper surface separation - microplane intersection projection - B-spline fitting", the method achieves aircraft skin boundary extraction based on scanned point clouds.

[0005] The technical solution of this invention:

[0006] A method for extracting the contour of irregular skinned point clouds by microplane intersection projection, comprising:

[0007] For the irregular skin point cloud data obtained by scanning, the voxel grid method is used for downsampling and outlier noise points are removed to reduce the subsequent computation and noise interference.

[0008] Principal component analysis (PCA) is performed on the preprocessed point cloud data to fit the principal plane of the point cloud. The rotation matrix between the principal plane and the original 3D Cartesian coordinate system A of the point cloud data is calculated, and the principal plane is rotated to the reference horizontal plane marked with Z=0 under the original 3D Cartesian coordinate system A. The Z coordinate of the point cloud can directly reflect the height relative to the principal plane. At this time, the gradient difference is calculated with the Z-axis as the height direction, without the need to calculate complex spatial angles, which greatly simplifies the calculation logic of local height gradient difference. All points are traversed, and the local height gradient difference in a certain neighborhood of each point is calculated. A local height gradient threshold is set, and points with gradient differences greater than the threshold are marked as end points.

[0009] After removing points whose distance from the end face exceeds a preset value, select suitable seed points on the upper surface, and use a region growth algorithm with the curvature difference within the growth range as the growth judgment condition to separate the upper surface part close to the end face.

[0010] Retrieve each end face point and the nearest point on the upper surface. Fit the end face point The equation M of the end face microplane is obtained by taking its k-neighbor points and fitting the corresponding nearest upper surface point. The equation N for the microplane of the upper surface is obtained by taking the k-neighbor points and the equation of the end face point. The projection point on the intersection line L of the two microplanes MN ;

[0011] The obtained set of all end face projection points {P} is the desired contour point set. The contour boundary of the part can be obtained by fitting a B-spline curve to the points in the point set.

[0012] As a further technical solution, the irregular dot cloud data is acquired by a laser scanner, and the irregular dot cloud data includes end face and upper surface data of the skin boundary portion.

[0013] As a further technical solution, voxel grid method for downsampling includes:

[0014] The three-dimensional space is divided into several voxel grids of uniform size. The centroid of each voxel grid is calculated as a unique representative point to achieve uniform downsampling of point cloud data, which effectively reduces the amount of subsequent calculations while preserving the overall geometric features.

[0015] As a further technical solution, the steps for calculating the rotation matrix include:

[0016] Estimate the unit normal vector of the point cloud data using principal component analysis. ;

[0017] To ensure that the principal plane of the point cloud is perpendicular to the Z-axis after rotation, the unit normal vector of the principal plane of the point cloud is used. Unit vector parallel to the Z-axis The cross product result is used as the rotation axis and rotate the axis Normalization; calculation and The angle between them is used as the rotation angle. ;

[0018] Using Rodrigues's rotation formula, based on the axis of rotation and rotation angle Generate rotation matrix ;

[0019]

[0020] in, It is a 3×3 identity matrix. To be with the axis of rotation The associated oblique symmetric matrix.

[0021] As a further technical solution, the calculation steps for the local height gradient difference are as follows:

[0022] For planar parts, the local height gradient difference at each point can be approximated as the standard deviation of the Z value in the k-neighborhood of that point.

[0023]

[0024] in Let Z be the Z coordinate of the i-th point in the k-neighborhood of this point; The mean Z-coordinate of all points in the k-neighborhood of this point;

[0025] For curved surface parts, considering the inherent Z-value variation on the upper surface, which might incorrectly identify the upper surface as an end face, the original 3D Cartesian coordinate system A origin O of the point cloud data is first transformed to the estimated curvature center of the surface, and then converted to a spherical coordinate system. At this point, the local height gradient difference of each point can be approximated as the radial distance from the point to the origin. Standard deviation;

[0026]

[0027] in , is the radial distance of the i-th point in the k-neighborhood of this point in the transformed coordinate system; This is the mean radial distance between the point and all points in its k-neighborhood;

[0028] The proportion of the original three-dimensional Cartesian coordinate system A origin O in the point cloud data is used to filter out the end face point cloud by setting a reasonable height gradient ratio threshold.

[0029] As a further technical solution, a region growing algorithm is used, with the curvature difference within the growing range as the growth criterion, to separate the upper surface portion near the end face, including:

[0030] First, calculate the local unit normal vector point by point, and calculate the mean vector of all local unit normal vectors. Set the direction of the mean vector as the reference direction pointing to the outer surface. Iterate through the local normal vectors of each point. If the dot product with the reference direction is less than zero, reverse the direction of the normal vector, thereby adjusting the direction of the local normal vectors of all points to point to the outer surface.

[0031] For each point, search its k-neighborhood, calculate the angle between the normal vector of the point and its neighboring points, and take the average of the above angles as the local curvature estimate of the point.

[0032]

[0033] in The desired local curvature; Let be the angle between the normal vector of the i-th point in the k-neighborhood and the mean vector;

[0034] In the point cloud, a seed point is specified on the upper surface. Growth starts from the seed point. The curvature difference between points in a certain neighborhood of the seed point is calculated. If the difference is less than the preset curvature threshold and the neighboring point has not been visited, the neighboring point is used as a new seed point. The next round of growth is carried out and judged. Finally, all seed points are the points on the upper surface.

[0035] The beneficial effects of this invention are:

[0036] 1. This invention is aimed at aircraft skin with a thickness of only a few millimeters. By separating the end face and the top surface from the scanned point cloud, dividing and fitting the micro-surface to find the intersection, and projecting the end face points onto the intersection line to obtain the skin contour, this method takes into account the assembly relationship between skins for contour extraction, which can provide a more accurate boundary contour for subsequent skin shape calibration or edge trimming.

[0037] 2. This invention only uses point clouds near the boundary, making the input and calculation process simple, and it also has the advantage of high speed when processing point clouds of large-sized skin parts. Attached Figure Description

[0038] Figure 1 The diagram shows a flowchart of a method for extracting the contour of irregular skinned point clouds based on microplane intersection projection according to the present invention.

[0039] Figure 2 The diagram shows the end face and upper surface separated from the part's point cloud.

[0040] Figure 3 The diagram shows a local microplane fitting to find the intersection and projecting the end face points onto the intersection line.

[0041] Figure 4 The image shown is the final extracted result of a certain skin boundary contour. Detailed Implementation

[0042] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings and technical solutions.

[0043] like Figure 1 As shown, an embodiment of the present invention provides a method for extracting the contour of irregular skinned point clouds based on microplane intersection projection, comprising the following steps:

[0044] Step 1: Collect point cloud data of the part using a laser scanner. The point cloud data should include the end face and upper surface data of the part's boundary. Preprocess the point cloud data of the part obtained by scanning.

[0045] Voxel mesh downsampling is performed by dividing the three-dimensional space into several voxel meshes of equal volume, calculating the centroid of each mesh as a unique representative point, and removing outliers to suppress noise interference, thus completing uniform downsampling of the point cloud and reducing the computational load of subsequent processes.

[0046] Step 2: Perform Principal Component Analysis (PCA) on the preprocessed point cloud data to obtain a reference plane, and calculate the unit normal vector of this plane. ,by Unit vector parallel to the Z-axis The cross product result is used as the rotation axis and will Normalization.

[0047] calculate and The angle between them is used as the rotation angle. ;

[0048] Using Rodrigues's rotation formula, based on the axis of rotation and the rotation angle to generate the rotation matrix :

[0049]

[0050] in, It is a 3×3 identity matrix. For vector The associated skew-symmetric matrix is ​​also called the cross product matrix.

[0051] Iterate through all points and calculate the local height gradient difference in the k-neighborhood of each point;

[0052] For planar parts, the standard deviation of the Z coordinates of each point in the k-neighborhood is used as the local height gradient difference of each point;

[0053] For curved surface parts, considering the inherent Z-value variation on the upper surface, which might incorrectly identify the upper surface as an end face, the origin of the coordinate system is first translated and transformed to the estimated curvature center of the surface. The point cloud representation is then converted to spherical coordinates, and the radial distances of each point and its k-neighbors to the origin are calculated. The standard deviation is used as the local height gradient difference;

[0054] Based on the proportion of the end face in the global point cloud, the end face point cloud is filtered out by setting a reasonable height gradient ratio threshold.

[0055] Step 3: After removing points whose distance from the end face exceeds the preset value, first calculate the local unit normal vector point by point, and then calculate the mean vector of all local unit normal vectors. Set the direction of the mean vector to the reference direction pointing to the outer surface.

[0056] Iterate through the normal vectors of each point. If the dot product with the reference direction is less than zero, reverse the normal vector, thereby adjusting the local normal vector direction of all points to point outwards.

[0057] For each point, search its k-neighborhood, calculate the angle between the normal vector of the point and its neighboring points, and take the average of the above angles as the local curvature estimate of the point.

[0058] In the point cloud, a seed point is specified on the upper surface. Growth begins from this seed point, calculating the curvature difference between points within a certain neighborhood of the seed point. If the difference is less than a preset curvature threshold and the neighboring point has not yet been visited, then that neighboring point is used as a new seed point for the next round of growth and evaluation, thereby achieving separation of the upper surface. The separation effect between the end face and the upper surface is shown below. Figure 2 As shown.

[0059] Step 4: Retrieve each end face point and the nearest point on the upper surface. Fit the end face point The equation M of the end face microplane is obtained by taking its k-neighbor points and fitting the corresponding nearest upper surface point. The equation N for the microplane of the upper surface is obtained by taking the k-neighbor points and the equation of the end face point. The projection point on the intersection line L of the two microplanes MN The fitting process is illustrated as follows: Figure 3 As shown.

[0060] Step 5: The projection point set {P} of all end face points is the desired contour point set. The contour boundary of the part can be obtained by fitting a B-spline curve to the points in the set. The final extraction effect of a certain skin boundary contour is shown below. Figure 4 As shown.

[0061] The above description is merely a preferred embodiment of the present invention; however, the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and its improved concepts, should be covered within the scope of protection of the present invention.

Claims

1. A method for extracting the contour of irregular skinned point clouds by microplane intersection projection, characterized in that, Includes the following steps: Step 1: For the irregular skin point cloud data obtained by scanning, the voxel grid method is used for downsampling and outlier noise points are removed to reduce the subsequent computation and noise interference. Step 2: Perform principal component analysis on the preprocessed point cloud data to fit the point cloud principal plane. Calculate the rotation matrix between the point cloud principal plane and the original 3D Cartesian coordinate system A of the point cloud data. Rotate the point cloud principal plane to the reference horizontal plane marked with Z=0 under the original 3D Cartesian coordinate system A. The Z coordinate of the point cloud directly reflects its height relative to the point cloud principal plane. At this point, the gradient difference is calculated with the Z-axis as the height direction, eliminating the need for additional complex spatial angle calculations and greatly simplifying the calculation logic of the local height gradient difference. Iterate through all points, calculate the local height gradient difference within a certain neighborhood for each point, and set a local height gradient threshold. Mark points with gradient differences greater than the threshold as end face point clouds. Step 3: After removing points whose distance from the end face exceeds the preset value, select suitable seed points on the upper surface and use the region growth algorithm, with the curvature difference within the growth range as the growth judgment condition, to separate the upper surface part close to the end face. Step 4: Retrieve each end face point and the nearest point on the upper surface. Fit the end face point The equation M of the end face microplane is obtained by taking its k-neighbor points and fitting the corresponding nearest upper surface point. The equation N for the microplane of the upper surface is obtained by taking the k-neighbor points and the equation of the end face point. The projection point on the intersection line L of the two microplanes MN ; Step 5: The set of projection points {P} of all end face points obtained in Step 4 is the desired contour point set. By performing B-spline curve fitting on the points in the contour point set, the contour boundary of the part can be obtained.

2. The method for extracting irregular skin point cloud contours based on microplane intersection projection according to claim 1, characterized in that, In step one, the irregular dot cloud data is acquired by a laser scanner. This irregular dot cloud data includes end face and upper surface data of the skin boundary portion.

3. The method for extracting irregular skin point cloud contours based on microplane intersection projection according to claim 1, characterized in that, In step one, the voxel grid method for downsampling involves dividing the three-dimensional space into several voxel grids of uniform size, calculating the centroid of each voxel grid as a unique representative point, and achieving uniform downsampling of the point cloud data.

4. The method for extracting irregular skin point cloud contours based on microplane intersection projection according to claim 1, characterized in that, In step two, the calculation steps for the rotation matrix are as follows: S1.1 Estimates the unit normal vector of point cloud data using principal component analysis. ; S1.2 To ensure that the principal plane of the point cloud is perpendicular to the Z-axis after rotation, the unit normal vector of the principal plane of the point cloud is used. Unit vector parallel to the Z-axis The cross product result is used as the rotation axis and rotate the axis Normalization; calculation and The angle between them is used as the rotation angle. ; S1.3 adopts Rodrigues's rotation formula, based on the axis of rotation. and rotation angle Generate rotation matrix ; ； in, It is a 3×3 identity matrix. To be with the axis of rotation The associated oblique symmetric matrix.

5. The method for extracting irregular skinned point cloud contours based on microplane intersection projection according to claim 1, characterized in that: In step two, the calculation steps for the local height gradient difference are as follows: S2.1 For planar parts, the standard deviation of the Z-coordinates of each point within its k-neighborhood is taken as the local height gradient difference of each point: ； in, Let Z be the Z coordinate of the i-th point in the k-neighborhood of this point; The mean Z-coordinate of all points in the k-neighborhood of this point; S2.2 For curved surface parts, considering the variation in Z-value on the upper surface, which might incorrectly identify the upper surface as an end face, the original three-dimensional Cartesian coordinate system A of the point cloud data is first translated and transformed to the estimated curvature center of the surface. The point cloud representation is then converted to spherical coordinates, and the radial distances of each point and its k-neighborhood points to the origin are calculated. The standard deviation is used as the local height gradient difference: ； in, , is the radial distance of the i-th point in the k-neighborhood of this point in the transformed coordinate system; This is the mean radial distance between the point and all points in its k-neighborhood; S2.3 Estimate the proportion of the end face point cloud in all processed point clouds, set a height gradient ratio threshold, and filter out the end face point cloud.

6. The method for extracting irregular skinned point cloud contours according to claim 1, characterized in that: In step three, the upper surface portion near the end face is separated using a region growing algorithm, with the curvature difference within the growing range as the growth criterion. The steps are as follows: S3.1 First, calculate the local unit normal vector point by point, and calculate the mean vector of all local unit normal vectors. Set the direction of the mean vector to the reference direction pointing to the outer surface. Iterate through the local normal vectors of each point. If the dot product with the reference direction is less than zero, reverse the direction of the normal vector and adjust the direction of the local normal vectors of all points to point to the outer surface. S3.2 For each point, search its k-neighborhood, calculate the angle between the normal vector of the point and its neighboring points, and take the average of the above angles as the local curvature estimate of the point. ； in, The desired local curvature; Let be the angle between the normal vector of the i-th point in the k-neighborhood and the mean vector; S3.3 Specify a seed point on the upper surface in the point cloud, start growth from the seed point, calculate the difference between points in a certain neighborhood of the seed point, if the curvature difference is less than the preset curvature threshold and the neighboring point has not been visited, then the neighboring point is used as a new seed point, and the next round of growth is carried out and judged. Finally, all seed points are the points on the upper surface.

Images