Edge recognition method for aircraft part point cloud based on multi-scale principal manifold and RANSAC algorithm

Abstract

The application discloses an airplane part point cloud edge recognition method based on a multi-scale principal manifold and a RANSAC algorithm, and comprises the following steps: S1, preprocessing the point cloud, including filtering and neighborhood division; S2, preliminarily calculating point cloud edge feature points; S3, taking the edge feature points obtained in the last step as the center, dividing a multi-scale ellipsoid neighborhood and performing point cloud interpolation; S4, constructing an ellipsoid point cloud principal manifold surface model and calculating multi-scale curvature values based on the principal manifold surface model; and S5, updating the point cloud edge feature points by using the RANSAC algorithm. The application has the following advantages when recognizing point cloud edge features: the multi-scale ellipsoid improves the robustness and noise resistance performance of the recognized point cloud edge; and the curvature calculation is based on the principal manifold surface of the ellipsoid point cloud, so that the calculation result is more accurate, and the fitting deviation caused by the discrete curvature calculation is reduced.

Description

Technical Field

[0001] This invention relates to a point cloud processing technology, and more particularly to a point cloud edge recognition technology, specifically a point cloud edge recognition method for aircraft parts based on multi-scale mainstream shape and RANSAC algorithm. Background Technology

[0002] To meet the higher precision machining and quality control requirements of aircraft parts, high-precision 3D scanning equipment has been widely used for aircraft part scanning. Accurate processing of the obtained point cloud data is a crucial step in aircraft digital measurement. Edge features are an important component of aircraft parts, playing a vital role in their assembly, modeling, and accurate representation.

[0003] This invention proposes a point cloud edge recognition method for aircraft parts based on multi-scale mainstream shape and EM algorithm. It fully considers the neighborhood information of the point cloud at multiple scales and uses differential geometry to accurately calculate the curvature of the point cloud, thereby overcoming the limitations of traditional methods in edge recognition. Summary of the Invention

[0004] The purpose of this invention is to provide a point cloud edge recognition method for aircraft parts based on multi-scale mainstream shape and RANSAC algorithm, so as to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a method for edge recognition of aircraft part point clouds based on multi-scale mainstream shape and RANSAC algorithm, wherein the method for edge recognition of aircraft part point clouds based on multi-scale mainstream shape and RANSAC algorithm is performed according to the following steps:

[0006] S1. Preprocess the point cloud;

[0007] S2. Preliminary calculation of point cloud edge feature points;

[0008] S3. Using the edge feature points obtained in the previous step as the center, divide the multi-scale ellipsoidal neighborhood and perform point cloud interpolation.

[0009] S4. Construct a mainstream curved surface model of the ellipsoidal point cloud and calculate multi-scale curvature values ​​based on the mainstream curved surface model;

[0010] S5. Update the point cloud edge feature points using the RANSAC algorithm.

[0011] As a preferred technical solution of the present invention, step S1, point cloud preprocessing, includes filtering and neighborhood division, using Gaussian filtering to denoise the point cloud, and constructing a Delaunay triangular mesh to divide the point cloud neighborhood.

[0012] As a preferred embodiment of the present invention, step S2, which involves the preliminary calculation of point cloud edge feature points, includes the following steps:

[0013] S2.1 Fit plane π. The points involved in the fitting include the first-order Delaunay neighborhood points Pij (j = 1, 2, ..., n) of the target point Pi, where n is the number of first-order Delaunay neighborhood points of point Pi. Project point Pij onto plane π to obtain the projection point Pip. Construct an auxiliary triangle, which consists of point Pi, neighborhood point Pij, and projection point Pip.

[0014] S2.2 Calculate the feature value E of the point cloud edge feature points according to the following formula:

[0015]

[0016] In the formula, h is the distance from point Pi to the projection point Pip, di is the distance from point Pij to the projection point Pip, and ei is an auxiliary calculation value, calculated as follows:

[0017]

[0018] The point cloud edge feature points are initially selected based on the magnitude of the feature value E.

[0019] As a preferred embodiment of the present invention, step S3, which involves dividing the multi-scale ellipsoidal neighborhood and performing point cloud interpolation, includes the following steps:

[0020] S3.1 takes the edge feature points selected in S2.2 as the center, performs principal component analysis on their first-order and second-order Delaunay neighborhood points, generates three principal axes λ1, λ2, and λ3, and constructs ellipsoids based on these axes, which are defined as "first-order ellipsoid" and "second-order ellipsoid" respectively.

[0021] S3.2 Perform point cloud upsampling between edge feature points and the first-order ellipsoid, with the sampling distance set to half the average distance of the first-order Delaunay neighborhood points; perform point cloud upsampling between the first-order and second-order ellipsoids, with the sampling distance set to the average of the average distance of the first-order Delaunay neighborhood points and the geodesic distance of the outer points of the second-order Delaunay neighborhood. After sampling, construct ellipsoids using step S3.1, defining them as "0.5-order ellipsoid" and "1.5-order ellipsoid".

[0022] As a preferred embodiment of the present invention, step S4, which involves constructing a mainstream curved surface model of the ellipsoidal point cloud and calculating multi-scale curvature values ​​based on the mainstream curved surface model, includes the following steps:

[0023] S4.1 Define the latitude and longitude directions of the four ellipsoids. The longitude direction is parallel to the λ3 axis, and the latitude direction is parallel to λ1 and λ2. Divide the grid along the latitude and longitude directions to form the initial main grid, including the grid point set V = {vm} and the edge set E = {en}.

[0024] S4.2 Defines the main shape control parameter C:

[0025]

[0026] In the formula, μ j It is a constant. The grid point set V is continuously updated with the goal of minimizing the control parameter C until it is less than a set threshold.

[0027] S4.3 calculates the principal curvature ki (i = 1, 2, 3, 4) of each ellipsoid point output from S4.2. Calculate the average curvature at each point:

[0028]

[0029] Whether a point is an edge point is determined by whether K is greater than a set threshold.

[0030] As a preferred embodiment of the present invention, step S5, which uses the RANSAC algorithm to update the edge feature points of the point cloud, includes the following steps:

[0031] S5.1 moves the initial point of the normal vector of all edge feature points output by S4.3 to the origin of the coordinate system, and uses the endpoint of the normal vector as the target point.

[0032] S5.2 Randomly select several groups of target points as interior points to fit a sphere. Use the sphere equation to test other target points. If the error between the target point and the sphere equation is less than a set value, then include it as an interior point. If the final number of interior points is greater than a set threshold, then output all interior points as the final edge feature points.

[0033] Compared with existing technologies, the beneficial effects of this invention are: this invention is a point cloud edge recognition method for aircraft parts based on multi-scale mainstream shape and RANSAC algorithm.

[0034] 1) Multi-scale ellipsoids improve the robustness and noise resistance of point cloud edge recognition;

[0035] 2) Curvature calculation is based on the mainstream curved surface of the ellipsoidal point cloud, and the calculation results are more accurate, reducing the fitting deviation caused by discrete curvature calculation. Attached Figure Description

[0036] Figure 1 This is a schematic diagram of the process of the present invention. Detailed Implementation

[0037] The technical solutions in the embodiments of the present invention have been clearly and completely described. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0038] Please see Figure 1 This invention provides a method for edge recognition of aircraft part point clouds based on multi-scale mainstream shape and RANSAC algorithm. The method for edge recognition of aircraft part point clouds based on multi-scale mainstream shape and RANSAC algorithm is performed according to the following steps:

[0039] S1. Preprocess the point cloud;

[0040] S2. Preliminary calculation of point cloud edge feature points;

[0041] S3. Using the edge feature points obtained in the previous step as the center, divide the multi-scale ellipsoidal neighborhood and perform point cloud interpolation.

[0042] S4. Construct a mainstream curved surface model of the ellipsoidal point cloud and calculate multi-scale curvature values ​​based on the mainstream curved surface model;

[0043] S5. Update the point cloud edge feature points using the RANSAC algorithm.

[0044] Furthermore, step S1, point cloud preprocessing, includes filtering and neighborhood partitioning. Gaussian filtering is used to denoise the point cloud, and Delaunay triangular meshes are constructed to partition the point cloud neighborhood.

[0045] Furthermore, step S2, which involves initially calculating the edge feature points of the point cloud, includes the following steps:

[0046] S2.1 Fit plane π. The points involved in the fitting include the first-order Delaunay neighborhood points Pij (j = 1, 2, ..., n) of the target point Pi, where n is the number of first-order Delaunay neighborhood points of point Pi. Project point Pij onto plane π to obtain the projection point Pip. Construct an auxiliary triangle, which consists of point Pi, neighborhood point Pij, and projection point Pip.

[0047] S2.2 Calculate the feature value E of the point cloud edge feature points according to the following formula:

[0048]

[0049] In the formula, h is the distance from point Pi to the projection point Pip, di is the distance from point Pij to the projection point Pip, and ei is an auxiliary calculation value, calculated as follows:

[0050]

[0051] The point cloud edge feature points are initially selected based on the magnitude of the feature value E.

[0052] Specifically, step S3, which involves dividing the multi-scale ellipsoidal neighborhood and performing point cloud interpolation, includes the following steps:

[0053] S3.1 takes the edge feature points selected in S2.2 as the center, performs principal component analysis on their first-order and second-order Delaunay neighborhood points, generates three principal axes λ1, λ2, and λ3, and constructs ellipsoids based on these axes, which are defined as "first-order ellipsoid" and "second-order ellipsoid" respectively.

[0054] S3.2 Perform point cloud upsampling between edge feature points and the first-order ellipsoid, with the sampling distance set to half the average distance of the first-order Delaunay neighborhood points; perform point cloud upsampling between the first-order and second-order ellipsoids, with the sampling distance set to the average of the average distance of the first-order Delaunay neighborhood points and the geodesic distance of the outer points of the second-order Delaunay neighborhood. After sampling, construct ellipsoids using step S3.1, defining them as "0.5-order ellipsoid" and "1.5-order ellipsoid".

[0055] Preferably, step S4, which involves constructing a mainstream curved surface model of the ellipsoidal point cloud and calculating multi-scale curvature values ​​based on the mainstream curved surface model, includes the following steps:

[0056] S4.1 Define the latitude and longitude directions of the four ellipsoids. The longitude direction is parallel to the λ3 axis, and the latitude direction is parallel to λ1 and λ2. Divide the grid along the latitude and longitude directions to form the initial main grid, including the grid point set V = {vm} and the edge set E = {en}.

[0057] S4.2 Defines the main shape control parameter C:

[0058]

[0059] In the formula, μ j It is a constant. The grid point set V is continuously updated with the goal of minimizing the control parameter C until it is less than a set threshold.

[0060] S4.3 calculates the principal curvature ki (i = 1, 2, 3, 4) of each ellipsoid point output from S4.2. Calculate the average curvature at each point:

[0061]

[0062] Whether a point is an edge point is determined by whether K is greater than a set threshold.

[0063] Step S5, which uses the RANSAC algorithm to update the point cloud edge feature points, includes the following steps:

[0064] S5.1 moves the initial point of the normal vector of all edge feature points output by S4.3 to the origin of the coordinate system, and uses the endpoint of the normal vector as the target point.

[0065] S5.2 Randomly select several groups of target points as interior points to fit a sphere. Use the sphere equation to test other target points. If the error between the target point and the sphere equation is less than a set value, then include it as an interior point. If the final number of interior points is greater than a set threshold, then output all interior points as the final edge feature points.

[0066] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for edge recognition of aircraft parts point clouds based on multi-scale mainstream shape and RANSAC algorithm, characterized in that, The method for edge recognition of aircraft parts point clouds based on multi-scale mainstream shape and RANSAC algorithm is performed according to the following steps: S1. Preprocess the point cloud; S2. Preliminary calculation of point cloud edge feature points; S3. Using the edge feature points obtained in the previous step as the center, divide the multi-scale ellipsoidal neighborhood and perform point cloud interpolation. S4. Construct a mainstream curved surface model of the ellipsoidal point cloud and calculate multi-scale curvature values ​​based on the mainstream curved surface model; S5. Update the point cloud edge feature points using the RANSAC algorithm; The point cloud preprocessing step S1 includes filtering and neighborhood partitioning. Gaussian filtering is used to denoise the point cloud, and Delaunay triangular meshes are constructed to partition the point cloud neighborhood. The preliminary calculation of point cloud edge feature points in step S2 includes the following steps: S2.1 Fit plane π. The points involved in the fitting include the first-order Delaunay neighborhood points Pij (j = 1, 2, ..., n) of the target point Pi, where n is the number of first-order Delaunay neighborhood points of point Pi. Project point Pij onto plane π to obtain the projection point Pip. Construct an auxiliary triangle, which consists of point Pi, neighborhood point Pij, and projection point Pip. S2.2 Calculate the feature value E of the point cloud edge feature points according to the following formula: In the formula, h is the distance from point Pi to the projection point Pip, di is the distance from point Pij to the projection point Pip, and ei is an auxiliary calculation value, calculated as follows: The point cloud edge feature points are initially selected based on the magnitude of the feature value E; Step S3, which involves dividing the multi-scale ellipsoidal neighborhood and performing point cloud interpolation, includes the following steps: S3.1 takes the edge feature points selected in S2.2 as the center, performs principal component analysis on their first-order and second-order Delaunay neighborhood points, generates three principal axes λ1, λ2, and λ3, and constructs ellipsoids based on these axes, which are defined as "first-order ellipsoid" and "second-order ellipsoid" respectively. S3.2 Perform point cloud upsampling between edge feature points and the first-order ellipsoid, with the sampling distance set to half the average distance of the first-order Delaunay neighborhood points; perform point cloud upsampling between the first-order and second-order ellipsoids, with the sampling distance set to the average of the average distance of the first-order Delaunay neighborhood points and the geodesic distance of the second-order Delaunay neighborhood perimeter points; after sampling, construct ellipsoids using step S3.1, defining them as "0.5-order ellipsoid" and "1.5-order ellipsoid" respectively; Step S4, which involves constructing a mainstream surface model of the ellipsoidal point cloud and calculating multi-scale curvature values ​​based on the mainstream surface model, includes the following steps: S4.1 Define the latitude and longitude directions of the four ellipsoids. The longitude direction is parallel to the λ3 axis, and the latitude direction is parallel to λ1 and λ2. Divide the grid along the latitude and longitude directions to form the initial main grid, including the grid point set V = {vm} and the edge set E = {en}. S4.2 Defines the main shape control parameter C: In the formula, μ j It is a constant; the grid point set V is continuously updated with the goal of minimizing the control parameter C until it is less than the set threshold; S4.3 Calculate the principal curvature ki (i = 1, 2, 3, 4) of each ellipsoid point output from S4.2; calculate the average curvature of each point: Determine whether a point is an edge point based on whether K is greater than a set threshold. Step S5, which uses the RANSAC algorithm to update the point cloud edge feature points, includes the following steps: S5.1 Moves the initial point of the normal vector of all edge feature points output by S4.3 to the origin of the coordinate system, and takes the endpoint of the normal vector as the target point; S5.2 Randomly select several groups of target points as interior points to fit a sphere. Use the sphere equation to test other target points. If the error between the target point and the sphere equation is less than a set value, then include it as an interior point. If the final number of interior points is greater than a set threshold, then output all interior points as the final edge feature points.

Images