A method for automatically segmenting multi-feature point cloud data of a complex curved surface standard part

Abstract

The application discloses a kind of complex curved surface standard piece's multi-feature point cloud data automatic segmentation method, belong to geometric parameter digitization measurement evaluation field.The application implementation method is: according to the information obtained by analytic MBD model, the corresponding relationship between each surface name coding, surface basic attribute, GD&T is established, and the surface feature library of each surface is established;The MBD model of complex curved surface standard piece is converted into the theoretical point cloud of complex curved surface standard piece, and the point of surface is subdivided according to dense, obtains the theoretical point cloud M nom , complete the file format conversion of MBD model;Based on the ICP improved algorithm of fusion geometry feature, complex curved surface standard piece measured point cloud is registered with theoretical point cloud;According to surface feature library, the segmentation of measured point and nearby point of surface point cloud is realized, and further point cloud segmentation result is carried out point cloud segmentation, and the measured point cloud of complex curved surface standard piece containing multiple geometric parameters is realized automatic high-precision segmentation.

Description

TECHNICAL FIELD

[0001] The present application relates to a kind of automatic segmentation method of point cloud data, belong to geometric parameter digitization measurement evaluation field. BACKGROUND

[0002] The present stage digitization measurement evaluation technology is still in the semi-automatic evaluation stage, can only artificial interactive step by step complete the whole process of parameter evaluation, especially when facing huge measured point cloud data, operation is more cumbersome, not suitable for single-piece production mode.Especially when measurement data appears special situation, the evaluation mode based on macro record batch processing reproduction often affects the accuracy of batch product data evaluation.Realize the digital transformation of geometric parameter measurement evaluation, not only need the digital transformation of measuring equipment, also need to be based on design digital model information to the measured data automatic processing evaluation, to meet the development needs of intelligent manufacturing.

[0003] At present, in actual engineering application, commercial geometric parameter evaluation software almost needs inspection personnel to interactively process measurement data, cannot realize automatic processing evaluation according to the MBD model of part, when facing complex point cloud data, processing work is heavy, and efficiency is not high.Therefore, automatic segmentation processing of measured point cloud data of part is an urgent problem to be solved in the digital transformation of geometric parameter measurement evaluation. SUMMARY

[0004] In view of the above problems of prior art, the main purpose of the present application is to provide a kind of automatic segmentation method of multi-feature point cloud data of complex curved surface standard parts, which can realize automatic segmentation of measured point cloud of complex curved surface standard parts containing multiple geometric parameters, can avoid the problem of slow efficiency of manual segmentation of measured point cloud data by inspection personnel, reduce the work intensity of inspection and measurement personnel, and can support automatic evaluation of geometric parameters and digitization of geometric parameter measurement evaluation.

[0005] To achieve the above purpose, the technical scheme of the embodiment of the present application is as follows:

[0006] The automatic segmentation method of multi-feature point cloud data of complex curved surface standard parts disclosed by the present application comprises the following steps:

[0007] Step one: apply the three-dimensional model based on model definition (Model Based Definition, MBD) of complex curved surface standard parts to the measured point cloud segmentation process, complete the automatic segmentation of each surface point cloud in the measured point cloud data of complex curved surface standard parts.

[0008] It is necessary to use the Data Exchange package in the OpenCascade geometry kernel library to read the solid model information in the MBD model of the complex surface standard part, and then use the OCAF organization framework in the geometry kernel library to use the XCAF package to read other relevant information in the MBD model of the complex surface standard part, including the coordinate system of the datum, annotation information and attribute information, so as to complete the reading of all information in the MBD model.

[0009] Step 2: Based on each surface in the solid model obtained in Step 1, create a corresponding directed bounding box for each surface.

[0010] Step 3: Based on all the information obtained in Step 1, establish the correspondence between the name code of each surface, the basic properties of the surface, and the geometric dimensions and tolerances (GD&T), thereby establishing the surface feature library for each surface.

[0011] Step 4: Convert the MBD model of the complex curved surface standard part into a theoretical point cloud of the complex curved surface standard part, and further subdivide and densify the points on the surface to obtain the theoretical point cloud M corresponding to the theoretical model. nom Complete the file format conversion of the MBD model.

[0012] Step 5: Perform point cloud registration based on the improved Iterative Closest Point (ICP) algorithm for fused geometric features, and register the measured point cloud of the complex curved surface standard part with the theoretical point cloud.

[0013] Step 5.1: Establish the measured point cloud M of the complex curved surface standard part act K-Dimensional Tree (KD-Tree);

[0014] Step 5.2: Convert the measured point cloud M of the complex curved surface standard part act Curvature-based voxel sampling is performed. The curvature is calculated as shown in formula (1), and the voxel sampling is as shown in formula (2), thus obtaining the measured point cloud M of the complex curved surface standard part. act_d ;

[0015]

[0016] In the formula, the point cloud q i The normal is n i The normals corresponding to the K-neighbor points are {n i1 ,n i2 ,…,n ik}, α i For each point q iThe mean of the angles between the normals to its K-neighbor points, expressed as the mean angle α. i The curvature of the point cloud is approximately approximated.

[0017]

[0018] In the formula, (x centroid ,y centroid ,z centroid (x) is the point obtained by voxel sampling. i ,y i ,z i ) represents the measured point cloud coordinates, and n represents the corresponding neighborhood number.

[0019] Step 5.3: Measured point cloud M of complex curved surface standard parts act_d With theoretical point cloud M nom The point cloud is registered using an improved Iterative Closest Point (ICP) algorithm based on fused geometric features. The registration error during the algorithm iteration process is improved as shown in formula (3), resulting in the measured point cloud M. act_d Required rotation and translation matrix M a1 .

[0020]

[0021] In the formula, e m is the iteration error corresponding to the iteration number m, C is the curvature difference of the corresponding point, θ is the normal vector difference of the corresponding point, and n is the number of neighborhood points of the corresponding point.

[0022] Step 5.4: Based on the rotation and translation matrix M a1 Measured point cloud M of a standard part with a complex curved surface act By performing spatial pose coordinate transformation, the measured point cloud M of the complex curved surface standard part is obtained. act_1 Complete point cloud registration.

[0023] Step Six: Based on the directed bounding box information corresponding to each surface established in Step Two, perform point cloud registration on the measured point cloud M of the complex surface standard part. act_1 The first point cloud segmentation is performed to automatically obtain the point cloud data of the surface and nearby points.

[0024] Step 7: Based on the surface feature library established in Step 3, perform point cloud segmentation again on the first point cloud segmentation result in Step 6, and automatically obtain the measured point cloud corresponding to each surface. That is, the measured point cloud of the complex surface standard part containing most geometric parameters is automatically segmented.

[0025] Step 7.1: Read the surface feature library established in Step 3, using a point on the surface from the basic surface properties as the seed. The constraints are the distance from neighboring points to the surface and the normal vector of the point. For basic surfaces, the distance is the distance from the point to the surface; for freeform surfaces, the distance is the distance from the point to the triangular facet in the freeform surface. The normal vector of the point is obtained by fitting a plane through local neighboring points.

[0026] Step 7.2: Based on the region-growing constraints in Step 7.1, the measured point cloud M of the complex surface standard part obtained in Step 5 is processed. act_1 The first point cloud segmentation result is used to perform automatic point cloud segmentation again to obtain the measured point cloud corresponding to each surface;

[0027] Step 7.3: To address the issue of over-segmentation or under-segmentation that may occur in Step 1, it is necessary to determine the boundary points of the surface. Therefore, the Crust algorithm is used to determine the boundary points of the surface. During the point cloud segmentation process, if all the boundary points of the surface are found, the point cloud segmentation process of the surface will automatically end, that is, the measured point cloud of a complex surface standard part containing most geometric parameters can be automatically segmented.

[0028] Beneficial effects:

[0029] 1. This invention discloses an automatic segmentation method for multi-feature point cloud data of complex curved surface standard parts. The MBD model of the complex curved surface standard parts is introduced into the process of processing the measured point cloud data of the complex curved surface standard parts. The features in the MBD model are used to automatically process the point cloud segmentation calculation process, changing the original parameter evaluation method of interactive semi-automatic segmentation of measured point cloud data. This method realizes automatic segmentation of measured point cloud data, which has the advantages of high segmentation efficiency and high accuracy. It can also improve the efficiency of measurement and evaluation by inspection personnel and reduce labor intensity.

[0030] 2. This invention discloses an automatic segmentation method for multi-feature point cloud data of complex curved surface standard parts. Based on all the information obtained from parsing, it establishes the correspondence between the name code of each surface, the basic attributes of the surface, and GD&T, thereby establishing a surface feature library for each surface. It converts the MBD model of the complex curved surface standard part into a theoretical point cloud of the complex curved surface standard part, and further subdivides and densifies the points of the surface to obtain the theoretical point cloud M corresponding to the theoretical model. nom The process involves: converting the MBD model to its file format; performing point cloud registration based on the improved ICP algorithm that integrates geometric features, registering the measured point cloud of the complex curved surface standard part with the theoretical point cloud; and, based on the surface feature library, segmenting the measured points and nearby points of the surface point cloud, further segmenting the point cloud segmentation results to automatically obtain the measured point cloud corresponding to each surface, thus achieving automatic segmentation of the measured point cloud of the complex curved surface standard part containing multiple geometric parameters.

[0031] 3. The present invention discloses an automatic segmentation method for multi-feature point cloud data of complex curved surface standard parts. It establishes a KD-Tree of measured point cloud, and then uses an improved ICP algorithm that integrates geometric features to perform point cloud registration. Geometric features refer to the curvature and normal vector of the point cloud. This method can solve the problem that ICP registration requires appropriate initial values, avoid two point cloud registration processes, and improve the efficiency of automatic segmentation of multi-feature point cloud data.

[0032] 4. The present invention discloses an automatic segmentation method for multi-feature point cloud data of complex curved surface standard parts, which can automatically segment the measured point cloud of complex curved surface standard parts containing a large number of geometric parameters, and can support the automated evaluation of geometric parameters and the digitization of geometric parameter measurement evaluation, providing a foundation for the digital measurement and evaluation of parts. Attached Figure Description

[0033] Figure 1 For solid models in MBD models of complex curved surface standard parts;

[0034] Figure 2 This is to establish a feature library for cylindrical surfaces in standard parts with complex curved surfaces;

[0035] Figure 3 Convert complex surface standard part MBD model files into surface point clouds and subdivide them into dense layers;

[0036] Figure 4 The registration results of theoretical point clouds and measured point clouds for standard parts with complex curved surfaces;

[0037] Figure 5 This is a point cloud containing a cylindrical component within a standard part with a complex curved surface.

[0038] Figure 6 To Figure 5 The result obtained by performing the first step of automatic point cloud segmentation on the cylindrical surface in the image;

[0039] Figure 7 To Figure 5 The result obtained by automatically segmenting the point cloud in the second step using the cylindrical surface in the middle;

[0040] Figure 8 This is a flowchart of an automatic segmentation method for multi-feature point cloud data of a complex curved surface standard part disclosed in this invention. Detailed Implementation

[0041] The present invention will now be described with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0042] like Figure 8As shown in the figure, this embodiment discloses an automatic segmentation method for multi-feature point cloud data of a complex curved surface standard part. In order to verify the feasibility of the method, the measured point cloud of a cylindrical surface in the complex curved surface standard part is selected as an example for detailed description.

[0043] 1. To automatically segment the measured point cloud data of the target cylindrical surface from the measured point cloud data of a complex curved surface standard part, it is first necessary to use the Data Exchange package in the OpenCascade geometry kernel library to read the solid model information in the MBD model of the complex curved surface standard part, such as... Figure 1 The image shows the solid model information of the complex curved surface standard part. Then, with the help of the OCAF organization framework in the geometry kernel library, the XCAF package is used to read the annotation dimension information of the cylindrical surface in the MBD model of the complex curved surface standard part, including the dimension information defined by the XCAFDoc_Dimension class, the geometric tolerance information defined by the XCAFDoc_GeomTolerance class, and the geometric tolerance datum defined by the XCAFDoc_Datum class. The MBD model is stored in a neutral file STEP that conforms to the AP242 standard.

[0044] 2. Based on the cylindrical surface dimension information read in step 1, such as... Figure 2 As shown, the correspondence between the name encoding of the cylindrical surface, the basic properties of the surface, the directed bounding box of the cylindrical surface, and GD&T is established and stored in the corresponding dictionary. The name encoding of the cylindrical surface can quickly find the relevant feature information of the cylindrical surface through the dictionary structure, which is convenient for driving the automatic segmentation of the measured point cloud data of the cylindrical surface.

[0045] 3. Using the `write_stl_file` function in the Opencascade geometry kernel library, the MBD model of the complex surface standard part is converted into the theoretical point cloud of the complex surface standard part. Then, using formulas (1) and (2), the points on all surfaces are subdivided and densified, such as... Figure 3 As shown, it contains the theoretical point cloud of the target cylindrical surface after subdivision and densification, providing a basis for point cloud registration.

[0046] For the new control term vertex

[0047]

[0048] In the formula, V m+1 It is a new control term vertex after m subdivisions, consisting of n vertices sharing an edge. Composition, i = 1, 2, ..., n],

[0049] For the new shared-edge vertices

[0050]

[0051] 4. The theoretical point cloud and measured point cloud of a complex curved surface standard part are registered together. The specific steps are as follows:

[0052] 1) Using a 3D structured light scanner, the measured point cloud data M of the complex curved surface standard part was obtained. act And establish the measured point cloud M of the complex curved surface standard part. act KD-Tree;

[0053] 2) The measured point cloud M of the complex curved surface standard part act Voxel sampling based on curvature is performed. The curvature is calculated as shown in formula (3), and the voxel sampling is as shown in formula (4). The measured point cloud M of the sampled complex curved surface standard part is obtained. act_d ;

[0054]

[0055] In the formula, the point cloud q i The normal is n i The normals corresponding to the K-neighbor points are {n i1 ,n i2 ,…,n ik}, α i For each point q i The mean of the angles between the normals to its K-neighbor points, expressed as the mean angle α. i The curvature of the point cloud is approximately approximated.

[0056]

[0057] In the formula, (x centroid ,y centroid ,z centroid (x) is the point obtained by voxel sampling. i ,y i ,z i ) represents the measured point cloud coordinates, and n represents the corresponding neighborhood number.

[0058] 3) The measured point cloud M of the complex curved surface standard part act_d With theoretical point cloud M nom Point cloud registration was performed using an improved ICP algorithm based on fused geometric features. The curvature and normal vector information of the point cloud were fused into the registration error formula of the ICP algorithm, as shown in formula (5). This improved the ICP registration algorithm, and the measured point cloud M was obtained after registration. act_d Required rotation and translation matrix M a .

[0059]

[0060] In the formula, e m is the iteration error corresponding to the iteration number m, C is the curvature difference of the corresponding point, θ is the normal vector difference of the corresponding point, and n is the number of neighborhood points of the corresponding point.

[0061] 4) Based on the rotation and translation matrix M a Measured point cloud M of a standard part with a complex curved surface act Perform spatial transformations, such as Figure 4 As shown, the measured point cloud M of the complex curved surface standard part after coordinate transformation is obtained. act_1 .

[0062] 5. Based on the registration and alignment of the theoretical and measured point clouds of the complex curved surface standard part in step 4, and according to the directed bounding box information of the cylindrical surface in the feature library established in step 2, such as... Figure 5 The box position information shown represents the measured point cloud M of a complex curved surface standard part. act_1 The first step, automatic segmentation, automatically obtains the measured point cloud of the cylindrical surface and nearby points, such as... Figure 6 As shown.

[0063] 6. Read the feature library of the cylindrical surface established in step 2 again, and take a point on the corresponding cylindrical surface as the seed. The constraints are the distance from the neighboring points to the surface and the normal vector of the point, as shown in formulas (6) and (7). Perform point cloud search for the cylindrical surface. At the same time, for the boundary points of the cylindrical surface, the Crust algorithm is used. When the seed has searched all the boundary points, the point cloud segmentation process ends. Figure 7 As shown, the measured point cloud data of the target cylindrical surface is automatically obtained. The entire point cloud segmentation process is automatically completed based on the relevant information of the cylindrical surface in the MBD model of the complex curved surface standard part.

[0064]

[0065] In the formula, (a,b,c) is the direction vector of the cylinder axis, (x0,y0,z0) is a point on the cylinder axis, and (x i ,y i ,z i () represents the measured point cloud.

[0066]

[0067] In the formula, (A,B,C) is the direction vector obtained by fitting the plane with the neighboring points of the measured point, and is used as the direction vector of the measured point.

[0068] 7. After automatically segmenting the measured point cloud corresponding to each surface in a complex curved standard part, the geometric parameters of the corresponding surface can be evaluated with the help of the datum, GD&T and other information in the MBD model. This will establish a digital inspection model that includes manufacturing deviations in form, position and size, and complete the corresponding qualification judgment, which can provide a foundation for realizing digital inspection of part geometric parameters and digital pre-assembly of parts.

[0069] The above detailed description further illustrates the purpose, technical solution, and beneficial effects of the invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. An automatic segmentation method for multi-feature point cloud data of a complex curved surface standard part, characterized in that: Includes the following steps, Step 1: Apply the Model Definition-Based 3D model of the complex curved surface standard part to the measured point cloud segmentation process to complete the automatic segmentation of each surface point cloud in the measured point cloud data of the complex curved surface standard part. Step 2: Based on each surface in the solid model obtained in Step 1, create a corresponding directed bounding box for each surface; Step 3: Based on all the information obtained in Step 1, establish the correspondence between the name code of each surface, the basic attributes of the surface, the geometric dimensions and the tolerance GD&T, thereby establishing the surface feature library for each surface; Step 4: Convert the MBD model of the complex curved surface standard part into a theoretical point cloud of the complex curved surface standard part, and subdivide and densify the points on the surface to obtain the theoretical point cloud M corresponding to the theoretical model. nom Complete the file format conversion of the MBD model; Step 5: Based on the improved Iterative Closest Point (ICP) algorithm that integrates geometric features, perform point cloud registration to register the measured point cloud of the complex curved surface standard part with the theoretical point cloud; To improve the registration error during the algorithm iteration process, as shown in formula (1), the measured point cloud M of the complex surface standard part after voxel sampling is obtained. act_d Required rotation and translation matrix M a1 ; In the formula, e m C(M) is the iteration error corresponding to the iteration number m. nom -M act_d ) is the curvature difference at corresponding points, θ(M) nom -M act_d ) is the difference in normal vectors of the corresponding points, and n is the number of neighborhood points of the corresponding point; Step Six: Based on the directed bounding box information corresponding to each surface established in Step Two, perform point cloud registration on the measured point cloud M of the complex surface standard part. act_1 Perform the first point cloud segmentation to automatically obtain the point cloud data of the surface and nearby points; Step 7: Based on the surface feature library established in Step 3, perform point cloud segmentation again on the first point cloud segmentation result in Step 6, and automatically obtain the measured point cloud corresponding to each surface, that is, achieve automatic segmentation of the measured point cloud of complex surface standard parts containing most geometric parameters.

2. The method for automatic segmentation of multi-feature point cloud data of a complex curved surface standard part as described in claim 1, characterized in that: The implementation method for step one is as follows: The solid model information in the MBD model of a complex surface standard part is read using the Data Exchange package in the OpenCascade geometry kernel library. Then, with the help of the OCAF organization framework in the geometry kernel library, the XCAF package is used to read other relevant information in the MBD model of the complex surface standard part, including the coordinate system of the datum, annotation information and attribute information, thus completing the reading of all information in the MBD model.

3. The method for automatic segmentation of multi-feature point cloud data of a complex curved surface standard part as described in claim 2, characterized in that: Step five is implemented as follows: Step 5.1: Establish the measured point cloud M of the complex curved surface standard part act K-dimensional tree; Step 5.2: Convert the measured point cloud M of the complex curved surface standard part act Curvature-based voxel sampling is performed to obtain the measured point cloud M of the complex curved surface standard part after voxel sampling. act_d ; Step 5.3: Measured point cloud M of the complex curved surface standard part after voxel sampling act_d With theoretical point cloud M nom The improved iterative nearest point (ICP) algorithm based on fused geometric features is used for point cloud registration. Step 5.4: Based on the rotation and translation matrix M a1 Measured point cloud M of a standard part with a complex curved surface act By performing spatial pose coordinate transformation, the measured point cloud M of the complex curved surface standard part after point cloud registration is obtained. act_1 Complete point cloud registration.

4. The automatic segmentation method for multi-feature point cloud data of a complex curved surface standard part as described in claim 3, characterized in that: Step seven is implemented as follows: Step 7.1: Read the surface feature library established in Step 3, using a point on the surface in the basic surface properties as a seed. The constraints are the distance from the neighboring points to the surface and the normal vector of the point. For basic surfaces, the distance is the distance from the point to the surface. For freeform surfaces, the distance is the distance from the point to the triangular facet in the freeform surface. The normal vector of the point is obtained by fitting a plane through local neighboring points. Step 7.2: Based on the constraints in Step 7.1, calculate the measured point cloud M of the complex curved surface standard part obtained in Step 5. act_1 The first point cloud segmentation result is used to perform automatic point cloud segmentation again to obtain the measured point cloud corresponding to each surface; Step 7.3: Determine the boundary points of the surface. Specifically, the Crust algorithm is used to determine the boundary points of the surface. During the point cloud segmentation process, if all the boundary points of the surface are found, the point cloud segmentation process of the surface will automatically end. That is, the measured point cloud of the complex surface standard part containing most geometric parameters can be automatically segmented.

Images