Buckle searching and positioning method based on 3D model point cloud sampling matching

Abstract

The invention relates to the technical field of computer vision and three-dimensional reconstruction, and discloses a buckle searching and positioning method based on 3D model point cloud sampling matching, and the method comprises the steps: obtaining a modeling file data set, carrying out the point cloud sampling of an automobile housing part and a target buckle STEP, and obtaining a modeling file data set; and calculating the size and the moving step length of the point cloud segmentation window according to the geometric dimension information of the target buckle. According to the buckle searching and positioning method based on 3D model point cloud sampling matching, point cloud segmentation based on a sliding window is carried out, and the point cloud of a workpiece model is cut by adopting the idea of a convolution sliding window so as to adapt to the large and small resolution of a buckle; setting of a self-adaptive matching threshold: automatically determining the size and the step length of a sliding window by calculating a minimum bounding box of a buckle, and ensuring the precision of a matching process; the method has an effective duplicate removal mechanism: in a matching result, the uniqueness of each target point cloud window is ensured by sorting matching degrees and removing adjacent windows, so that the accuracy of the matching result is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of computer vision and three-dimensional reconstruction, and in particular to a buckle search and positioning method based on 3D model point cloud sampling and matching. Background Art

[0002] In the modern automotive design and manufacturing process, the use of CAD (computer-aided design) models has become a standard practice. STEP (Standard for the Exchange of Product Model Data) stands for the product model data exchange standard. Since it became an ISO international standard (ISO10303) in 1992, STEP files, as a universal text intermediate exchange format for 3D CAD, have played an important role in geometric data interaction. In particular, STEP files are a universal format for describing BRep (boundary representation) model data, which can accurately represent the geometry and topology information of parts, making it possible to exchange and share them between different CAD systems. With the popularization of digital design, component identification and analysis based on STEP digital models have become increasingly important.

[0003] As an important connector of automotive parts, snap-fit ​​has specific geometric shapes and sizes. Its accurate representation in CAD models is crucial for subsequent assembly and quality control. Traditional snap-fit ​​detection methods usually rely on manual review or simple geometric comparison, which may lead to recognition errors or omissions in complex models.

[0004] With the development of computer vision and point cloud processing technology, automatic buckle recognition technology based on 3D models has gradually emerged. By sampling and matching the point cloud of STEP digital models, the location and number of buckles can be more effectively identified. However, the current technology still faces some challenges, especially in multi-buckle models, where windows with similar matching degrees may cause the target at the same location to be recorded repeatedly, thus affecting the accuracy of the recognition results.

[0005] Therefore, developing an efficient buckle recognition method that can perform accurate geometric matching based on STEP digital models and effectively handle duplicate record problems has become an urgent need for the industry. This method can not only improve the buckle recognition accuracy, but also improve the automation level of the entire design and manufacturing process.

[0006] Therefore, a buckle search and positioning method based on 3D model point cloud sampling and matching is proposed. Summary of the invention

[0007] Technical issues solved In view of the shortcomings of the prior art, the present invention provides a buckle search and positioning method based on 3D model point cloud sampling and matching, which has the ability to obtain the precise position and posture of the buckle in the target part model, thereby providing data support for the downstream task of extracting and displaying partial information of the buckle features in the part model. The buckle detection method using point cloud segmentation and ICP alignment has the advantages of accurate recognition and positioning of the buckle, and solves the problem that windows with similar matching degrees may cause targets at the same position to be recorded repeatedly, thereby affecting the accuracy of the recognition results, and the problem that the existing method may cause recognition errors or omissions in complex models.

[0008] Technical Solution To achieve the above object, the present invention provides the following technical solution: a buckle search and positioning method based on 3D model point cloud sampling and matching, comprising the following steps: Step 1: Obtain the modeling file dataset and perform point cloud sampling on the automobile shell parts and target buckle STEP; Step 2: Calculate the size and moving step of the point cloud segmentation window according to the geometric size information of the target buckle; Step 3: Perform ICP registration on the point cloud in the segmented window of the component point cloud and the target buckle point cloud to obtain the spatial transformation matrix including displacement and rotation, as well as the registration matching degree information; Step 4: Filter and retain the window point clouds with matching degrees higher than the threshold according to the matching degree, and remove duplicate recognition information according to the window position and matching degree.

[0009] Preferably, in step 2, the minimum bounding box of the point cloud data of the target buckle is first calculated.

[0010] Preferably, the maximum three-dimensional size of the bounding box is then obtained, and the size is defined as the side length of the sliding window, and half of the size is defined as the moving step length of the sliding window.

[0011] Preferably, in step three, a cube space with the same side length as the sliding window is first constructed according to the side length of the window in step two, and then the point cloud is cut by sliding the window in the component point cloud space from the minimum vertex to the maximum vertex using a convolution sliding window.

[0012] Preferably, the ICP matching is then performed with the buckle point cloud model in each sliding window. The ICP algorithm calculates the best transformation matrix through iterative optimization, so that the point cloud in the window can be accurately aligned with the buckle model, and records the matching degree, transformation matrix and window position coordinates of each window.

[0013] Preferably, in step 4, a threshold is given to screen the matching data obtained in S3 to eliminate low matching data.

[0014] Preferably, the filtered data are sorted from large to small according to the matching degree, the window with the highest matching degree is selected as a benchmark, and all windows adjacent to the window are eliminated.

[0015] Compared with the prior art, the present invention provides a buckle search and positioning method based on 3D model point cloud sampling and matching, which has the following beneficial effects: 1. The buckle search and positioning method based on 3D model point cloud sampling and matching is based on the point cloud segmentation of the sliding window: the point cloud of the workpiece model is cut to adapt to the size resolution of the buckle by adopting the idea of ​​convolution sliding window; With adaptive matching threshold setting: by calculating the minimum bounding box of the buckle, the size and step length of the sliding window are automatically determined to ensure the accuracy of the matching process; It has an effective deduplication mechanism: in the matching results, the matching degree is sorted and adjacent windows are eliminated to ensure the uniqueness of each target point cloud window, thereby improving the accuracy of the matching results.

[0016] 2. This buckle search and positioning method based on 3D model point cloud sampling and matching significantly improves the accuracy of buckle detection by combining the high-dimensional information of point cloud data and the precise matching of ICP algorithm. Compared with traditional methods, it can more effectively identify the true position and posture of each buckle and reduce false detection and missed detection. It can be widely used in automotive parts of different sizes and types and has good generalization. This allows it to play a role in various production environments, thus providing customized solutions for different manufacturing companies.

[0017] 3. The buckle search and positioning method based on 3D model point cloud sampling and matching reduces unnecessary calculations and repeated matching, and optimizes the use of computing resources by setting reasonable matching thresholds and deduplication mechanisms; it makes the detection process faster and more efficient in large-scale production environments, and can provide a stable data source for subsequent operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 This is a schematic diagram illustrating the matching degree and the sliding window coordinates in two-dimensional space of the present invention; Figure 2 The present invention is a schematic diagram of the process of searching and locating the buckle of the automobile shell component based on 3D model point cloud sampling and matching. DETAILED DESCRIPTION

[0019] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0020] See also Figure 1-Figure 2 ; A buckle search and positioning method based on 3D model point cloud sampling and matching includes the following steps: Step 1: Obtain the modeling file dataset and perform point cloud sampling on the automobile shell parts and target buckle STEP; Step 2: Calculate the size and moving step of the point cloud segmentation window according to the geometric size information of the target buckle; First, calculate the minimum bounding box of the point cloud data of the target buckle; Then, the maximum three-dimensional size of the bounding box is obtained, and the size is defined as the side length of the sliding window, and half of the size is defined as the moving step length of the sliding window; Step 3: Perform ICP registration on the point cloud in the segmented window of the component point cloud and the target buckle point cloud to obtain the spatial transformation matrix including displacement and rotation, as well as the registration matching degree information; First, a cube space with the same side length is constructed as a sliding window according to the side length of the window in step 2. Then, the point cloud is cut by sliding the window in the component point cloud space from the minimum vertex to the maximum vertex using a convolution sliding window. Then, in each sliding window, ICP matching is performed with the buckle point cloud model. The ICP algorithm calculates the best transformation matrix through iterative optimization, so that the point cloud in the window can be accurately aligned with the buckle model, and records the matching degree, transformation matrix and window position coordinates of each window; Step 4: Filter and retain the window point clouds with matching degrees higher than the threshold according to the matching degree, and remove duplicate recognition information according to the window position and matching degree; The matching data obtained by S3 is screened by a given threshold to eliminate low matching data; The filtered data is sorted from large to small according to the matching degree, and the window with the highest matching degree is selected as the benchmark, and all windows adjacent to the window are eliminated.

[0021] Experimental example: The following steps are involved: Step 1: Point cloud data preparation Extract part model data from STEP files and perform point cloud sampling processing on them.

[0022] Example of STEP file data content: Automobile shell parts STEP file data: ... “ DATA; #5=PRODUCT('L002610151_AC_back panel trim ASM','','',(#2)); #1=APPLICATION_CONTEXT('configuration controlled 3D design of mechanical parts and assemblies') ; #14=PRODUCT_DEFINITION('',' ',#6,#3); #16=SECURITY_CLASSIFICATION(' ',' ',#15) ; #15=SECURITY_CLASSIFICATION_LEVEL('unclassified') ; #46=CARTESIAN_POINT(' ',(0.,0.,0.)); #51=CARTESIAN_POINT('Control Point',(1443.144007,-141.2848955,392.8323995)); #52=CARTESIAN_POINT('Control Point',(1443.14367497,-140.483701658,392.903720393)); #53=CARTESIAN_POINT('Control Point',(1443.14463017,-139.682486544,392.974808268)); #54=CARTESIAN_POINT('Control Point',(1443.14684183,-138.881253062,393.045668898)); #55=CARTESIAN_POINT('Control Point',(1443.15031167,-138.080004555,393.116302802)); #56 = CARTESIAN_POINT('Control Point', (1443.15499682, -137.278743057, 393.186717818)); #57 = CARTESIAN_POINT('Control Point', (1443.16090289, -136.477467705, 393.256914195)); #58 = CARTESIAN_POINT('Limit', (1443.144007, -141.2848955, 392.8323995)); #59 = CARTESIAN_POINT('Limit', (1443.16090289, -136.477467705, 393.256914195)); ...... Snap model STEP file data: DATA; #5 = PRODUCT('Back panel Bottom Structure for Screw', '', '', (#2)); #1 = APPLICATION_CONTEXT('configuration controlled 3D design of mechanical parts and assemblies'); #14 = PRODUCT_DEFINITION('', '', #6, #3); #16 = SECURITY_CLASSIFICATION('', '', #15); #15 = SECURITY_CLASSIFICATION_LEVEL('unclassified'); #47 = CARTESIAN_POINT(' ', (0., 0., 0.)); #52 = CARTESIAN_POINT('Axis2P3D Location', (1465.66355977, -475.000068777, 321.897457821)); #57=CARTESIAN_POINT('Line Origine',(1447.97729323,-489.000137078,296.638840022)); #61=CARTESIAN_POINT('Vertex',(1448.85093114,-489.000133704,297.886524262)); #63=CARTESIAN_POINT('Vertex',(1447.10365532,-489.000140452,295.391155781)); #67=CARTESIAN_POINT('Control Point',(1448.85093114,-489.000133704,297.886524262)); #68=CARTESIAN_POINT('Control Point',(1448.95533116,-489.000133301,298.035622945)); #69=CARTESIAN_POINT('Control Point',(1449.0597229,-488.97345598,298.184709821)); #70=CARTESIAN_POINT('Control Point',(1449.1611496,-488.919848715,298.329562202)); Step 2: The longest dimension of the buckle is 98.329, the side length of the sliding window is set to 98.329, and the sliding step is set to 49.165; Step 3: Perform ICP registration on the point cloud segmented in the sliding window and the buckle to obtain the feature information with high matching degree of all buckles in the part: Buckle 1 Sliding window position: [1419.85892063 -513.86841113 265.71488722] Transformation matrix: [[ 9.99529054e-01 3.03754052e-02 -4.35958121e-031.92121879e+01] [-3.05595193e-02 9.98210528e-01 -5.13989967e-02 5.57553250e+01] [2.79051451e-03 5.15080173e-02 9.98668682e-01 2.23739244e+01] [0.00000000e+00 0.00000000e+00 0.00000000e+00 1.00000000e+00]] Fitness: 1.0 Centroid: [1459.90608998 -470.02526845 318.35788794] Buckle 2 Sliding window position: [1419.85892063 -366.37477042 265.71488722] Transformation matrix: [[ 9.99977109e-01 4.35656080e-03 -5.17702378e-035.94600845e+00] [-4.36296734e-03 9.99989730e-01 -1.22684513e-03 9.15692363e+00] [5.17162579e-03 1.24940423e-03 9.99985847e-01 -5.02870502e+00] [0.00000000e+00 0.00000000e+00 0.00000000e+00 1.00000000e+00]] fitness: 0.9998 Centroid: [1459.7970909 -315.92322065 318.43735958] Buckle 3 Sliding window position: [1567.35256135 -563.03295804 659.03126246] Transformation matrix: [[ 9.89334023e-01 -1.27161937e-01 7.10495089e-02 -9.71832217e+01] [7.09693379e-02 8.46741166e-01 5.27250179e-01 -5.53546337e+02] [-1.27206698e-01 -5.16584204e-01 8.46734442e-01 1.94517207e+01] [0.00000000e+00 0.00000000e+00 0.00000000e+00 1.00000000e+00]] fitness: 0.7108 Centroid: [1606.01517233 -501.58903628 682.28546142]] Step 4: Sort them based on the fitness, and remove the adjacent window information starting from the window position with the highest matching degree, and finally retain two buckle features, which are buckle 1 and buckle 2 above.

[0023] The beneficial effects of the present invention are as follows: the buckle search and positioning method based on 3D model point cloud sampling and matching is based on the point cloud segmentation of the sliding window: the point cloud of the workpiece model is cut to adapt to the size resolution of the buckle by adopting the idea of ​​convolution sliding window; With adaptive matching threshold setting: by calculating the minimum bounding box of the buckle, the size and step length of the sliding window are automatically determined to ensure the accuracy of the matching process; It has an effective deduplication mechanism: in the matching results, the matching degree is sorted and adjacent windows are eliminated to ensure the uniqueness of each target point cloud window, thereby improving the accuracy of the matching results. By combining the high-dimensional information of the point cloud data and the precise matching of the ICP algorithm, the accuracy of buckle detection is significantly improved. Compared with traditional methods, it can more effectively identify the true position and posture of each buckle, reduce false detection and missed detection; it can be widely used in automotive parts of different sizes and types, and has good generalization. This allows it to play a role in various production environments, thereby providing customized solutions for different manufacturing companies. By setting reasonable matching thresholds and deduplication mechanisms, unnecessary calculations and repeated matching are reduced, and the use of computing resources is optimized; in large-scale production environments, the detection process is faster and more efficient, and a stable data source can be provided for subsequent operations.

[0024] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A buckle search and positioning method based on 3D model point cloud sampling and matching, characterized in that: The following steps are involved: Step 1: Obtain the modeling file dataset and perform point cloud sampling on the automobile shell parts and target buckle STEP; Step 2: Calculate the size and moving step of the point cloud segmentation window according to the geometric size information of the target buckle; Step 3: Perform ICP registration on the point cloud in the segmented window of the component point cloud and the target buckle point cloud to obtain the spatial transformation matrix including displacement and rotation, as well as the registration matching degree information; Step 4: Filter and retain the window point clouds with matching degrees higher than the threshold according to the matching degree, and remove duplicate recognition information according to the window position and matching degree.

2. The buckle search and positioning method based on 3D model point cloud sampling and matching according to claim 1 is characterized in that: In the step 2, the minimum bounding box of the point cloud data of the target buckle is first calculated.

3. The buckle search and positioning method based on 3D model point cloud sampling and matching according to claim 2 is characterized in that: Then, the maximum three-dimensional size of the bounding box is obtained, and the size is defined as the side length of the sliding window, and half of the size is defined as the moving step length of the sliding window.

4. The buckle search and positioning method based on 3D model point cloud sampling and matching according to claim 3 is characterized in that: In the step three, a cube space with the same side length is first constructed as a sliding window according to the side length of the window in step two, and then the point cloud is cut by sliding the window in the component point cloud space from the minimum vertex to the maximum vertex using a convolution sliding window.

5. The buckle search and positioning method based on 3D model point cloud sampling and matching according to claim 4 is characterized in that: Then, in each sliding window, ICP matching is performed with the buckle point cloud model. The ICP algorithm calculates the best transformation matrix through iterative optimization, so that the point cloud in the window can be accurately aligned with the buckle model, and records the matching degree, transformation matrix and window position coordinates of each window.

6. The buckle search and positioning method based on 3D model point cloud sampling and matching according to claim 5, characterized in that: In the step 4, a threshold is given to screen the matching data obtained in S3 to eliminate low matching data.

7. The buckle search and positioning method based on 3D model point cloud sampling and matching according to claim 6, characterized in that: The filtered data is sorted from large to small according to the matching degree, and the window with the highest matching degree is selected as the benchmark, and all windows adjacent to the window are eliminated.

Images