A method for identifying cross section of engineering component based on rotation optimization and chamfer distance

Abstract

This invention discloses a method for identifying the cross-section of engineering components based on rotation optimization and Chamfer distance, belonging to the field of Building Information Modeling and Intelligent Construction. This method solves the problems of semantic information loss and difficulty in identifying cross-section types after BIM / CAD models are converted to a general 3D mesh format, providing a high-fidelity geometric model with standard cross-section attributes for the simulation preprocessing engine. This method does not rely on the parametric information of the original model; it only relies on 3D mesh geometric data. Through extension direction analysis and continuous optimization matching based on differentiable transformation and Chamfer distance, it achieves high-precision, robust identification and standardized matching of typical component cross-section types such as beams, columns, and supports.

Description

Technical Field

[0001] This invention belongs to the field of Building Information Modeling (BIM) and intelligent construction, and more specifically, relates to a method for identifying the cross-section of engineering components based on rotation optimization and Chamfer distance. Background Technology

[0002] In the digital design and simulation process of civil engineering structures, design models typically originate from various heterogeneous software systems, including Building Information Modeling (BIM) platforms (such as Revit), structural analysis software (such as ETABS and MIDAS), and general CAD or 3D modeling tools (such as Rhino and SketchUp). When these models are imported into the simulation preprocessing system, they often exist in general geometric formats (such as OBJ, IFC, and STEP), lacking clear semantic information about the components. Among these, the "semantic deficiency" of section type severely restricts the subsequent automated processing such as mesh generation, boundary condition application, and physical property assignment, becoming a major bottleneck for the efficiency and accuracy of simulation preprocessing.

[0003] Furthermore, in the design and construction of engineering structures such as buildings, bridges, and industrial plants, components such as beams, columns, supports, and walls typically have regular or semi-regular cross-sectional types (such as I-shaped, rectangular, circular, and T-shaped). Accurately identifying the cross-sectional type of components and matching it with a standard cross-section library is a crucial prerequisite for achieving automated design, drawing, quantity calculation, and construction in BIM modeling, construction detailing, component prefabrication, and intelligent inspection stages.

[0004] In existing technologies, the identification of component cross-sections mainly employs the following methods: 1. Manual Assignment or Attribute Reading: In BIM software, component cross-sectional information is usually manually assigned by the user during modeling or directly read through family / type attributes. This method relies on modeling standardization and is not applicable when the model lacks attributes or is a reverse-engineered model (such as one generated from point clouds or OBJ meshes). II. Cross-section inference based on geometric rules: This method infers the possible cross-section type of a component by analyzing its bounding box, principal direction, or symmetry. However, this method is sensitive to noise and struggles to handle non-standard or variable cross-section components. III. Image Contour Matching Method: This method projects the component cross-section as a two-dimensional contour, extracts boundary points, and then compares it with a standard template library based on invariant moment methods such as Hu moments and Zernike moments, or optimal point set matching methods (such as the Hungarian algorithm). Currently, this type of method lacks sufficient discriminative power for optimal complex or similar cross-sections (such as H-beams and box-sections).

[0005] In recent years, although some studies have attempted to introduce deep learning for cross-section classification, it relies on a large amount of labeled data, has limited generalization ability, and is difficult to interpret the matching process, making it difficult to promote in practical engineering. In addition, existing matching algorithms often ignore the globally optimal matching between point pairs when dealing with the optimal correspondence between contour point sets, resulting in a high mismatch rate. Summary of the Invention

[0006] In view of the above-mentioned defects or improvement needs of the existing technology, the present invention provides an engineering component cross-section identification method based on rotation optimization and Chamfer distance, which can achieve high-precision, high-robust identification and standardized matching of component cross-section types.

[0007] To achieve the above objectives, according to a first aspect of the present invention, a method for identifying the cross-section of engineering components based on rotation optimization and Chamfer distance is provided, comprising: S1. Establish a template library containing K component templates; save the two-dimensional contour point set of each component template to represent each component template; The cross-sectional types of each component template are different, and the components are engineering components; component template k is obtained through... Characterization; , , , These are the first, second, and third points of the two-dimensional contour point set of component template k, respectively. Two-dimensional contour points, Let k be the number of two-dimensional contour points of component template k, where k∈{1,2,…,K}; S2, proceed according to the preset direction and preset angle step size. right J uniform discrete rotation samplings are performed to obtain J rotated component templates. ;in, The component template is obtained by the j-th rotation, and the range of the rotation sampling angle is... , ; S3 extracts all components from the three-dimensional mesh model file of the civil engineering structure, and processes each component as a target component to obtain its cross-section recognition result; The process includes: calculating the extension direction of the target component; equally dividing the target component into three intercepting planes (top, middle, and bottom) perpendicular to the extension direction; finding the intersection of these intercepting planes with the three-dimensional mesh of the target component to obtain a closed three-dimensional cross-sectional loop; and orthogonally projecting the three-dimensional cross-sectional loop onto a two-dimensional plane to obtain a two-dimensional contour point set of the target component. Initialize the affine transformation matrix M, and apply M to the templates of each rotated component. The target template is obtained by performing an affine transformation. ,Will and The Chamfer distance between the target component and each rotated component template is used as the loss function. A gradient optimization algorithm is used for backpropagation to update M until the loss converges or the maximum number of iterations is reached. The Chamfer distance at this point is then saved as the distance between the target component and each rotated component template. The matching score is used as the matching score between the target component and the component template k. The lowest matching score is used as the matching score between the target component and the component template k. The matching scores between the target component and each component template are obtained by traversing each component template. The cross-section type of the component template corresponding to the lowest matching score is used as the cross-section recognition result of the target component.

[0008] According to a second aspect of the present invention, an electronic device is provided, comprising: a computer-readable storage medium and a processor; The computer-readable storage medium is used to store executable instructions; The processor is configured to read executable instructions stored in the computer-readable storage medium and execute the method as described in the first aspect.

[0009] According to a third aspect of the invention, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions for causing a processor to perform the method as described in the first aspect.

[0010] According to a fourth aspect of the invention, a computer program product is provided, comprising a computer program or instructions that, when executed by a processor, implement the method described in the first aspect.

[0011] In summary, compared with the prior art, the above-described technical solutions conceived by this invention can achieve the following beneficial effects: This invention provides a method for identifying the cross-section of engineering components based on rotation optimization and Chamfer distance. This method solves the problems of semantic information loss and difficulty in identifying cross-section types after BIM / CAD models are converted to a common 3D mesh format (such as OBJ). It provides a high-fidelity geometric model with standard cross-section attributes for the simulation preprocessing engine. This method does not rely on the parametric information of the original model; it only uses 3D mesh geometric data. Through extension direction analysis and continuous optimization matching based on differentiable transformation and Chamfer distance, it achieves high-precision and robust identification and standardized matching of typical component cross-section types such as beams, columns, and supports. It has the following advantages: 1. Does not rely on the original modeling semantics: Only requires general 3D mesh input, suitable for typical scenarios where the semantics of BIM / CAD models are lost after format conversion, effectively supporting the unified processing of multi-source design models; 2. High robustness and accuracy: Through extension direction correction and orthogonal projection, the geometric fidelity of cross-section extraction is significantly improved; 3. The optimal matching of the contour point set is achieved by using differentiable transformation and continuous optimization of Chamfer distance, which has strong robustness to rotation, scaling and local noise. 4. Tightly coupled simulation preprocessing workflow: The identified section types can be directly used for component classification, attribute assignment, and selection of geometric simplification rules; 5. High scalability: The template library can be flexibly expanded to support new or non-standard cross sections, and the algorithm framework can also be migrated to other geometric representations such as point clouds or NURBS surfaces.

[0012] In summary, the method provided by this invention offers key technical support for the automated conversion of multi-source design models of civil engineering structures into simulation-ready models, and is an important component of realizing an integrated intelligent pre-processing system for "design-simulation". It is applicable to the automatic identification, classification, and standardized matching of building components such as steel and concrete structures in BIM models or 3D reconstruction models. Attached Figure Description

[0013] Figure 1 A flowchart of the engineering component cross-section identification method based on rotation optimization and Chamfer distance provided in this embodiment of the invention; Figure 2 This is a database relationship structure diagram for building a template library provided in an embodiment of the present invention. Detailed Implementation

[0014] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0015] Existing cross-section recognition methods suffer from the following shortcomings: They rely on original attribute information; after exporting most BIM / CAD models to a common format (such as OBJ), semantic information such as component type and cross-section parameters is lost, leaving only the geometric mesh. Traditional methods cannot recover the cross-section type from pure geometric data, leading to extensive manual intervention in simulation preprocessing. Cross-section recognition also exhibits poor robustness; existing methods based on bounding boxes, principal component analysis (PCA), or simple projection struggle to accurately extract the true extension direction of components, especially when dealing with braces, variable cross-section beams, or complex connection nodes, easily resulting in misjudgments. Furthermore, matching accuracy and efficiency are insufficient; when comparing cross-section contours with standard template libraries, image similarity or local feature matching is often used, without considering the globally optimal correspondence between contour point sets, leading to low matching accuracy in the presence of noise, rotation, scaling, or inconsistent point counts.

[0016] To address the problems of existing cross-section classification methods, this invention employs rotation optimization and a global distance metric for point sets for shape matching. The Chamfer distance, the sum of the nearest neighbor distances from all points on one contour to another contour, is naturally robust to inconsistencies in point counts and local missing points, and has been widely applied in target detection and shape retrieval. However, existing methods often fix the template pose and do not fully consider the arbitrary rotational degrees of freedom of the cross-section in the projected coordinate system, making the matching results sensitive to initial alignment.

[0017] Based on this, embodiments of the present invention provide a method for identifying the cross-section of engineering components based on rotation optimization and Chamfer distance, such as... Figure 1 As shown, it includes: S1. Establish a template library containing K component templates; save the two-dimensional contour point set (i.e., two-dimensional contour section point set) of each component template to represent each component template; The cross-sectional types of each component template are different, and the components are engineering components; component template k is obtained through... Characterization; , , , These are the first, second, and third points of the two-dimensional contour point set of component template k, respectively. A two-dimensional contour point, represented by x,y coordinates; Let k be the number of two-dimensional contour points of component template k, where k∈{1,2,…,K}.

[0018] In step S1, a standard section template library is constructed, including: pre-establishing a standard template library covering common civil engineering structure geometric component section types (such as rectangular, circular, I-beam, box girder, T-beam, angle steel, etc.); each template generates a normalized two-dimensional contour point set in the same way. It stores its geometric parameters and section type labels.

[0019] S2, proceed according to the preset direction and preset angle step size. right Perform J discrete rotation sampling operations to obtain J rotated component templates. , Let be the component template obtained by the j-th rotation; each rotated component template is also represented by its two-dimensional contour point set.

[0020] In step S2, for each template , and put it in Follow the preset direction and preset angle step size. Perform multiple rotations, that is, for Discrete rotation sampling is performed to obtain a series of rotated templates. .

[0021] S3 extracts all components from the three-dimensional mesh model file of the civil engineering structure, and processes each component as a target component to obtain its cross-section recognition result; The process includes: calculating the extension direction of the target component; equally dividing the target component into three intercepting planes (top, middle, and bottom) perpendicular to the extension direction; finding the intersection of these intercepting planes with the three-dimensional mesh of the target component to obtain a closed three-dimensional cross-sectional loop; and orthogonally projecting the three-dimensional cross-sectional loop onto a two-dimensional plane to obtain a two-dimensional contour point set of the target component. Initialize the affine transformation matrix M, and apply M to the templates of each rotated component. The target template is obtained by performing an affine transformation. ,Will and The Chamfer distance between the target component and each rotated component template is used as the loss function. A gradient optimization algorithm is used for backpropagation to update M until the loss converges or the maximum number of iterations is reached. The Chamfer distance at this point is then saved as the distance between the target component and each rotated component template. The matching score is used as the matching score between the target component and the component template k. The lowest matching score is used as the matching score between the target component and the component template k. The matching scores between the target component and each component template are obtained by traversing each component template. The cross-section type of the component template corresponding to the lowest matching score is used as the cross-section recognition result of the target component.

[0022] Specifically, step S3 includes: S31, 3D mesh model and identification of candidate components.

[0023] It receives 3D mesh models of civil engineering structures in common formats such as OBJ and GLB; based on geometric connectivity and aspect ratio characteristics, it automatically segments geometric components (such as beams, columns, and supports) as engineering components to be matched.

[0024] S32, calculate the extension direction of each engineering component.

[0025] For each candidate component, principal component analysis (PCA) or minimum bounding cylinder fitting method is used to calculate its principal axis direction as the extrusion direction; further, the consistency of local curvature and end face normal is checked to correct the directional deviation caused by mesh noise or interference from connection nodes, ensuring that the extension direction accurately reflects the extension trend of the component.

[0026] S33, cut a section along the extension direction of each engineering component to be matched and project it into a two-dimensional profile to construct a two-dimensional profile point set of each engineering component to be matched.

[0027] For each engineering component to be matched, three cutting planes perpendicular to the extension direction are equally divided into the upper, middle, and lower parts of the target component. The intersection of these cutting planes with the three-dimensional mesh of the component is calculated to obtain a closed three-dimensional cross-sectional loop (for an effective component, the manifold watertight splicing structure of its triangular facets determines that the intersection line obtained by any cross-sectional plane with it must be a closed loop). This cross-sectional loop is orthogonally projected onto a two-dimensional plane to generate a two-dimensional point set in a normalized coordinate system, denoted as the two-dimensional contour point set of the engineering component to be matched. .

[0028] S34, based on continuous optimization matching of differentiable transformation and Chamfer distance.

[0029] Initialize the affine transformation matrix Perform an affine transformation on each rotated template. Using one of the rotated templates... For example, After affine transformation, the target point set is obtained. ; Calculate the two-dimensional contour point set of the engineering component to be matched and Chamfer distance between Using L as the loss function, the affine transformation matrix is ​​updated via backpropagation using a gradient optimization algorithm. Continue this process until the loss converges or the maximum number of iterations is reached; save the minimum Chamfer distance as a template. The matching score is calculated. All templates and their rotation angles are iterated through, and the template with the smallest Chamfer distance is selected as the final matching result for the engineering component to be matched. The corresponding section type is then output.

[0030] S35 writes back the successfully matched section type, size parameters, and coordinate system information to the original component geometric model, generating a structured component object with complete semantics, which can be called by subsequent geometric simplification, topology correction, and simulation preprocessing modules.

[0031] Preferably, the affine transformation matrix is... ;in, These are the stretching factors in the x and y directions, respectively. These are the stretch amounts in the x and y directions, respectively; during initialization, .

[0032] Preferably, M is used for each rotated component template. Performing affine transformations includes: for Any two-dimensional point q=(x,y) in the array T Represent its homogeneous coordinates as After affine transformation, the two-dimensional point becomes .

[0033] Preferably, the preset direction is clockwise or counterclockwise.

[0034] Preferably, step S3 further includes: if the lowest matching score among the matching scores of the target component and each component template obtained by traversing each component template is higher than a preset threshold, then the matching result of the target component is that there is no component template in the template library that matches the target component.

[0035] Understandably, if the lowest matching score among the matching scores of the target component and each component template obtained by traversing each component template is higher than the preset threshold, it means that there is no component template in the template library that matches the target component, and the current matching result is only the best match in the template library, not a reasonable match.

[0036] Preferably, the two-dimensional contour point sets of each component template and each target component are normalized two-dimensional contour point sets.

[0037] In summary, the method provided by this invention combines stretching direction analysis, cross-sectional profile extraction, and cross-sectional matching based on Chamfer distance and rotation optimization matching to achieve cross-sectional matching of engineering components. It does not rely on the original model attributes, can automatically extract the extension direction and cross-sectional profile of components from general three-dimensional mesh models (such as OBJ format), and the cross-sectional matching method based on point set energy, alignment rotation attitude, and overall geometric similarity as the optimization objective can support the automation requirements of component identification and standardization in intelligent construction.

[0038] The method provided by the present invention will be further illustrated below with a specific example.

[0039] Using a section of a bridge (provided in GLB format) as input, the method provided by this invention automatically identifies the cross-sectional types of typical components such as box girders, cross bracing, and pier caps. The identification process includes the following steps: (1) Component standard template library.

[0040] Obtain the parameters for the corresponding cross-section type from the atlas or actual bridge model. Starting from the top left corner of the cross-section, calculate the coordinates of each inflection point counterclockwise based on the parameters, and then generate the template. The contour points are stored in an embedded database.

[0041] (2) Extract and separate components from the input.

[0042] The input model is a GLB file containing only geometric information, with approximately 15,000 triangles. Based on GLB file specifications and connected component analysis, individual components are extracted from the overall model. In this example, 124 components were extracted.

[0043] (3) Calculate the extension direction of the component.

[0044] First, geometric primitives are extracted, and the coordinates of all vertices are extracted from the input 3D mesh model. and edge connection relationship Each edge Represent the connection between two vertices; then calculate the set of edge vectors for each edge. Calculate its direction vector , forming a set of edge vectors Principal component analysis obtains the vertex set of candidate principal axis pairs. Principal component analysis and geometric mean method were used to calculate the centroids of the vertices. And construct a decentralized covariance matrix Based on this, the eigenvalues ​​and eigenvectors are calculated, and the eigenvalues ​​are sorted from largest to smallest to obtain the three orthogonal principal axis directions. These correspond to the length, width, and height directions, respectively; for each candidate principal axis Calculate all valid edge vectors (length greater than the threshold). )and The absolute value of the cosine of the included angle:

[0045] Count the number of edges that satisfy the following two conditions: first, approximately parallel edges; (In this example) Secondly, approximately perpendicular sides: (In this embodiment) ); Calculate the "structural consistency ratio" under this axis:

[0046] in This represents the total number of effective edges. A higher proportion indicates that the direction is more likely to be the extension direction of the member (because typical beam-column structures have a large number of parallel edges along the axial direction and a large number of perpendicular edges in the cross-sectional direction); finally, select the direction that makes... Largest candidate axis As the final extension direction .

[0047] (4) Extract the component cross section.

[0048] First, perform scalar projection on all vertices of the component along the extension direction to obtain the minimum value of the projection. With the maximum value This defines the effective length of the component in the axial direction. And calculate the midpoint of the projection. Subsequently, the projected coordinates of the three candidate cut-off locations were determined. The selection of 0.49 as the scaling factor aims to ensure that the cutting plane is located within the main segment of the component and can capture the cross-sections of the middle and both segments.

[0049] For each target projection value, it is converted into a cutoff point in three-dimensional space: which is based on the geometric center of the component. Based on the extension direction Adjust to the corresponding projection position to obtain the origin of the three intercepting planes. Each cutting plane is uniquely determined by the origin and the extrusion direction (as the plane normal).

[0050] Next, the intersection of these three planes with the 3D mesh of the component is performed sequentially to obtain the spatial cross-section intersection lines. For valid 3D cross-sections (i.e., those capable of forming closed 3D cross-section loops), they are further orthogonally projected onto a local 2D coordinate system perpendicular to the extrusion direction, and plotted as an image. The contour point set is then extracted from the image. The profile preserves the geometric features of the cross-section and serves as input for subsequent matching.

[0051] (5) Continuous optimization matching based on differentiable transformation and Chamfer distance.

[0052] Obtaining an effective two-dimensional cross-sectional profile Next, it will be matched with a high-precision standard template library. First, the contour is preprocessed and normalized, and the set of points of the cross-sectional contour to be matched (denoted as the target contour) is geometrically normalized. Then, each template in the standard cross-sectional template library is traversed. For each template, the following operations are performed: 1) Within the range of 0° to 360°, perform discrete rotation sampling with a preset angle step size to generate a series of template instances of initial postures. ; 2) For each template in each rotational orientation, initiate the continuous optimization process: First, define the affine transformation matrix. Including rotation angle Stretch factor in the x / y direction and translation amount . The initial value is set to the unit transformation (i.e. , (This is a 3D identity matrix). Then, an affine transformation is performed on the template under each rotational orientation to obtain a series of templates. A component template in Taking affine transformation as an example, let's denote the point set obtained after the affine transformation as... .

[0053] for Any two-dimensional point q=(x,y) in the array T Represent its homogeneous coordinates as After affine transformation, the two-dimensional point becomes . After all points in the transformation are completed, the component module after affine transformation is obtained. .

[0054] 3) Construct the Chamfer distance as the optimization objective function, which is defined as the sum of bidirectional average nearest neighbor distances:

[0055] Where p is a point in P, for The point in the middle.

[0056] 4) Employ iterative optimization algorithms (such as gradient descent or parameter fitting) to continuously adjust and transform... To minimize Stop when the loss function converges or the maximum number of iterations is reached.

[0057] 5) Optimal matching result selection: Record the minimum Chamfer distance value obtained by each template across all rotation angles and optimization iterations. After traversing the entire template library, select the template with the minimum loss and a loss below a preset threshold as the final matching result, and output its cross-section type.

[0058] (6) For example Figure 2 As shown, the successfully matched section type, size parameters, and coordinate system information are written back to the original component geometric model to generate a structured component object with complete semantics, which can be called by subsequent geometric simplification, topology correction, and simulation preprocessing modules.

[0059] This invention provides an electronic device, including: a computer-readable storage medium and a processor; The computer-readable storage medium is used to store executable instructions; The processor is configured to read executable instructions stored in the computer-readable storage medium and execute the method as described in any of the above embodiments.

[0060] This invention provides a computer-readable storage medium storing computer instructions that cause a processor to perform the method described in any of the above embodiments.

[0061] This invention provides a computer program product, including a computer program or instructions, which, when executed by a processor, implement the method described in any of the above embodiments.

[0062] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for identifying the cross-section of engineering components based on rotation optimization and Chamfer distance, characterized in that, include: S1. Establish a template library containing K component templates; save the two-dimensional contour point set of each component template to represent each component template; The cross-sectional types of each component template are different, and the components are engineering components; component template k is obtained through... Characterization; , , , These are the 1st, 2nd, and 3rd points of the two-dimensional contour point set of component template k, respectively. Two-dimensional contour points, Let k be the number of two-dimensional contour points of component template k, where k∈{1,2,…,K}; S2, proceed according to the preset direction and preset angle step size. right J uniform discrete rotation samplings are performed to obtain J rotated component templates. ;in, The component template is obtained by the j-th rotation, and the range of the rotation sampling angle is... , ; S3 extracts all components from the three-dimensional mesh model file of the civil engineering structure, and processes each component as a target component to obtain its cross-section recognition result; The process includes: calculating the extension direction of the target component; equally dividing the target component into three intercepting planes (top, middle, and bottom) perpendicular to the extension direction; finding the intersection of these intercepting planes with the three-dimensional mesh of the target component to obtain a closed three-dimensional cross-sectional loop; and orthogonally projecting the three-dimensional cross-sectional loop onto a two-dimensional plane to obtain a two-dimensional contour point set of the target component. Initialize the affine transformation matrix M, and apply M to the templates of each rotated component. The target template is obtained by performing an affine transformation. ,Will and The Chamfer distance between the target component and each rotated component template is used as the loss function. A gradient optimization algorithm is used for backpropagation to update M until the loss converges or the maximum number of iterations is reached. The Chamfer distance at this point is then saved as the distance between the target component and each rotated component template. The matching score is used as the matching score between the target component and the component template k. The lowest matching score is used as the matching score between the target component and the component template k. The matching scores between the target component and each component template are obtained by traversing each component template. The cross-section type of the component template corresponding to the lowest matching score is used as the cross-section recognition result of the target component.

2. The method as described in claim 1, characterized in that, Affine transformation matrix ;in, These are the stretching factors in the x and y directions, respectively. These are the stretch amounts in the x and y directions, respectively; during initialization, .

3. The method as described in claim 2, characterized in that, Use M to template each rotated component Affine transformations include: for Any two-dimensional point q=(x,y) in the array T Represent its homogeneous coordinates as After affine transformation, the two-dimensional point becomes .

4. The method as described in claim 1, characterized in that, The preset direction is either clockwise or counterclockwise.

5. The method as described in claim 1, characterized in that, Step S3 further includes: if the lowest matching score among the matching scores of the target component and each component template obtained by traversing each component template is higher than a preset threshold, then the matching result of the target component is that there is no component template in the template library that matches the target component.

6. The method as described in claim 1, characterized in that, The two-dimensional contour point sets of each component template and each target component are normalized two-dimensional contour point sets.

7. An electronic device, characterized in that, include: Computer-readable storage media and processors; The computer-readable storage medium is used to store executable instructions; The processor is configured to read executable instructions stored in the computer-readable storage medium and execute the method as described in any one of claims 1-6.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing a processor to perform the method as described in any one of claims 1-6.

9. A computer program product, comprising a computer program or instructions, characterized in that, When the computer program or instructions are executed by a processor, they implement the method as described in any one of claims 1-6.

Images