Data matching method, apparatus, device, storage medium, and computer program product

Abstract

The application provides a data matching method and device, equipment, a storage medium and a computer program product. The method comprises the following steps: obtaining a first point cloud set of a to-be-matched object at a system level and a second point cloud set of the to-be-matched object at a device level by using a digital twin model of a traction system; determining first local plane domain information corresponding to the first point cloud set and second local plane domain information corresponding to the second point cloud set; determining a first extended transformation matrix corresponding to the first point cloud set and a second extended transformation matrix corresponding to the second point cloud set according to the first local plane domain information and the second local plane domain information; performing pose transformation on the first point cloud set and the second point cloud set based on the first extended transformation matrix and the second extended transformation matrix, to obtain a third point cloud set after the pose transformation of the first point cloud set and a fourth point cloud set after the pose transformation of the second point cloud set; and performing data matching on the to-be-matched object at the system level and the device level according to the third point cloud set and the fourth point cloud set.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a data matching method, apparatus, device, storage medium, and computer program product. Background Technology

[0002] From the device level and component level to the system level, different scales have corresponding requirements for the level of refinement of device model construction, and the modeling and simulation software used will also be different, resulting in differences in the spatial position (pose) and the number of model meshes (point cloud density) of the same device at different scale levels. Summary of the Invention

[0003] This application provides a data matching method, apparatus, device, storage medium, and computer program product that can achieve accurate matching of the same component at the system level and device level.

[0004] To achieve the above objectives, the technical solution of this application embodiment is implemented as follows:

[0005] Firstly, this application proposes a data matching method, the method comprising:

[0006] The first point cloud at the system level and the second point cloud at the device level are obtained using the digital twin model of the traction system; the point cloud distribution density and pose information of the first point cloud and the second point cloud are different.

[0007] Determine the first local planar domain information corresponding to the first point set and the second local planar domain information corresponding to the second point set, respectively;

[0008] Based on the first local planar domain information and the second local planar domain information, determine the first extended transformation matrix corresponding to the first point cloud and the second extended transformation matrix corresponding to the second point cloud, respectively.

[0009] Based on the first extended transformation matrix and the second extended transformation matrix, the pose transformations of the first point cloud and the second point cloud are performed respectively to obtain the third point cloud after the pose transformation of the first point cloud and the fourth point cloud after the pose transformation of the second point cloud.

[0010] The object to be matched is matched at the system level and the device level based on the third point cloud and the fourth point cloud.

[0011] Secondly, this application proposes a data matching device, the device comprising:

[0012] The acquisition unit is used to acquire the first point cloud at the system level and the second point cloud at the device level of the object to be matched using the digital twin model of the traction system; the point cloud distribution density and pose information of the first point cloud and the second point cloud are different.

[0013] The first determining unit is used to determine the first local planar domain information corresponding to the first point set and the second local planar domain information corresponding to the second point set, respectively.

[0014] The second determining unit is used to determine the first extended transformation matrix corresponding to the first point cloud and the second extended transformation matrix corresponding to the second point cloud based on the first local plane domain information and the second local plane domain information, respectively.

[0015] The pose transformation unit is used to perform pose transformation on the first point cloud and the second point cloud based on the first extended transformation matrix and the second extended transformation matrix respectively, to obtain the third point cloud after the pose transformation of the first point cloud and the fourth point cloud after the pose transformation of the second point cloud.

[0016] A data matching unit is used to perform data matching on the object to be matched at the system level and the device level based on the third point cloud and the fourth point cloud.

[0017] Thirdly, this application proposes a data matching device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of any of the methods described above.

[0018] Fourthly, this application proposes a storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the methods described above.

[0019] Fifthly, this application proposes a computer program product, including a computer program that, when executed by a processor, implements the steps of any of the methods described above.

[0020] This application proposes a data matching method, apparatus, device, storage medium, and computer program product. The method includes: acquiring a first point cloud at the system level and a second point cloud at the device level for an object to be matched using a digital twin model of a traction system; the point cloud distribution density and pose information of the first and second point clouds are different; determining the first local planar domain information corresponding to the first point cloud and the second local planar domain information corresponding to the second point cloud; determining the first extended transformation matrix corresponding to the first point cloud and the second extended transformation matrix corresponding to the second point cloud based on the first and second local planar domain information; performing pose transformation on the first and second point clouds based on the first and second extended transformation matrices to obtain a third point cloud after pose transformation of the first point cloud and a fourth point cloud after pose transformation of the second point cloud; and performing data matching on the object to be matched at the system level and the device level based on the third and fourth point clouds. Using the above implementation scheme, the first point cloud data of the object to be matched at the system level is sequentially transformed through the first local planar domain information and the first extended transformation matrix to obtain the third point cloud data. The second point cloud data of the object to be matched at the device level is sequentially transformed through the second local planar domain information and the second extended transformation matrix to obtain the fourth point cloud data. Based on the third and fourth point cloud data after the pose transformation, data matching of the object to be matched is performed at the system level and the device level, realizing accurate matching of the same component (the object to be matched) at the system level and the device level. Attached Figure Description

[0021] Figure 1 A flowchart illustrating a data matching method provided in an embodiment of this application;

[0022] Figure 2 A schematic diagram of the structure of an exemplary digital twin model of a traction system provided in an embodiment of this application;

[0023] Figure 3 This is an exemplary point cloud diagram of a model of an object to be matched at the system and device levels, provided for embodiments of this application.

[0024] Figure 4 A schematic diagram of an exemplary system-level and device-level local planar region provided for embodiments of this application;

[0025] Figure 5 A schematic diagram of an exemplary system-level and device-level primary rotating local plane domain of a target object provided in an embodiment of this application;

[0026] Figure 6 A schematic diagram illustrating an exemplary preliminary rotational transformation of an object to be matched, provided for an embodiment of this application;

[0027] Figure 7 A schematic diagram of an exemplary system-level and device-level secondary rotational local planar domain of a target object provided in this application embodiment;

[0028] Figure 8 A schematic diagram of a secondary rotational transformation of an object to be matched, provided in an embodiment of this application;

[0029] Figure 9 A schematic diagram of the point cloud corresponding to the final pose of an object to be matched at the system level and device level, provided for an embodiment of this application;

[0030] Figure 10 This is a schematic diagram of the structure of a data matching device provided in an embodiment of this application;

[0031] Figure 11 This is a schematic diagram of the structure of a data matching device provided in the application embodiment. Detailed Implementation

[0032] In order to gain a more detailed understanding of the features and technical content of the embodiments of this application, the implementation of the embodiments of this application will be described in detail below with reference to the accompanying drawings. The accompanying drawings are for reference and illustration only and are not intended to limit the embodiments of this application.

[0033] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0034] In the following description, references to "some embodiments" are made, which describe a subset of all possible embodiments. However, it is understood that "some embodiments" may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict. It should also be noted that the terms "first," "second," etc., used in the embodiments of this application are only used to distinguish similar objects and do not represent a specific ordering of objects. It is understood that "first," "second," etc., may be interchanged in a specific order or sequence where permitted, so that the embodiments of this application described herein can be implemented in an order other than that illustrated or described herein.

[0035] With the rapid development of rail transit traction systems, their reliability assessment and intelligent monitoring technologies also require attention. Digital twins are a virtual representation technology of actual physical systems based on data and machine learning, enabling the simulation of operational characteristics and parameter monitoring of actual physical systems. Providing multi-level digital twin simulation models that meet different monitoring needs, combined with practical application scenarios, to achieve real-time visual monitoring of components under different line operating conditions has become an important foundation for the digital and intelligent monitoring and operation and maintenance of rail transit traction systems.

[0036] From the device level and component level to the system level, different scales require different levels of refinement in device model construction, and the modeling and simulation software used will also differ. This leads to variations in the spatial position (pose) and mesh quantity (point cloud density) of the same device at different scale levels. To more comprehensively evaluate, analyze, and monitor the device's status, accurate matching of point cloud data at different scales is necessary. While systematic 3D local feature extraction and matching of local features one by one yield a series of matching feature point pairs, mismatches can occur when dealing with local feature data exhibiting uneven density or outlier interference, affecting the data matching results.

[0037] Based on this, embodiments of this application provide a data matching method. Figure 1 A flowchart illustrating a data matching method provided in an embodiment of this application; as shown Figure 1 As shown, the method includes:

[0038] S101. Using the digital twin model of the traction system, obtain the first point cloud at the system level and the second point cloud at the device level for the object to be matched; the density and pose information of the point cloud distribution of the first point cloud and the second point cloud are different.

[0039] It should be noted that the digital twin model of the traction system is obtained by sequentially connecting the vehicle dynamics model, the lumped parameter equivalent digital circuit model, and the multi-field reduced-order model. For ease of understanding, Figure 2 A schematic diagram of the structure of an exemplary digital twin model of a traction system provided in this application embodiment; as shown Figure 2 As shown, the digital twin model is obtained by sequentially connecting the vehicle dynamics model, the lumped parameter equivalent digital circuit model, and the multi-field reduced-order model.

[0040] It should be noted that the object to be matched can be referred to as a monitoring component of the traction system. The specific selection of the monitoring component can be determined according to the actual situation and is not limited here. In practical applications, the system level can also be called the system-level, and the device level can also be called the device-level. Using the digital twin model of the traction system, the first point cloud of the object to be matched at the system level and the second point cloud at the device level are obtained respectively. This can be understood as using the data of the object to be matched under the same working conditions as the input of the digital twin model, and outputting the first point cloud of the object to be matched at the system level and the second point cloud at the device level. Among them, the density and pose information of the point cloud distribution of the first point cloud and the second point cloud are different, and the physical fields of the first point cloud and the second point cloud are similar, including but not limited to temperature field, stress field, flow field, etc.

[0041] In practical applications, the first point cloud can be denoted as g(x,y,z,u). syetem The second point can be denoted as g(x,y,z,u). device The physical field can be denoted as u.

[0042] S102. Determine the first local planar domain information corresponding to the first point cloud and the second local planar domain information corresponding to the second point cloud.

[0043] It should be noted that the first local planar domain information and the second local planar domain information can be understood as local planar domains of the first point cloud and the second point cloud that have similar physical characteristics in spatial distribution and satisfy geometric constraints. In practical applications, the first local planar domain information can be denoted as Zone. plane-system The second local planar domain information can be denoted as Zone. plane-device .

[0044] In this embodiment of the application, the process of determining the first local planar domain information corresponding to the first point cloud set and the second local planar domain information corresponding to the second point cloud set specifically includes: extracting the first point cloud physical feature parameters corresponding to the first point cloud set and the second point cloud physical feature parameters corresponding to the second point cloud set using a point cloud feature extraction algorithm; classifying the first point cloud set and the second point cloud set according to the first point cloud physical feature parameters, the second point cloud physical feature parameters and a clustering algorithm to obtain the first local planar domain information and the second local planar domain information.

[0045] It should be noted that, in practical applications, the point cloud feature extraction algorithm can be denoted as... Clustering algorithms can be denoted as: Point cloud feature extraction algorithm Clustering algorithms This can be understood as a point cloud feature extraction optimization function, which can be denoted as: Point cloud feature extraction algorithms are used to extract the physical feature parameters of the first point cloud set and the second point cloud set, respectively. This can be understood as constructing a mathematical model of how a physical quantity changes over time to obtain the physical feature parameters of the first and second point clouds. In practical applications, when the physical quantity is related to temperature, the constructed mathematical model includes U(t) = U0 + (U1 - U0)e kt Or U(t) = U0 + (U1 - U0)(1 - e kt The mathematical model is defined as follows: U(t) = U0 + U1, where U(t) is the physical quantity corresponding to a monotonic interval of the time series point cloud, t is the time variable, k is the physical characteristic parameter corresponding to each point, and U0 and U1 are the initial or final values ​​of the physical quantity corresponding to a monotonic interval of the time series point cloud, respectively. The physical characteristic parameters of the point cloud can be obtained through optimization algorithms such as linear regression or least squares. When the physical quantity is independent of temperature, the constructed mathematical model includes U(t) = U0 + U1 + U2 + U3 + U4 + U5 + U6 + U7 + U8 + U9 + U1 k +k1t+k2t 2 +k3t 3 Where U(t) is the physical quantity corresponding to a monotonic interval of the time series point set, t is the time variable, and k1~k3 are the physical characteristic parameters corresponding to each point. k The offset is used; the physical feature parameters of the point cloud can be obtained through optimization algorithms such as the least squares method.

[0046] It should be noted that the clustering algorithm can be any clustering algorithm, and no limitation is made here. As an example, the clustering algorithm includes the KMeans clustering algorithm and the Euclidean clustering algorithm based on a distance threshold. Based on the physical feature parameters of the first point cloud, the physical feature parameters of the second point cloud, and the clustering algorithm, the first and second point clouds are classified respectively to obtain the first and second local planar domain information. This can be understood as follows: based on the physical feature parameters of the first and second point clouds, the KMeans clustering algorithm is used to perform preliminary classification of the first and second point clouds, obtaining local regions with similar physical characteristics in spatial distribution between the system-level simplified model and the device-level detailed model. These local regions may simultaneously contain planar and non-planar regions. Further secondary classification is performed using the Euclidean clustering algorithm based on a distance threshold, ultimately obtaining the first and second local planar domain information with significant planar features.

[0047] The solution in this application embodiment uses clustering algorithms (including KMeans clustering algorithm and Euclidean clustering algorithm based on distance threshold) to perform preliminary and secondary classification of the first point cloud and the second point cloud, thereby achieving hierarchical processing from coarse segmentation to fine segmentation.

[0048] S103. Determine the first extended transformation matrix corresponding to the first point cloud and the second extended transformation matrix corresponding to the second point cloud based on the first local planar domain information and the second local planar domain information, respectively.

[0049] It should be noted that, in practical applications, the first extended transformation matrix can be denoted as M. system The second extended transformation matrix can be denoted as M. device .

[0050] In this embodiment of the application, the process of determining the first translation vector corresponding to the first point set and the second translation vector corresponding to the second point set based on the first local plane domain information and the second local plane domain information specifically includes: constructing a first primary rotation matrix corresponding to the first point set and a second primary rotation matrix corresponding to the second point set based on the first local plane domain information and the second local plane domain information; using the first primary rotation matrix and the second primary rotation matrix to obtain the first primary rotation local plane domain information corresponding to the first local plane domain information and the second primary rotation local plane domain information corresponding to the second local plane domain information; and constructing a first primary rotation local plane domain matrix and a second primary rotation local plane domain matrix respectively based on the first primary rotation local plane domain information and the second primary rotation local plane domain information. Construct the first-order rotation matrix corresponding to the first point set and the second-order rotation matrix corresponding to the second point set; use the first-order rotation matrix and the second-order rotation matrix to obtain the first-order rotation local plane domain information corresponding to the first primary rotation local plane domain information and the second-order rotation local plane domain information corresponding to the second primary rotation local plane domain information, respectively; determine the first translation vector and the second translation vector based on the first-order rotation local plane domain information and the second-order rotation local plane domain information; determine the first extended transformation matrix and the second extended transformation matrix according to the first primary rotation matrix, the second primary rotation matrix, the first-order rotation matrix, the second-order rotation matrix, the first translation vector, and the second translation vector.

[0051] It should be noted that the first primary rotation matrix can be understood as the first preliminary rotation matrix; the second primary rotation matrix can be understood as the second preliminary rotation matrix. In practical applications, the first primary rotation matrix can be denoted as R. system1 The second primary rotation matrix can be denoted as R. device1 .

[0052] In this embodiment, the process of constructing a first primary rotation matrix corresponding to a first point cloud and a second primary rotation matrix corresponding to a second point cloud based on first local planar domain information and second local planar domain information specifically includes: constructing a first normal vector of the point cloud transformation reference plane; using a first dimensionality reduction algorithm to obtain multiple first eigenvalues ​​and multiple first eigenvectors corresponding to the first local planar domain information, and multiple second eigenvalues ​​and second eigenvectors corresponding to the second local planar domain information; searching for the smallest first target eigenvalue from the multiple first eigenvalues, and searching for the first target eigenvector corresponding to the first target eigenvalue from the multiple first eigenvectors; searching for the smallest second target eigenvalue from the multiple second eigenvalues, and searching for the second target eigenvector corresponding to the second target eigenvalue from the multiple second eigenvectors; determining the first rotation angle information and the first rotation axis information between the first target eigenvector and the first normal vector, and the second rotation angle information and the second rotation axis information between the second target eigenvector and the first normal vector; and constructing the first primary rotation matrix and the second primary rotation matrix using the first rotation angle information, the first rotation axis information, the second rotation angle information, and the second rotation axis information.

[0053] It should be noted that the reference plane for point cloud transformation includes the first normal vector and the first tangent vector; the first normal vector is the normal vector of the reference plane, and in practical applications, the first normal vector can be denoted as Vector. ref1 .

[0054] It should be noted that the first dimensionality reduction algorithm can be understood as any dimensionality reduction algorithm, and no limitation is made here. As an example, the first dimensionality reduction algorithm can be the Principal Components Analysis (PCA) algorithm. Using the first dimensionality reduction algorithm to obtain multiple first eigenvalues ​​and multiple first eigenvectors corresponding to the first local planar domain information, and multiple second eigenvalues ​​and multiple second eigenvectors corresponding to the second local planar domain information, can be understood as calculating multiple first eigenvalues ​​and multiple first eigenvectors corresponding to the first local planar domain information, and multiple second eigenvalues ​​and multiple second eigenvectors corresponding to the second local planar domain information, using the point cloud normal vector calculation function and the first dimensionality reduction algorithm. The point cloud normal vector calculation function is...

[0055] It should be noted that the first target eigenvalue is the smallest eigenvalue among multiple first eigenvalues, and the second target eigenvalue is the smallest eigenvalue among multiple second eigenvalues. The first target eigenvector is the vector corresponding to the first target eigenvalue, which can also be understood as the normal vector corresponding to the first target eigenvalue; the second target eigenvector is the vector corresponding to the second target eigenvalue, which can also be understood as the normal vector corresponding to the second target eigenvalue. In practical applications, the smallest eigenvalue can be denoted as k. minThe first target feature vector can be denoted as Vector. targ-system1 The second target feature vector can be denoted as Vector. targ-device1 .

[0056] It should be noted that the first rotation angle information and the first rotation axis information can be understood as the minimum rotation angle information and rotation axis information between the first target feature vector and the first normal vector; the second rotation angle information and the second rotation axis information can be understood as the minimum rotation angle information and rotation axis information between the second target feature vector and the first normal vector. Determining the first rotation angle information and the first rotation axis information between the first target feature vector and the first normal vector, and the second rotation angle information and the second rotation axis information between the second target feature vector and the first normal vector, can be understood as determining the first rotation angle information and the first rotation axis information based on the vector dot product and cross product operations between the first target feature vector and the first normal vector; and determining the second rotation angle information and the second rotation axis information based on the vector dot product and cross product operations between the second target feature vector and the first normal vector. In practical applications, the first rotation angle information can be denoted as angle. rotate-system1 The first axis of rotation information can be denoted as axis. rotate-system1 The second rotation angle information can be denoted as angle. rotate-dvice1 The second rotation axis information can be denoted as axis. rotate-dvice1 .

[0057] It should be noted that constructing the first and second primary rotation matrices using the first rotation angle information, the first rotation axis information, the second rotation angle information, and the second rotation axis information can be understood as using a point cloud rotation matrix calculation function and a Rodrigues rotation or quaternion transformation algorithm to construct the first and second primary rotation matrices based on the first rotation angle information, the first rotation axis information, the second rotation angle information, and the second rotation axis information. The point cloud rotation matrix calculation function can be denoted as...

[0058] In this embodiment, the first primary rotation matrix and the second primary rotation matrix are used to obtain the first primary rotated local plane domain information corresponding to the first local plane domain information, and the second primary rotated local plane domain information corresponding to the second local plane domain information, respectively. This can be understood as applying the first primary rotation matrix and the second primary rotation matrix to the first local plane domain information and the second local plane domain information to obtain the first primary rotated local plane domain information and the second primary rotated local plane domain information. In practical applications, the first primary rotated local plane domain information and the second primary rotated local plane domain information are parallel to the point cloud reference plane. The first primary rotated local plane domain information can be denoted as Zone. plane-system-rotate1 The second primary rotational local plane domain information can be denoted as Zone.plane-device-rotate1 .

[0059] In this embodiment, the process of constructing the first-level rotation matrix corresponding to the first point cloud and the second-level rotation matrix corresponding to the second point cloud based on the first primary rotation plane domain information and the second primary rotation plane domain information specifically includes: constructing the first tangent vector of the point cloud transformation reference plane; using a first dimensionality reduction algorithm to obtain multiple third eigenvalues ​​and multiple third eigenvectors corresponding to the first primary rotation local plane domain information, and multiple fourth eigenvalues ​​and multiple fourth eigenvectors corresponding to the second primary rotation local plane domain information; finding the third target eigenvalue with the largest value from the multiple third eigenvalues, and finding the third target eigenvector corresponding to the third target eigenvalue from the multiple third eigenvectors; finding the fourth target eigenvalue with the largest value from the multiple fourth eigenvalues, and finding the fourth target eigenvector corresponding to the fourth target eigenvalue from the multiple fourth eigenvectors; determining the third rotation angle information and the third rotation axis information between the third target eigenvector and the first tangent vector, and the fourth rotation angle information and the fourth rotation axis information between the fourth target eigenvector and the first tangent vector; and constructing the first-level rotation matrix and the second-level rotation matrix using the third rotation angle information, the third rotation axis information, the fourth rotation angle information, and the fourth rotation axis information.

[0060] It should be noted that the point cloud transformation reference plane includes the first normal vector and the first tangent vector; the first tangent vector is any basis vector in the reference plane. In practical applications, the first tangent vector can be denoted as Vector. ref2 .

[0061] It should be noted that the first dimensionality reduction algorithm can be understood as any dimensionality reduction algorithm, and no limitation is made here. As an example, the first dimensionality reduction algorithm can be the Principal Component Analysis (PCA) algorithm. Using the first dimensionality reduction algorithm to obtain multiple third eigenvalues ​​and multiple third eigenvectors corresponding to the first primary local rotating plane domain information, and multiple fourth eigenvalues ​​and multiple fourth eigenvectors corresponding to the second primary local rotating plane domain information, can be understood as calculating multiple third eigenvalues ​​and multiple third eigenvectors corresponding to the first primary local rotating plane domain information, and multiple fourth eigenvalues ​​and multiple fourth eigenvectors corresponding to the second primary local rotating plane domain information, using the point cloud tangent vector calculation function and the first dimensionality reduction algorithm. The point cloud tangent vector calculation function is as follows:

[0062] It should be noted that the third target eigenvalue is the largest eigenvalue among multiple third eigenvalues, and the fourth target eigenvalue is the largest eigenvalue among multiple fourth eigenvalues. The third target eigenvector is the vector corresponding to the third target eigenvalue, which can also be understood as the tangent vector corresponding to the third target eigenvalue; the fourth target eigenvector is the vector corresponding to the fourth target eigenvalue, which can also be understood as the tangent vector corresponding to the fourth target eigenvalue. In practical applications, the largest eigenvalue can be denoted as k. max The third target feature vector can be denoted as Vector. targ-system2 The fourth target feature vector can be denoted as Vector. targ-device2 .

[0063] It should be noted that the third rotation angle information and the third rotation axis information can be understood as the minimum rotation angle information and rotation axis information between the third target feature vector and the first tangent vector; the fourth rotation angle information and the fourth rotation axis information can be understood as the minimum rotation angle information and rotation axis information between the fourth target feature vector and the first tangent vector. Determining the third rotation angle information and the third rotation axis information between the third target feature vector and the first tangent vector, and the fourth rotation angle information and the fourth rotation axis information between the fourth target feature vector and the first tangent vector, can be understood as determining the third rotation angle information and the third rotation axis information based on the vector dot product and cross product operations between the third target feature vector and the first tangent vector; and determining the fourth rotation angle information and the fourth rotation axis information based on the vector dot product and cross product operations between the fourth target feature vector and the first tangent vector. In practical applications, the third rotation angle information can be denoted as angle. rotate-system2 The third axis of rotation information can be denoted as axis. rotate-system2 The fourth rotation angle information can be denoted as angle. rotate-dvice2 The fourth rotation axis information can be denoted as axis. rotate-dvice2 .

[0064] It should be noted that constructing the first-order and second-order rotation matrices using the third rotation angle information, third rotation axis information, fourth rotation angle information, and fourth rotation axis information can be understood as using a point cloud rotation matrix calculation function and a Rodriguez rotation or quaternion transformation algorithm to construct the first-order and second-order rotation matrices based on the third rotation angle information, third rotation axis information, fourth rotation angle information, and fourth rotation axis information. The point cloud rotation matrix calculation function can be denoted as...

[0065] In this embodiment, the first-level rotation matrix and the second-level rotation matrix are used to obtain the first-level local plane rotation information corresponding to the first primary rotation local plane information, and the second-level local plane rotation information corresponding to the second primary rotation local plane information, respectively. This can be understood as applying the first-level rotation matrix and the second-level rotation matrix to the first primary rotation local plane information and the second primary rotation local plane information to obtain the first-level rotation local plane information and the second-level rotation local plane information. In practical applications, the first-level rotation local plane information and the second-level rotation local plane information are parallel to the tangent vector of the point cloud reference plane. The first-level rotation local plane information can be denoted as Zone. plane-system-rotate2 The second-order rotated local planar domain information can be denoted as Zone. plane-device-rotate2 .

[0066] In this embodiment, the first translation vector can be denoted as T. system The second translation vector can be denoted as T. device The process of determining the first translation vector and the second translation vector based on the first-level and second-level local plane domain information specifically includes: determining the first centroid information corresponding to the first-level local plane domain information and the second centroid information corresponding to the second-level local plane domain information; and determining the first translation vector and the second translation vector based on the first centroid information and the second centroid information.

[0067] It should be noted that the first centroid information corresponding to the first-level local plane rotation information and the second centroid information corresponding to the second-level local plane rotation information are determined respectively; the first translation vector and the second translation vector are determined based on the first centroid information and the second centroid information. This can be understood as calculating the centroids of the first-level local plane rotation information and the second-level local plane rotation information to obtain the first centroid information and the second centroid information; and determining the first translation vector and the second translation vector based on the first centroid information and the second centroid information.

[0068] In this embodiment, the first extended transformation matrix and the second extended transformation matrix are determined based on the first primary rotation matrix, the second primary rotation matrix, the first intermediate rotation matrix, the second intermediate rotation matrix, the first translation vector, and the second translation vector. This can be understood as follows: the first extended transformation matrix of the system-level model point cloud is constructed based on the first translation vector, the first primary rotation matrix, and the first intermediate rotation matrix; and the second extended transformation matrix of the device-level model point cloud is constructed based on the second translation vector, the second primary rotation matrix, and the second intermediate rotation matrix.

[0069] S104. Based on the first extended transformation matrix and the second extended transformation matrix, perform pose transformation on the first point cloud and the second point cloud respectively to obtain the third point cloud after pose transformation of the first point cloud and the fourth point cloud after pose transformation of the second point cloud.

[0070] It should be noted that the third point cloud is the point cloud obtained by transforming the pose of the first point cloud; the fourth point cloud is the point cloud obtained by transforming the pose of the second point cloud. In practical applications, the third point cloud can be denoted as q. system The fourth point, gathering, can be denoted as q. device .

[0071] It should be noted that the pose transformations of the first and second point clouds based on the first and second extended transformation matrices, respectively, yield a third point cloud after the pose transformation of the first point cloud and a fourth point cloud after the pose transformation of the second point cloud. This can be understood as follows: the pose transformation of the first point cloud using the first extended transformation matrix yields the third point cloud; and the pose transformation of the second point cloud using the second extended transformation matrix yields the fourth point cloud. Specifically, the pose transformation of the first point cloud using the first extended transformation matrix to obtain the third point cloud can be illustrated by converting each point in the first point cloud into a point cloud in homogeneous coordinates, and then applying the first extended transformation matrix to the point cloud after homogeneous coordinate expansion. Similarly, the pose transformation of the second point cloud using the second extended transformation matrix to obtain the fourth point cloud can be illustrated by converting each point in the second point cloud into a point cloud in homogeneous coordinates, and then applying the second extended transformation matrix to the point cloud after homogeneous coordinate expansion.

[0072] S105. Based on the third and fourth point cloud sets, perform data matching on the objects to be matched at the system and device levels.

[0073] It should be noted that data matching of the traction system monitoring components is performed at both the system and device levels based on the third and fourth points of the cloud collection.

[0074] The method in this application embodiment achieves precise matching of a multi-level digital twin model of the traction system by transforming the pose of the first point cloud and the second point cloud, and performs a comprehensive assessment and analysis of the health status of the monitoring devices.

[0075] To facilitate understanding, specific examples are provided here to illustrate the data matching methods described above.

[0076] 1. Establish system-level and device-level full-range field visualization digital twin models for monitoring components of traction systems in rail transit, urban rail transit, and other fields. Extract the three-dimensional physical field time-series point cloud g(x,y,z,u) of the simplified system-level model and detailed device-level model of the monitoring components of the traction system under the same operating conditions. syetem and g(x,y,z,u) device Where g(x,y,z,u) syetem and g(x,y,z,u) device The density and initial pose of the points are different, but the physical field u of the point cloud is similar. The physical field u includes, but is not limited to, temperature field, stress field, flow field, etc. Figure 3 This application provides an exemplary point cloud diagram of a model of an object to be matched at both the system and device levels; for example... Figure 3 As shown, Figure 3 (a) Schematic diagram of the point cloud of the model at the system level. Figure 3 (b) is a schematic diagram of the point cloud of the model at the device level.

[0077] 2. Establish an optimization function for point cloud feature extraction. By combining point cloud physical feature extraction algorithms with clustering algorithms, the above g(x,y,z,u) is obtained. syetem and g(x,y,z,u) device Local planar regions Zones that have similar physical characteristics in spatial distribution and satisfy geometric constraints plane-system and Zone plane-device Specifically, this can include: optimizing the point cloud feature extraction function. middle:

[0078] First, through point cloud physical feature extraction algorithm A mathematical model of how physical quantities change over time is established to obtain the physical feature parameters of the point cloud, where:

[0079] When the physical quantity is related to temperature, the mathematical model U(t) = U0 + (U1 - U0)e is constructed. kt Or U(t) = U0 + (U1 - U0)(1 - e kt In this context, U(t) represents the physical quantity corresponding to a monotonic interval of the time series set, t is the time variable, k is the physical characteristic parameter corresponding to each point, and U0 and U1 are the initial or final values ​​of the physical quantity corresponding to a monotonic interval of the time series set, respectively. When the monotonic interval of the time series set is a decreasing interval, U(t) = U0 + (U1 - U0)e kt In the mathematical model, U0 represents the final value of the physical quantity corresponding to the decreasing interval of the time series set, and U1 represents the initial value of the physical quantity corresponding to the decreasing interval of the time series set. When a monotonic interval of the time series set is an increasing interval, U(t) = U0 + (U1 - U0)(1 - e^(-t / t)) is used.kt The mathematical model is as follows: U0 is the initial value of the physical quantity corresponding to the increasing interval of the time series point cloud, and U1 is the final value of the physical quantity corresponding to the increasing interval of the time series point cloud. The physical feature parameters of the point cloud can be obtained through optimization algorithms such as linear regression or least squares.

[0080] When the physical quantity is independent of temperature, a polynomial regression mathematical model U(t) = U is constructed. k +k1t+k2t 2 +k3t 3 Where U(t) is the physical quantity corresponding to a monotonic interval of the time series point set, t is the time variable, and k1~k3 are the physical characteristic parameters corresponding to each point. k The offset is used; the physical feature parameters of the point cloud can be obtained through optimization algorithms such as the least squares method.

[0081] Secondly, based on the physical characteristic parameters obtained above, a clustering algorithm is used. To perform similar feature segmentation, the KMeans clustering algorithm is first used to segment the three-dimensional physical field temporal point clouds g(x,y,z,u) of the system-level simplified model and the device-level detailed model. syetem and g(x,y,z,u) device Preliminary classification identifies local regions with similar physical characteristics in spatial distribution between the system-level simplified model and the device-level detailed model. These local regions may simultaneously contain planar and non-planar regions. Further secondary classification using a Euclidean clustering algorithm based on distance thresholds ultimately yields local planar regions (Zones) with significant planar features. plane-system and Zone plane-device This achieves hierarchical processing from coarse segmentation to fine segmentation. Figure 4 A schematic diagram of an exemplary system-level and device-level local planar region is provided for embodiments of this application; as shown below. Figure 4 As shown, Figure 4 (a) is a schematic diagram of a local planar domain at the system level (i.e., system level); Figure 4 (b) is a schematic diagram of a local planar region at the device level (i.e., device level).

[0082] 3. Construct the point cloud plane normal vector alignment transformation function The initial rotation matrix R of the above system-level model and device-level model is obtained. system1 and R device1 To realize the local planar region Zone plane-system and Zone plane-device Parallel to the pose of the point cloud reference plane.

[0083] Specifically, this can include: aligning the transformation function with the normal vectors of the point cloud plane. middle:

[0084] First, establish a reference plane for point cloud transformation, where Vector ref1 Vector is the normal vector of the reference plane. ref2 Let be any basis vector (tangent vector) in the reference plane;

[0085] Secondly, the function is calculated using the point cloud normal vectors. Using the principal component analysis (PCA) algorithm, the local planar regions corresponding to the above system-level model and device-level model are calculated. plane-system and Zone plane-device The smallest eigenvalue k min The corresponding vector (normal vector) targ-system1 and Vector targ-device1 ;

[0086] Then, the Vector is determined through vector dot product and cross product operations. targ-system1 and Vector ref1 Minimum rotation angle and rotation axis between rotate-system1 and axis rotate-system1 and Vector targ-dvice1 and Vector ref1 Minimum rotation angle and rotation axis between rotate-dvice1 and axis rotate-dvice1 Based on the minimum rotation angle and rotation axis, a function is calculated using the point cloud rotation matrix. The initial rotation matrix R is constructed using the Rodriguez rotation or quaternion transformation algorithm to create the system-level and device-level models. system1 and R device1 ;

[0087] Finally, R system1 and R device1 Acts on Zone plane-system and Zone plane-device Generate the aligned rotated local plane region Zone plane-system-rotate1 and Zone plane-device-rotate1 . Figure 5 This application provides an exemplary schematic diagram of a primary rotating local plane domain at the system and device levels for a target object; as shown in the embodiments of this application. Figure 5 As shown, Figure 5 (a) is a schematic diagram of the primary rotational local plane domain of the system-level object to be matched; Figure 5 (b) is a schematic diagram of the primary rotating local plane domain of the object to be matched at the device level. Figure 6 This application provides an exemplary schematic diagram of the initial rotational transformation of an object to be matched, as shown in the embodiments of this application. Figure 6 As shown.

[0088] 4. Construct the point cloud planar tangent vector alignment transformation function The final rotation matrix R of the above system-level model and device-level model is obtained. system2 and R device2 To realize the rotation of the local plane region Zone plane-system-rotate1 and Zone plane-device-rotate1 The pose is parallel to the tangent vector of the point cloud reference plane.

[0089] Specifically, this can include: a transformation function for aligning tangent vectors in the point cloud plane. middle:

[0090] First, calculate the function using the point cloud tangent vector. Using the principal component analysis (PCA) algorithm, the local planar regions corresponding to the above system-level model and device-level model are calculated. plane-system-rotate1 and Zone plane-device-rotate1 The largest eigenvalue k max The corresponding vector (tangent vector) targ-system2 and Vector targ-device2 ;

[0091] Then, the Vector is determined through vector dot product and cross product operations. targ-system2 And the reference plane tangent vector Vector ref2 Minimum rotation angle and rotation axis between rotate-system2 and axis rotate-system2 and Vector targ-dvice2 and Vector ref2 Minimum rotation angle and rotation axis between rotate-dvice2 and axis rotate-dvice2 Based on the minimum rotation angle and rotation axis, a function is calculated using the point cloud rotation matrix. The final rotation matrix R of the system-level model and device-level model is constructed using the Rodriguez rotation or quaternion transformation algorithm. system2 and R device2 ;

[0092] Finally, R system2 and R device2 Acts on Zone plane-system-rotate1 and Zone plane-device-rotate1 Generate the aligned rotated local plane region Zone plane-system-rotate2 and Zone plane-device-rotate2 . Figure 7 This application provides an exemplary schematic diagram of a system-level and device-level secondary rotational local planar region of a target object; as shown in the embodiments of this application. Figure 7 As shown, Figure 7 (a) is a schematic diagram of the system-level secondary rotational local planar domain of the object to be matched; Figure 7 (b) is a schematic diagram of the device-level secondary rotational local planar domain of the object to be matched. Figure 8 This is a schematic diagram of a secondary rotation transformation of an object to be matched, provided in an embodiment of this application. Figure 8 As shown.

[0093] 5. The Zone of Rotation Plane for the above system-level and device-level models. plane-system-rotate2 and Zone plane-device-rotate2 The centroid of the cloud is used to determine the translation vector T. system and T device .

[0094] 6. Construct point cloud transformation functions Obtain the extended transformation matrix M of the above system-level model and device-level model. system and M device This enables the matching and alignment of point clouds between system-level and device-level models.

[0095] Specifically, this can include: in point cloud transformation functions First, based on the rotation matrix R of the above system-level model... system1 R system2 Translation vector T system Construct the extended transformation matrix M of the point cloud of the system-level model. system =[R 11 ,R 12 ,R 13 ;R 21 ,R 22 ,R 23 ;R 31 ,R 32 ,R 33 ;T x ,T y ,T z The matrix dimension is 4×3, where the composite rotation matrix R is composed of R... system1 and R system2 Composed by matrix multiplication, R = R system1 ·R system2 Secondly, each point p in the system-level model point cloud... i-system (x i ,y i ,z i Convert p to homogeneous coordinate form i-system '(x i ,y i ,z i ,1); Finally, the extended transformation matrix M system Acting on the point cloud p after homogeneous coordinate expansion system ', thus obtaining the transformed system-level model point cloud q system =Msystem ·p system '.

[0096] Similarly, the extended transformation matrix M of the point cloud of the device-level model is constructed. device Perform extended coordinate homogeneous transformation p device ', thus obtaining the transformed device-level model point cloud q device =M device ·p device '. Figure 9 A schematic diagram of the point cloud corresponding to the final pose of the object to be matched at the system and device levels, provided as an embodiment of this application; Figure 9 As shown, Figure 9 (a) is a schematic diagram of the point cloud corresponding to the final pose of the object to be matched at the system level; Figure 9 (b) is a schematic diagram of the point cloud corresponding to the final pose of the object to be matched at the device level.

[0097] 7. Based on steps 2 to 6, it is possible to quickly and accurately match point clouds of monitoring models at the system level and device level with different densities and initial poses, providing technical support for hierarchical status monitoring of system-level product digital twins.

[0098] This application provides a data matching device. Figure 10 This is a schematic diagram of the structure of a data matching device provided in an embodiment of this application; as shown below. Figure 10 As shown, the data matching device 1000 includes:

[0099] The acquisition unit 1001 is used to acquire the first point cloud at the system level and the second point cloud at the device level of the object to be matched using the digital twin model of the traction system; the point cloud distribution density and pose information of the first point cloud and the second point cloud are different.

[0100] The first determining unit 1002 is used to determine the first local plane domain information corresponding to the first point set and the second local plane domain information corresponding to the second point set, respectively.

[0101] The second determining unit 1003 is used to determine the first extended transformation matrix corresponding to the first point cloud and the second extended transformation matrix corresponding to the second point cloud based on the first local plane domain information and the second local plane domain information, respectively.

[0102] Pose transformation unit 1004 is used to perform pose transformation on the first point cloud and the second point cloud based on the first extended transformation matrix and the second extended transformation matrix respectively, to obtain a third point cloud after pose transformation of the first point cloud and a fourth point cloud after pose transformation of the second point cloud.

[0103] The data matching unit 1005 is used to perform data matching on the object to be matched at the system level and the device level based on the third point cloud and the fourth point cloud.

[0104] Optionally, the second determining unit 1003 is further configured to: construct a first primary rotation matrix corresponding to the first point cloud and a second primary rotation matrix corresponding to the second point cloud based on the first local plane domain information and the second local plane domain information, respectively; obtain the first primary rotation local plane domain information corresponding to the first local plane domain information and the second primary rotation local plane domain information corresponding to the second local plane domain information using the first primary rotation local plane domain information and the second primary rotation local plane domain information, respectively; construct a first-order rotation matrix corresponding to the first point cloud and a second-order rotation matrix corresponding to the second point cloud based on the first primary rotation local plane domain information and the second primary rotation local plane domain information, respectively; obtain the first-order rotation local plane domain information corresponding to the first primary rotation local plane domain information and the second-order rotation local plane domain information corresponding to the second primary rotation local plane domain information, respectively using the first-order rotation matrix and the second-order rotation matrix; determine a first translation vector and a second translation vector based on the first-order rotation local plane domain information and the second-order rotation local plane domain information; and determine the first extended transformation matrix and the second extended transformation matrix based on the first primary rotation matrix, the second primary rotation matrix, the first-order rotation matrix, the second-order rotation matrix, the first translation vector, and the second translation vector.

[0105] Optionally, the first determining unit 1002 is further configured to extract the first point cloud physical feature parameters corresponding to the first point cloud set and the second point cloud physical feature parameters corresponding to the second point cloud set using a point cloud feature extraction algorithm; and to classify the first point cloud set and the second point cloud set according to the first point cloud physical feature parameters, the second point cloud physical feature parameters and a clustering algorithm to obtain the first local planar domain information and the second local planar domain information.

[0106] Optionally, the second determining unit 1003 is further configured to: construct a first normal vector of the point cloud transformation reference plane; use a first dimensionality reduction algorithm to obtain multiple first eigenvalues ​​and multiple first eigenvectors corresponding to the first local planar domain information, and multiple second eigenvalues ​​and second eigenvectors corresponding to the second local planar domain information; search for the first target eigenvalue with the smallest value from the multiple first eigenvalues, and search for the first target eigenvector corresponding to the first target eigenvalue from the multiple first eigenvectors; search for the second target eigenvalue with the smallest value from the multiple second eigenvalues, and search for the second target eigenvector corresponding to the second target eigenvalue from the multiple second eigenvectors; determine the first rotation angle information and the first rotation axis information between the first target eigenvector and the first normal vector, and the second rotation angle information and the second rotation axis information between the second target eigenvector and the first normal vector; and construct the first primary rotation matrix and the second primary rotation matrix using the first rotation angle information, the first rotation axis information, the second rotation angle information, and the second rotation axis information.

[0107] Optionally, the second determining unit 1003 is further configured to: construct a first tangent vector of the point cloud transformation reference plane; obtain multiple third eigenvalues ​​and multiple third eigenvectors corresponding to the first primary rotational local plane domain information, and multiple fourth eigenvalues ​​and multiple fourth eigenvectors corresponding to the second primary rotational local plane domain information using a first dimensionality reduction algorithm; find the third target eigenvalue with the largest value from the multiple third eigenvalues, and find the third target eigenvector corresponding to the third target eigenvalue from the multiple third eigenvectors; find the fourth target eigenvalue with the largest value from the multiple fourth eigenvalues, and find the fourth target eigenvector corresponding to the fourth target eigenvalue from the multiple fourth eigenvectors; determine the third rotation angle information and the third rotation axis information between the third target eigenvector and the first tangent vector, and the fourth rotation angle information and the fourth rotation axis information between the fourth target eigenvector and the first tangent vector; and construct the first-level rotation matrix and the second-level rotation matrix using the third rotation angle information, the third rotation axis information, the fourth rotation angle information, and the fourth rotation axis information.

[0108] Optionally, the second determining unit 1003 is further configured to determine the first centroid information corresponding to the first-level rotational local plane domain information and the second centroid information corresponding to the second-level rotational local plane domain information; and to determine the first translation vector and the second translation vector based on the first centroid information and the second centroid information.

[0109] Optionally, the digital twin model of the traction system is obtained by sequentially connecting the vehicle dynamics model, the lumped parameter equivalent digital circuit model, and the multi-field reduced-order model.

[0110] This application also provides a data matching device; Figure 11 A schematic diagram of the structure of a data matching device provided in the application embodiment; as shown Figure 11 As shown, the data matching device 1100 includes a processor 1101 and a memory 1103. Optionally, the data matching device 1100 may also include a communication bus 1102.

[0111] In specific embodiments, the processor 1101 can be at least one of the following: Application Specific Integrated Circuit (ASIC), Digital Signal Processor (DSP), Digital Signal Processing Device (DSPD), Programmable Logic Device (PLD), Field Programmable Gate Array (FPGA), CPU, controller, microcontroller, and microprocessor. It is understood that for different devices, the electronic device used to implement the above processor function can also be other types, and this embodiment does not specifically limit it.

[0112] In this embodiment, the communication bus 1102 is used to realize the connection communication between the processor 1101 and the memory 1103; when the processor 1101 executes the running program stored in the memory 1103, it implements the following data matching method:

[0113] The first point cloud at the system level and the second point cloud at the device level are obtained using the digital twin model of the traction system; the point cloud distribution density and pose information of the first point cloud and the second point cloud are different.

[0114] Determine the first local planar domain information corresponding to the first point set and the second local planar domain information corresponding to the second point set, respectively;

[0115] Based on the first local planar domain information and the second local planar domain information, determine the first extended transformation matrix corresponding to the first point cloud and the second extended transformation matrix corresponding to the second point cloud, respectively.

[0116] Based on the first extended transformation matrix and the second extended transformation matrix, the pose transformations of the first point cloud and the second point cloud are performed respectively to obtain the third point cloud after the pose transformation of the first point cloud and the fourth point cloud after the pose transformation of the second point cloud.

[0117] The object to be matched is matched at the system level and the device level based on the third point cloud and the fourth point cloud.

[0118] Furthermore, the processor 1101 is also configured to: construct a first primary rotation matrix corresponding to the first point cloud and a second primary rotation matrix corresponding to the second point cloud based on the first local plane domain information and the second local plane domain information, respectively; obtain the first primary rotation local plane domain information corresponding to the first local plane domain information and the second primary rotation local plane domain information corresponding to the second local plane domain information using the first primary rotation local plane domain information and the second primary rotation local plane domain information, respectively; construct a first-order rotation matrix corresponding to the first point cloud and a second-order rotation matrix corresponding to the second point cloud based on the first primary rotation local plane domain information and the second primary rotation local plane domain information, respectively; obtain the first-order rotation local plane domain information corresponding to the first primary rotation local plane domain information and the second-order rotation local plane domain information corresponding to the second primary rotation local plane domain information, respectively using the first-order rotation matrix and the second-order rotation matrix; determine a first translation vector and a second translation vector based on the first-order rotation local plane domain information and the second-order rotation local plane domain information; and determine the first extended transformation matrix and the second extended transformation matrix based on the first primary rotation matrix, the second primary rotation matrix, the first-order rotation matrix, the second-order rotation matrix, the first translation vector, and the second translation vector.

[0119] Furthermore, the processor 1101 is also used to extract the first point cloud physical feature parameters corresponding to the first point cloud set and the second point cloud physical feature parameters corresponding to the second point cloud set using a point cloud feature extraction algorithm; and to classify the first point cloud set and the second point cloud set according to the first point cloud physical feature parameters, the second point cloud physical feature parameters and a clustering algorithm to obtain the first local planar domain information and the second local planar domain information.

[0120] Furthermore, the processor 1101 is also used to construct a first normal vector of the point cloud transformation reference plane; to obtain multiple first eigenvalues ​​and multiple first eigenvectors corresponding to the first local plane domain information, and multiple second eigenvalues ​​and second eigenvectors corresponding to the second local plane domain information using a first dimensionality reduction algorithm; to find the first target eigenvalue with the smallest value from the multiple first eigenvalues, and to find the first target eigenvector corresponding to the first target eigenvalue from the multiple first eigenvectors; to find the second target eigenvalue with the smallest value from the multiple second eigenvalues, and to find the second target eigenvector corresponding to the second target eigenvalue from the multiple second eigenvectors; to determine the first rotation angle information and the first rotation axis information between the first target eigenvector and the first normal vector, and the second rotation angle information and the second rotation axis information between the second target eigenvector and the first normal vector; and to construct the first primary rotation matrix and the second primary rotation matrix using the first rotation angle information, the first rotation axis information, the second rotation angle information, and the second rotation axis information.

[0121] Furthermore, the processor 1101 is also used to construct a first tangent vector of the point cloud transformation reference plane; to obtain multiple third eigenvalues ​​and multiple third eigenvectors corresponding to the first primary rotational local plane domain information, and multiple fourth eigenvalues ​​and multiple fourth eigenvectors corresponding to the second primary rotational local plane domain information using a first dimensionality reduction algorithm; to find the third target eigenvalue with the largest value from the multiple third eigenvalues, and to find the third target eigenvector corresponding to the third target eigenvalue from the multiple third eigenvectors; to find the fourth target eigenvalue with the largest value from the multiple fourth eigenvalues, and to find the fourth target eigenvector corresponding to the fourth target eigenvalue from the multiple fourth eigenvectors; to determine the third rotation angle information and the third rotation axis information between the third target eigenvector and the first tangent vector, and the fourth rotation angle information and the fourth rotation axis information between the fourth target eigenvector and the first tangent vector; and to construct the first-level rotation matrix and the second-level rotation matrix using the third rotation angle information, the third rotation axis information, the fourth rotation angle information, and the fourth rotation axis information.

[0122] Furthermore, the processor 1101 is also configured to determine the first centroid information corresponding to the first-level rotational local plane domain information and the second centroid information corresponding to the second-level rotational local plane domain information; and to determine the first translation vector and the second translation vector based on the first centroid information and the second centroid information.

[0123] Furthermore, the digital twin model of the traction system is obtained by sequentially connecting the vehicle dynamics model, the lumped parameter equivalent digital circuit model, and the multi-field reduced-order model.

[0124] This application provides a storage medium storing a computer program thereon. The computer-readable storage medium stores one or more programs, which can be executed by one or more processors. The computer program implements the data matching method described above.

[0125] Based on the above embodiments, this application provides a computer program product, including a computer program that can be executed by one or more processors, and the computer program implements the data matching method described above.

[0126] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0127] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this disclosure, in essence, or the part that contributes to the related technology, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes several instructions to cause an image display device (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of this disclosure.

[0128] The above description is merely a preferred embodiment of this application and is not intended to limit the scope of protection of this application.

Claims

1. A data matching method, characterized in that, The method includes: The first point cloud at the system level and the second point cloud at the device level are obtained using the digital twin model of the traction system; the point cloud distribution density and pose information of the first point cloud and the second point cloud are different. Determine the first local planar domain information corresponding to the first point set and the second local planar domain information corresponding to the second point set, respectively; Based on the first local planar domain information and the second local planar domain information, determine the first extended transformation matrix corresponding to the first point cloud and the second extended transformation matrix corresponding to the second point cloud, respectively. Based on the first extended transformation matrix and the second extended transformation matrix, the pose transformations of the first point cloud and the second point cloud are performed respectively to obtain the third point cloud after the pose transformation of the first point cloud and the fourth point cloud after the pose transformation of the second point cloud. Based on the third point cloud and the fourth point cloud, the object to be matched is matched at the system level and the device level. The step of determining the first extended transformation matrix corresponding to the first point cloud and the second extended transformation matrix corresponding to the second point cloud based on the first local plane domain information and the second local plane domain information respectively includes: Based on the first local planar domain information and the second local planar domain information, construct the first primary rotation matrix corresponding to the first point cloud and the second primary rotation matrix corresponding to the second point cloud, respectively. The first primary rotation matrix and the second primary rotation matrix are used to obtain the first primary rotation local plane domain information corresponding to the first local plane domain information and the second primary rotation local plane domain information corresponding to the second local plane domain information, respectively. Based on the first primary rotational local plane domain information and the second primary rotational local plane domain information, construct the first-level rotation matrix corresponding to the first point cloud and the second-level rotation matrix corresponding to the second point cloud, respectively. The first-level rotation matrix and the second-level rotation matrix are used to obtain the first-level rotation local plane domain information corresponding to the first primary rotation local plane domain information and the second-level rotation local plane domain information corresponding to the second primary rotation local plane domain information, respectively. The first translation vector and the second translation vector are determined based on the first-level rotational local plane domain information and the second-level rotational local plane domain information. The first extended transformation matrix and the second extended transformation matrix are determined based on the first primary rotation matrix, the second primary rotation matrix, the first first-level rotation matrix, the second-level rotation matrix, the first translation vector, and the second translation vector. The step of constructing the first primary rotation matrix corresponding to the first point cloud and the second primary rotation matrix corresponding to the second point cloud based on the first local plane domain information and the second local plane domain information respectively includes: Construct the first normal vector of the point cloud transformation reference plane; The first dimensionality reduction algorithm is used to obtain multiple first feature values ​​and multiple first feature vectors corresponding to the first local planar domain information, as well as multiple second feature values ​​and second feature vectors corresponding to the second local planar domain information. Find the first target feature value with the smallest value from the plurality of first feature values, and find the first target feature vector corresponding to the first target feature value from the plurality of first feature vectors; find the second target feature value with the smallest value from the plurality of second feature values, and find the second target feature vector corresponding to the second target feature value from the plurality of second feature vectors; Determine the first rotation angle information and the first rotation axis information between the first target feature vector and the first normal vector, and the second rotation angle information and the second rotation axis information between the second target feature vector and the first normal vector; The first primary rotation matrix and the second primary rotation matrix are constructed using the first rotation angle information, the first rotation axis information, the second rotation angle information, and the second rotation axis information.

2. The method according to claim 1, characterized in that, The step of determining the first local plane domain information corresponding to the first point cloud and the second local plane domain information corresponding to the second point cloud includes: The physical feature parameters of the first point cloud corresponding to the first point cloud set and the physical feature parameters of the second point cloud corresponding to the second point cloud set are extracted using a point cloud feature extraction algorithm. The first point cloud set and the second point cloud set are classified according to the physical feature parameters of the first point cloud set, the physical feature parameters of the second point cloud set, and the clustering algorithm, respectively, to obtain the first local planar domain information and the second local planar domain information.

3. The method according to claim 1, characterized in that, The step of constructing the first-order rotation matrix corresponding to the first point cloud and the second-order rotation matrix corresponding to the second point cloud based on the first primary rotation local plane domain information and the second primary rotation local plane domain information respectively includes: Construct the first tangent vector of the point cloud transformation reference plane; The first dimensionality reduction algorithm is used to obtain multiple third feature values ​​and multiple third feature vectors corresponding to the first primary rotational local plane domain information, as well as multiple fourth feature values ​​and multiple fourth feature vectors corresponding to the second primary rotational local plane domain information. Find the third target feature value with the largest value from the plurality of third feature values, and find the third target feature vector corresponding to the third target feature value from the plurality of third feature vectors; find the fourth target feature value with the largest value from the plurality of fourth feature values, and find the fourth target feature vector corresponding to the fourth target feature value from the plurality of fourth feature vectors; Determine the third rotation angle information and the third rotation axis information between the third target feature vector and the first tangent vector, and the fourth rotation angle information and the fourth rotation axis information between the fourth target feature vector and the first tangent vector; The first-level rotation matrix and the second-level rotation matrix are constructed using the third rotation angle information, the third rotation axis information, the fourth rotation angle information, and the fourth rotation axis information.

4. The method according to claim 1, characterized in that, The step of determining the first translation vector and the second translation vector based on the first-level rotational local plane domain information and the second-level rotational local plane domain information includes: The first centroid information corresponding to the first-level rotated local plane domain information and the second centroid information corresponding to the second-level rotated local plane domain information are determined respectively. The first translation vector and the second translation vector are determined based on the first centroid information and the second centroid information.

5. The method according to claim 1, characterized in that, The digital twin model of the traction system is obtained by sequentially connecting the vehicle dynamics model, the lumped parameter equivalent digital circuit model, and the multi-field reduced-order model.

6. A data matching device, characterized in that, The device includes: The acquisition unit is used to acquire the first point cloud at the system level and the second point cloud at the device level of the object to be matched using the digital twin model of the traction system; the point cloud distribution density and pose information of the first point cloud and the second point cloud are different. The first determining unit is used to determine the first local planar domain information corresponding to the first point set and the second local planar domain information corresponding to the second point set, respectively. The second determining unit is used to determine the first extended transformation matrix corresponding to the first point cloud and the second extended transformation matrix corresponding to the second point cloud based on the first local plane domain information and the second local plane domain information, respectively. The pose transformation unit is used to perform pose transformation on the first point cloud and the second point cloud based on the first extended transformation matrix and the second extended transformation matrix respectively, to obtain the third point cloud after the pose transformation of the first point cloud and the fourth point cloud after the pose transformation of the second point cloud. A data matching unit is used to perform data matching on the object to be matched at the system level and the device level based on the third point cloud and the fourth point cloud. The second determining unit is further configured to: construct a first primary rotation matrix corresponding to the first point set and a second primary rotation matrix corresponding to the second point set based on the first local plane domain information and the second local plane domain information, respectively; obtain first primary rotation local plane domain information corresponding to the first local plane domain information and second primary rotation local plane domain information corresponding to the second local plane domain information using the first primary rotation local plane domain information and the second primary rotation local plane domain information, respectively; construct a first-order rotation matrix corresponding to the first point set and a second-order rotation matrix corresponding to the second point set based on the first primary rotation local plane domain information and the second primary rotation local plane domain information, respectively; obtain first-order rotation local plane domain information corresponding to the first primary rotation local plane domain information and second-order rotation local plane domain information corresponding to the second primary rotation local plane domain information using the first-order rotation matrix and the second-order rotation matrix, respectively; determine a first translation vector and a second translation vector based on the first-order rotation local plane domain information and the second-order rotation local plane domain information; and determine the first extended transformation matrix and the second extended transformation matrix based on the first primary rotation matrix, the second primary rotation matrix, the first-order rotation matrix, the second-order rotation matrix, the first translation vector, and the second translation vector. The second determining unit is further configured to: construct a first normal vector of the point cloud transformation reference plane; use a first dimensionality reduction algorithm to obtain multiple first eigenvalues ​​and multiple first eigenvectors corresponding to the first local planar domain information, and multiple second eigenvalues ​​and second eigenvectors corresponding to the second local planar domain information; search for the first target eigenvalue with the smallest value from the multiple first eigenvalues, and search for the first target eigenvector corresponding to the first target eigenvalue from the multiple first eigenvectors; search for the second target eigenvalue with the smallest value from the multiple second eigenvalues, and search for the second target eigenvector corresponding to the second target eigenvalue from the multiple second eigenvectors; determine the first rotation angle information and the first rotation axis information between the first target eigenvector and the first normal vector, and the second rotation angle information and the second rotation axis information between the second target eigenvector and the first normal vector; and construct the first primary rotation matrix and the second primary rotation matrix using the first rotation angle information, the first rotation axis information, the second rotation angle information, and the second rotation axis information.

7. A data matching device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements the steps of the method according to any one of claims 1 to 5.

8. A storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the steps of the method according to any one of claims 1 to 5.

9. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the steps of the method according to any one of claims 1 to 5.

Images