A rigid body pose calibration method, system, device and medium

Abstract

This application provides a rigid body pose calibration method, system, device, and medium, relating to the fields of computer vision and precision measurement technology. It acquires the mechanical installation information of a rigid body calibration object and locks the object based on this information, obtaining the locking spindle direction information and mechanical origin information. Then, based on these information, it projects the point cloud information of the rigid body calibration object to obtain its corresponding projected coordinate information. Finally, based on the projected coordinate information and the preset target pose information, it performs pose calibration processing to obtain the pose calibration result. This solves the problems of low calibration accuracy, poor repeatability, and slow convergence speed caused by rotation center drift and degree-of-freedom coupling in existing technologies, improving calibration accuracy and stability.

Description

Technical Field

[0001] This application relates to the fields of computer vision and precision measurement technology, and more specifically, to a rigid body pose calibration method, system, device, and medium. Background Technology

[0002] In industrial automation, robot hand-eye calibration, and precision assembly, rigid body pose calibration is a core step in ensuring high-precision alignment of equipment. Its accuracy directly determines the subsequent assembly quality and system operational stability. Existing technologies typically employ general iterative nearest-point algorithms or point cloud registration algorithms based on singular value decomposition for reverse calibration calculations. When dealing with objects with physical constraints or rigid connections, these methods often assume the object is a free rigid body, setting the rotation center as the point cloud centroid, and without applying effective constraints in any of the six degrees of freedom. Due to limited sensor field of view or object structure occlusion in real-world industrial scenarios, the point cloud data acquired through scanning is often incomplete. In this case, the point cloud centroid fluctuates significantly with changes in the scanning range. Existing general algorithms perform rotation calculations around this fluctuating centroid, causing the rotation center to drift and resulting in spurious deviations in concentricity calculations. This deviation can cause serious assembly errors in the installation of high-precision industrial equipment (such as the docking of wind turbine towers and flange assembly), and may even lead to vibrations or structural stress concentrations during equipment operation, affecting the long-term safety and reliability of the equipment.

[0003] Therefore, how to solve the problem of decreased calibration accuracy caused by the drift of the rotation center has become a key challenge that urgently needs to be overcome in the current technical field. Summary of the Invention

[0004] In view of this, this application provides a rigid body pose calibration method, system, device and medium, which aims to solve the problem of rotation center drift caused by the fluctuation of point cloud centroid with the scanning range in the prior art, and improve the accuracy of rigid body pose calibration.

[0005] Firstly, this application provides a method for rigid body pose calibration, including: Obtain the mechanical installation information of the rigid body calibration object; Based on the mechanical installation information, the rigid body calibration object is locked to obtain the locking spindle direction information and mechanical origin information corresponding to the rigid body calibration object; Based on the locked spindle direction information and mechanical origin information, the point cloud information of the rigid body calibration object is projected and transformed to obtain the projected coordinate information corresponding to the rigid body calibration object; Based on the projection coordinate information, and combined with the preset target pose information corresponding to the rigid body calibration object, pose calibration processing is performed to obtain the pose calibration processing result of the rigid body calibration object.

[0006] Optionally, this application performs locking processing on the rigid body calibration object based on the mechanical installation information to obtain the locking spindle direction information and mechanical origin information corresponding to the rigid body calibration object, including: Based on the mechanical installation information, the axis is determined to obtain the locking spindle direction information corresponding to the rigid body calibration object; Based on the locked spindle direction information and combined with the mechanical installation information, the rotation center is fixed to obtain the mechanical origin information of the rigid body calibration object.

[0007] Through the above scheme, this application accurately determines the locking spindle direction and mechanical origin of the rigid body calibration object by fixing the axis and fixing the point, providing an accurate reference for subsequent pose calibration.

[0008] Optionally, this application performs projection transformation on the point cloud information corresponding to the rigid body calibration object based on the locked spindle direction information and the mechanical origin information to obtain the projected coordinate information corresponding to the rigid body calibration object, including: Obtain the point cloud information of the rigid body calibration object; Based on the mechanical origin information, coordinate transformation is performed on the point cloud information to obtain the coordinate transformation information corresponding to the point cloud information; Based on the coordinate transformation information and combined with the locked principal axis direction information, projection dimensionality reduction is performed to obtain the projected coordinate information.

[0009] By employing the above-described scheme, this application can more effectively process point cloud data and obtain accurate projected coordinate information by first performing coordinate transformation and then projection dimensionality reduction.

[0010] Optionally, based on the locked spindle direction information and the mechanical origin information, the point cloud information corresponding to the rigid body calibration object is projected and transformed to obtain the projected coordinate information of the rigid body calibration object, including: Obtain the point cloud information of the rigid body calibration object; Based on the locked principal axis direction information, the point cloud information is dimensionality-reduced by projection to obtain the dimensionality-reduced projection information corresponding to the point cloud information. Based on the mechanical origin information, coordinate transformation is performed on the reduced-dimensional projection information to obtain the projected coordinate information.

[0011] Through the above scheme, this application provides another effective way of processing point cloud data by performing dimensionality reduction projection and then coordinate transformation to obtain accurate projection coordinate information. The dimensionality reduction projection also reduces the amount of data processing computation and improves the pose calibration response efficiency.

[0012] Optionally, this application performs pose calibration processing based on the projected coordinate information and the preset target pose information corresponding to the rigid body calibration object, to obtain the pose calibration processing result of the rigid body calibration object, including: The dimensionality reduction solution is obtained by using the projected coordinate information and the target pose information to obtain the target rotation increment corresponding to the rigid body calibration object; Based on the target rotation increment, the rotation matrix corresponding to the rigid body calibration object is updated to obtain the pose calibration result.

[0013] Through the above scheme, this application achieves precise pose control of the rigid body calibration object by solving the problem in a reduced dimension and updating the rotation matrix corresponding to the rigid body calibration object based on the target rotation increment, thereby improving the accuracy of pose calibration.

[0014] Optionally, this application uses projected coordinate information and target pose information for dimensionality reduction to obtain the target rotation increment corresponding to the rigid body calibration object, including: Using projected coordinate information, the rotation weight coefficients corresponding to each feature point of the rigid body calibration object are determined; The local rotation increment of each feature point is determined using the projected coordinate information and the target pose information. Based on the local rotation increment of each feature point and the rotation weight coefficient corresponding to each feature point, a weighted sum is performed to obtain the total target increment. The weights of each feature point are obtained by weighting the rotation weight coefficients corresponding to each feature point. The target rotation increment is obtained by dividing the sum of the target increments by the sum of the weights.

[0015] Through the above scheme, this application can more accurately calculate the target rotation increment by weighted summarization and weighted processing, thereby improving the accuracy of pose calibration processing.

[0016] Optionally, this application obtains the mechanical installation information of the rigid body calibration object, including: determining the rigid body calibration object; and, if the rigid body calibration object is an object or equipment, determining the equipment installation information and mechanical design information corresponding to the object or equipment as the mechanical installation information.

[0017] Through the above scheme, this application clarifies the method for obtaining mechanical installation information, ensuring the accuracy of the subsequent calibration process.

[0018] Secondly, this application provides a rigid body pose calibration system, comprising: The mechanical installation information acquisition module is used to acquire the mechanical installation information of the rigid body calibration object; The locking processing module is used to perform locking processing on the rigid body calibration object based on the mechanical installation information, and obtain the locking spindle direction information and mechanical origin information corresponding to the rigid body calibration object; The projection conversion module is used to perform projection conversion on the point cloud information of the rigid body calibration object based on the locked spindle direction information and the mechanical origin information to obtain the projection coordinate information corresponding to the rigid body calibration object; The pose calibration processing module is used to perform pose calibration processing based on the projected coordinate information and the preset target pose information corresponding to the rigid body calibration object, so as to obtain the pose calibration result of the rigid body calibration object.

[0019] Thirdly, this application provides a rigid body pose calibration device, including: a processor, a communication interface, a memory, and a communication bus; wherein, the processor, the communication interface, and the memory communicate with each other through the communication bus; the memory is used to store computer programs; and the processor is used to execute the program stored in the memory to implement the steps of the rigid body pose calibration method as described in any one of the first aspects of this application.

[0020] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the rigid body pose calibration method as described in any one of the first aspects of this application.

[0021] In summary, the rigid body pose calibration method, system, device, and storage medium provided in this application acquire the mechanical installation information of the rigid body calibration object and lock the rigid body calibration object based on the mechanical installation information to obtain the locking spindle direction information and mechanical origin information corresponding to the rigid body calibration object. This forcibly anchors the rotation center of the pose calibration from the easily fluctuating centroid of the point cloud to the preset physical coordinate origin, fundamentally eliminating the problem of rotation center drift caused by changes in the scanning range. Then, based on the locking spindle direction information and mechanical origin information, the point cloud information of the rigid body calibration object is projected and transformed to obtain the projected coordinate information corresponding to the rigid body calibration object. Finally, based on the projected coordinate information... By combining the preset target pose information corresponding to the rigid body calibration object for pose calibration processing, the pose calibration algorithm always rotates around a fixed mechanical origin during the calculation process and is limited by the preset locked principal axis direction, avoiding false deviations in concentricity calculation caused by centroid fluctuations. Compared with the existing technology that assumes the object is a free rigid body and the rotation center fluctuates with the centroid of the point cloud, the technical solution of this application significantly improves the calibration accuracy and stability. It is particularly suitable for rigid body pose calibration scenarios with limited sensor field of view and incomplete point cloud data, and can meet the stringent requirements of high-precision industrial equipment installation and precision measurement. It has the advantages of improving the accuracy, repeatability and convergence speed of rigid body pose calibration. Attached Figure Description

[0022] Figure 1 A flowchart illustrating the steps of a rigid body pose calibration method provided in this application embodiment; Figure 2 A structural block diagram of a rigid body pose calibration system provided in this application embodiment; Figure 3 This is a structural block diagram of a rigid body pose calibration device provided in an embodiment of this application. Detailed Implementation

[0023] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0024] In industrial automation, robot hand-eye calibration, and precision assembly, rigid body pose calibration is a core step. For example, in robotic automated production lines, the relative pose between the robotic arm's end effector and the workpiece to be grasped needs to be accurately calibrated to ensure smooth production processes and consistent product quality. Existing traditional pose calibration methods typically rely on iterative nearest-point algorithms or point cloud registration algorithms based on singular value decomposition. These algorithms perform reasonably well when handling general problems, but their inherent theoretical assumption—treating the object to be calibrated as a free rigid body with six degrees of freedom in three-dimensional space—exposes serious limitations when facing engineering scenarios with explicit physical constraints.

[0025] A typical problem is the drift of the rotation center. Existing traditional algorithms typically use the centroid of the acquired point cloud data as the rotation center for iterative calculations. However, in real-world operating environments, due to factors such as sensor field-of-view limitations, partial occlusion, or incomplete scan data, the acquired point cloud often only represents a portion of the object. For example, when scanning a circular flange, only a segment of the arc's point cloud may be acquired. In this case, the centroid calculated by the algorithm will be located at the geometric center of this arc, rather than the true physical center of the entire flange. As the scanning angle or range changes, this calculated centroid will fluctuate continuously, causing the concentricity deviation calculated based on this centroid to contain spurious components, severely affecting the accuracy and repeatability of the calibration results.

[0026] Another significant challenge is the coupling of degrees of freedom. For objects with a rotating structure, rotation about an axis is coupled with translation in the circumferential direction. For example, a cylindrical tower or flange has its rotational motion about its own central axis and its circumferential motion mathematically coupled. Existing traditional algorithms, during the calibration and optimization process, cannot distinguish between these two types of motion and may incorrectly compensate for translational errors in the circumferential direction by adjusting the axial tilt angle of the object. While this compensation method may mathematically lead to convergence of the objective function, it is physically incorrect, ultimately resulting in the calibrated axes of the two components being non-parallel, violating basic mechanical assembly requirements.

[0027] Furthermore, when the rotation axis of the object to be calibrated is nearly parallel to an axis of the coordinate system, the traditional method of using Euler angles to describe attitude encounters the gimbal lock problem. This is a mathematical singularity that can cause severe oscillations or even complete failure in the solution process, making the calibration process unstable.

[0028] In order to overcome the problems of low calibration accuracy, poor repeatability and slow convergence speed caused by neglecting prior physical knowledge in the existing technology, this application proposes a new rigid body pose calibration method. One of the core ideas of this method is that it no longer treats the calibration object as an unconstrained free rigid body, but makes full use of its known mechanical installation information. By locking the rigid body calibration object based on its mechanical installation information, the principal axis direction and mechanical origin of the rigid body calibration object are locked. Through a double forced locking method, the rotation center of the pose calibration is forcibly anchored to a preset physical coordinate point (i.e., the mechanical origin coordinate point) rather than the data centroid, thereby eliminating the influence of centroid fluctuations caused by incomplete scan data on the calibration accuracy. This fundamentally constrains the solution space of the algorithm. Based on the locked principal axis direction information and mechanical origin information obtained after the locking process, the point cloud information of the rigid body calibration object is projected and transformed. This transforms a complex, ill-posed free matching problem into a simple, well-state, low-dimensional fine-tuning problem. Finally, based on the projected coordinate information and the preset target pose information corresponding to the rigid body calibration object, pose calibration is performed. This effectively solves the problems of low calibration accuracy, poor repeatability, and slow convergence speed caused by rotation center drift and degree-of-freedom coupling in the prior art, and improves the calibration accuracy and stability.

[0029] Reference Figure 1 The diagram illustrates a step-by-step flowchart of a rigid body pose calibration method provided in an embodiment of this application. Figure 1 As shown in the embodiments of this application, the rigid body pose calibration method specifically includes the following steps: Step 110: Obtain the mechanical installation information of the rigid body calibration object; Step 120: Lock the rigid body calibration object according to the mechanical installation information to obtain the locking spindle direction information and mechanical origin information corresponding to the rigid body calibration object; Step 130: Based on the locked spindle direction information and mechanical origin information, perform projection transformation on the point cloud information of the rigid body calibration object to obtain the projection coordinate information corresponding to the rigid body calibration object; Step 140: Based on the projection coordinate information and combined with the preset target pose information corresponding to the rigid body calibration object, pose calibration processing is performed to obtain the pose calibration processing result of the rigid body calibration object.

[0030] As can be seen, the rigid body pose calibration method provided in this application, after obtaining the mechanical installation information of the rigid body calibration object, locks the rigid body calibration object based on the mechanical installation information. By introducing the mechanical installation information to lock the rigid body calibration object, the principal axis direction and mechanical origin of the rigid body calibration object can be locked. Based on the principal axis direction information and mechanical origin information obtained after the locking process, the point cloud information of the rigid body calibration object is projected and transformed to obtain the corresponding projected coordinate information of the rigid body calibration object. Finally, based on the projected coordinate information and combined with the preset target pose information corresponding to the rigid body calibration object, pose calibration processing is performed. This ensures that the pose calibration processing always rotates around the fixed mechanical origin and is limited by the preset locked principal axis direction, avoiding false deviations in concentricity calculation caused by centroid fluctuations. The pose calibration result of the rigid body calibration object is obtained, effectively solving the problems of low calibration accuracy, poor repeatability, and slow convergence speed caused by rotation center drift and degree of freedom coupling in the prior art, thus improving calibration accuracy and stability.

[0031] In this context, a rigid body calibration object refers to a rigid object whose spatial position and attitude need to be determined during the pose calibration process. This rigid object is treated as an undeformable entity during calibration, and the relative positions between its internal points remain unchanged. For example, a rigid object can refer to an object with physical constraints or rigid connections, such as a wind turbine tower, a flange assembly, or a multi-sensor total station or other rotating structure equipment. Mechanical installation information refers to prior knowledge related to the physical installation state and mechanical design characteristics of the rigid body calibration object, including but not limited to the geometric center position, principal axis direction vector, rigid connection relationships between components, and reference positioning data during on-site installation recorded in the equipment's design drawings. This application does not impose limitations on this aspect.

[0032] Specifically, the rigid body pose calibration method provided in this application first executes step 110 to obtain the mechanical installation information of the rigid body calibration object. The purpose of this step is to preload prior mechanical knowledge related to the calibration object, providing a data foundation for subsequent locking processing. For example, in a specific implementation scenario, taking the installation calibration of a wind turbine tower as an example, the mechanical installation information includes the design main axis direction of the tower, the theoretical geometric center coordinates of the bottom flange of the tower, and other information. This mechanical installation information can be obtained from the equipment's design drawings, manufacturing specifications, or on-site installation records. This application embodiment does not impose specific restrictions on the method of obtaining the mechanical installation information.

[0033] In one specific optional implementation of this application, the process of obtaining the mechanical installation information of a rigid body calibration object may include: first, determining the rigid body calibration object; specifically, the object to be calibrated can be determined as the calibration object; then, it can be determined whether the calibration object is a rigid body calibration object by judging whether the object has a fixed rotation axis, so that the calibration object with a fixed rotation axis is determined as a rigid body calibration object; and according to the type of the rigid body calibration object, its corresponding mechanical design parameters and equipment rigid installation information can be obtained as the mechanical installation information of the rigid body calibration object. For example, it can be determined whether the rigid body calibration object is a preset object device, so that it is necessary to obtain the preset equipment installation information and mechanical design information corresponding to the object device, so that the preset equipment installation information and mechanical design information corresponding to the object device are used as the mechanical installation information of the rigid body calibration object. Thus, when the rigid body calibration object is an object device, the equipment installation information and mechanical design information corresponding to the object device can be determined as the mechanical installation information of the rigid body calibration object.

[0034] Among them, equipment installation information can refer to information related to the installation of the object equipment, while mechanical design information refers to mechanical design parameter information related to the object equipment itself. For example, in the case of the object equipment being a tower flange, equipment installation information can refer to the actual installation position and installation posture of the tower flange, while mechanical design information can refer to the geometric design parameters of the tower flange itself.

[0035] After obtaining the mechanical installation information of the rigid body calibration object, this embodiment of the application proceeds to step 120, which involves locking the rigid body calibration object based on the mechanical installation information to obtain the locking spindle direction information and mechanical origin information corresponding to the rigid body calibration object. This step is the core of this application's solution to the rotation center drift problem. A constraint space is established through a dual forced locking strategy to forcibly lock those degrees of freedom that do not physically exist. Specifically, the locking process of the rigid body calibration object based on the mechanical installation information in this embodiment of the application includes two levels of operations: axis fixing and point fixing.

[0036] Specifically, in this embodiment, the fixed axis operation can set the preset fixed principal axis direction corresponding to the rigid body calibration object as the projection axis direction. That is, the preset fixed principal axis of the rigid body calibration object (such as the fixed rotation principal axis of the rigid body calibration object) is aligned with a certain axis of the world coordinate system (such as the Z axis), thereby restricting the rotational degree of freedom of the rigid body calibration object to the neighborhood of the preset principal axis direction vector. The principal axis direction vector can be generated based on the fixed principal axis direction as the locked principal axis direction information corresponding to the rigid body calibration object, so that the point cloud information of the rigid body calibration object can be dimensionality reduced and projected according to the locked principal axis direction information, thereby reducing the projection calculation of pose calibration. Among them, the fixed principal axis preset by the rigid body calibration object refers to the fixed rotation principal axis preset by the rigid body calibration object. For example, when the rigid calibration object to be calibrated is two rigidly connected objects, the direction of the conjugate center line of the two rigidly connected objects can be used as the fixed principal axis direction preset by the rigid body calibration object. Through the axis setting operation, the direction of the conjugate center line of the two rigidly connected objects is determined as the fixed projection axis direction, so as to set the preset principal axis direction of the rigid body calibration object as the fixed projection axis direction and realize the locking of the rotation principal axis direction.

[0037] Furthermore, this embodiment of the application can obtain the mechanical origin information of the rigid body calibration object through fixed-point operation. Specifically, this embodiment of the application can set the rotation point of the rigid calibration object as the mechanical origin of the rigid calibration object through fixed-point operation, that is, the rotation center of the pose calibration is forcibly anchored to the preset mechanical origin, rather than the data centroid. This directly solves the problems of rotation center drift and degree of freedom coupling, ensuring the stability and accuracy of the calibration process.

[0038] Optionally, in this embodiment of the application, the rigid body calibration object is locked based on the mechanical installation information to obtain the locking spindle direction information and mechanical origin information corresponding to the rigid body calibration object. Specifically, it may include two sub-steps: first, the axis is determined based on the mechanical installation information to obtain the locking spindle direction information corresponding to the rigid body calibration object; second, the rotation center is determined based on the locking spindle direction information and the mechanical installation information to obtain the mechanical origin information of the rigid body calibration object.

[0039] Specifically, this embodiment performs latching processing based on the acquired mechanical installation information. First, a axis-fixing operation is performed to obtain the locked principal axis direction information. Specifically, in the axis-fixing operation, the specific mechanical structure of the rigid body calibration object (such as rigid connections, coaxial, or concurrent mechanical structures) can be determined based on the mechanical installation information. Then, based on the mechanical installation information and the specific mechanical structure of the rigid body calibration object, a preset calibration algorithm is used to forcibly restrict the rotational degrees of freedom of the rigid body calibration object to the neighborhood of the preset principal axis direction vector. Taking a wind turbine tower as an example, since the tower has a clear vertical structural feature, the algorithm can define the preset principal axis direction vector d_theo corresponding to the rigid body calibration object as [0, 0, 1] based on a pre-constructed world coordinate system and the mechanical installation information. This means that in all subsequent pose calculations, the algorithm is forced to only calculate rotations around this fixed Z-axis, completely prohibiting the calculation of any rotational components around the X or Y axes. In other words, the original pitch and yaw rotational degrees of freedom are completely locked at the algorithm level. Regardless of the cause of any slight tilt in the fitted axial vector of the point cloud data collected on-site, such as a fitted axial vector of [0.01, -0.02, 0.999], the algorithm will directly ignore these two small deviations of 0.01 and -0.02, and will directly rotate the matrix... The first two columns (corresponding to the X and Y axes) are fixed as orthogonal bases in the horizontal plane, while only the third column (Z axis) is allowed to rotate in the horizontal plane, thus forcibly constraining the rotation to the pure [0, 0, 1] axis. In this way, the problem of axial tilting errors caused by degree-of-freedom coupling is completely eliminated.

[0040] In its implementation, the algorithm no longer searches for rotations in arbitrary directions like traditional algorithms. Instead, based on the acquired mechanical installation information, it locks the theoretical principal axis direction vector d_theo of the object / equipment as a unique and unchangeable principal axis direction vector. This locking of the principal axis direction information achieves dimensionality reduction of the degrees of freedom. That is, by locking the direction, it forces the principal axis direction to align with a certain axis (such as the Z-axis) of the world coordinate system. This reduces the original three rotational degrees of freedom (pitch, yaw, and roll) to dimensionality. For example, pitch and yaw are completely locked, leaving only the small rotation around the principal axis (roll angle) as the variable to be determined. Thus, a complex three-dimensional (3D) free matching problem can be transformed into a simple two-dimensional (2D) plane or one-dimensional (1D) line fine-tuning problem. This axial manifold constraint mechanism ensures that the rotation calculation of the algorithm always revolves around a fixed axis, regardless of how the actual scanned point cloud data fluctuates. This avoids the problem of non-parallel axes caused by degree-of-freedom coupling. Furthermore, the calculation range can be forcibly constrained by the direction locking mechanism, thereby reducing the amount of pose calibration calculation and accelerating the pose calibration efficiency.

[0041] After the axis determination is completed, the rotation center is fixed to obtain the mechanical origin information. Specifically, in the rotation center fixing operation, the traditional method of rotating around the centroid of the point cloud is abandoned. Instead, based on locking the spindle direction information and combining it with the mechanical installation information, a mechanical origin anchoring mechanism is introduced. This forces the rotation center of the transformed model to be set as the mechanical origin, thus anchoring the rotation center to a preset physical coordinate point (i.e., the mechanical origin coordinate point) rather than the data centroid. This eliminates the deviation between the calculated centroid point and the actual center point required for rigid coaxial connection, thereby solving the rotation failure problem caused by the difference between the calculated centroid point and the actual center point required for rigid coaxial connection in the existing technology. Specifically, during rigid rotation, the rotation center point is crucial. If the rotation center point is calculated using the centroid, any reverse calibration will cause data deviation due to instrument errors, resulting in the calculated origin coordinates not being the true mechanical origin coordinates. When the mechanical origin coordinates are incorrect, even a small angular deviation will amplify the linear error, leading to a decrease in calibration accuracy. In this embodiment, after locking the spindle direction, the locked spindle direction is determined as a fixed projection axis direction based on the locked spindle direction information. Based on the fixed projection axis direction, the rigid rotation point of the rigid calibration object is determined as the mechanical origin, thereby determining the preset physical coordinate point information corresponding to the rigid rotation point as the mechanical origin information of the rigid body calibration object.

[0042] As an example of this application, assuming the preset physical coordinate information corresponding to the rigid rotation point of the rigid calibration object is [100, 50, 0], the mechanical origin coordinates O of the rigid calibration object can be locked to [100, 50, 0] through the rotation center fixing operation. Thus, regardless of whether the point cloud data subsequently scanned for the rigid calibration object is the complete point cloud information or a partial point cloud information of the rigid calibration object, the rotation center will always be this locked, fixed mechanical origin coordinates [100, 50, 0], rather than the coordinates of the centroid of the point cloud calculated by the scan. That is, in subsequent point cloud data calculations, all rotational transformations of the point cloud will revolve around this anchored mechanical origin coordinates, which is fundamentally different from the traditional method of rotating around the centroid of the point cloud. In this way, even if the laser scanner only collects a partial point cloud information of the rigid calibration object, the algorithm will still calculate the rotation around the preset mechanical origin coordinates corresponding to the entire rigid calibration object, rather than around the calculated centroid. This fundamentally solves the problem of rotation center drift, ensuring the stability and accuracy of the calibration results.

[0043] After completing the locking process of the rigid body calibration object, this embodiment of the application performs projection transformation on the point cloud information of the rigid body calibration object based on the determined locking main axis direction information and mechanical origin information, that is, executes step 130, thereby obtaining the projection coordinate information corresponding to the rigid body calibration object. This step involves data projection and dimensionality reduction. For example, dimensionality reduction can reduce 3D data to 2D planar data for optimization. Taking the Z-axis as an example, if the principal axis is locked to the Z-axis of the world coordinate system, the projection principal axis is forcibly aligned with the Z-axis of the world coordinate system. According to the preset projection formula P_proj=[x_rel,y_rel], the 3D point cloud coordinates can be projected onto the XY plane perpendicular to the Z-axis. The Z-coordinate component is ignored, resulting in the 2D projection coordinates [x_rel,y_rel]. These 2D projection coordinates [x_rel,y_rel] can be used as dimensionality reduction projection information for subsequent pose calibration. This transforms the complex 3D registration problem into a much simpler 2D planar matching problem, significantly reducing computational complexity and instability, and avoiding mathematical problems such as gimbal locking.

[0044] Where x_rel is the X component of the relative coordinates and y_rel is the Y component of the relative coordinates, the relative coordinates refer to the translation coordinates of the 3D point cloud coordinates relative to the mechanical origin coordinates. For example, assuming the mechanical origin coordinates of the rigid calibration object are locked at [100, 50, 0], and the 3D coordinates of the point cloud data of a certain point of the rigid calibration object are [102, 50, 100], through point-fixing operations, the feature point corresponding to the point cloud data can be translated to the relative coordinates based on the mechanical origin. For example, by subtracting the mechanical origin coordinates [102, 50, 0] from the 3D coordinates [102, 50, 100] of the point cloud data, the translated relative coordinates [2, 0, 100] can be obtained. Subsequently, dimensionality reduction projection can be performed based on the relative coordinates [2, 0, 100] to obtain the corresponding 2D projected coordinates [2, 0]. The 2D projected coordinates [2, 0] can be used as dimensionality reduction projection information for subsequent pose calibration processing, thus achieving dimensionality reduction optimization.

[0045] In one optional embodiment of this application, the point cloud information corresponding to the rigid body calibration object is projected and transformed based on the locked main axis direction information and the mechanical origin information to obtain the projected coordinate information corresponding to the rigid body calibration object. Specifically, this may include: acquiring the point cloud information of the rigid body calibration object; performing coordinate transformation on the point cloud information based on the mechanical origin information to obtain the coordinate transformation information corresponding to the point cloud information; and performing projection dimensionality reduction based on the coordinate transformation information and the locked main axis direction information to obtain the projected coordinate information.

[0046] Specifically, after completing the locking process of the rigid body calibration object, this embodiment of the application can first obtain the point cloud information of the rigid body calibration object. This point cloud information can refer to the original point cloud information collected by the three-dimensional scanning device, specifically including each point cloud data obtained by scanning, such as a point cloud data set of N points {P_i|i=1,2,...,N}, where each point P_i is represented as a three-dimensional vector [x_i,y_i,z_i]. Subsequently, based on the mechanical origin information determined in the previous step, coordinate transformation is performed on the point cloud information to translate the original coordinates of each point cloud data in the point cloud information to a relative coordinate system with the mechanical origin as the reference. Specifically, the operation can be to subtract the three-dimensional coordinates of the mechanical origin from the original three-dimensional coordinates of each point cloud data. The geometric meaning of this operation is to translate the entire point cloud data in three-dimensional space so that the locked mechanical origin coordinates coincide with the origin coordinates [0, 0, 0] of the world coordinate system. After this transformation, all point cloud data are converted into relative coordinates with the mechanical origin as a reference. These relative coordinates can then be used as the coordinate transformation information for the point cloud data. Finally, based on this transformed coordinate information and the locked principal axis direction information, projection dimensionality reduction is performed to achieve dimensionality reduction projection.

[0047] For example, if the locked principal axis is the Z-axis, then the projection dimensionality reduction operation simply discards all Z-coordinate components of the point cloud data and retains only its X and Y coordinate components. For instance, if the three-dimensional coordinates of the original point cloud data corresponding to a rigid body calibration object are [1002.1, 500.2, 100.5], and the corresponding locked mechanical origin coordinates are [1000.0, 500.0, 100.0], after coordinate transformation, it can be determined that the three-dimensional relative coordinates corresponding to the original point cloud data become [2.1, 0.2, 0.5]. After projection dimensionality reduction, the final projected coordinate information is the two-dimensional coordinates [2.1, 0.2].

[0048] As can be seen, in this embodiment, after obtaining the point cloud information of the rigid body calibration object, coordinate transformation is performed before projection dimensionality reduction. That is, coordinate translation is performed first to establish a local coordinate system centered on the mechanical origin, thereby effectively suppressing the long-distance offset noise that may exist in the original point cloud data, making the subsequent projection dimensionality reduction more stable in numerical calculation. At the same time, since the translation operation is completed before projection, the projection plane naturally passes through the mechanical origin, ensuring that all two-dimensional registration operations are performed around this fixed anchor point, fundamentally eliminating the risk of rotation center drift.

[0049] In another optional embodiment of this application, after obtaining the point cloud information of the rigid body calibration object, the obtained point cloud information can first be dimensionally reduced and projected based on the locked principal axis direction information. Then, based on the mechanical origin information, the dimensionally reduced projection information obtained after dimensionality reduction and projection is subjected to coordinate transformation. By performing dimensionality reduction and projection first, the three-dimensional point cloud data is immediately compressed into low-dimensional data, thereby significantly reducing the computational load of subsequent coordinate transformation. This is particularly suitable for rigid body pose calibration scenarios with large amounts of point cloud data and high real-time requirements. At the same time, performing coordinate transformation in low-dimensional space avoids complex matrix operations in three-dimensional space, reducing the consumption of computing resources.

[0050] Optionally, in this embodiment, based on the locked principal axis direction information and the mechanical origin information, a projection transformation is performed on the point cloud information corresponding to the rigid body calibration object to obtain the projected coordinate information corresponding to the rigid body calibration object. Specifically, this may include: acquiring the point cloud information of the rigid body calibration object; and performing dimensionality reduction projection on the point cloud information based on the locked principal axis direction information. That is, after acquiring the original three-dimensional point cloud information, the original three-dimensional point cloud information is directly dimensionality reduced and projected based on the locked principal axis direction information to convert the original three-dimensional point cloud information into two-dimensional point cloud information. That is, without performing three-dimensional translation, all coordinate components of the point cloud data in the locked principal axis direction are directly discarded, and the original 3D coordinates are converted into 2D coordinates to obtain the dimensionality reduction projection information corresponding to the point cloud information. This dimensionality reduction projection information includes the coordinate information of all point cloud data projected onto the plane. For example, when the locked principal axis is the Z-axis, the X, Y, and Z coordinates of all point cloud data are directly converted into X, Y, and Z coordinates. The Y-coordinate is used to obtain a two-dimensional reduced-dimensional projection information, which serves as the reduced-dimensional projection information corresponding to the point cloud information. Then, based on the mechanical origin information, coordinate transformation is performed on this two-dimensional reduced-dimensional projection information. Specifically, the coordinate transformation of the reduced-dimensional projection information corresponding to the point cloud information is performed by subtracting the two-dimensional coordinates of the mechanical origin on the projection plane from the coordinates of each two-dimensional point in the reduced-dimensional projection information, thereby obtaining the final projection coordinate information.

[0051] For example, assuming the 3D coordinates of a scanned original point cloud data are [1002.1, 500.2, 100.5], based on the locked principal axis direction information, the corresponding 2D coordinates of the original point cloud data can be directly projected to [1002.1, 500.2] through dimensionality reduction projection. If the locked machine origin coordinates are [1000.0, 500.0, 100.0], its corresponding projected coordinates are [1000.0, 500.0]. By subtracting the projected coordinates of the machine origin [1000.0, 500.0] from the 2D coordinates of the original point cloud data [1002.1, 500.2], the final 2D projected coordinate information with the machine origin as the rotation center can be obtained, which is also [2.1, 0.2].

[0052] As can be seen, this application can first perform coordinate transformation and then projection dimensionality reduction, or it can first perform data projection and dimensionality reduction and then coordinate transformation. In the end, it can obtain the projection coordinate information with the machine origin as the rotation center, so that the pose calibration process can be performed based on the projection coordinate information, thereby reducing the computational complexity of the pose calibration process.

[0053] Specifically, after obtaining the projection coordinate information, this embodiment of the application can perform pose calibration processing based on the obtained projection coordinate information and the preset target pose information to accurately solve the pose deviation corresponding to the rigid body calibration object in the dimensionality-reduced two-dimensional plane. For example, the projection coordinate information and the target pose information can be used to perform dimensionality reduction to obtain the target rotation increment corresponding to the rigid body calibration object. Then, the pose matrix of the rigid body calibration object can be updated based on the target rotation increment to achieve accurate calibration of the rigid body pose.

[0054] The target pose information refers to the pose data that a rigid body calibration object should have under a predefined ideal state. It can usually be obtained from design drawings or ideal models. Specifically, it can include pre-set target point cloud data. For example, the target pose information can be a set of M target point cloud data {Q_j|j=1,2,...,M}, where each target point data Q_j can be represented as a two-dimensional coordinate [q_x,q_y] in the projection plane, or as a three-dimensional coordinate [q_x,q_y,q_z] in three-dimensional space. This application does not limit this.

[0055] Optionally, in this embodiment of the application, pose calibration processing is performed based on the projection coordinate information and the preset target pose information corresponding to the rigid body calibration object to obtain the pose calibration processing result of the rigid body calibration object. Specifically, it may include the following sub-steps: Sub-step 1401: Use the projected coordinate information and the target pose information to perform dimensionality reduction solution to obtain the target rotation increment corresponding to the rigid body calibration object; Sub-step 1402: Based on the target rotation increment, update the rotation matrix corresponding to the rigid body calibration object to obtain the pose calibration result of the rigid body calibration object.

[0056] Specifically, in order to calculate the pose deviation of the rigid body calibration object in low-dimensional space and achieve accurate pose calibration, this embodiment of the application, after obtaining the projection coordinate information corresponding to the rigid body calibration object, firstly uses the projection coordinate information and target pose information to perform dimensionality reduction solution to obtain the target rotation increment corresponding to the rigid body calibration object. This target rotation increment refers to the rotation angle calculated during the rigid body pose calibration process, used to adjust the attitude of the rigid body calibration object. This rotation angle represents the amount by which the rigid body calibration object needs to rotate around its locking principal axis so that the projection of the scanned point cloud corresponding to the rigid body calibration object best coincides with the projection of the target point cloud. This ensures that the current pose of the rigid body calibration object and the preset target pose are optimally aligned on the two-dimensional projection plane, thereby making the current pose of the rigid body calibration object the same as the preset target pose.

[0057] Optionally, in this embodiment, the projected coordinate information and the target pose information are used for dimensionality reduction to obtain the target rotation increment corresponding to the rigid body calibration object. Specifically, this may include: using the projected coordinate information to determine the rotation weight coefficient corresponding to each feature point of the rigid body calibration object; using the projected coordinate information and the target pose information to determine the local rotation increment of each feature point; performing weighted summation based on the local rotation increment of each feature point and the rotation weight coefficient corresponding to each feature point to obtain the total target increment; performing weighted processing based on the rotation weight coefficient corresponding to each feature point to obtain the total weight corresponding to each feature point; and dividing the total target increment by the total weight to obtain the target rotation increment. Specifically, to further improve the accuracy and robustness of the solution, the embodiments of this application employ a weighted averaging strategy in calculating the target rotation increment. First, using projected coordinate information, the rotation weight coefficient corresponding to each feature point of the rigid body calibration object is determined. Here, a feature point refers to a cloud point corresponding to a certain projected coordinate in the projected coordinate information. Considering that feature points farther from the rotation center are more sensitive to changes in the rotation angle, their contribution to pose calibration should be greater. Therefore, in an optional embodiment of this application, the square of the distance from each feature point to the mechanical origin on the two-dimensional projection plane can be used as its rotation weight coefficient. This means that the farther the distance between the feature point and the locked mechanical origin, the larger its corresponding rotation weight coefficient.

[0058] Specifically, for each feature point, its projection coordinates can be used to calculate the magnitude of the feature point on the two-dimensional projection plane from the machine origin. For example, if the coordinates of the machine origin are (x0, y0), the two-dimensional projection coordinates (x_i, y_i) of feature point i can be used, calculated according to the preset weighted least squares circle formula. Calculate the magnitude of each feature point i on the two-dimensional projection plane from the mechanical origin. Since feature points farther from the center of rotation are more sensitive to changes in rotation angle, the rotation weight coefficients corresponding to these feature points... It is directly proportional to the square of its distance from the mechanical origin on the two-dimensional projection plane. For example, the square of the magnitude of the feature point's distance from the mechanical origin on the two-dimensional projection plane can be expressed as... Set as the rotation weight coefficient corresponding to this feature point, i.e. The rotation weighting coefficient is the square of the straight-line distance from the feature point's coordinates to the machine origin. This ensures that distant feature points dominate subsequent angle calculations, thereby improving the stability of global angle calculations.

[0059] Next, the local rotation increment of each feature point is determined using the projected coordinate information and the target pose information. Specifically, for each scanned feature point, the projected coordinates corresponding to the feature point can be extracted from the projected coordinate information to construct the scan vector V_source_i from the machine origin to the feature point. The target point cloud data corresponding to the feature point can be extracted from the target pose information to determine the target coordinates corresponding to the feature point. Then, the target vector V_target_i from the machine origin to the feature point can be determined using the target coordinates. Subsequently, the rotation angle can be calculated using a preset cross product formula based on the target vector V_target_i and the scan vector V_source_i, and the calculated rotation angle is determined as the local rotation increment of the feature point. For example, if the target vector V_target_i = [a,b] and the scan vector V_source_i = [x,y], then the rotation angle is determined according to the preset cross product formula. By performing calculations, the angle between the two vectors can be calculated, and this angle can be used as the rotation angle corresponding to the feature point, thus serving as the local rotation increment of the feature point. .

[0060] Then, based on the local rotation increments of each feature point and their corresponding rotation weight coefficients, a weighted sum is performed to obtain a total target increment. Specifically, this can be achieved using a preset weighted average formula. The increments of all feature points are summed, and the specific calculation method is to sum the local rotation increments of each feature point. Multiply by its rotation weighting factor To increment the local rotation of each feature point Its rotation weighting coefficient The product of these products determines the increment for each feature point. Then, all products are summed to obtain the total target increment for all feature points. Simultaneously, it can also be based on the rotation weight coefficients corresponding to each feature point. Perform weighted processing to obtain the sum of weights for all feature points. That is, simply sum the rotation weight coefficients of all feature points; finally, use the sum of the target increments. Divide by the total weight Thus, the globally optimal target rotation increment is obtained. This target rotation increment The calculation results integrate information from all feature points and are weighted according to their importance, thus exhibiting high accuracy and stability.

[0061] After calculating the target rotation increment, the rotation matrix corresponding to the current pose of the rigid body calibration object is updated based on this target rotation increment to obtain a rotation update matrix. This updated rotation matrix is ​​then used to update the pose matrix of the rigid body calibration object, resulting in the pose calibration result. Specifically, in this embodiment, an incremental rotation matrix representing a small rotation can be constructed based on the target rotation increment and multiplied by the current rotation matrix to obtain an updated rotation matrix. This updated rotation matrix serves as the rotation update matrix. Subsequently, the rigid body calibration object can be controlled to perform actual physical rotation based on the updated rotation matrix, i.e., the pose of the rigid body calibration object is updated and controlled based on the rotation update matrix, thereby obtaining the pose calibration result.

[0062] As a specific implementation of this application, after calculating the target rotation increment, the pose update can be calculated in the tangent space using Lie algebra iterative correction. For example, if rotation components around the X and Y axes are prohibited, and only the rotation component around the Z axis (depending on the locked principal axis) is retained, the calculated target rotation increment can be used as the basis for the update. Construct an update vector ξ=[0,0,Δθ_final] (note that the rotation components of the X and Y axes are forced to be 0) to calculate the target rotation increment. The pose update is converted into an update vector ξ, which is then used as the pose update quantity. Based on this pose update quantity, an exponential mapping can be used to convert it into a true rotation update matrix R_update. For example, the rotation update matrix R_update can be calculated using the update vector ξ according to the preset exponential mapping formula R_update=exp(ξ). The exponential mapping can be implemented using the Rodriguez formula; for small angles, the exponential mapping can be simplified to R_update≈I+ξ_skew, where I is the identity matrix and ξ_skew is an antisymmetric matrix constructed based on the update vector ξ. Subsequently, the calculated rotation update matrix R_update is multiplied and superimposed with the current rotation matrix R_old of the rigid body calibration object. After superposition, a manifold projection constraint step is performed. That is, after each iteration update, the calculated rotation matrix is ​​forcibly projected back to the "constrained manifold" set in the first step using a pre-built algorithm, ensuring that the axial direction never deviates from the theoretical value and the mechanical origin never drifts during the iterative approximation of the optimal solution. The specific operation is as follows: It checks whether there are rotation components around non-principal axes in the rotation update matrix after each update. If so, these illegal rotation components are discarded, and the rotation angles in the non-principal axes corresponding to the rotation matrix are forcibly reset to zero. Only the rotation components around the locked principal axis are retained for updating. Then, the reassigned matrix is ​​orthogonalized to obtain the final rotation update matrix. This forced projection method of discarding, correcting, and reflowing ensures that each iteration does not deviate from the preset physical and mechanical constraints.

[0063] As an example of this application, suppose it is calculated that rotation around the X-axis is needed to "assemble" the point cloud. If it is detected that "rotation around the X-axis" is prohibited, that is, when the X-axis direction is a non-principal axis direction of the rigid calibration object, such as when the principal axis direction of the rigid calibration object is locked to the Z-axis direction, the rotation component of the X-axis can be directly discarded, and only the rotation component around the Z-axis is retained for updating, resulting in the rotation update matrix R_update. Subsequently, based on this rotation update matrix R_update, according to the rotation matrix update formula: R_new = R_update·R_old, the rotation update matrix R_update is multiplied and superimposed with the current rotation matrix R_old of the rigid body calibration object. That is, the rotation update matrix R_update is superimposed on the current rotation matrix R_old of the rigid body calibration object, so that the pose matrix of the rigid body calibration object is updated using the rotation update matrix R_update. The target rotation matrix R_new obtained after superposition and update can be used as the new pose matrix to calibrate and control the pose of the rigid body calibration object, and the pose calibration result is obtained.

[0064] Through the aforementioned series of steps, the rigid body pose calibration method provided in this application, after obtaining the mechanical installation information of the rigid body calibration object, locks the rigid body calibration object based on the mechanical installation information. This utilizes prior mechanical knowledge to forcibly lock the principal axis and mechanical origin of the rigid body calibration object, effectively setting the rotation center of the rigid body calibration object to a preset physical coordinate point (i.e., the mechanical origin coordinates) rather than the data centroid. This avoids the problem of centroid fluctuation with the scanning range caused by incomplete point cloud data, fundamentally eliminating the phenomenon of rotation center drift. Simultaneously, through principal axis locking, the rotational degrees of freedom are restricted to a preset value using an axial manifold constraint mechanism. Within the neighborhood of the principal axis direction vector, hard locking of invalid degrees of freedom is achieved, avoiding degree-of-freedom coupling and gimbal locking problems. After obtaining the locked principal axis direction information and mechanical origin information corresponding to the rigid body calibration object, the point cloud information of the rigid body calibration object is projected and transformed based on the locked principal axis direction information and mechanical origin information. Finally, the pose calibration process is performed by combining the target pose information, so as to successfully transform the complex three-dimensional six-degree-of-freedom registration problem into a robust and efficient two-dimensional single-degree-of-freedom fine-tuning problem. This fundamentally solves the problems of rotation center drift, degree-of-freedom coupling and gimbal locking of traditional algorithms, and significantly improves the accuracy, stability and speed of rigid body pose calibration under physical constraints.

[0065] Furthermore, in this embodiment, after obtaining the point cloud information of the rigid body calibration object, the dimensionality reduction projection of the obtained point cloud information is performed based on the locked principal axis direction information. This transforms the three-dimensional point cloud registration problem into a two-dimensional plane fine-tuning problem, significantly reducing computational complexity and improving the convergence speed and real-time performance of the algorithm. Coordinate transformation of the obtained point cloud information can be performed based on the fixed mechanical origin information, such as using the fixed mechanical origin anchoring transformation formula. The relative coordinates of each feature point are obtained by subtracting the projected coordinates of the mechanical origin from the projected coordinates of each feature point in the point cloud information. The pose rotation matrix of the rigid body calibration object is iteratively optimized based on the relative coordinates of each feature point. After each iteration, the rotation matrix is ​​forcibly projected back to the constrained manifold, ensuring that the axial direction never deviates from the locked principal axis direction and the rotation center does not drift throughout the optimization process. This effectively solves the problems of low calibration accuracy, poor repeatability, and slow convergence speed caused by rotation center drift and degree-of-freedom coupling in the prior art, improving the accuracy and stability of rigid body pose calibration.

[0066] It should be noted that, for the sake of simplicity, the method embodiments are described as a series of actions. However, those skilled in the art should know that the embodiments of this application are not limited to the described order of actions, because according to the embodiments of this application, some steps may be performed in other orders or simultaneously.

[0067] Figure 2 This is a structural block diagram of a rigid body pose calibration system provided in an embodiment of this application. Figure 2 As shown, the rigid body pose calibration system provided in this application embodiment may specifically include the following modules: Mechanical installation information acquisition module 210 is used to acquire mechanical installation information of rigid body calibration object; The locking processing module 220 is used to lock the rigid body calibration object according to the mechanical installation information, and obtain the locking spindle direction information and mechanical origin information corresponding to the rigid body calibration object; The projection conversion module 230 is used to perform projection conversion on the point cloud information of the rigid body calibration object based on the locked spindle direction information and the mechanical origin information, so as to obtain the projection coordinate information corresponding to the rigid body calibration object. The pose calibration processing module 240 is used to perform pose calibration processing based on the projection coordinate information and the preset target pose information corresponding to the rigid body calibration object, so as to obtain the pose calibration result of the rigid body calibration object.

[0068] Optionally, the locking processing module 220 may specifically include the following sub-modules: The axis-fixing submodule is used to fix the axis based on the mechanical installation information and obtain the locking spindle direction information corresponding to the rigid body calibration object; The positioning submodule is used to determine the rotation center based on the locked spindle direction information and the mechanical installation information, so as to obtain the mechanical origin information of the rigid body calibration object.

[0069] In an optional embodiment of this application, the projection conversion module 230 may specifically include the following sub-modules: The point cloud information acquisition submodule is used to acquire the point cloud information of the rigid body calibration object; The first coordinate transformation submodule is used to perform coordinate transformation on the point cloud information based on the mechanical origin information to obtain the coordinate transformation information corresponding to the point cloud information; The first projection dimensionality reduction submodule is used to perform projection dimensionality reduction based on the coordinate transformation information and the locked principal axis direction information to obtain the projection coordinate information.

[0070] In another optional embodiment of this application, the projection conversion module 230 may specifically include the following sub-modules: The point cloud information acquisition submodule is used to acquire the point cloud information of the rigid body calibration object; The second dimension reduction projection submodule is used to perform dimension reduction projection on the point cloud information based on the locked principal axis direction information to obtain the dimension reduction projection information corresponding to the point cloud information. The second coordinate transformation submodule is used to perform coordinate transformation on the reduced-dimensional projection information based on the mechanical origin information to obtain the projected coordinate information.

[0071] Optionally, the pose calibration processing module 240 may specifically include the following sub-modules: The dimension reduction solution submodule is used to perform dimension reduction solution using the projected coordinate information and the target pose information to obtain the target rotation increment corresponding to the rigid body calibration object; The rotation update submodule is used to update the rotation matrix corresponding to the rigid body calibration object based on the target rotation increment, so as to obtain the pose calibration processing result.

[0072] Optionally, the dimensionality reduction solution submodule includes the following units: The rotation weight coefficient determination unit is used to determine the rotation weight coefficient corresponding to each feature point of the rigid body calibration object using the projection coordinate information. The local rotation increment determination unit is used to determine the local rotation increment of each feature point using the projected coordinate information and the target pose information; The target increment sum determination unit is used to perform weighted summation based on the local rotation increment of each feature point and the rotation weight coefficient corresponding to each feature point to obtain the target increment sum; The weight sum determination unit is used to perform weighted processing based on the rotation weight coefficients corresponding to each feature point to obtain the weight sum corresponding to each feature point; The target rotation increment determination unit is used to obtain the target rotation increment by dividing the sum of the target increments by the sum of the weights.

[0073] Optionally, the mechanical installation information acquisition module includes: The rigid body calibration object determination submodule is used to determine the rigid body calibration object; The mechanical installation information acquisition submodule is used to determine the equipment installation information and mechanical design information corresponding to the object equipment as the mechanical installation information when the rigid body calibration object is an object device.

[0074] It should be noted that the rigid body pose calibration system provided in this application embodiment can execute the rigid body pose calibration method provided in any of the above embodiments of this application, and has the corresponding functions and beneficial effects of the execution method.

[0075] In practical implementation, the aforementioned rigid body pose calibration system can be integrated into electronic devices such as personal computers and servers. This allows the electronic device to function as a rigid body pose calibration device. It can acquire the mechanical installation information of the rigid body calibration object and lock the object based on this information, obtaining the locking spindle direction information and mechanical origin information. This forces the rotation center of the pose calibration from the easily fluctuating centroid of the point cloud to a preset physical coordinate origin, fundamentally eliminating the problem of rotation center drift caused by changes in the scanning range. Then, based on the locking spindle direction information and mechanical origin information, the system can perform pose calibration on the rigid body. The point cloud information of the rigid body calibration object is projected and transformed to obtain the projected coordinate information corresponding to the rigid body calibration object. Based on the projected coordinate information, combined with the preset target pose information corresponding to the rigid body calibration object, pose calibration processing is performed. This ensures that the pose calibration algorithm always rotates around a fixed mechanical origin during the calculation process and is limited by the preset locked principal axis direction, avoiding false deviations in concentricity calculation caused by centroid fluctuations. This improves the accuracy and stability of rigid pose calibration, meets the stringent requirements of high-precision industrial equipment installation and precision measurement, and has the advantages of improving the accuracy, repeatability, and convergence speed of rigid body pose calibration.

[0076] like Figure 3 As shown, this application embodiment also provides a rigid body pose calibration device 300, which includes: a processor 111, a communication interface 112, a memory 113, and a communication bus 114, wherein the processor 111, the communication interface 112, and the memory 113 communicate with each other through the communication bus 114; the memory 113 is used to store computer programs; when the processor 310 executes the computer program stored in the memory 320, it implements the steps of the rigid body pose calibration method provided in any of the foregoing method embodiments of this application, including: obtaining... Obtain the mechanical installation information of the rigid body calibration object; perform locking processing on the rigid body calibration object based on the mechanical installation information to obtain the locking spindle direction information and mechanical origin information corresponding to the rigid body calibration object; perform projection transformation on the point cloud information of the rigid body calibration object based on the locking spindle direction information and mechanical origin information to obtain the projection coordinate information corresponding to the rigid body calibration object; perform pose calibration processing based on the projection coordinate information and the preset target pose information corresponding to the rigid body calibration object to obtain the pose calibration processing result of the rigid body calibration object.

[0077] This application also provides a computer-readable storage medium storing a computer program thereon, characterized in that, when the computer program is executed by a processor, it implements the steps of the rigid body pose calibration method as described in any of the above method embodiments of this application.

[0078] It should be noted that the various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For system, device, and storage medium embodiments, since they are basically similar to the method embodiments, the descriptions are relatively simple, and relevant parts can be referred to the descriptions of the method embodiments.

[0079] In this document, relational terms such as “first” and “second” are used merely to distinguish one entity or operation from another, without necessarily requiring or implying any such actual relationship or order between these entities or operations. Furthermore, 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. Without further limitation, 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 said element.

[0080] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0081] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for rigid body pose calibration, characterized in that, The method includes: Obtain the mechanical installation information of the rigid body calibration object; Based on the mechanical installation information, the rigid body calibration object is locked to obtain the locking spindle direction information and mechanical origin information corresponding to the rigid body calibration object; Based on the locked spindle direction information and the mechanical origin information, the point cloud information of the rigid body calibration object is projected and transformed to obtain the projected coordinate information corresponding to the rigid body calibration object; Based on the projected coordinate information, and combined with the preset target pose information corresponding to the rigid body calibration object, pose calibration processing is performed to obtain the pose calibration processing result of the rigid body calibration object.

2. The rigid body pose calibration method according to claim 1, characterized in that, The process of locking the rigid body calibration object based on the mechanical installation information to obtain the locking spindle direction information and mechanical origin information corresponding to the rigid body calibration object includes: Based on the mechanical installation information, the axis is determined to obtain the locking spindle direction information corresponding to the rigid body calibration object; Based on the locked spindle direction information and the mechanical installation information, the rotation center is fixed to obtain the mechanical origin information of the rigid body calibration object.

3. The rigid body pose calibration method according to claim 1, characterized in that, The step of projecting and transforming the point cloud information corresponding to the rigid body calibration object based on the locked spindle direction information and the mechanical origin information to obtain the projected coordinate information corresponding to the rigid body calibration object includes: Obtain the point cloud information of the rigid body calibration object; Based on the mechanical origin information, coordinate transformation is performed on the point cloud information to obtain the coordinate transformation information corresponding to the point cloud information; Based on the coordinate transformation information and combined with the locked principal axis direction information, projection dimensionality reduction is performed to obtain the projection coordinate information.

4. The rigid body pose calibration method according to claim 1, characterized in that, The step of projecting and transforming the point cloud information corresponding to the rigid body calibration object based on the locked spindle direction information and the mechanical origin information to obtain the projected coordinate information of the rigid body calibration object includes: Obtain the point cloud information of the rigid body calibration object; Based on the locked principal axis direction information, the point cloud information is subjected to dimensionality reduction projection to obtain the dimensionality reduction projection information corresponding to the point cloud information. Based on the mechanical origin information, coordinate transformation is performed on the reduced-dimensional projection information to obtain the projected coordinate information.

5. The rigid body pose calibration method according to claim 1, characterized in that, The step of performing pose calibration processing based on the projected coordinate information and the preset target pose information corresponding to the rigid body calibration object to obtain the pose calibration processing result of the rigid body calibration object includes: The target rotation increment corresponding to the rigid body calibration object is obtained by using the projected coordinate information and the target pose information for dimensionality reduction solution. Based on the target rotation increment, the rotation matrix corresponding to the rigid body calibration object is updated to obtain the pose calibration result.

6. The rigid body pose calibration method according to claim 5, characterized in that, The step of using the projected coordinate information and the target pose information to perform dimensionality reduction to obtain the target rotation increment corresponding to the rigid body calibration object includes: Using the projected coordinate information, the rotation weight coefficient corresponding to each feature point of the rigid body calibration object is determined; Using the projected coordinate information and the target pose information, the local rotation increment of each feature point is determined; Based on the local rotation increment of each feature point and the rotation weight coefficient corresponding to each feature point, a weighted sum is performed to obtain the total target increment. The weights of each feature point are obtained by weighting the rotation weight coefficients corresponding to each feature point. The target rotation increment is obtained by dividing the sum of the target increments by the sum of the weights.

7. The rigid body pose calibration method according to any one of claims 1 to 6, characterized in that, The acquisition of the mechanical installation information of the rigid body calibration object includes: Determine the rigid body calibration object; When the rigid body calibration object is an object device, the equipment installation information and mechanical design information corresponding to the object device are determined as the mechanical installation information.

8. A rigid body pose calibration system, characterized in that, include: The mechanical installation information acquisition module is used to acquire the mechanical installation information of the rigid body calibration object; The locking processing module is used to perform locking processing on the rigid body calibration object based on the mechanical installation information, and obtain the locking spindle direction information and mechanical origin information corresponding to the rigid body calibration object; The projection conversion module is used to perform projection conversion on the point cloud information of the rigid body calibration object based on the locked spindle direction information and the mechanical origin information to obtain the projection coordinate information corresponding to the rigid body calibration object; The pose calibration processing module is used to perform pose calibration processing based on the projected coordinate information and the preset target pose information corresponding to the rigid body calibration object, so as to obtain the pose calibration result of the rigid body calibration object.

9. A rigid body posture calibration device, characterized in that, include: The processor, the communication interface, the memory, and the communication bus; wherein the processor, the communication interface, and the memory communicate with each other through the communication bus. The memory is used to store computer programs; When the processor executes the program stored in the memory, it implements the steps of the rigid body pose calibration method as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the rigid body pose calibration method as described in any one of claims 1-7.

Images