Optical measurement and spatial pose registration method for machining features of a complex-shaped workpiece

Abstract

The application discloses a kind of optical measurement and space pose registration methods of complex shape workpiece processing feature, belong to optical measurement technical field technical field, utilize non-uniform deformation compensation field and constrain state point cloud is corrected as free state equivalent point cloud, according to deformation sensitive parameter distribution registration weight, reduce the interference of local elastic deformation to overall pose solution, compared with the way that existing force state point cloud is directly rigidly registered with free state CAD model, the application can reduce the risk that clamping or support deformation is misjudged as pose deviation, so that the position, profile and topography detection result of processing feature can more accurately reflect the actual geometric deviation of workpiece in free state.

Description

Technical Field

[0001] This invention relates to the field of optical measurement technology, and in particular to an optical measurement and spatial pose registration method for machining features of complex-shaped workpieces. Background Technology

[0002] With the development of manufacturing technologies in aerospace, precision molds, energy equipment, and complex structural components, complex-shaped workpieces such as thin-walled parts, slender parts, overhanging parts, and complex curved surfaces are increasingly widely used in high-end manufacturing. These workpieces typically have characteristics such as thin walls, low structural stiffness, complex local features, and high machining accuracy requirements. After machining, it is necessary to perform three-dimensional inspection on machining features such as holes, slots, edges, and curved surface contours to determine their dimensional accuracy, geometrical accuracy, and machining quality. Compared to contact measurement methods, optical three-dimensional measurement technologies such as structured light scanning, laser scanning, and binocular vision have advantages such as non-contact operation, high efficiency, and wide coverage, and have been gradually applied to the inspection of machining features of complex workpieces. Typically, optical measurement equipment acquires three-dimensional point cloud data of the workpiece surface. Subsequently, this point cloud data needs to be spatially registered with the workpiece's CAD design model to complete machining error analysis, quality evaluation, and compensation machining within the design coordinate system.

[0003] However, in actual measurement processes, complex-shaped workpieces are often under clamping, support, or other constraints. The workpiece is subject to clamping forces, support forces, contact forces, torques, and even its own weight, resulting in elastic deformation. This deformation is particularly pronounced and exhibits localized non-uniformity in thin-walled regions, slender overhangs, locally weak stiffness regions, and free ends far from support points. Existing point cloud registration methods typically focus on the rigid pose deviation between the measured point cloud and the CAD model, solving for rotation and translation through feature matching, ICP, or their improved algorithms. However, these methods struggle to distinguish between rigid pose deviations in the measured point cloud and non-rigid deformation deviations caused by constraint forces. When directly registering the point cloud obtained under constraints with the free-state CAD model, local deformation regions can easily interfere with the overall pose estimation, leading to distorted registration results and consequently affecting the accuracy of machining feature detection results.

[0004] Therefore, it is necessary to design an optical measurement and spatial pose registration method for machining features of complex-shaped workpieces to solve the above problems. Summary of the Invention

[0005] This invention overcomes the shortcomings of the prior art and provides an optical measurement and spatial pose registration method for machining features of complex-shaped workpieces.

[0006] To achieve the above objectives, the technical solution adopted by this invention is: an optical measurement and spatial pose registration method for machining features of complex-shaped workpieces, comprising the following steps:

[0007] Step S1: Obtain the optical three-dimensional point cloud, constraint force data, and workpiece feature data of the workpiece under constraint.

[0008] Step S2: Based on the constraint force data and workpiece characteristic data, determine the deformation-sensitive area and its deformation-sensitive parameters;

[0009] Step S3: Generate a non-uniform deformation compensation field based on the deformation-sensitive region and its deformation-sensitive parameters;

[0010] Step S4: Use a non-uniform deformation compensation field to perform deformation compensation on the optical three-dimensional point cloud to obtain a free-state equivalent point cloud.

[0011] Step S5: Determine the registration weights based on the deformation sensitivity parameters, and perform weighted spatial pose registration between the free state equivalent point cloud and the free state CAD model.

[0012] In a preferred embodiment of the present invention, in step S1, the constraint force data includes at least one of clamping force, supporting force, contact force, and torque data;

[0013] Workpiece feature data includes at least one of geometric parameters, material parameters, and constraint position parameters;

[0014] The geometric parameters include at least one of the following: wall thickness, length, diameter, aspect ratio, radius of curvature, overhang length, and local cross-sectional dimensions;

[0015] Material parameters include at least one of elastic modulus, Poisson's ratio, and density;

[0016] The constraint position parameters include at least one of the clamping point position, support point position, and force direction.

[0017] In a preferred embodiment of the present invention, in step S2, determining the deformation-sensitive region and its deformation-sensitive parameters includes:

[0018] Based on the workpiece feature data and the constraint force data, the registration area of ​​the workpiece to be tested is divided into multiple candidate areas;

[0019] For each candidate region, the deformation sensitivity parameter of the candidate region is calculated based on at least two of the structural dimension parameters, material stiffness parameters, constraint location parameters, and stress parameters of the candidate region.

[0020] Candidate regions whose deformation-sensitive parameters meet preset sensitivity conditions are identified as deformation-sensitive regions.

[0021] In a preferred embodiment of the present invention, the deformation sensitive parameter includes at least one of regional deformation sensitivity, predicted deformation, local stiffness characterization value, compensation resolution parameter and deformation compensation confidence.

[0022] The deformation-sensitive region includes at least one of the following: thin-walled region, slender overhang region, locally weak stiffness region, curvature abrupt change region, free end region far from clamping point or support point, and region where the predicted deformation is greater than a preset deformation threshold.

[0023] In a preferred embodiment of the present invention, in step S3, generating a non-uniform deformation compensation field includes:

[0024] Based on the workpiece feature data, constraint force data, and deformation sensitive parameters, a model matching feature vector of the workpiece to be tested is constructed.

[0025] Based on the model matching feature vector, a target deformation prediction model that matches the workpiece to be tested is determined from a pre-established finite element deformation model library.

[0026] Based on the target deformation prediction model, calculate the three-dimensional deformation distribution of the workpiece under constraint relative to the free state;

[0027] A non-uniform deformation compensation field is generated based on the three-dimensional deformation distribution.

[0028] In a preferred embodiment of the present invention, the finite element deformation model library includes multiple finite element deformation models, each of which is used to characterize the correspondence between workpiece geometric parameters, material parameters, constraint conditions, force data and three-dimensional deformation of the workpiece surface;

[0029] The model matching feature vector includes at least two of the following: wall thickness, aspect ratio, overhang length, local cross-sectional dimensions, material elastic modulus, clamping point location, support point location, force magnitude, and force direction.

[0030] In a preferred embodiment of the present invention, when there is no finite element deformation model in the finite element deformation model library that perfectly matches the workpiece to be tested, the target deformation prediction model is determined in the following manner:

[0031] Based on the model matching feature vector, select multiple similar finite element deformation models from the finite element deformation model library;

[0032] Calculate the parameter similarity between the workpiece under test and each similar finite element deformation model;

[0033] The interpolation weights of similar finite element deformation models are determined based on parameter similarity.

[0034] The three-dimensional deformation fields output by each similar finite element deformation model are fused according to the interpolation weights to generate a target deformation prediction model adapted to the workpiece to be tested.

[0035] In a preferred embodiment of the present invention, in step S3, the non-uniform deformation compensation field includes a first compensation field corresponding to the deformation-sensitive region and a second compensation field corresponding to the non-deformation-sensitive region.

[0036] The first compensation field has higher deformation prediction accuracy, spatial resolution, and compensation weight than the second compensation field.

[0037] The first compensation field is used to perform key deformation compensation on measuring points in the deformation-sensitive area, while the second compensation field is used to perform simplified deformation compensation or low-weight deformation compensation on measuring points in the non-deformation-sensitive area.

[0038] In a preferred embodiment of the present invention, step S5 includes:

[0039] Based on the deformation sensitivity parameters and deformation compensation reliability of the region to which the measurement point belongs in the free state equivalent point cloud, the registration weight of each measurement point is determined.

[0040] Construct a weighted registration objective function based on the registration weights;

[0041] The weighted registration objective function is optimized to obtain the rotation and translation parameters of the free-state equivalent point cloud relative to the free-state CAD model.

[0042] The spatial pose transformation result is obtained based on the rotation and translation parameters.

[0043] In a preferred embodiment of the present invention, after obtaining the spatial pose transformation result, the method further includes:

[0044] Calculate the registration residual between the free-state equivalent point cloud and the free-state CAD model;

[0045] The reliability of deformation compensation is updated based on the registration residual;

[0046] The registration weights are adjusted based on the updated deformation compensation confidence, and the non-uniform deformation compensation field is corrected.

[0047] Spatial pose registration is re-executed using the adjusted registration weights and the corrected non-uniform deformation compensation field until the preset convergence condition is met.

[0048] This invention addresses the shortcomings of the prior art and has the following beneficial effects:

[0049] (1) This invention obtains optical three-dimensional point cloud, constraint force data and workpiece feature data under constraint state, and generates non-uniform deformation compensation field. Since the force will cause elastic displacement of the deformation sensitive area with spatial position change, and the displacement cannot be eliminated by simple rotation and translation, the non-uniform deformation compensation field is used to correct the constraint state point cloud to the free state equivalent point cloud. The registration weight is allocated according to the deformation sensitive parameters to reduce the interference of local elastic deformation on the overall pose solution. Compared with the existing method of directly rigidly registering the force state point cloud with the free state CAD model, this invention can reduce the risk of clamping or support deformation being misjudged as pose deviation, so that the position, contour and shape detection results of the processing features can more accurately reflect the actual geometric deviation of the workpiece in the free state.

[0050] (2) The present invention determines the deformation-sensitive region and its deformation-sensitive parameters based on the constraint force data and workpiece feature data, so that the deformation compensation is concentrated on the local area that has a greater impact on the registration error. Since the deformation-sensitive region is more likely to bend, twist or deflect under the same force conditions, its measurement point offset is usually significantly greater than that of the high stiffness region. By distinguishing the degree of stress deformation in different regions through the deformation-sensitive parameters, the same compensation strategy can be avoided for the entire point cloud. Compared with the existing global uniform compensation or only rigid registration, the present invention reduces the pulling effect of a small number of large deformation measurement points on the overall pose solution, and reduces the possibility of incorrect rotation or incorrect translation in the registration result.

[0051] (3) The present invention generates a non-uniform deformation compensation field based on the deformation-sensitive region and its deformation-sensitive parameters, and makes the compensation accuracy, spatial resolution or compensation weight of the deformation-sensitive region higher than that of the non-deformation-sensitive region. The measuring points in the deformation-sensitive region are more affected by the constraint force, and their local displacement will directly change the spatial position of the processing features such as holes, slots, edges and curved surfaces in the point cloud. Therefore, more refined deformation correction is required. Taking advantage of the small force displacement in the non-deformation-sensitive region, the simplified compensation can maintain the stability of the original geometric constraint in the region. Compared with the method of using the same compensation accuracy for all regions, the present invention obtains more effective displacement correction in the large deformation region, while reducing the additional deviation caused by over-compensation in the high stiffness region. Attached Figure Description

[0052] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0053] Figure 1 This is a flowchart of a preferred embodiment of the present invention;

[0054] Figure 2 This is a flowchart of the feedback correction process based on registration residuals of the present invention. Detailed Implementation

[0055] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0056] This invention is applicable to optical three-dimensional inspection scenarios of machining features of complex-shaped workpieces, especially thin-walled parts, slender parts, overhanging parts, complex curved surface parts, and workpieces with local weak stiffness.

[0057] During the inspection process, such workpieces usually need to be constrained by clamps, supports, positioning parts or clamping parts. After the workpiece is subjected to clamping force, supporting force, contact force or torque, the thin-walled area, the overhanging area and the free end area far away from the support point are prone to bending, torsion or deflection deformation.

[0058] The point cloud obtained by the optical measuring equipment in this state is not the geometric shape of the workpiece in its free state, but includes the local elastic displacement caused by the constraint force. If the point cloud is directly rigidly registered with the free state CAD model, the local large deformation area will be treated as an ordinary registration point to participate in the solution of rotation and translation parameters, thereby causing the overall pose matrix to be pulled by the local deformation residual.

[0059] Based on this, the present invention identifies deformation-sensitive areas by constraining force data and workpiece feature data, generates a non-uniform deformation compensation field, and performs weighted spatial pose registration after compensation, so that the constraint deformation error is weakened before entering the pose solution and further suppressed in the registration solution.

[0060] The error addressed in this invention is not ordinary random measurement noise, but rather a local elastic displacement formed by clamping force, support force, contact force, or torque transmitted through the workpiece structure. This local elastic displacement has a clear mechanical origin and is related to wall thickness, overhang length, local stiffness, clamping point position, support point position, and force direction. When this local elastic displacement is superimposed on the optical 3D point cloud, it creates a local spatial residual between the point cloud and the CAD model. If this residual is treated the same as ordinary pose residual in the rigid registration objective function, it will change the solution direction of the rotation matrix and translation vector. This invention constructs a suppression mechanism for this error propagation path through deformation-sensitive region identification, non-uniform deformation compensation, weighted registration based on deformation-sensitive parameters, and residual feedback correction.

[0061] Example 1:

[0062] like Figure 1 and Figure 2 As shown, an optical measurement and spatial pose registration method for machining features of complex-shaped workpieces includes the following steps:

[0063] Step S1: Obtain the optical three-dimensional point cloud, constraint force data, and workpiece feature data of the workpiece under constraint.

[0064] Step S2: Based on the constraint force data and workpiece characteristic data, determine the deformation-sensitive area and its deformation-sensitive parameters;

[0065] Step S3: Generate a non-uniform deformation compensation field based on the deformation-sensitive region and its deformation-sensitive parameters;

[0066] Step S4: Use a non-uniform deformation compensation field to perform deformation compensation on the optical three-dimensional point cloud to obtain a free-state equivalent point cloud.

[0067] Step S5: Determine the registration weights based on the deformation sensitivity parameters, and perform weighted spatial pose registration between the free-state equivalent point cloud and the free-state CAD model:

[0068] In this embodiment, the constrained state refers to the state in which the workpiece under test is subjected to external clamping, support, positioning or pressing during the optical measurement process.

[0069] Optical 3D point cloud refers to the set of 3D coordinates of a workpiece surface obtained through structured light scanning, line laser scanning, area laser scanning, binocular vision measurement, or multi-view vision measurement.

[0070] Constraint force data is used to describe the external mechanical forces acting on the workpiece under measurement conditions, while workpiece characteristic data is used to describe the structural and material basis for the workpiece's resistance to deformation.

[0071] The basic concept of this embodiment is to incorporate the stress conditions and structural features of the workpiece under constraints into the point cloud registration process. Traditional point cloud registration typically assumes that there are mainly rigid pose differences between the measured point cloud and the CAD model, and therefore aligns them through rotation and translation. For workpieces with complex shapes, especially thin-walled areas, overhanging areas, and areas with local weak stiffness, this assumption is not entirely valid.

[0072] These regions will experience spatially related elastic displacements under clamping forces, supporting forces, or torques. These elastic displacements cannot be eliminated by simple rigid body rotation and translation. In this embodiment, regions prone to deformation are first identified based on force and structural information. Then, non-uniform deformation compensation is performed on these regions. During the registration process, weights are assigned based on deformation sensitivity parameters, so that the pose matrix is ​​determined more by regions with smaller deformations or higher compensation reliability.

[0073] In this embodiment, the deformation-sensitive parameter is not simply a label used to identify the region type, but rather an intermediate technical quantity that is continuously used in the subsequent generation of compensation fields and determination of registration weights.

[0074] Specifically, the deformation-sensitive parameter is used to determine whether a certain area needs high-resolution compensation, to affect the spatial density of the compensation field in that area, to participate in the registration weight calculation of the measurement points in that area, and can also be used in the residual feedback process to determine whether the compensation reliability of that area needs to be updated.

[0075] Thus, a continuous data association is formed between deformation-sensitive region identification, non-uniform compensation, and weighted registration, rather than sequential processing steps that are independent of each other.

[0076] The difficulty to be solved in implementing this invention is that the deformation caused by the constraint force is not a uniform displacement of the entire workpiece. When the same clamping force is applied to areas with different wall thicknesses, different overhang lengths or different support positions, it will produce elastic displacements with different directions and amplitudes.

[0077] If only the point cloud is translated or rotated as a whole, local deformation cannot be eliminated; if the same compensation amount is applied to all regions, new compensation deviations are easily introduced in high-stiffness regions. This embodiment establishes regional differences by using deformation-sensitive regions and deformation-sensitive parameters, and then applies different compensation methods to different regions through a non-uniform deformation compensation field, thereby providing a more stable point cloud basis for subsequent weighted registration.

[0078] Example 2:

[0079] The technical solution in this embodiment is a further refinement based on the above embodiment 1.

[0080] In step S1, acquiring the optical three-dimensional point cloud, constraint force data, and workpiece feature data of the workpiece under constraint means simultaneously obtaining the three-dimensional morphology of the workpiece surface, the external load on the workpiece, and the geometric and material information of the workpiece itself while the workpiece is kept in the actual detection clamping or support posture.

[0081] This step is not about collecting point clouds separately, but about providing force and structural evidence for subsequent determination of the source of local offsets in the point cloud.

[0082] The constraint force data includes at least one of the following: clamping force, support force, contact force, and torque data.

[0083] The clamping force is generated by the clamping jaws, clamping blocks, flexible clamping units, or clamping mechanisms.

[0084] The supporting force can be generated by the support block, support pin, support seat or auxiliary support mechanism.

[0085] The contact force comes from the positioning surface, positioning pin, pressure block, adsorption surface or limiting component.

[0086] Torque data is generated by the existence of a lever arm of the applied force relative to the workpiece reference point, or it can be directly measured by a six-dimensional force sensor.

[0087] Workpiece feature data includes at least one of geometric parameters, material parameters, and constraint position parameters.

[0088] The geometric parameters include at least one of the following: wall thickness, length, diameter, aspect ratio, radius of curvature, overhang length, and local cross-sectional dimensions.

[0089] Material parameters include at least one of elastic modulus, Poisson's ratio, and density.

[0090] The constraint position parameters include at least one of the clamping point position, support point position, and force direction.

[0091] In this embodiment, step S1 can be performed in the following manner.

[0092] Step S11: Install the workpiece to be measured on the testing fixture, so that it is in a clamping state, a supporting state, or a state in which clamping and supporting work together, consistent with the actual measurement.

[0093] For thin-walled curved parts, the clamping position can be set near the process edge, reinforcing rib, or positioning boss; for slender workpieces, the support position can be set at both ends or in the middle support area; for workpieces with holes, grooves, or complex contours, the constraint position can be set corresponding to the process datum or inspection datum.

[0094] Step S12: Collect constraint force data using force sensors installed in the clamp, support, or contact area.

[0095] The force sensor can be a thin-film pressure sensor, a piezoelectric force sensor, a strain gauge force sensor, or a six-dimensional force sensor.

[0096] For workpieces with multiple clamping or multiple support points, the force values ​​of each clamping or support point can be recorded separately, and each force value can be associated with its position in the workpiece coordinate system and saved.

[0097] If the sensor can only output unidirectional pressure, then that pressure can be used as a local normal contact force. If the sensor can output multi-directional forces and torques, then the multi-directional forces and torques can be used together as constraint force data.

[0098] When collecting constraint force data, a steady-state judgment process needs to be set up before optical scanning begins.

[0099] Specifically, when the change in clamping force, supporting force, or contact force is less than a preset fluctuation threshold within several consecutive sampling periods, the workpiece is determined to be in a stable constraint state, and the force data under this stable constraint state is used as the input for subsequent deformation-sensitive area identification and compensation field generation.

[0100] This process can prevent the impact load at the moment of clamping or the transient force value during the fine adjustment of the fixture from entering the deformation prediction process, so that the compensation field corresponds to the stress state during the actual scanning.

[0101] Step S13: Use an optical three-dimensional measurement device to collect point cloud data of the processing feature area of ​​the workpiece to be measured and its adjacent areas.

[0102] The machining feature area may include at least one of holes, slots, steps, edges, curved contours, end faces, assembly surfaces, and positioning surfaces.

[0103] The scanning range preferably covers the processing feature itself and its surrounding transition area, because the curvature changes, edge contours and adjacent support areas near the processing feature can provide geometric constraints for subsequent registration.

[0104] Step S14: Obtain workpiece feature data from CAD models, process documents, material databases, or manually entered information.

[0105] Wall thickness can be obtained from the distance between relative curved surfaces in the CAD model; overhang length can be determined based on the distance from the region to the nearest support or clamping point; local cross-sectional dimensions can be obtained through cross-sectional analysis; elastic modulus, Poisson's ratio, and density can be obtained from the material database corresponding to the material grade. For workpieces with different material combinations, material parameters should be recorded according to material zones.

[0106] Step S15: Perform corresponding processing on the optical 3D point cloud, constraint force data, and workpiece feature data.

[0107] Corresponding processing includes coordinate unification, regional correspondence, and data collection time period correspondence.

[0108] Coordinate unification is achieved through fixture reference surfaces, positioning holes, calibration blocks, or calibration balls, ensuring that the force sensor position, point cloud coordinates, and CAD model are in a mutually convertible coordinate relationship.

[0109] Region correspondence refers to mapping each workpiece feature data to a specific location in the region to be registered.

[0110] The corresponding acquisition time period refers to selecting force data that matches the point cloud scanning time period, avoiding the use of transient force values ​​during the clamping process to replace force values ​​under stable measurement conditions.

[0111] During the coordinate unification process, the fixture coordinate system is used as the intermediate coordinate system.

[0112] The installation position of the force sensor is first calibrated to the fixture coordinate system. The optical 3D point cloud is transformed to the fixture coordinate system through the reference ball, reference hole or reference surface on the fixture. The CAD model is then established to correspond with the fixture positioning reference through the process reference.

[0113] Through this intermediate coordinate system, the position of the constraint force data, the position of the measurement point in the optical 3D point cloud, and the surface position of the CAD model can be mapped to the same spatial relationship. This process provides a basis for subsequent determination of the distance between a measurement point and the clamping or support point, the direction of the force, and the local stiffness.

[0114] Through step S1 above, subsequent processing no longer only deals with isolated point cloud coordinates, but simultaneously obtains the possible force and structural sources corresponding to the point cloud offset. Thus, when identifying local deformation, it can distinguish between high-stiffness and low-stiffness regions, near-support and far-support regions, and lightly loaded and heavily loaded regions, providing basic data for subsequent determination of deformation-sensitive areas.

[0115] In step S2, based on the constraint force data and workpiece characteristic data, the deformation-sensitive region and its deformation-sensitive parameters are determined. This means finding the region in the registration area that is more likely to produce local elastic displacement after being subjected to constraint force, and calculating parameters for this region that can characterize its deformation tendency and compensation requirements.

[0116] In this embodiment, the deformation-sensitive region may include at least one of the following: a thin-walled region, a slender overhanging region, a region with local weak stiffness, a region with abrupt curvature change, a free end region far from the clamping point or support point, and a region where the predicted deformation is greater than a preset deformation threshold.

[0117] Deformation sensitivity parameters may include at least one of the following: regional deformation sensitivity, predicted deformation, local stiffness characterization value, compensation resolution parameter, and deformation compensation confidence.

[0118] Regional deformation sensitivity is used to indicate the probability of displacement amplification in a certain region after being affected by constraint force; predicted deformation is used to indicate the three-dimensional displacement amplitude that the region may produce under the current constraint force; local stiffness characterization value is used to indicate the ability of the region to resist deformation; compensation resolution parameter is used to determine the density of the compensation field in the region; deformation compensation reliability is used to indicate the reliability of the compensation result in the registration solution.

[0119] In this embodiment, step S2 can be performed in the following manner.

[0120] Step S21: Based on the surface partitioning, machining feature boundaries, finite element mesh or point cloud distribution of the workpiece CAD model, divide the registration area of ​​the workpiece to be tested into multiple candidate areas.

[0121] Candidate regions can correspond to a CAD patch, a machining feature neighborhood, a curved mesh block, or a point cloud clustering region.

[0122] For local machining features such as holes, grooves, and edges, the machining feature itself and its surrounding area are considered as a candidate region; for continuous curved surfaces, candidate regions are divided according to curvature changes or grid scale.

[0123] The scale for dividing the candidate region is determined based on the average spacing of the point cloud, the size of the processing features, and local curvature changes. If the candidate region is too large, thin-walled and high-stiffness regions may be included in the same region, resulting in the averaged deformation sensitivity of the region; if the candidate region is too small, point cloud noise or missing local scans may have an excessive impact on the deformation-sensitive parameters. Therefore, the candidate region should preferably cover a sufficient number of measurement points, and its boundaries should be consistent with the boundaries of the processing features, abrupt curvature changes, or structural cross-sectional changes as much as possible.

[0124] Step S22: For each candidate region, extract the structural dimension parameters, material stiffness parameters, constraint location parameters, and force parameters of that region.

[0125] Structural dimensional parameters include wall thickness, local cross-sectional dimensions, overhang length, or radius of curvature. Material stiffness parameters may include elastic modulus and Poisson's ratio.

[0126] The constraint position parameters include the distance from the candidate region to the nearest clamping or support point, and the positional relationship of the candidate region relative to the force direction.

[0127] The force parameters include the clamping force, supporting force, contact force, or equivalent torque acting near the region.

[0128] Step S23: Perform dimension consistency processing on the parameters of the candidate region.

[0129] Since the units for wall thickness, force, length, and material stiffness are different, this embodiment can perform dimensionless processing on each parameter before calculating the deformation-sensitive parameters.

[0130] The purpose of dimensionless evaluation is to enable different physical quantities to be included in the same evaluation system, rather than simply adding parameters of different units together.

[0131] In one specific implementation, the wall thickness influence term can be determined based on the ratio of the region reference wall thickness to the current region wall thickness, so that regions with smaller wall thicknesses correspond to larger wall thickness influence terms; the overhang length influence term can be determined based on the ratio of the distance from the current region to the nearest support point or clamping point to the maximum distance within the region to be registered, so that regions far from the constraint point correspond to larger overhang influence terms; the force influence term can be determined based on the ratio of the equivalent load corresponding to the current region to the maximum equivalent load in all candidate regions; the local stiffness influence term can be determined based on the reciprocal of the region's unit load displacement or equivalent stiffness, so that low-stiffness regions correspond to larger stiffness influence terms; and the constraint distance influence term can be determined based on the spatial distance from the region's center of gravity to the nearest clamping point or support point.

[0132] Through the above processing, all influencing terms are converted into dimensionless quantities, which can avoid calculation distortion caused by the direct superposition of parameters with different units.

[0133] Step S24: Calculate the deformation sensitivity parameters based on the structural dimensions, material stiffness, constraint locations, and stress states of the candidate regions. The deformation sensitivity of the regions related to the core inventive points of this invention is expressed mathematically as follows: .

[0134] in, This represents the regional deformation sensitivity of the j-th candidate region; This indicates the effect of wall thickness; the smaller the wall thickness, the larger this term. This indicates the effect of overhang length; the longer the overhang, the larger this item becomes. This indicates the effect of force; the greater the equivalent force, the larger this term. This represents the local stiffness influence term; the lower the local stiffness, the larger this term. This term represents the influence of the distance from the candidate region to the clamping or support point; the greater the distance, the larger the term. , , , , Let be the weighting coefficient, and satisfy: .

[0135] All of the above-mentioned influencing factors are dimensionless. For thin-walled workpieces, this can be improved. and The value of [value] allows wall thickness and local stiffness to account for a larger proportion in the sensitivity assessment; for slender cantilever components, it can improve [the sensitivity assessment]. and The corresponding weights make the effects of cantilever length and distance from the support more pronounced; for clamping scenarios with significant force changes, this can improve... The value of is determined based on the workpiece type, inspection accuracy, and historical simulation results.

[0136] Local stiffness influence term It is obtained by simplifying mechanical calculations or finite element unit load calculations.

[0137] When using the finite element method with a unit load, a unit normal load is applied to the j-th candidate region, and the average displacement of that region is calculated. The larger the average displacement under a unit load, the lower the local stiffness of that region. The larger the value, the better.

[0138] When using simplified mechanical calculations, for approximate plate and shell structures, the local bending capacity is estimated by combining the elastic modulus, wall thickness, and regional characteristic length; for approximate beam structures, the local bending capacity is estimated by combining the cross-sectional inertial characteristics and overhang length.

[0139] This local stiffness influence term is used to incorporate the structure's resistance to deformation into the assessment of deformation-sensitive areas, rather than relying solely on the magnitude of the force to determine whether a region needs compensation.

[0140] Step S25: Determine the candidate regions whose deformation sensitivity parameters meet the preset sensitivity conditions as deformation sensitive regions.

[0141] The preset sensitivity conditions are: regional deformation sensitivity is greater than a preset sensitivity threshold, predicted deformation is greater than a preset deformation threshold, and local stiffness is lower than a preset stiffness threshold.

[0142] Through the above step S2, this embodiment can identify the regions that are prone to carrying constraint deformation errors into the pose solution before registration, avoiding treating all point cloud measurement points as equally reliable, and also avoiding the influence of a small number of large deformation regions on the overall pose matrix during the subsequent registration process.

[0143] For example, when the position detection error of the machining feature is small and the optical measurement repeatability is close to the allowable error, the predicted deformation threshold is set to the same or slightly higher order of magnitude as the measurement repeatability, so that compensation can be performed before the local elastic displacement significantly exceeds the machining tolerance.

[0144] If the predicted deformation of a certain area is much smaller than the measurement noise level, then even if there is slight elastic displacement in that area, it should not be considered as a key compensation area, so as to avoid introducing new computational disturbances during the compensation process. Therefore, a correspondence is established between preset sensitivity conditions and detection accuracy.

[0145] In step S3, generating a non-uniform deformation compensation field based on the deformation-sensitive region and its deformation-sensitive parameters refers to calculating the three-dimensional displacement compensation amount required to correct each measuring point in the optical three-dimensional point cloud from a constrained state to an equivalent free state, based on the location, stress state, and local deformation resistance of the deformation-sensitive region. This compensation field varies with spatial location and is not a compensation amount in a single direction or with a single amplitude.

[0146] In this embodiment, step S3 can be performed in the following manner.

[0147] Step S31: Based on the workpiece feature data, constraint force data, and deformation sensitive parameters, construct the model matching feature vector of the workpiece to be tested.

[0148] The model matching feature vector includes at least two of the following: wall thickness, aspect ratio, overhang length, local cross-sectional dimensions, material elastic modulus, clamping point location, support point location, force magnitude, and force direction.

[0149] Model matching feature vectors are used to find models in the finite element deformation model library that are similar to the structure and stress state of the workpiece under test.

[0150] Before entering the matching calculation, all parameters in the model matching feature vector are uniformly converted into dimensionless parameters.

[0151] For parameters with large numerical ranges, such as the elastic modulus and the magnitude of force, normalization can be performed using the maximum and minimum values ​​in the model library. For directional parameters, such as the direction of force, direction angles or direction cosines can be used. For positional parameters, such as the positions of clamping points and support points, normalized coordinates relative to the workpiece's reference coordinate system can be used. After this processing, the model matching process can simultaneously consider geometry, material, force, and constraint positions, without the influence of other parameters being masked by the large numerical scale of one parameter.

[0152] Step S32: Based on the model matching feature vector, determine the target deformation prediction model that matches the workpiece to be tested from the pre-established finite element deformation model library.

[0153] The finite element deformation model library includes multiple finite element deformation models. Each finite element deformation model is used to characterize the correspondence between workpiece geometric parameters, material parameters, constraints, force data and the three-dimensional deformation of the workpiece surface.

[0154] Finite element deformation models can be established through offline simulation, and the parameter range, constraints, load conditions, and surface deformation distribution corresponding to the model can be stored in the model library.

[0155] The finite element deformation model library can be built offline following the procedure below.

[0156] First, select typical workpiece types as modeling objects. Typical workpiece types include at least one of thin-walled plate shell parts, slender beam parts, thin-walled curved surface parts, perforated thin-walled parts, and partially overhanging structural parts.

[0157] A three-dimensional model is established based on the geometric parameters of each typical workpiece, and the elastic modulus, Poisson's ratio, and density are set according to the actual material of the workpiece.

[0158] Set boundary conditions according to the actual clamping, support or positioning method, and apply loads of different magnitudes and directions at the clamping point, support point or contact area.

[0159] After completing the finite element solution, the three-dimensional displacement of the workpiece surface nodes under the constraint state relative to the free state is extracted, and the three-dimensional displacement, along with the corresponding geometric parameters, material parameters, constraint position parameters, and force parameters, is stored in the finite element deformation model library.

[0160] Therefore, each finite element deformation model in the model library is not a separate geometric model, but a complete mapping unit containing input parameters, boundary conditions, and surface deformation output results.

[0161] During the establishment of the finite element deformation model library, typical working conditions can be parametrically extended. For example, under the same workpiece type, different wall thicknesses, different overhang lengths, different clamping point positions, different support point positions, and different clamping force magnitudes can be set to form multiple model nodes.

[0162] For each model node, its model matching feature vector and surface three-dimensional deformation field are recorded. During on-site inspection, even if the parameters of the workpiece to be tested are not completely consistent with a certain model node, the deformation prediction result of the current workpiece can be obtained by interpolation and fusion of multiple similar models.

[0163] When a finite element deformation model exists in the model library that perfectly matches or meets the matching requirements of the workpiece to be tested, the model is directly used as the target deformation prediction model.

[0164] When no perfectly matching model exists in the model library, multiple similar finite element deformation models can be selected based on the model matching feature vectors and then interpolated and fused. The interpolation weights related to the core inventive points of this invention can be determined in the following way: .

[0165] in, This represents the interpolation weight of the m-th similar finite element deformation model; This represents the parametric distance between the workpiece under test and the m-th nearest finite element deformation model; represents the distance scale parameter; k represents the number of similar finite element deformation models participating in the interpolation fusion.

[0166] The smaller the parameter distance, the closer the model is to the workpiece in terms of geometry, material properties, and stress conditions, and the greater its interpolation weight. The sum of all interpolation weights is 1.

[0167] Parameter distance The model is determined based on the differences in dimensionless parameters in each dimension of the matching feature vector. For numerical parameters such as wall thickness, aspect ratio, overhang length, material elastic modulus, clamping force, and support force, the normalized difference between the workpiece and the model nodes is calculated; for spatial parameters such as clamping point position and support point position, the normalized spatial distance is calculated; and for the force direction, the difference in the directional angle is calculated.

[0168] The differences between the parameters can be combined into a parameter distance according to the preset weights. This parameter distance reflects the degree of similarity between the workpiece under test and the model nodes in the model library in terms of structure, material and constraint conditions. Therefore, it can be used as the basis for calculating the interpolation weights.

[0169] After determining the interpolation weights, the three-dimensional deformation fields output by multiple similar finite element deformation models are fused: .

[0170] in, This represents the three-dimensional deformation at position x on the surface of the workpiece to be measured. This represents the three-dimensional deformation of the m-th similar finite element deformation model at the corresponding position; This represents the interpolation weights corresponding to the m-th similar model.

[0171] If different models have different surface node distributions, the deformation fields of each model are first mapped to the same surface position through feature point correspondence, surface projection, or normalized surface parameters before fusion.

[0172] For thin-walled curved parts, the center of the positioning hole, the edge contour point, the curvature extreme point or the process reference point are selected as the corresponding feature points, and a surface mapping relationship is established between the similar finite element deformation model and the workpiece to be tested.

[0173] For slender parts, normalized axial coordinates are established along the length direction, and deformation field mapping is performed in combination with the circumferential position of the cross section.

[0174] For thin-walled parts with openings, the hole center, hole edge, and adjacent curved surface regions are used as local mapping constraints. Through this mapping method, the three-dimensional deformation fields output by multiple similar finite element deformation models can correspond to the same surface region of the workpiece under test, avoiding the erroneous fusion of displacements at different spatial locations.

[0175] Step S33: Calculate the three-dimensional deformation distribution of the workpiece under constraint relative to the free state based on the target deformation prediction model.

[0176] This three-dimensional deformation distribution is used to describe the spatial displacement of various positions on the workpiece surface caused by clamping, supporting, or contact forces.

[0177] For deformation-sensitive areas, it is preferable to use higher density compensation nodes and higher accuracy deformation prediction; for non-deformation-sensitive areas, lower density compensation nodes or regional average compensation methods are used.

[0178] Step S34: Generate a non-uniform deformation compensation field based on the three-dimensional deformation distribution.

[0179] The non-uniform deformation compensation field includes a first compensation field corresponding to the deformation-sensitive region and a second compensation field corresponding to the non-deformation-sensitive region.

[0180] The first compensation field has higher deformation prediction accuracy, spatial resolution, and compensation weight than the second compensation field.

[0181] The first compensation field is used to perform key deformation compensation on measuring points in the deformation-sensitive area, while the second compensation field is used to perform simplified deformation compensation or low-weight deformation compensation on measuring points in the non-deformation-sensitive area.

[0182] The spacing between compensation nodes in the first compensation field is determined based on deformation-sensitive parameters.

[0183] The larger the deformation sensitivity parameter, the higher the possibility of local elastic displacement in the region after being affected by the constraint force. Therefore, a smaller compensation node spacing and a higher compensation resolution should be used in this region.

[0184] In non-deformation-sensitive areas, the spacing between compensation nodes can be appropriately increased, or a regional average compensation amount can be used. This differentiated setting allows the compensation calculation to focus on areas where deformation has a significant impact on registration, while avoiding additional disturbances in high-stiffness areas due to excessively refined compensation.

[0185] In this embodiment, the difference between the first compensation field and the second compensation field is not whether compensation is performed, but rather the difference in the level of precision of compensation.

[0186] In deformation-sensitive areas, local elastic displacement has a significant impact on the spatial position of holes, slots, edges, and curved surface profiles, thus requiring finer compensation nodes and higher compensation weights.

[0187] In non-deformation-sensitive regions, local displacements are small, and these regions often provide relatively stable geometric constraints for pose registration, so excessive correction is not advisable. Non-uniform compensation methods can reduce undercompensation in large deformation regions and overcompensation in high-stiffness regions.

[0188] In step S4, deformation compensation is performed on the optical three-dimensional point cloud using a non-uniform deformation compensation field to obtain a free-state equivalent point cloud. This means that the coordinates of the point cloud under constrained conditions are spatially shifted according to the compensation field. The point cloud compensation relationship related to the core inventive point of this invention can be expressed as follows: .

[0189] in, This represents the i-th measurement point in the optical 3D point cloud under constrained conditions. This indicates the non-uniform deformation compensation field at the measuring point. The corresponding three-dimensional compensation amount; This represents the equivalent point of the free state after compensation.

[0190] If the target deformation prediction model outputs the displacement of the constrained state relative to the free state, the compensation amount is opposite to the direction of that displacement; if the target deformation prediction model outputs the displacement that is pushed back from the constrained state to the free state, it is directly used as the compensation amount.

[0191] In this embodiment, step S4 can be performed in the following manner.

[0192] Step S41: Establish the spatial correspondence between each measurement point in the optical 3D point cloud and the non-uniform deformation compensation field. This correspondence can be obtained through CAD surface projection, nearest neighbor lookup, mesh cell positioning, or machining feature region mapping.

[0193] In one specific implementation, the optical 3D point cloud under constraint is first coarsely positioned using a fixture reference or process reference, so that the point cloud and the CAD model are approximately corresponding.

[0194] Project each measurement point onto the surface of the CAD model to determine the CAD patch, finite element surface element, or compensation field mesh element in which it is located.

[0195] If the measuring point is located between multiple compensation nodes, the three-dimensional compensation amount corresponding to the measuring point is obtained by interpolation based on the local coordinates of the measuring point in the unit.

[0196] If the measuring point is located at the edge of the machining feature or at a sudden change in curvature, compensation nodes within the same feature area should be used for interpolation to avoid averaging compensation amounts across edges or hole boundaries.

[0197] Step S42: Read or interpolate the corresponding compensation amount for each measurement point.

[0198] For deformation-sensitive regions, surface interpolation, finite element shape function interpolation, or radial basis interpolation are used to obtain the compensation amount; for non-deformation-sensitive regions, neighborhood averaging compensation or low-resolution compensation are used.

[0199] When using finite element shape function interpolation, after the measuring point is projected onto the finite element surface element, the compensation amount is obtained by interpolating the displacements of adjacent nodes based on its position within the element.

[0200] When using surface interpolation, several compensation nodes are selected in the neighborhood of the measurement point, and the interpolation weights are determined based on the spatial distance and the consistency of the surface normal.

[0201] When the neighborhood of the measuring point crosses the edge of the hole, the edge of the edge, or the boundary of the curvature change, the interpolation nodes should be restricted to come from the same continuous surface region to avoid the compensation amounts of different deformation modes from mixing with each other.

[0202] Step S43: Correct the coordinates of the measuring points according to the compensation amount to obtain the equivalent point cloud in the free state. This process only changes the spatial position of the measuring points and does not change the basic topological relationship of machining features such as holes, slots, edges, and curved surface contours in the point cloud.

[0203] Step S44: Perform a compensation continuity check on the compensated point cloud. If the compensation amount of a certain measurement point changes abruptly relative to its neighboring measurement points, and the measurement point is not located in a region of actual edges, sharp corners, or curvature abrupt changes, then smoothing is performed using the neighboring compensation amount. The purpose is to remove abnormal compensation values ​​caused by model mapping or measurement noise, rather than smoothing the processed features themselves.

[0204] Compensation continuity checks can be performed on a region-by-region basis, rather than uniformly smoothing the entire point cloud.

[0205] Within the deformation-sensitive area, check whether there is an abnormal reversal of the compensation direction of adjacent measuring points, and whether the compensation amplitude deviates significantly from the median value of the neighborhood; for non-deformation-sensitive areas, check whether the compensation amplitude exceeds the simplified compensation range allowed for that area.

[0206] If the abnormal compensation point is located near the edge of a hole, groove, or edge, it should be determined whether it is a real geometric change in conjunction with the CAD feature boundary to avoid misprocessing the real machining feature as a compensation abnormality.

[0207] In step S4, the spatial offset caused by local elastic displacement in the constrained state point cloud is reduced. When the compensated free state equivalent point cloud participates in registration again, the interference of local deformation regions on the overall pose solution is reduced.

[0208] In step S5, the registration weights are determined based on the deformation-sensitive parameters, and the free-state equivalent point cloud and the free-state CAD model are registered with weighted spatial pose. This means that in the registration solution, different weights are assigned according to the deformation state and compensation reliability of the region to which the measurement point belongs, so that different measurement points contribute differently to the rotation and translation parameters.

[0209] In this embodiment, step S5 can be performed in the following manner.

[0210] Step S51: Determine the registration weight of each measurement point based on the deformation sensitivity parameters and deformation compensation reliability of the region to which the measurement point belongs in the free-state equivalent point cloud. The registration weight related to the core inventive point of this invention can be expressed as: .

[0211] in, This represents the registration weight of the i-th measurement point; This indicates the reliability of deformation compensation in the area to which the measuring point belongs; This indicates the deformation-sensitive parameter of the area to which the measuring point belongs; This represents the sensitivity inhibition coefficient.

[0212] The higher the reliability of deformation compensation, the more suitable the measuring point is to participate in pose solving; the larger the deformation sensitivity parameter, the more obvious the influence of local elastic deformation on the measuring point, so its weight is appropriately reduced.

[0213] In actual use, Limiting the weights within a preset range avoids instability in registration caused by excessively high or low weights for individual measurement points.

[0214] The purpose of the aforementioned weight calculation is to control the contribution of the residual at each measurement point to the pose matrix solution. For regions with large deformation-sensitive parameters and low compensation reliability, even if a large residual still exists after compensation, the role of this residual in the registration objective function will be reduced, thereby decreasing its influence on the rotation matrix and translation vector.

[0215] For regions with small deformation, high compensation reliability, or stable geometric constraints, the measurement points can maintain a high weight, continuing to provide reliable constraints for the overall pose solution. Therefore, this invention does not simply eliminate deformation-sensitive regions, but rather limits their error propagation capability while retaining some of their geometric information.

[0216] Step S52: Construct a weighted registration objective function based on the registration weights. The weighted spatial pose registration related to the core inventive point of this invention can be expressed as: .

[0217] in, The rotation matrix is ​​represented by t; the translation vector is represented by t. This represents the i-th measurement point in the free-state equivalent point cloud; In a free-state CAD model, the relationship between Corresponding model points; represents the registration weight of the i-th measurement point; n represents the number of measurement points participating in the registration. This objective function makes measurement points with higher compensation confidence or lower deformation sensitivity contribute more to the pose solution, and weakens the pulling effect of high deformation or low confidence regions on the overall pose matrix.

[0218] Step S53: Optimize the weighted registration objective function to obtain the rotation and translation parameters of the free state equivalent point cloud relative to the free state CAD model.

[0219] Weighted ICP, point-to-surface weighted ICP, feature-constraint-based weighted registration, or robust least squares registration can be used. The specific registration algorithm used does not affect the core idea of ​​this invention, as long as the registration weights determined based on deformation-sensitive parameters are used in the registration solution.

[0220] Step S54: Obtain the spatial pose transformation result based on the rotation parameters and translation parameters.

[0221] The spatial pose transformation result is used to map the compensated point cloud to the coordinate system of the free-state CAD model, thereby enabling the detection of machining feature positions, contours, and shape deviations.

[0222] After obtaining the spatial pose transformation results, the point cloud of the machining feature area is mapped to the CAD model coordinate system, and the deviation of the machining hole center position, the groove width deviation, the edge contour deviation, the surface normal deviation, or the contour line deviation is calculated.

[0223] Since the registration process has weakened the influence of the constraint deformation residual on the pose matrix, the proportion of elastic deformation error caused by clamping or support in the deviation results is reduced, which is more conducive to identifying geometric errors caused by the processing process itself.

[0224] Through step S5, this embodiment solves the problem that local residuals may still influence the pose matrix after compensation. Even though non-uniform compensation has weakened the constraint deformation, some deformation-sensitive areas may still have uncertain residuals due to model errors, sensor noise, or simplified boundary conditions.

[0225] Weighted registration prevents these residuals from participating in pose solving in the same way as the residuals in the stable region, thereby reducing the transfer of local deformation errors to rotation and translation parameters.

[0226] Example 3:

[0227] This embodiment further illustrates the residual feedback correction mechanism based on the above embodiments. This mechanism is used to update the deformation compensation reliability and adjust the registration weights or correct the non-uniform deformation compensation field after the initial weighted spatial pose registration, according to the residual distribution between the free-state equivalent point cloud and the free-state CAD model.

[0228] In this embodiment, the technical motivation for residual feedback is that although the finite element deformation model library and the non-uniform compensation field can weaken the constrained deformation, the actual fixture contact state, workpiece surface roughness, local boundary conditions, and sensor errors may still cause deviations between the compensation results and the actual deformation. If a large residual continues to appear after compensation in a certain deformation-sensitive area, it indicates that the reliability of the compensation in that area should be reduced, or the compensation field in that area needs to be slightly modified. If the measurement points in that area are still used for registration with the original weights, the residual will continue to influence the pose matrix.

[0229] In this embodiment, the residual feedback is not simply used to reduce the registration residual value, but to determine the source of the residual.

[0230] When the residual direction in a certain region is stable and consistent with the compensation direction in that region, the residual may come from insufficient compensation amplitude or deviation in compensation direction. In this case, the non-uniform deformation compensation field can be slightly corrected.

[0231] When the residual direction is chaotic, there are many discrete points, or the residual is concentrated at reflective or occluded edges in a certain area, the residual may originate from scanning noise, missing point clouds, or incorrect corresponding points. In this case, it is not advisable to correct the compensation field; instead, the registration weight of that area should be reduced. This judgment rule can prevent measurement noise from being mistakenly compensated as deformation model error.

[0232] In this embodiment, after obtaining the spatial pose transformation result, the following steps can be performed.

[0233] Calculate the registration residual between the free-state equivalent point cloud and the free-state CAD model. For the measurement points... Its residual can be expressed as: .

[0234] in, This represents the spatial residual of the i-th measurement point relative to the corresponding point in the CAD model after compensation and registration. t represents the currently obtained spatial pose transformation parameters; For the corresponding point on the CAD model.

[0235] For a candidate region or deformation-sensitive region, the residual level of the region is statistically analyzed based on the residuals of each measuring point within that region. The residual level of the region can be characterized by the mean residual, the root mean square residual, or the maximum residual.

[0236] The deformation compensation confidence level is updated based on the regional residual. The confidence level update related to the core inventive point of this invention can be expressed as: .

[0237] in, This indicates the reliability of the deformation compensation after the j-th region is updated; This indicates the reliability of the deformation compensation in the area before the update; This indicates the registration residual level in this region; This represents the residual scaling parameter.

[0238] The larger the residual, the more significant the inconsistency between the compensation result and the CAD model in that area, and the greater the decrease in its reliability. This method can feed back the spatial deviation information after registration into the next round of weight calculation.

[0239] Regional residual level It is determined by the root mean square value of the residuals at the measuring points within the region.

[0240] When the residual directions of most measurement points in the region are similar and the residuals are mainly distributed along the surface normal direction of the region, it indicates that there may be a compensation amplitude deviation in the region. When the residual directions in the region are dispersed and the residuals are concentrated at isolated points or scan edges, it indicates that there is more likely to be measurement noise or corresponding point error in the region.

[0241] By introducing a residual direction consistency judgment, the residual feedback is made to make processing choices not only based on the residual magnitude, but also on the residual spatial distribution characteristics.

[0242] Adjust the registration weights based on the updated deformation compensation confidence, or correct the non-uniform deformation compensation field.

[0243] When the residual is large in a certain region but the residual direction is stable, a small correction is made to the compensation field of that region; when the residual distribution has no stable direction or there is obvious discrete noise, it is preferable to reduce the registration weight of that region, which can avoid mistaking measurement noise as deformation compensation error.

[0244] Specifically, the following judgment rule is set: if the residual level of the j-th region is greater than the preset residual threshold, and the consistency of the residual direction in the region is greater than the preset consistency threshold, then it is determined that there is a correctable deviation in the compensation field in the region, and the non-uniform deformation compensation field in the region is corrected in small steps.

[0245] If the residual level of the j-th region is greater than the preset residual threshold, but the consistency of the residual direction does not meet the preset consistency threshold, then it is determined that the residual of the region mainly comes from measurement noise, local reflection, occlusion or unstable corresponding points, and the registration weight of the region is reduced without correcting the compensation field.

[0246] If the regional residual level of the j-th region is less than or equal to the preset residual threshold, then the compensation field and registration weight of that region remain unchanged, or only a small confidence recovery is performed.

[0247] When the compensation field needs to be modified, a small-step modification is performed on the original compensation field. The compensation field modification related to the core inventive point of this invention can be expressed as: .

[0248] in, This represents the compensation amount after correction at the measuring point location p; This indicates the amount of compensation before the correction; This represents the average residual of the j-th region; This indicates the step size for adjustment.

[0249] The correction step size should not be too large to avoid excessive correction in a single step causing oscillations in the compensation field. This correction method is only used for small iterations of the compensation field and is not used to replace the physical prediction basis of the initial non-uniform deformation compensation field.

[0250] Correction step size The correction step size is determined based on the regional deformation compensation confidence level. For regions with high confidence but still stable residuals, a relatively small but effective correction step size is used; for regions with low confidence or unstable residual directions, the correction step size should be reduced or no compensation field correction should be performed to prevent noise errors from being written into the subsequent point cloud compensation process through compensation field correction in low confidence regions.

[0251] Spatial pose registration is re-executed using the adjusted registration weights and the corrected non-uniform deformation compensation field until the preset convergence condition is met.

[0252] The convergence condition is one of the following: the change in residual between two adjacent registrations is less than a preset residual threshold, the change in spatial pose transformation parameters between two adjacent spatial pose transformations is less than a preset pose threshold, or the number of iterations reaches a preset upper limit.

[0253] Through the residual feedback mechanism in this embodiment, the present invention forms a closed-loop processing method consisting of deformation-sensitive region identification, non-uniform deformation compensation, weighted spatial pose registration, and residual feedback correction.

[0254] This closed loop ensures that the deformation compensation result is no longer a one-time output, but can be adjusted for reliability based on the spatial residuals after registration. Compared to simply compensating first and then registering, this embodiment can reduce the cumulative impact of model prediction errors in multiple rounds of registration.

[0255] This embodiment illustrates the basis for selecting the finite element deformation model library and related parameters in this invention.

[0256] Finite element deformation model libraries can be built through offline simulation.

[0257] When building the model library, multiple finite element deformation models are constructed based on typical workpiece structures, typical material parameters, and typical constraint conditions.

[0258] Each model is configured with boundary conditions corresponding to the actual clamping or support method, and clamping forces, support forces, or moments of different magnitudes and directions are set. After the simulation is completed, the three-dimensional deformation distribution of the workpiece surface, the corresponding geometric parameters, material parameters, constraint conditions, and force data are stored together.

[0259] The model library is not meant to replace on-site measurements, but rather to quickly provide a deformation prediction basis that matches the current workpiece and current constraint state during on-site inspection.

[0260] The establishment of a finite element deformation model library may include the following steps: establishing a typical workpiece geometric model; setting the elastic modulus, Poisson's ratio, and density according to the workpiece material; setting clamping constraints, support constraints, and contact boundaries according to the actual fixture structure; applying clamping force, support force, contact force, or torque according to a preset load combination; solving the three-dimensional displacement of the workpiece surface nodes under each load combination; and storing the three-dimensional displacement, model matching feature vector, load conditions, and boundary conditions in association.

[0261] When it is difficult to accurately model the stiffness of the contact area, the compensation reliability of the model can be corrected by actual measurement calibration data without changing the overall method flow of the present invention.

[0262] The parameter range in the model library is determined according to the type of workpiece being tested.

[0263] For thin-walled parts, different wall thicknesses, different radii of curvatures, and different clamping positions should be covered; for slender parts, different aspect ratios, different support spacings, and different overhang lengths should be covered; for complex curved surface parts, different local curvatures, different opening structures, and different support layouts should be covered. In this way, the model library can reflect the influence of different structural parameters and stress conditions on surface deformation.

[0264] Before using the model library, calibration is performed using a small number of typical workpieces.

[0265] During calibration, typical workpieces are placed in both free and constrained states, and surface measurement data under both states are acquired and compared with the constrained deformation predicted by the finite element model.

[0266] If the predicted deformation of a certain type of fixture or contact boundary is generally too large or too small, the compensation reliability or deformation output coefficient of the model can be corrected.

[0267] This calibration process is used to reduce the effects of fixture contact stiffness, sensor installation errors, and simplification of actual boundary conditions.

[0268] The deformation sensitivity threshold, compensation confidence threshold, and registration weight range can be set according to the measurement accuracy requirements.

[0269] If the workpiece machining feature tolerance is small, a stricter threshold should be used for deformation-sensitive area identification to prevent small constraint deformations from entering the pose solution; if the detection object is a workpiece with general precision, the threshold can be appropriately relaxed.

[0270] Registration weights should not be set to completely exclude a certain type of measurement point, because important processing features, such as hole edges and ridges, may still be contained in the deformation-sensitive area. A more suitable approach is to appropriately reduce the weights based on the deformation-sensitive parameters and compensation confidence, so that they can still provide local geometric constraints without affecting the overall pose matrix.

[0271] The data involved in this invention mainly includes industrial testing data, workpiece CAD data, material parameter data, and fixture stress data, and does not require personal identification information, sensitive personal information, or biometric information as necessary input.

[0272] Based on the preferred embodiments of the present invention described above, those skilled in the art can make various changes and modifications without departing from the inventive concept. The technical scope of this invention is not limited to the contents of the specification, but must be determined according to the scope of the claims.

Claims

1. A method for optical measurement and spatial pose registration of machining features of complex-shaped workpieces, characterized in that, Includes the following steps: Step S1: Obtain the optical three-dimensional point cloud, constraint force data, and workpiece feature data of the workpiece under constraint. Step S2: Based on the constraint force data and workpiece characteristic data, determine the deformation-sensitive area and its deformation-sensitive parameters; Step S3: Generate a non-uniform deformation compensation field based on the deformation-sensitive region and its deformation-sensitive parameters; Step S4: Use a non-uniform deformation compensation field to perform deformation compensation on the optical three-dimensional point cloud to obtain a free-state equivalent point cloud. Step S5: Determine the registration weights based on the deformation sensitivity parameters, and perform weighted spatial pose registration between the free state equivalent point cloud and the free state CAD model.

2. The optical measurement and spatial pose registration method for machining features of complex-shaped workpieces according to claim 1, characterized in that: In step S1, the constraint force data includes at least one of clamping force, support force, contact force, and torque data; Workpiece feature data includes at least one of geometric parameters, material parameters, and constraint position parameters; The geometric parameters include at least one of the following: wall thickness, length, diameter, aspect ratio, radius of curvature, overhang length, and local cross-sectional dimensions; Material parameters include at least one of elastic modulus, Poisson's ratio, and density; The constraint position parameters include at least one of the clamping point position, support point position, and force direction.

3. The optical measurement and spatial pose registration method for machining features of complex-shaped workpieces according to claim 1, characterized in that: In step S2, determining the deformation-sensitive region and its deformation-sensitive parameters includes: Based on the workpiece feature data and the constraint force data, the registration area of ​​the workpiece to be tested is divided into multiple candidate areas; For each candidate region, the deformation sensitivity parameter of the candidate region is calculated based on at least two of the structural dimension parameters, material stiffness parameters, constraint location parameters, and stress parameters of the candidate region. Candidate regions whose deformation-sensitive parameters meet preset sensitivity conditions are identified as deformation-sensitive regions.

4. The optical measurement and spatial pose registration method for machining features of complex-shaped workpieces according to claim 3, characterized in that: The deformation-sensitive parameters include at least one of regional deformation sensitivity, predicted deformation, local stiffness characterization value, compensation resolution parameter, and deformation compensation confidence. The deformation-sensitive region includes at least one of the following: thin-walled region, slender overhang region, locally weak stiffness region, curvature abrupt change region, free end region far from clamping point or support point, and region where the predicted deformation is greater than a preset deformation threshold.

5. The optical measurement and spatial pose registration method for machining features of complex-shaped workpieces according to claim 1, characterized in that: In step S3, generating a non-uniform deformation compensation field includes: Based on the workpiece feature data, constraint force data, and deformation sensitive parameters, a model matching feature vector of the workpiece to be tested is constructed. Based on the model matching feature vector, a target deformation prediction model that matches the workpiece to be tested is determined from a pre-established finite element deformation model library. Based on the target deformation prediction model, calculate the three-dimensional deformation distribution of the workpiece under constraint relative to the free state; A non-uniform deformation compensation field is generated based on the three-dimensional deformation distribution.

6. The optical measurement and spatial pose registration method for machining features of complex-shaped workpieces according to claim 5, characterized in that: The finite element deformation model library includes multiple finite element deformation models. Each finite element deformation model is used to characterize the correspondence between workpiece geometric parameters, material parameters, constraint conditions, force data and the three-dimensional deformation of the workpiece surface. The model matching feature vector includes at least two of the following: wall thickness, aspect ratio, overhang length, local cross-sectional dimensions, material elastic modulus, clamping point location, support point location, force magnitude, and force direction.

7. The optical measurement and spatial pose registration method for machining features of complex-shaped workpieces according to claim 5, characterized in that: When there is no finite element deformation model in the finite element deformation model library that perfectly matches the workpiece to be measured, the target deformation prediction model is determined in the following way: Based on the model matching feature vector, select multiple similar finite element deformation models from the finite element deformation model library; Calculate the parameter similarity between the workpiece under test and each similar finite element deformation model; The interpolation weights of similar finite element deformation models are determined based on parameter similarity. The three-dimensional deformation fields output by each similar finite element deformation model are fused according to the interpolation weights to generate a target deformation prediction model adapted to the workpiece to be tested.

8. The optical measurement and spatial pose registration method for machining features of complex-shaped workpieces according to claim 1, characterized in that: In step S3, the non-uniform deformation compensation field includes a first compensation field corresponding to the deformation-sensitive region and a second compensation field corresponding to the non-deformation-sensitive region. The first compensation field has higher deformation prediction accuracy, spatial resolution, and compensation weight than the second compensation field. The first compensation field is used to perform key deformation compensation on measuring points in the deformation-sensitive area, while the second compensation field is used to perform simplified deformation compensation or low-weight deformation compensation on measuring points in the non-deformation-sensitive area.

9. The optical measurement and spatial pose registration method for machining features of complex-shaped workpieces according to claim 1, characterized in that: Step S5 includes: Based on the deformation sensitivity parameters and deformation compensation reliability of the region to which the measurement point belongs in the free state equivalent point cloud, the registration weight of each measurement point is determined. Construct a weighted registration objective function based on the registration weights; The weighted registration objective function is optimized to obtain the rotation and translation parameters of the free-state equivalent point cloud relative to the free-state CAD model. The spatial pose transformation result is obtained based on the rotation and translation parameters.

10. The optical measurement and spatial pose registration method for machining features of complex-shaped workpieces according to claim 9, characterized in that: After obtaining the spatial pose transformation results, the following is also included: Calculate the registration residual between the free-state equivalent point cloud and the free-state CAD model; The reliability of deformation compensation is updated based on the registration residual; The registration weights are adjusted based on the updated deformation compensation confidence, and the non-uniform deformation compensation field is corrected. Spatial pose registration is re-executed using the adjusted registration weights and the corrected non-uniform deformation compensation field until the preset convergence condition is met.

Images