Workpiece machining path compensation method and system based on multi-vision cooperative positioning

Abstract

The application provides a workpiece machining path compensation method and system based on multi-vision cooperative positioning, and relates to the technical field of workpiece machining. The method comprises the following steps: establishing a compensation reference coordinate system of the workpiece machining path; determining a theoretical machining path of the workpiece according to a CAD model and a process plan of the workpiece, and mapping the theoretical machining path to the compensation reference coordinate system; performing global rough compensation on the theoretical machining path to generate a global compensation amount and a global compensation path; performing local compensation on the global compensation path according to a local point cloud to obtain a local compensation amount and a local compensation path; performing flexibility compensation on the local compensation path according to cutting parameters of the workpiece to obtain a flexibility compensation amount; performing compensation amount fusion on the global compensation amount, the local compensation amount and the flexibility compensation amount in the compensation reference coordinate system to generate a final compensation path; and executing the final compensation path. The application significantly improves the execution accuracy, stability and adaptability to complex machining conditions of the machining path.

Description

Technical Field

[0001] This invention relates to the field of workpiece processing technology, and in particular to a workpiece processing path compensation method and system based on multi-vision collaborative positioning. Background Technology

[0002] The multi-vision collaborative localization workpiece processing path compensation method refers to a class of methods that comprehensively utilize visual sensors of different scales and field of view (such as global vision, local vision or multi-level vision systems) to obtain the spatial posture and local feature information of the workpiece in the automated processing scenario of large or complex workpieces, and dynamically correct the path before or during the robot executes the preset processing path.

[0003] With the increasing application of large and complex workpieces in the manufacturing of aviation, shipbuilding, wind power and new energy equipment, their huge geometric size, significant clamping deformation and random placement posture make it difficult for traditional processing methods that rely on single vision positioning or fixed tooling reference to guarantee stable processing accuracy and production efficiency. At the same time, robot processing tasks have put forward higher requirements for the accuracy of local feature recognition and the consistency of the overall path, making it impossible to independently meet the dual requirements of high precision and high robustness by relying solely on global vision or local vision.

[0004] However, existing path compensation methods only compensate for a single error source and lack the ability to calibrate a unified coordinate system across machine tools, robots, and vision systems, making it difficult to fuse multi-source data. At the same time, traditional methods cannot accurately characterize the local true shape of the workpiece and do not consider the machine tool flexibility deformation caused by cutting loads, resulting in insufficient response of the compensation path to local deviations and dynamic errors, making it difficult to obtain accurate machining trajectories, thus limiting machining accuracy and stability. Summary of the Invention

[0005] In view of the shortcomings of the prior art, the purpose of this invention is to provide a workpiece machining path compensation method based on multi-vision collaborative positioning. This method can solve the problems of existing path compensation methods that only compensate for a single error source and lack the ability to calibrate a unified coordinate system across machine tools, robots, and vision systems, making it difficult to fuse multi-source data. At the same time, traditional methods cannot accurately characterize the local true shape of the workpiece and do not consider the machine tool flexibility deformation caused by cutting loads, resulting in insufficient response of the compensation path to local deviations and dynamic errors, making it difficult to obtain accurate machining trajectories, and thus limiting the machining accuracy and stability.

[0006] A first aspect of this invention proposes a workpiece machining path compensation method based on multi-vision collaborative positioning, comprising:

[0007] S1: Establish a compensation reference coordinate system for the workpiece machining path through multi-coordinate system joint calibration.

[0008] S2: Based on the CAD model and process plan of the workpiece, determine the theoretical machining path of the workpiece and map the theoretical machining path to the compensation reference coordinate system.

[0009] S3: Perform global coarse compensation on the theoretical processing path to generate global compensation amount and global compensation path.

[0010] S4: Collect the local point cloud of the workpiece, and perform local compensation on the global compensation path based on the local point cloud to obtain the local compensation amount and the local compensation path.

[0011] S5: Based on the cutting parameters of the workpiece, perform compliance compensation on the local compensation path to obtain the compliance compensation amount.

[0012] S6: Under the compensation reference coordinate system, the global compensation amount, local compensation amount and flexibility compensation amount are fused to generate the final compensation path.

[0013] S7: Execute the final compensation path.

[0014] A second aspect of this invention provides a workpiece processing path compensation system based on multi-vision collaborative positioning, comprising a processor and a memory.

[0015] The memory stores a program or instructions that can run on the processor, and when the program or instructions are executed by the processor, they implement the steps of the workpiece machining path compensation method based on multi-vision collaborative positioning as described in the first aspect.

[0016] A third aspect of the present invention provides a readable storage medium on which a program or instructions are stored, which, when executed by a processor, implement the steps of the workpiece machining path compensation method based on multi-vision collaborative positioning as described in the first aspect.

[0017] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0018] In this embodiment of the invention, multi-coordinate system joint calibration enables spatial alignment of the machine tool, robot, and multiple vision systems under a unified coordinate reference, providing a stable and consistent reference basis for machining path compensation calculation. Global coarse compensation corrects workpiece clamping and overall pose deviations, ensuring the machining path remains consistent with the actual workpiece position in the machine tool coordinate system. Local compensation finely corrects machining errors caused by the actual local geometry of the workpiece, improving machining accuracy for complex surfaces and irregular areas. Flexibility compensation, combined with cutting parameters, predictively compensates for elastic deformations generated by the machine tool and tool system under machining loads. An adaptive fusion strategy weighted integrates multiple compensation quantities to generate a continuous, smooth, and directly applicable compensated machining path for machine tool execution, significantly improving the execution accuracy, stability, and adaptability to complex machining conditions. This makes it suitable for machining scenarios involving high-precision and complex curved surface workpieces. Attached Figure Description

[0019] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Throughout the drawings, the same reference numerals denote the same parts. Obviously, the drawings described below are merely some embodiments of the present invention, and those skilled in the art can obtain other drawings based on these drawings without any creative effort.

[0020] Figure 1 This is a flowchart illustrating a workpiece processing path compensation method based on multi-vision collaborative positioning provided in an embodiment of the present invention.

[0021] Figure 2 This is a schematic diagram of a workpiece processing path compensation system based on multi-vision collaborative positioning provided in an embodiment of the present invention. Detailed Implementation

[0022] To enable those skilled in the art to better understand the technical solutions in the embodiments of the present invention, the technical solutions 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, not all embodiments. It should be understood that these descriptions are merely exemplary and are not intended to limit the scope of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0023] The following description, in conjunction with the accompanying drawings, details the workpiece processing path compensation method based on multi-vision collaborative positioning provided by the present invention through specific embodiments and application scenarios.

[0024] Reference manual attached Figure 1The diagram shows a flowchart of a workpiece processing path compensation method based on multi-vision collaborative positioning provided by an embodiment of the present invention.

[0025] This invention provides a workpiece machining path compensation method based on multi-vision collaborative positioning, which may include the following steps:

[0026] S1: Establish a compensation reference coordinate system for the workpiece machining path through multi-coordinate system joint calibration.

[0027] In this embodiment of the invention, by joint calibration of multiple coordinate systems, the spatial mapping relationship between machine tool, robot, global vision, local vision and workpiece model can be accurately established, so that all measurement data and path information are expressed in the same reference coordinate system. This not only significantly reduces the propagation of spatial errors between multi-sensor systems, but also ensures that the calculation of subsequent global compensation, local compensation and compliance compensation is based on a unified and highly reliable spatial foundation, thereby improving the accuracy and stability of machining path compensation as a whole.

[0028] In one possible implementation, S1 specifically includes:

[0029] S101: Collect calibration point data in multiple coordinate systems, including the machine tool coordinate system, robot base coordinate system, global vision sensor coordinate system, local vision sensor coordinate system, and workpiece theoretical model coordinate system.

[0030] Specifically, to achieve a unified spatial representation between the multi-vision system and the machine tool system, this embodiment of the invention first places a calibration plate with known geometric features in several poses within the machine tool's machining area. The machine tool probe, the end effector corresponding to the robot's base coordinate system, the global vision sensor, and the local vision sensor respectively acquire the three-dimensional coordinates of the same feature point on the calibration plate in the machine tool coordinate system, the robot's base coordinate system, the global vision sensor coordinate system, and the local vision sensor coordinate system. Simultaneously, based on the preset reference feature point position in the workpiece theoretical model, the coordinate data of that feature point in the workpiece theoretical model coordinate system is obtained, thereby forming a multi-source calibration point dataset covering five coordinate systems: machine tool, robot, global vision, local vision, and workpiece theoretical model.

[0031] S102: The homogeneous transformation matrix between each coordinate system is determined by the least squares rigid body registration algorithm, and the set of homogeneous transformation matrices is obtained.

[0032] The Least-Squares Rigid Registration algorithm is an algorithm that, given corresponding point pairs in two coordinate systems, solves for the optimal rotation matrix and translation vector by minimizing the spatial distance error between corresponding points, so that a point set is rigidly aligned to another point set in three-dimensional space (without deformation).

[0033] In this embodiment of the invention, the homogeneous transformation matrix between each coordinate system is obtained by the least squares rigid body registration algorithm, which can obtain the optimal rotation and translation parameters under noise interference and ensure the accurate and reliable spatial relationship between multiple coordinate systems.

[0034] Let any two coordinate systems be Σ i and Σ j , Σ i and Σ j It can be the machine tool coordinate system Σ M Global visual sensor coordinate system Σ G Σ coordinate system of local vision sensor L Robot base coordinate system Σ R and the workpiece theoretical model coordinate system Σ W .

[0035] Using the following formula for calculating the homogeneous transformation matrix, the computer bed coordinate system is Σ. M Global visual sensor coordinate system Σ G Σ coordinate system of local vision sensor L Robot base coordinate system Σ R and the workpiece theoretical model coordinate system Σ W The homogeneous transformation matrices between each pair of homogeneous transformation matrices are combined to obtain the set of homogeneous transformation matrices.

[0036] The formula for calculating the homogeneous transformation matrix is ​​as follows:

[0037]

[0038] in, Indicates from coordinate system Σ j To the machine tool coordinate system Σ i The optimal rotation matrix, Indicates the machine Σ from the external coordinate system j To the machine tool coordinate system Σ i The optimal translation vector, where argmin represents minimizing, This indicates that the same physical calibration point k is in the machine tool coordinate system Σ i The measured position This indicates that the same physical calibration point k is in the coordinate system Σ j The measured position, R, represents the position from the coordinate system Σ.j To the machine tool coordinate system Σ i The rotation matrix, t represents the rotation from the coordinate system Σ. j To the machine tool coordinate system Σ i The translation vector, Indicates from coordinate system machine Σ j To the machine tool coordinate system Σ i The homogeneous transformation matrix, It represents the set of all rotation matrices in three-dimensional space.

[0039] Specifically, the homogeneous transformation matrix is ​​composed of a rotation matrix and a translation vector, used to describe the complete three-dimensional rigid body transformation relationship of one coordinate system relative to another. In this embodiment of the invention, by performing least-squares fitting calculations on the coordinates of the same physical calibration point in each coordinate system, the optimal rotation matrix and translation vector can be obtained, and they are combined into a homogeneous transformation matrix. The rotation matrix is ​​used to characterize the attitude change of the external coordinate system in space, and the translation vector is used to characterize the spatial position offset of the external coordinate system relative to the machine tool coordinate system.

[0040] S103: Construct a multi-coordinate system graph and a closed-loop set based on the set of homogeneous transformation matrices.

[0041] Specifically, using the compensation reference coordinate system Σ C (Preferred coordinate system: machine tool coordinate system Σ) M Using Σ as a reference, the machine tool coordinate system is... M Global visual sensor coordinate system Σ G Σ coordinate system of local vision sensor L Robot base coordinate system Σ R and the workpiece theoretical model coordinate system Σ W Treating them as nodes in the graph, using the homogeneous transformation matrix As the measurement values ​​of the edges, a multi-coordinate system connection graph is constructed, and several closed paths are selected in this graph that connect the nodes sequentially and eventually return to the starting node, denoted as the closed-loop set. And assign weight coefficients to each closed loop. This is used to characterize the importance and confidence of the closed-loop constraint, so as to achieve the global consistency optimization objective of making the composite transformation as close as possible to the unit transformation on all closed loops.

[0042] S104: Based on multi-coordinate system diagrams and closed-loop sets, construct the objective function with the goal of minimizing the closed-loop residuals:

[0043]

[0044] Where min represents minimization, and J represents the objective function. Indicates the closed-loop weights. Represents the logarithmic function. This means multiplying the transformation matrices corresponding to each edge in the closed loop c in the order of the edges. Represents the coordinate system Σ i Relative to the compensation reference coordinate system Σ C The homogeneous transformation matrix, Indicates from coordinate system machine Σ j To the machine tool coordinate system Σ i The homogeneous transformation matrix, Represents the coordinate system Σ j Relative to the compensation reference coordinate system Σ C The homogeneous transformation matrix, Represents the square of the L2 norm. -1 This indicates the inverse operation.

[0045] S105: Solve the objective function to obtain multiple optimal homogeneous transformation matrices of the global vision sensor coordinate system, the machine local vision sensor coordinate system, the robot base coordinate system, and the workpiece theoretical model coordinate system relative to the machine tool coordinate system.

[0046] S106: Based on each optimal homogeneous transformation matrix, map the robot base coordinate system, global vision sensor coordinate system, local vision sensor coordinate system, and workpiece theoretical model coordinate system to the machine tool coordinate system to establish a compensation reference coordinate system.

[0047] S2: Based on the CAD model and process plan of the workpiece, determine the theoretical machining path of the workpiece and map the theoretical machining path to the compensation reference coordinate system.

[0048] The CAD model of a workpiece refers to a three-dimensional digital geometric model of the workpiece constructed using computer-aided design software. It accurately describes the workpiece's shape, dimensions, surface features, hole structure, and other machining-related geometric information, serving as the fundamental data source for machining path planning, simulation, and quality inspection. Process planning refers to the overall planning process of determining the required tool types, machining sequence, cutting methods, feed parameters, and stepover settings based on the workpiece's geometric characteristics, material properties, and machining requirements. It guides the generation of reasonable, efficient tool motion trajectories that meet machining accuracy requirements.

[0049] For example, after generating the theoretical processing path and mapping the compensation reference coordinate system, this embodiment of the invention can further introduce a three-dimensional contour scanning mechanism based on laser triangulation or structured light principles to obtain the true surface morphology of the workpiece. Specifically, a line laser projection module projects laser stripes onto the workpiece surface, and a high-resolution camera collects the deformation pattern of the laser stripes on the workpiece surface caused by geometric undulations. The three-dimensional coordinates of each sampling point on the workpiece surface are calculated through phase demodulation or a triangulation model, thereby forming the spatial point cloud data of the workpiece. For large workpieces (such as wind turbine shrouds several meters or tens of meters in length), this invention can use a strategy of synchronous scanning along a guide rail with multiple three-dimensional laser contour cameras to achieve full coverage acquisition of the entire workpiece. By using the pre-calibrated extrinsic parameters of each camera, the multi-segment scan point clouds are uniformly converted to the same global scanning coordinate system to construct a complete and consistent three-dimensional model of the true surface of the workpiece, providing a more reliable geometric reference for subsequent global and local compensation steps.

[0050] In this embodiment of the invention, by generating a theoretical processing path based on the CAD model and process planning and mapping it to a unified compensation reference coordinate system, it can be ensured that subsequent compensation calculations are carried out within a consistent spatial framework, so that the initial processing trajectory has accurate geometric basis and a unified expression form, thereby improving the overall coordination and reliability of processing planning and error compensation.

[0051] In one possible implementation, S2 specifically includes:

[0052] S201: Extract the machining geometry features of the workpiece from the CAD model. Specifically, the machining geometry features include: the workpiece's surface contour, boundary lines, holes, and freeform surfaces.

[0053] Among them, the surface profile refers to the overall shape boundary of the workpiece, which is composed of continuous smooth curved surfaces, and is an important geometric baseline for generating machining paths. Boundary lines refer to the transition lines between curved or flat surfaces on the workpiece, the outer contour lines, or the boundary constraint lines of the machining area, used to define the machining range. Hole positions refer to the geometric location and dimensional characteristics of holes on the workpiece that need to be drilled, milled, or positioned, and are important feature points for machining positioning and path planning. Freeform surfaces refer to three-dimensional smooth surfaces with complex curvature changes that cannot be expressed by simple analytical formulas, and are the most challenging geometric regions in high-precision machining.

[0054] For example, when extracting machining geometric features from large composite material workpieces such as wind turbine fairings and guide fairings, geometric features in the CAD model can be identified and extracted based on 3D scanned point clouds. Specifically, firstly, statistical filtering and radius filtering are applied to the scanned point cloud to reduce noise and remove isolated noise points and outliers. Then, based on the feature types in the CAD model, macroscopic structural features corresponding to the actual workpiece surface are extracted, such as the blade root end face contour, leading or trailing edge boundary lines, feature sections at the maximum chord length, and manually placed marker points. In terms of algorithm implementation, a region growing method based on normal changes can be used to extract local planar features, or an edge detection method based on curvature abrupt changes can be used to extract feature lines on complex curved surfaces. Through the above processing, geometric features consistent with the CAD model can be obtained from the actual workpiece point cloud.

[0055] In this embodiment of the invention, by completely extracting machining geometric features such as surface contours, boundary lines, hole positions, and freeform surfaces from the CAD model, it can be ensured that subsequent tool position generation and path planning fully reflect the real structural characteristics of the workpiece, making the theoretical machining path highly consistent with the actual machining requirements, thereby significantly improving machining accuracy and path rationality.

[0056] S202: Based on the process plan, the machining geometry features are discretized to generate a theoretical tool position sequence.

[0057] Specifically, based on the machining geometry of the workpiece, and in accordance with the process planning requirements (tool type, cutting method, step distance parameters, and machining sequence set in the process planning), the spatial coordinates of several discrete tool positions are calculated using methods such as equidistant segmentation, curve interpolation, surface isoparametric sampling, or adaptive mesh subdivision. The normal direction of each tool position is determined in conjunction with the tool posture model, thereby generating a theoretical tool position sequence containing position and posture information.

[0058] S203: Generate the theoretical machining path based on the theoretical tool position sequence.

[0059] Specifically, based on the arrangement order of the theoretical tool position sequence, the three-dimensional spatial coordinates and tool normal information of each tool position are organized and structured. A continuous machining trajectory is formed through path interpolation, curve fitting, or sequential connection. At the same time, according to the requirements of the machining process, necessary tool posture parameters and trajectory attributes are added to each tool position, thereby generating a complete theoretical machining path in the coordinate system of the workpiece theoretical model.

[0060] S204: Based on the homogeneous transformation matrix between the workpiece theoretical model coordinate system and the machine tool coordinate system, map the theoretical machining path to the compensation reference coordinate system.

[0061] Specifically, the three-dimensional coordinates of each tool position point in the theoretical machining path and the corresponding tool normal direction are transformed by the rotation and translation parameters in the homogeneous transformation matrix, respectively, so as to transform them from the workpiece theoretical model coordinate system to the compensation reference coordinate system, thereby obtaining the theoretical machining path expressed in the unified coordinate system.

[0062] S3: Perform global coarse compensation on the theoretical processing path to generate global compensation amount and global compensation path.

[0063] In this embodiment of the invention, by performing global coarse compensation on the theoretical machining path, the global spatial offset caused by clamping error, overall pose deviation or initial alignment error of the machine tool can be comprehensively corrected based on the registration results between the workpiece scan point cloud and the CAD model. The path can be corrected as a whole before machining, so that the compensated trajectory accurately corresponds to the real position and posture of the workpiece on a macro scale, thereby significantly reducing the burden of subsequent local compensation and flexibility compensation, and improving the overall machining accuracy and stability.

[0064] In one possible implementation, S3 specifically includes:

[0065] S301: Obtain the scanned point cloud and CAD point cloud of the workpiece in the compensated reference coordinate system.

[0066] S302: Determine the initial homogeneous matrix between the scanned point cloud and the CAD point cloud through KPM registration.

[0067] KPM registration is a point cloud initial registration method based on key point extraction and key point feature matching. It extracts key points with stable geometric meaning from the source point cloud and the target point cloud, establishes a correspondence, and solves the spatial transformation that minimizes the error of the corresponding point pair, thereby achieving coarse alignment of the two point clouds.

[0068] In one possible implementation, S302 specifically includes:

[0069] S3021: The ISS algorithm is used to detect feature points in both the scanned point cloud and the CAD point cloud, resulting in the feature point set of the scanned point cloud and the feature point set of the CAD point cloud.

[0070] Specifically, when using the ISS algorithm for feature point detection, the first step is to construct a neighborhood point set for any point in the cloud data (scanned point cloud or CAD point cloud), and then calculate the centroid and local covariance matrix of the neighborhood points: , Where p represents a point in the point cloud data, Let q represent the set of neighboring points of point p, and let q represent the neighborhood set. any point in, Point The centroid of the neighborhood point coordinates.

[0071] Subsequently, the covariance matrix Eigenvalue decomposition yields three eigenvalues, which are then sorted by size as follows: Finally, according to Threshold and and Candidate feature points are selected based on proportional constraints, and a local maximum suppression strategy is used to retain the most significant points among the candidate points, forming the scanned point cloud feature point set and the CAD point cloud feature point set, respectively.

[0072] in, Threshold and and The specific proportional constraints are as follows:

[0073] (1) Points that exceed a preset threshold are marked as candidate feature points.

[0074] (2) To avoid instability in principal component analysis, for conditions that satisfy... Points that are not selected are discarded, and the remaining points are marked as candidate feature points. This is a preset ratio threshold.

[0075] S3022: Perform FPFH feature descriptor extraction on the scanned point cloud feature point set and the CAD point cloud feature point set respectively to obtain the scanned point cloud feature vector and the CAD point cloud feature vector;

[0076] in,

[0077] Specifically, the method for extracting FPFH feature descriptors is as follows:

[0078]

[0079] in, This represents the FPFH feature descriptor corresponding to point p, where p represents a point in the current target point cloud. Point Within the neighborhood of There are *k* neighboring points, where *k* represents the total number of neighboring points. Point Statistics of local geometric features formed by the neighborhood and all points within it. Representing neighborhood points SPFH feature histogram, Indicate neighboring points Distance weights.

[0080] The above formulas are used to generate corresponding feature vector sets for all feature points in the scanned point cloud and CAD point cloud, respectively, to characterize the local geometric properties of each feature point.

[0081] S3023: With the goal of minimizing the feature vector of the scanned point cloud and the feature vector of the CAD point cloud, determine the set of corresponding points between the feature vector of the scanned point cloud and the feature vector of the CAD point cloud;

[0082] Specifically, after obtaining the feature vector set of the scanned point cloud and the feature vector set of the CAD point cloud, with the goal of minimizing the feature distance between the two, a matching vector with the minimum Euclidean distance is searched in the feature vector set of the CAD point cloud for each feature vector in the scanned point cloud. A one-to-one correspondence between the scanned feature points and the CAD feature points is established through this feature similarity criterion. The resulting set of matching pairs is the set of corresponding points between the two types of feature points.

[0083] S3024: Based on the corresponding point set, determine the initial homogeneous matrix between the scanned point cloud and the CAD point cloud using the least squares rigid body registration algorithm.

[0084] S303: Based on the initial homogeneous matrix, determine the global compensation transformation matrix through ICP registration.

[0085] Specifically, after obtaining the initial homogeneous matrix between the scanned point cloud and the CAD point cloud... Then, the scanned point cloud will be... The point cloud is transformed to the neighborhood of the CAD point cloud, and the corresponding point set is constructed using the Iterative Closest Point (ICP) algorithm. By minimizing the point-to-point residual between the reference point cloud and the transformed scanned point cloud, the incremental matrix of the rigid body transformation is iteratively solved, so that the overall registration error gradually converges, thereby obtaining the final homogeneous transformation matrix representing the actual clamping posture of the workpiece relative to the theoretical posture. , which serves as the global compensation transformation matrix.

[0086] For example, to more robustly obtain the global compensation transformation matrix in the machining of large curved workpieces, a point cloud-model registration method based on rigid body transformation optimization is used to solve for the macroscopic attitude deviation between the actual scanned point cloud and the theoretical workpiece model. Specifically, this problem can be modeled as finding the optimal rotation matrix that minimizes the error between the scanned point cloud and the theoretical model. With translation vector Essentially, this is a registration optimization problem between point sets or between point sets and a model. First, given an initial homogeneous matrix, the ICP iterative nearest-point algorithm is used to establish the nearest-point correspondence between each point in the scanned point cloud and the theoretical model. Then, singular value decomposition (SVD) or a quaternion-based least squares solution method is used to obtain the rigid body transformation that minimizes the mean square error of the corresponding point pairs, and this transformation is applied to the scanned point cloud. The above correspondence construction and transformation solution process can be repeated until the registration error is below a preset threshold or the maximum number of iterations is reached, thereby obtaining a globally consistent fine registration result. For ultra-large workpieces with large initial deviations (such as fan shrouds), this invention can introduce a coarse registration strategy before ICP fine registration. For example, feature edges, feature surfaces, or marker points extracted in the second step can be used for feature-assisted initial alignment, or a robust initial transformation matrix can be obtained using the sampling consistency method of the Fast Point Feature Histogram (FPFH) descriptor, to avoid ICP getting trapped in local optima. By combining coarse and fine registration as described above, a high-precision global rigid body transformation matrix can be obtained, which is used to characterize the pose deviation of the theoretical 3D model coordinate system relative to the actual scanned global coordinate system, and serves as the final compensation transformation matrix for the global compensation step of this invention.

[0087] S304: Based on the global compensation transformation matrix, perform a global rigid body transformation on the theoretical tool positions of the theoretical machining path to obtain a global coarse compensation tool position sequence.

[0088] S305: Calculate the global compensation amount based on the global coarse compensation tool position sequence.

[0089] S306: Generate a global compensation path based on the global compensation amount.

[0090] Specifically, after obtaining the global compensation transformation matrix, the theoretical tool positions in the theoretical machining path under the compensation reference coordinate system are updated uniformly according to the rigid body transformation relationship. That is, the global compensation transformation matrix is ​​used to perform a left multiplication operation on each tool position to make the theoretical tool positions... Converted to compensated tool position By performing this coordinate transformation point by point on the entire theoretical tool position sequence, a global coarse compensation tool position sequence that reflects the actual clamping posture deviation of the workpiece is obtained. Subsequently, by comparing the tool position difference before and after compensation, the corresponding global compensation amount is calculated to characterize the displacement compensation requirement of each tool position in three-dimensional space. Finally, the machining path is reconstructed based on the global coarse compensation tool position sequence and its compensation amount, thereby generating a global compensation path that can be directly used for machining execution, realizing global compensation for the overall clamping error of the workpiece.

[0091] S4: Collect the local point cloud of the workpiece, and perform local compensation on the global compensation path based on the local point cloud to obtain the local compensation amount and the local compensation path.

[0092] Among them, local point cloud refers to high-density three-dimensional point cloud data collected for local areas of the workpiece (usually the area that the tool is about to process or the critical processing area). It is usually acquired in real time or semi-real time on the machine tool by equipment such as structured light, laser scanning, line laser, and depth camera, and is used to reflect the actual local geometry of the workpiece or the local clamping deformation.

[0093] In this embodiment of the invention, by acquiring local point clouds of the workpiece and performing secondary local compensation on the global compensation path based on these point clouds, local shape deviations, clamping micro-deformations, or roughing residual errors in the key machining areas of the workpiece can be effectively captured. This allows for further fine correction of the tool path on the basis of global compensation, making the compensation path closer to the real workpiece surface in local areas, improving machining accuracy and surface consistency, and significantly enhancing the adaptability of the compensation strategy to complex working conditions.

[0094] Specifically, to further improve the alignment accuracy of the processing path in local areas, this embodiment of the invention first uses a local vision sensor to acquire local point clouds of the workpiece's processing area and performs benchmarking processing on these point clouds to eliminate posture deviations and coordinate drift. Then, the benchmarked local point cloud is registered with the theoretical local surface point cloud obtained from the CAD model: by establishing a regular grid in the corresponding area of ​​the theoretical surface, a theoretical height field and a measured height field are constructed respectively, and a two-dimensional affine transformation model including rotation, scaling, and translation is introduced. The optimal registration parameters are solved with the goal of maximizing the cross-correlation between the theoretical height field and the affine-transformed measured height field, thereby obtaining a registered local point cloud aligned with the theoretical surface. Based on this, by subtracting the registered local point cloud from the theoretical local surface point cloud on the same regular grid, a height deviation field for the local area of ​​the workpiece is constructed. The deviation field is then subjected to outlier removal and smoothing filtering to generate a smooth local compensation field. Finally, the tool positions in the global compensation path are projected onto the local compensation field to obtain the local compensation amount of each tool position, and the position is corrected along the tool normal direction to obtain a local compensation path consistent with the local real shape, thereby realizing fine compensation of the machining trajectory in the local area.

[0095] In one possible implementation, S4 specifically includes:

[0096] S401: Collect local point cloud data of the workpiece and perform benchmarking processing on the local point cloud data.

[0097] In this embodiment of the invention, since the acquisition process is affected by sensor attitude, installation deviation and environmental conditions, the original point cloud usually has spatial drift, tilt error and noise interference. Therefore, after the data acquisition is completed, isolated noise points are first removed based on point cloud normal estimation and local plane fitting. Then, coordinate benchmarking processing of the point cloud is performed by means of main direction alignment, key plane alignment and other methods to achieve standardization of point cloud attitude and consistency of coordinate system.

[0098] For example, in scenarios requiring high-precision 3D positioning of target cutting lines or localized machining areas, a 3D line laser contour camera can be used to perform high-resolution scanning of a specified local area of ​​the workpiece. Specifically, the line laser projection module forms laser stripes on the workpiece surface, and the camera acquires in real time the deformation images generated by the stripes under the geometric undulations of the workpiece. Based on a triangulation model, a high-precision 3D point cloud of the local area is calculated. Since the local scanning camera is usually mounted on the robot's end effector, the acquired raw point cloud is initially located in the end-tool coordinate system. This invention can utilize a pre-calibrated tool coordinate system-compensation reference coordinate system transformation relationship to uniformly map the point cloud to the compensation reference coordinate system. Subsequently, by performing benchmarking processing steps such as filtering, normal estimation, and plane calibration on the point cloud, standardized local point cloud data for local topography analysis and local compensation field construction can be obtained.

[0099] S402: Register the local point cloud after benchmarking with the theoretical local surface point cloud of the workpiece.

[0100] In one possible implementation, S402 specifically includes:

[0101] S4021: Extract theoretical local surface point cloud.

[0102] S4022: Establish a regular mesh on the local region corresponding to the theoretical local surface point cloud.

[0103] Specifically, after obtaining the theoretical local surface point cloud, the spatial distribution range of the point cloud in the compensation reference coordinate system is used as the boundary, and the corresponding local area is divided into two-dimensional parameterized regions according to the preset grid resolution. By uniformly sampling along the main direction of the local surface, a grid structure composed of regular grid nodes is constructed, so that the grid can cover the entire area of ​​the theoretical local surface point cloud in space.

[0104] S4023: On a regular grid, the theoretical local surface point cloud is resampled to obtain the theoretical height field.

[0105] Specifically, based on the constructed regular grid, the theoretical local surface point cloud is projected onto the grid coordinate system, and neighborhood search and interpolation resampling are performed on the point cloud according to the spatial position of the grid nodes. The theoretical surface height value of each grid node along the preset normal direction is calculated, thereby forming a theoretical height field that corresponds one-to-one with the regular grid. The height field can express the three-dimensional geometric features of the theoretical local surface in the form of a regular grid, providing a unified data expression structure for subsequent node-by-node error comparison and local compensation amount construction between the scanned local point cloud and the theoretical surface.

[0106] S4024: Interpolate the benchmarked local point cloud onto a regular grid to obtain the measured height field.

[0107] Specifically, in this embodiment of the invention, to achieve comparability between theoretical surfaces and measured point clouds within a unified spatial domain, a regular two-dimensional grid is first constructed on the spatial region covered by the theoretical local surface point cloud as a unified sampling framework for the height field. Then, the theoretical local surface point cloud is projected and resampled onto the regular grid. The theoretical height field defined at each grid node is obtained through interpolation or surface fitting, thereby forming a regularized description of the theoretical surface. At the same time, the local point cloud after benchmarking is mapped to the same regular grid according to its planar projection, and the corresponding measured height field is calculated through spatial interpolation, thus achieving a consistent numerical expression of the theoretical surface and the local point cloud in the same regular grid coordinate system.

[0108] S4025: Construct a two-dimensional affine transformation model for measuring the height field.

[0109] Specifically, on a regular grid The theoretical height field is known to be The measured height field is The purpose of constructing a two-dimensional affine transformation model is to measure the height field. Apply a two-dimensional affine transformation that includes rotation, scaling, and translation. Specifically, it transforms the mesh points... Transform to the position of the measurement height field coordinate system Then, at the transformed position, the measured height field is interpolated to obtain the affine transformed measured height field.

[0110] The two-dimensional affine transformation model is specifically as follows:

[0111]

[0112]

[0113]

[0114] in, Indicates in the parameter The measured height field after affine transformation under the action of the action, Represents the scale matrix. and These represent the scale factors in the x and y directions, respectively. Represents a planar rotation matrix. The measured height field is represented by the rotation angle relative to the theoretical height field, and (Δx, Δy) represents the two-dimensional translation used to compensate for the local coordinate origin offset. This represents the overall translation of the measured height field relative to the theoretical height field in the x-direction after the affine transformation. This represents the overall translation of the measured height field relative to the theoretical height field in the y-direction after the affine transformation. Represents the first position of a regular two-dimensional mesh in the x-direction. Each sampling coordinate, Represents the first position of a regular two-dimensional mesh in the y-direction. There are 3 sampling coordinates, where sin represents the sine operator and cos represents the cosine operator.

[0115] S4026: With the goal of maximizing the cross-correlation between the theoretical height field and the measured height field after two-dimensional affine transformation, determine the optimal rotation angle, optimal scale factor, and optimal translation amount of the measured height field relative to the theoretical height field.

[0116] Specifically, after constructing the theoretical and measured height fields, with spatial consistency between the two as the optimization objective, the measured height field is used as the object to be registered. An affine transformation model including rotation angle, scale factor, and translation is applied to it in a two-dimensional plane. Through traversal or iterative search, the cross-correlation index between the transformed measured height field and the theoretical height field is maximized. When the cross-correlation function reaches its maximum value, the corresponding rotation angle, scale factor, and translation are used as the optimal affine registration parameters of the measured height field relative to the theoretical height field. These parameters are used to compensate for attitude differences and scale shifts in local regions, thereby enhancing the accuracy and robustness of subsequent local error field construction.

[0117] S4027: Based on the optimal rotation angle, optimal scale factor, and optimal translation amount, perform a 3D registration transformation on the benchmarked local point cloud to obtain the registered local point cloud.

[0118] Specifically, the two-dimensional affine transformation composed of optimal parameters is mapped to the corresponding three-dimensional space. The planar distribution of the local point cloud is scaled and corrected by the scaling matrix, the overall pose of the local point cloud is adjusted by the rotation matrix, and the planar projection position of the local point cloud is translated and compensated by the translation vector. This enables the point cloud to achieve registration effect of consistent scale, consistent orientation and consistent position in three-dimensional space. The local point cloud after this three-dimensional registration transformation can form a one-to-one correspondence with the theoretical local surface point cloud in the regular grid coordinate system.

[0119] S403: Construct a local deviation field based on the registered local point cloud and the theoretical local surface point cloud.

[0120] It should be noted that the local deviation field refers to the two-dimensional function distribution describing the height difference between the actual measured surface and the theoretical surface of a local area of ​​the workpiece under a unified regular grid coordinate system. In this embodiment of the invention, the local deviation field is the difference between the registered measured height field and the theoretical height field.

[0121] S404: Suppress noise and remove outliers in the local deviation field to obtain a smooth local compensation field.

[0122] S405: Based on the smooth local compensation field, perform local compensation on the global compensation path to obtain the local compensation amount and the local compensation path.

[0123] Specifically, the planar projection positions of each tool position in the global compensation path are mapped to the regular grid of the smooth local compensation field, and the corresponding compensation value of each tool position in the local compensation field is obtained by interpolation. Then, the compensation value is superimposed on the tool position coordinates of the original global compensation path along the normal direction corresponding to the tool posture to obtain the locally compensated tool position sequence, thereby forming the corrected local compensation path. The local compensation path can refine the global tool position according to the local real shape of the workpiece, and is finally used to generate a compensated machining trajectory that better meets the actual machining requirements, realizing local high-precision adjustment of the machining path.

[0124] S5: Based on the cutting parameters of the workpiece, perform compliance compensation on the local compensation path to obtain the compliance compensation amount.

[0125] In one possible implementation, S5 specifically includes:

[0126] S501: Based on the cutting parameters, estimate the cutting load at each tool position in the local compensation path to obtain the cutting force feedforward sequence.

[0127] In this embodiment of the invention, by performing flexibility compensation on the local compensation path according to the cutting parameters, the elastic deformation of the machine tool structure and tool system caused by the cutting load during the machining process can be actively predicted and offset, making the actual cutting trajectory of the tool closer to the theoretical path. This not only effectively offsets the deformation error of the machine tool structure and spindle tool system under stress, but also avoids the oscillation effect under high-speed feed conditions, making the final tool trajectory more in line with the theoretical path, thereby significantly improving machining accuracy, surface quality and machining process stability.

[0128] Specifically, based on cutting parameters (including tool type, material properties, axial depth of cut, radial width of cut, feed rate, and spindle speed), the cutting forces at each tool position point in the local compensation path are modeled and calculated. Based on a preset empirical cutting force model or mechanical simulation model, the three-dimensional cutting load at each tool position point in the tool coordinate system is estimated and uniformly mapped to the compensation reference coordinate system, thus forming a feedforward sequence of cutting forces arranged sequentially along the machining path. The three-dimensional cutting load is specifically: , , as well as Indicates the first The cutting force components at each tool position point along the X, Y, and Z directions of the tool coordinate system.

[0129] S502: Determine the six-dimensional cutting load vector based on the cutting force feedforward sequence:

[0130]

[0131]

[0132] in, Indicates the first Effective cutting load at each tool position point Indicates the first The three-dimensional cutting force at each tool position point is measured by a force sensor. Indicates the tool position point The cutting force fusion weight, Indicates the first The six-dimensional cutting load vector at each tool position point This indicates the effective three-dimensional cutting load. The arrangement and splicing operations are used to organize the data into the first three elements containing the force components of the six-dimensional cutting load vector. Indicates the first At each tool position point, the position vector of the tool's force-bearing point relative to the tool reference point. Indicates the first The cutting torque components generated by each tool position point around the X-axis of the tool coordinate system Indicates the first The cutting torque components generated by each tool position point around the Y-axis of the tool coordinate system Indicates the first The cutting torque components generated by each tool position point about the Z-axis of the tool coordinate system, where T represents transpose. Indicates the first The cutting force feedforward sequence for each tool position.

[0133] S503: Establish the machine tool compliance model.

[0134] Specifically, the machine tool compliance model is used to characterize the elastic response of the machine tool structure and spindle-tool system under cutting loads. The machine tool compliance model is as follows:

[0135]

[0136]

[0137]

[0138] in, Indicates the machine tool's attitude The equivalent compliance matrix formed by superimposing the structural compliance and the tool compliance. Indicates the machine tool's attitude The end Jacobian matrix below describes the mapping relationship between small joint displacements and tool tip displacements, K. q Indicates the machine tool's attitude The equivalent stiffness matrix of each motion axis reflects the elastic characteristics of each axis. This represents the overall compliance of the spindle-tool holder-tool system, reflecting the elastic deflection of the tool body. In this embodiment of the invention, to obtain the end-position compliance for path compensation, the first three rows of the overall compliance matrix are extracted as a translation compliance submatrix. This represents the first three rows of the equivalent compliance matrix. Indicates the machine tool's attitude Below, the positional shift at the end due to flexibility, Indicates the machine tool posture The resulting three-dimensional compliance mapping matrix is ​​used to map the six-dimensional cutting load vector into a three-dimensional compliance displacement at the end of the tool.

[0139] S504: Input the six-dimensional cutting load vector into the machine tool compliance model and output the compliance displacement vector of each tool position in the local compensation path.

[0140] S505: Project the compliance displacement vector onto the normal vector of the machining tool to obtain the effective compliance compensation component.

[0141] Specifically, for the first in the local compensation path There are several tool points, and their compliance displacement vector is given by... The normal vector of the machining tool is The feed direction vector is The embodiments of the present invention project the compliance displacement vector onto the normal vector of the machining tool in two ways.

[0142] When the primary goal during machining is to improve contour accuracy or depth of cut accuracy, a projection method along the tool normal is used to project the compliance displacement vector onto the tool normal direction to extract key compensation components.

[0143]

[0144] in, Indicates the first Effective compliance compensation component for each cutter site Indicates the first The compliance displacement vector of each tool point. Indicates the first The tool normal vector at each tool position point.

[0145] When the tool feed rate is high or there is a risk of compensating oscillation along the feed direction during machining, the feed direction vibration suppression projection method is adopted to remove the displacement component along the tool feed direction in the compliance displacement vector, and only retain its effective compensation component in the feed normal plane:

[0146]

[0147] in, Indicates the first The tool feed direction vector at each tool position point.

[0148] S506: Generate the compliance compensation amount based on the effective compliance compensation component.

[0149] Specifically, for the first in the local compensation path Each cutter point, its effective compliance compensation component As the desired tool offset value, and to correct the position of the theoretical tool position point in a manner opposite to the compensation direction, the compliance compensation amount is defined as... This allows the tool to be pre-biased along the error direction during actual machining to counteract the elastic deformation of the tool tip caused by the machine tool's compliance. Then, the compliance compensation amount is combined and superimposed with the global compensation amount and the local compensation amount to form the compliance compensation result used for the final machining path generation.

[0150] S6: Under the compensation reference coordinate system, the global compensation amount, local compensation amount and flexibility compensation amount are fused to generate the final compensation path.

[0151] The final compensation path refers to the sequence of processing trajectory data, which includes the spatial location of path points and their corresponding compensation amounts, generated by the computer after digitizing the original processing path data.

[0152] In this embodiment of the invention, by fusing global compensation, local compensation, and compliance compensation in a unified compensation reference coordinate system, it is possible to simultaneously consider the overall workpiece posture error, local geometric shape deviation, and dynamic elastic deformation during processing. This ensures that the three types of compensation information are consistent in space, coordinated in direction, and adaptively weighted in value. Through comprehensive optimization of multi-source compensation, a high-quality compensation path is ultimately generated that possesses both global accuracy and local precision, and can offset processing mechanical errors, significantly improving the stability, reliability, and final processing accuracy of the processing.

[0153] In one possible implementation, S6 specifically includes:

[0154] S601: Construct a multi-source compensation feature descriptor based on the global compensation amount, local compensation amount, and flexibility compensation amount.

[0155] S602: Based on the multi-source compensation feature descriptor, calculate the adaptive reliability index of the global compensation amount, local compensation amount, and flexibility compensation amount respectively.

[0156] Specifically, the global compensation reliability is calculated based on the overall point cloud registration residual, feature matching consistency, and coverage integrity. The local compensation reliability is calculated based on the local height field registration error, scale factor deviation, and local surface complexity. The compliance compensation reliability is calculated based on the cutting force estimation stability, compliance model uncertainty, and historical compensation deviation feedback. Each reliability index, after normalization, is used as input to the attention weights for adaptive weighting of different compensation amounts in subsequent fusion stages.

[0157] S603: Determine the multi-source compensation attention weights based on the adaptive credibility index.

[0158] S604: Based on the multi-source compensation attention weights, the global compensation amount, local compensation amount, and flexibility compensation amount are vector-fused to obtain the fused compensation amount.

[0159] S605: The fusion compensation amount is superimposed onto the theoretical processing path under the compensation reference coordinate system to obtain the final compensation path.

[0160] Specifically, to simultaneously consider the overall workpiece clamping posture error, local geometric deviations, and flexibility elastic deformation during machining under a unified compensation reference coordinate system, thereby generating a final machining path that is numerically continuous, directionally reasonable, and kinematically executable, global compensation, local compensation, and flexibility compensation are fused from multiple sources. First, multi-source compensation feature descriptors are constructed based on the amplitude variations, directional consistency, and local surface complexity of the three types of compensation at each tool position, and adaptive reliability indices for global, local, and flexibility compensation are calculated accordingly. Then, corresponding multi-source compensation attention weights are determined based on each reliability index to reflect the effectiveness of different compensation types at the current tool position. These attention weights are then used to perform vector-weighted fusion of the three types of compensation to obtain a fused compensation amount that coordinates global consistency, local fit, and machining mechanical response. Finally, the fused compensation amount is superimposed onto the theoretical machining path under the compensation reference coordinate system to form the final compensation path, achieving coordinated optimization and unified execution of global, local, and flexibility compensations under the same spatial reference.

[0161] S7: Execute the final compensation path.

[0162] For example, when performing path compensation, embodiments of the present invention can combine the process characteristics of high-pressure water jet cutting to achieve linkage control between the path and process parameters. Specifically, the robot controller drives the end effector to move along the corrected trajectory according to the compensated tool position sequence, and at the same time, combined with a preset process database, it controls key parameters affecting cutting quality, such as water pressure. Target distance Cutting speed and abrasive flow rate The system can adaptively adjust its speed. For example, when the path enters a region with significant curvature in the composite material, the system can automatically reduce the cutting speed. Or dynamically adjust the target distance This avoids increased cut taper and deteriorated surface roughness due to excessive speed or target distance deviation. The above process parameters can be embedded into the final execution path as curves using robot offline programming software, and fine-tuned during operation according to actual posture conditions, thereby ensuring the consistency and stability of the entire cutting trajectory.

[0163] In summary, the embodiments of the present invention construct a unified compensation reference coordinate system, perform global attitude compensation, local geometric refinement, and compliance compensation calculation based on the load model on the machining path data, and combine an adaptive fusion mechanism of multi-source compensation quantities to achieve collaborative error correction of path data in the dimensions of macroscopic position, local morphology, and dynamic deformation. This can significantly improve the consistency and adaptability of the generated compensation path data in representing the real spatial state of the workpiece, and comprehensively optimize the final path data in terms of spatial continuity, local fit, and numerical stability. Thus, it provides a reliable data foundation for the generation and application of path data for complex curved surfaces and high-precision parts, and has good engineering application value.

[0164] It should be noted that the method described in this invention is applicable to a variety of processing or cutting technologies, including but not limited to tool cutting, waterjet cutting, laser cutting and plasma cutting, etc. This invention does not limit the specific cutting method.

[0165] For example, the multi-vision collaborative positioning and multi-stage path data compensation method described in this invention can be applied to path data generation and processing scenarios for large composite material components such as wind turbine fairings and guide fairings. Related application systems can be deployed in intelligent flexible processing equipment platforms with large workspace coverage capabilities. Such processing equipment typically employs a dual-rail, dual-robot structure to meet the spatial coverage requirements of large-sized components. Furthermore, the processing equipment can be equipped with an ultra-high-pressure water jet cutting unit (including a high-pressure pump set, an automatic sand supply device, and a sand-water recovery system) as the operation execution mechanism to support cutting applications for composite material components.

[0166] In this type of application environment, the system can acquire overall and local spatial data of the workpiece through various visual perception units. Among them, the global visual perception unit is used to realize the large-scale initial positioning and model matching of the workpiece, and the local three-dimensional visual perception unit is used to collect high-precision spatial data of the target area to form local point cloud data, providing data input for global compensation calculation of path data, local deviation field construction and flexibility compensation calculation.

[0167] Based on the aforementioned sensing and execution hardware platform, the path data processing algorithms in this embodiment of the invention, such as multi-coordinate system joint calibration, point cloud registration, local topography analysis, and multi-source compensation fusion, can operate stably in industrial application environments, achieving high-precision path data compensation and adaptive optimization generation for large composite material components, thereby providing a reliable data foundation for path data applications based on processes such as waterjet cutting.

[0168] Reference manual attached Figure 2 The diagram shows a schematic of a workpiece processing path compensation system based on multi-vision collaborative positioning provided by an embodiment of the present invention.

[0169] This invention provides a workpiece processing path compensation system 20 based on multi-vision collaborative positioning, including: a processor 201 and a memory 202;

[0170] The memory 202 stores programs or instructions that can run on the processor 201. When the program or instructions are executed by the processor 201, they implement the steps of the above-mentioned workpiece processing path compensation method based on multi-vision collaborative positioning and achieve the same technical effect. To avoid repetition, the present invention will not elaborate further.

[0171] It should be understood that the processor 201 in this embodiment of the invention may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0172] It should also be understood that the memory 202 in the embodiments of the present invention can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of random access memory are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct memory bus RAM (DR RAM).

[0173] The above embodiments can be implemented, in whole or in part, by software, hardware (such as circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0174] It should be understood that, in various embodiments of the present invention, the sequence number of each process does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0175] Those skilled in the art will 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, or a combination of computer software and electronic hardware. 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 implementations should not be considered beyond the scope of this invention.

[0176] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices, apparatuses, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0177] In the several embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0178] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0179] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0180] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0181] This invention provides a readable storage medium comprising: storing a program or instructions on the readable storage medium, wherein when the program or instructions are executed by a processor, the program or instructions implement the steps of the above-described workpiece processing path compensation method based on multi-vision collaborative positioning, and can achieve the same technical effect. To avoid repetition, this invention will not elaborate further.

[0182] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the embodiments of the present invention, and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the protection scope of the present invention.

Claims

1. A workpiece machining path compensation method based on multi-vision collaborative positioning, characterized in that, include: S1: Establish a compensation reference coordinate system for the workpiece machining path through multi-coordinate system joint calibration; Specifically, S1 includes: S101: Collect calibration point data in multiple coordinate systems, wherein the coordinate systems specifically include: machine tool coordinate system, robot base coordinate system, global vision sensor coordinate system, local vision sensor coordinate system and workpiece theoretical model coordinate system; S102: Determine the homogeneous transformation matrix between each coordinate system using the least squares rigid body registration algorithm to obtain a set of homogeneous transformation matrices; S103: Construct a multi-coordinate system graph and a closed-loop set based on the homogeneous transformation matrix set; S104: Based on the multi-coordinate system diagram and the closed-loop set, construct an objective function with the goal of minimizing the closed-loop residual; S105: Solve the objective function to obtain multiple optimal homogeneous transformation matrices of the global vision sensor coordinate system, the machine tool local vision sensor coordinate system, the robot base coordinate system, and the workpiece theoretical model coordinate system relative to the machine tool coordinate system. S106: Based on each of the optimal homogeneous transformation matrices, map the robot base coordinate system, the global vision sensor coordinate system, the local vision sensor coordinate system, and the workpiece theoretical model coordinate system to the machine tool coordinate system to establish the compensation reference coordinate system; S2: Based on the CAD model and process plan of the workpiece, determine the theoretical machining path of the workpiece, and map the theoretical machining path to the compensation reference coordinate system; S3: Perform global coarse compensation on the theoretical processing path to generate global compensation amount and global compensation path; Specifically, S3 includes: S301: Obtain the scanned point cloud and CAD point cloud of the workpiece under the compensation reference coordinate system; S302: Determine the initial homogeneous matrix between the scanned point cloud and the CAD point cloud through KPM registration; S303: Based on the initial homogeneous matrix, determine the global compensation transformation matrix through ICP registration; S304: Based on the global compensation transformation matrix, perform a global rigid body transformation on the theoretical tool position of the theoretical machining path to obtain a global coarse compensation tool position sequence; S305: Calculate the global compensation amount based on the global coarse compensation tool position sequence; S306: Generate the global compensation path based on the global compensation amount; S4: Collect the local point cloud of the workpiece, and perform local compensation on the global compensation path based on the local point cloud to obtain the local compensation amount and the local compensation path. S5: Based on the cutting parameters of the workpiece, perform compliance compensation on the local compensation path to obtain the compliance compensation amount; S6: Under the compensation reference coordinate system, the global compensation amount, the local compensation amount, and the flexibility compensation amount are fused to generate the final compensation path; S7: Execute the final compensation path.

2. The workpiece machining path compensation method based on multi-vision collaborative positioning according to claim 1, characterized in that, S2 specifically includes: S201: Extract the machining geometry features of the workpiece from the CAD model, wherein the machining geometry features specifically include: the surface contour, boundary lines, hole positions, and freeform surfaces of the workpiece; S202: Based on the process plan, the machining geometry features are discretized to generate a theoretical tool position sequence; S203: Generate the theoretical machining path based on the theoretical tool position sequence; S204: Based on the homogeneous transformation matrix between the workpiece theoretical model coordinate system and the machine tool coordinate system, map the theoretical machining path to the compensation reference coordinate system.

3. The workpiece machining path compensation method based on multi-vision collaborative positioning according to claim 1, characterized in that, Specifically, S302 includes: S3021: Using the ISS algorithm, feature point detection is performed on the scanned point cloud and the CAD point cloud respectively to obtain the feature point set of the scanned point cloud and the feature point set of the CAD point cloud. S3022: Perform FPFH feature descriptor extraction on the scanned point cloud feature point set and the CAD point cloud feature point set respectively to obtain the scanned point cloud feature vector and the CAD point cloud feature vector; S3023: With the goal of minimizing the feature vector of the scanned point cloud and the feature vector of the CAD point cloud, determine the set of corresponding points between the feature vector of the scanned point cloud and the feature vector of the CAD point cloud; S3024: Based on the corresponding point set, determine the initial homogeneous matrix between the scanned point cloud and the CAD point cloud using the least squares rigid body registration algorithm.

4. The workpiece machining path compensation method based on multi-vision collaborative positioning according to claim 1, characterized in that, S4 specifically includes: S401: Collect a local point cloud of the workpiece and perform benchmarking processing on the local point cloud; S402: Register the local point cloud after benchmarking with the theoretical local surface point cloud of the workpiece; S403: Construct a local deviation field based on the registered local point cloud and the theoretical local surface point cloud; S404: Perform noise suppression and outlier removal on the local deviation field to obtain a smooth local compensation field; S405: Based on the smooth local compensation field, perform local compensation on the global compensation path to obtain the local compensation amount and the local compensation path.

5. The workpiece machining path compensation method based on multi-vision collaborative positioning according to claim 4, characterized in that, Specifically, S402 includes: S4021: Extract theoretical local surface point cloud; S4022: Establish a regular mesh on the local region corresponding to the theoretical local surface point cloud; S4023: Resample the theoretical local surface point cloud on the regular grid to obtain the theoretical height field; S4024: Interpolate the benchmarked local point cloud onto the regular grid to obtain the measured height field; S4025: Construct a two-dimensional affine transformation model of the measured height field; S4026: With the goal of maximizing the cross-correlation between the theoretical height field and the measured height field after two-dimensional affine transformation, determine the optimal rotation angle, optimal scale factor, and optimal translation amount of the measured height field relative to the theoretical height field; S4027: Based on the optimal rotation angle, the optimal scale factor, and the optimal translation amount, perform a three-dimensional registration transformation on the benchmarked local point cloud to obtain the registered local point cloud.

6. The workpiece machining path compensation method based on multi-vision collaborative positioning according to claim 1, characterized in that, S5 specifically includes: S501: Based on the cutting parameters, estimate the cutting load at each tool position in the local compensation path to obtain the cutting force feedforward sequence; S502: Determine the six-dimensional cutting load vector based on the cutting force feedforward sequence; S503: Establish a machine tool compliance model; S504: Input the six-dimensional cutting load vector into the machine tool compliance model and output the compliance displacement vector of each tool position in the local compensation path; S505: Project the compliance displacement vector onto the normal vector of the machining tool to obtain the effective compliance compensation component; S506: Generate the compliance compensation amount based on the effective compliance compensation component.

7. The workpiece machining path compensation method based on multi-vision collaborative positioning according to claim 1, characterized in that, S6 specifically includes: S601: Construct a multi-source compensation feature descriptor based on the global compensation amount, the local compensation amount, and the flexibility compensation amount; S602: Based on the multi-source compensation feature descriptor, calculate the adaptive reliability index of the global compensation amount, the local compensation amount, and the flexibility compensation amount respectively; S603: Determine the multi-source compensation attention weights based on the adaptive credibility index; S604: Based on the multi-source compensation attention weights, perform vector fusion on the global compensation amount, the local compensation amount, and the flexibility compensation amount to obtain the fused compensation amount; S605: The fusion compensation amount is superimposed onto the theoretical processing path under the compensation reference coordinate system to obtain the final compensation path.

8. A workpiece machining path compensation system based on multi-vision collaborative positioning, characterized in that, include: Processor and memory; The memory stores programs or instructions that can run on the processor, which, when executed by the processor, implement the steps of the workpiece machining path compensation method based on multi-vision collaborative positioning as described in any one of claims 1 to 7.

Images