On-line rapid precision measurement method for complex free-form surface part

Abstract

The invention relates to an online rapid precision measurement device and method for a complex free-form surface part, the measurement device comprises a six-degree-of-freedom industrial mechanical arm, an AGV, a vision measurement device and a multi-axis numerical control machine tool, the six-degree-of-freedom industrial mechanical arm is installed on the AGV, and the vision measurement device is installed at the tail end of the six-degree-of-freedom industrial mechanical arm. According to the measurement method, on the basis of optical imaging, a computer vision technology and a geometric quantity measurement principle are combined, a machined workpiece is directly measured on a numerical control machine tool, the workpiece does not need to be taken down and then installed on a measurement platform for measurement, and comprehensive precision online real-time control over the manufacturing process of the complex free-form surface part can be directly achieved; secondary clamping errors are avoided, the technical bottleneck that the current measurement mode is low in efficiency, changeable in measurement basis and insufficient in self-adaptive calibration flexibility is broken through, and the method is an inevitable development trend of current industrial intelligent measurement.

Description

Technical Field

[0001] The invention relates to the technical field of numerical control machining of complex free-form surface parts, and in particular to an online rapid and precise measurement method for complex free-form surface parts. Background Art

[0002] With the rapid development of industries such as aerospace, national defense, vehicles, power and machinery, the performance requirements for high-end equipment are getting higher and higher, and a large number of high-performance complex free-form surface parts with high processing difficulty and demanding performance indicators have emerged. As a key equipment in modern manufacturing, multi-axis CNC machine tools integrate computer control, high-performance servo drive and precision processing technology. They are an important indicator of the development level and product quality of a country's equipment manufacturing industry, and can achieve efficient, precise and automated processing of complex free-form surface parts.

[0003] In the CNC machining process of complex free-form surface parts, it is very important to complete the online rapid and precise measurement of the machined parts, realize the closed-loop control of the parts machining accuracy, and ensure the product quality and production capacity of the parts.

[0004] The traditional measurement method is to remove the workpiece from the CNC machine tool, install it on the measuring platform for manual measurement or use special instruments for measurement, which will bring about the problem of secondary clamping and repositioning, and multiple processing and measurements are required to obtain accurate results. In addition, manual measurement usually uses basic measuring tools such as calipers, vernier calipers, and outside micrometers, but this method has large errors and subjectivity. Special instrument measurement uses instruments such as projectors, ring measuring instruments, and three-dimensional coordinate measuring machines for measurement. Although the accuracy is relatively high, the measurement time is long and the cost is high.

[0005] Moreover, for complex free-form surface parts, due to the complex structure of the workpiece, various and changeable postures, and complicated measurement parameters, there are prone to measurement blind spots, which makes it difficult to meet the high-efficiency, high-precision, and online real-time measurement requirements of workpiece CNC machining. Summary of the invention

[0006] The present invention discloses an online fast and precise measurement method for complex free-form surface parts, aiming to solve the technical problems that in the CNC machining process of complex free-form surface parts, traditional measurement methods are difficult to effectively complete the online fast and precise measurement of the machined parts, realize closed-loop control of the parts machining accuracy, and ensure the part product quality and production capacity.

[0007] The present invention proposes an online fast and precise measuring device for complex free-form surface parts. The measuring device comprises a six-degree-of-freedom industrial robot arm, an AGV, a visual measuring device and a multi-axis CNC machine tool. The six-degree-of-freedom industrial robot arm is installed on the AGV, and the visual measuring device is installed at the end of the six-degree-of-freedom industrial robot arm. The visual measuring device comprises a sinusoidal grating projector and a multi-eye camera. The sinusoidal grating projector is located at the center of the end of the six-degree-of-freedom industrial robot arm, and the multi-eye cameras are evenly distributed around the projector at intervals, and the optical axes of the cameras are inclined toward the optical axis of the projector; the robot arm is used to guide the visual measuring device to complete the image information collection of the processed parts from multiple visions.

[0008] Furthermore, the CNC machining of the complex free-form surface parts is completed by a multi-axis CNC machine tool, which includes a CNC system, and the CNC system is wirelessly connected to a six-degree-of-freedom industrial robot arm and an AGV through a control panel.

[0009] Furthermore, the multi-eye camera is a four-eye camera.

[0010] Furthermore, it also includes a calibration plate, which is installed on the protective door of the multi-axis CNC machine tool.

[0011] Furthermore, the calibration plate is a circular target calibration plate with a fixed-spacing pattern array and a black background and white circles.

[0012] An online fast and precise measurement method for complex free-form surface parts includes the following steps:

[0013] S1, 3D reconstruction of 3D point cloud data of processed parts;

[0014] S2, accurate registration of 3D point clouds;

[0015] S3. Precision measurement of complex free-form surface parts;

[0016] S4. Online assessment of machining accuracy of machined parts.

[0017] Furthermore, step S1 is specifically as follows:

[0018] S1.1. Calibration of visual measurement device system

[0019] S1.2. Control the movement of the six-degree-of-freedom industrial robot arm to drive the visual measurement device to collect images of complex free-form surface parts, continuously change the position and posture of the visual measurement device, and collect multiple sets of complex free-form surface part images;

[0020] S1.3, 3D reconstruction point cloud data

[0021] After the industrial robot arm guides the visual measurement device to complete the multi-visual and multi-directional image information acquisition of the processed parts, the external parameter matrix and the absolute phase information of the parts to be measured can be used to complete the conversion from the acquired fringe image (two-dimensional pixel coordinates) to the three-dimensional point cloud of the parts (three-dimensional space coordinates), and complete the three-dimensional reconstruction of the part point cloud data.

[0022] S1.4. Optimize the quality of reconstructed point cloud data

[0023] In the process of 3D point cloud reconstruction, there is a DMA (Discontinuity-induced Measurement Artefact) error in the area where the reflectivity of the part is discontinuous when the projector is projecting, which will cause the final reconstructed point cloud to have a large error. First, the sub-pixel edge detection method is used to determine the sub-pixel boundary of the reflectivity discontinuous area, and the phase interpolation calculation is performed through the phase of the nearby unaffected area to correct the phase error and obtain higher-precision point cloud data.

[0024] Furthermore, step S1.1 is specifically as follows:

[0025] S1.1.1. After the processing of the complex free-form surface parts is completed, the protective door of the CNC machine tool is opened, and the CNC system communicates with the AGV to control the six-degree-of-freedom industrial robot arm to move to the vicinity of the CNC machine tool. The six-degree-of-freedom industrial robot arm drives the visual measurement device to collect images of the calibration plate, and continuously changes the position and posture of the visual measurement device to collect multiple sets of calibration plate images;

[0026] S1.1.2. Take the center of the first circle in the upper left corner of the calibration plate as the origin, horizontally to the right as the x-axis direction, vertically downward as the y-axis direction, and determine the z-axis direction by the right-hand rule to complete the establishment of the world coordinate system, and obtain the relationship between the world coordinates of the key points on the calibration plate and their corresponding points in the pixel coordinate system:

[0027]

[0028] In the formula, (X W , Y W , Z W ) represents an arbitrary point in the world coordinate system, (u, v) represents the corresponding point in the camera pixel coordinate system, dx and dy represent the length of each pixel of the camera in the X direction and Y direction respectively; (u0, v0) is the coordinate of the pixel at the intersection of the camera optical axis and the imaging plane; f is the focal length of the lens, and the matrix R and vector T are the rotation matrix and translation vector of the camera coordinate system transformed to the world coordinate system respectively.

[0029] S1.1.3. Based on multiple sets of data, the monotonic matrix Hc between the two coordinate systems is calculated to obtain the internal and external parameters and distortion parameters of the camera;

[0030] S1.1.4. Use the LM (Levenberg Marquardt) optimization algorithm to update and optimize the internal and external parameters and distortion parameters of the camera.

[0031] S1.1.5. Map the key point coordinates in the camera pixel coordinate system to the projector pixel coordinate system, establish the relationship between the camera pixel coordinate system and the projector pixel coordinate system, and thus obtain the external parameter matrix of the camera, projector and world coordinate system.

[0032]

[0033] In the formula, (u p , v p ) represents the coordinate value in the projector pixel coordinate system, φ C Indicates the conversion factor.

[0034] Furthermore, step S2 is specifically as follows:

[0035] S2.1 Preprocessing and denoising of 3D point clouds

[0036] S2.2. Rough registration of 3D point cloud

[0037] S2.3. Precise registration of 3D point clouds

[0038] Furthermore, step S2.1 is specifically as follows:

[0039] S2.1.1. Read four sinusoidal fringe images of the same frequency and set the shadow threshold;

[0040] S2.1.2, traverse the gray value of each pixel in each fringe image, and determine whether the gray value of each pixel at different times is always less than or equal to the shadow threshold. If this condition is met, it is a shadow area;

[0041] S2.1.3, making the shadow area into a mask image as a shadow mask;

[0042] S2.1.4, during 3D reconstruction, when indexing a pixel in the mask, mark the point as a shadow noise point;

[0043] S2.1.5. After removing the noise points, the point cloud composed of the remaining points is the point cloud without shadow noise.

[0044] Furthermore, step S2.2 is specifically as follows:

[0045] After obtaining noise-free point cloud data, due to the huge number of point clouds, it is necessary to downsample the point clouds without losing the part morphology, reduce the data volume, speed up the processing speed, and complete the rough registration of the part point clouds under the four camera perspectives, so that the poses of the point clouds under multiple perspectives are basically consistent. The specific process is as follows:

[0046] Use the octree data structure to describe the part point cloud, find the point in the voxel that is closest to the voxel center, and use this point as the sampling point of the current voxel to achieve downsampling;

[0047] Obtain the homogeneous transformation relationship between the projector and the four camera coordinate systems;

[0048]

[0049] Where M c =[R c ,T c ] is the homogeneous transformation matrix from the world coordinate system to the camera coordinate system, M p =[R p ,T p ] is the homogeneous transformation matrix from the world coordinate system to the projector coordinate system, X c , X p and X w They are the points in the camera, projector and world coordinate systems respectively.

[0050] The point clouds in the four camera coordinate systems are transformed into the projector coordinate system to complete the rough registration of the part point cloud, providing a more accurate initial registration value for the subsequent fine registration of the point cloud.

[0051] Furthermore, step S2.3 is specifically as follows:

[0052] S2.3.1. Represent the associated points in the two point clouds to be registered with point sets P and Q, and calculate the homogeneous transformation matrix from set Q to set P:

[0053]

[0054] In the formula, p i =[p xi ,p yi ,p zi ] is the coordinate of the i-th point in the set P, q i =[q xi ,q yi ,q zi ] is the coordinate of the i-th point in the set Q, p i and q i is the associated point, R is the rotation matrix and T is the translation vector.

[0055] S2.3.2 Calculation and The optimal solution of can be calculated according to the geometric relationship between the associated point sets P and Q as follows:

[0056] S2.3.3. According to the obtained rotation matrix and translation vectors Perform homogeneous coordinate transformation on the original part point cloud to complete the registration of the original point cloud and the target point cloud;

[0057] S2.3.4. Fuse the point clouds from multiple perspectives, stitch the two frames of point clouds together, compensate for the missing point clouds due to overexposure at each perspective, and achieve precise alignment of the part point clouds.

[0058] Furthermore, step S3 is specifically as follows:

[0059] S3.1, Point cloud segmentation of complex free-form surface parts;

[0060] S3.2, construction of three-dimensional triangular mesh model of parts;

[0061] S3.3. Three-dimensional precision measurement of parts.

[0062] Furthermore, step S3.1 is specifically as follows:

[0063] S3.1.1. Input the part point cloud data, and first perform octree voxel coarse division on the point cloud according to its density; then, set the minimum voxel size threshold d min， Fine voxel segmentation of point cloud;

[0064] S3.1.2. In the finely divided octree voxels, according to the idea of ​​principal component analysis, the covariance matrix C is constructed using the point set Pi = [xi yi zi] in each voxel.

[0065]

[0066] In the formula, is the center of mass of the voxel.

[0067] S3.1.3, the matrix C is a symmetric positive definite matrix, and its eigenvalue decomposition is performed.

[0068]

[0069] Where λ1, λ2, λ3 are the three eigenvalues ​​of the matrix C, and e1, e2, e3 are the eigenvectors in the three directions of the corresponding point set.

[0070] S3.1.4, calculate the curvature δ, The curvature parameter value is used to determine the point cloud characteristics of the point cloud target, and the part target point cloud segmentation based on curvature constraints is realized.

[0071] Furthermore, step S3.2 is specifically as follows:

[0072] S3.2.1. Input the target point cloud data, represent the point cloud data with an octree, subdivide it so that each sampling point is a leaf node with a depth of D, and calculate the vector field of node N

[0073]

[0074] In the formula, q is a point in the point cloud set Q, α N,q is the weighting coefficient, is the normal vector information of q, F N (q) is the node function of node N.

[0075] S3.2.2. By function The vector v consisting of the Laplace operator projected on the function space solves the function Complete the Poisson equation The solution of

[0076] Where Δ is the Laplace operator, represents the indicator function corresponding to the surface, is the divergence operator.

[0077] S3.2.3. After solving the Poisson equation, the marching cube method MC is used to extract the isosurface, complete the surface reconstruction of the part point cloud, and construct a three-dimensional triangular mesh model of the part surface point cloud.

[0078] Furthermore, step S3.3 is specifically as follows:

[0079] S3.3.1. Slice the triangular mesh model of the part to obtain the cross-sectional data of the part, and divide the cross-sectional data into different point sets to obtain the cross-sectional point set data of various irregular target parts.

[0080] S3.3.2. Establish a characterization model based on the multivariate nonlinear optimal regression of the Marquardt method to fit the cross-sectional geometry of the component and calculate the measurement parameters of the component such as radius, chord length, thickness, etc.

[0081] S3.3.3. Based on the minimum inclusion area theory, a target cross-section error assessment model is established to complete the entire process of target precision measurement and quality evaluation.

[0082] Further, step S4 is specifically as follows:

[0083] The measured parameters of the machined parts are compared with the standard model template. If the parts machining accuracy requirements are met, the parts are qualified and the next part is processed by CNC.

[0084] If the part processing accuracy requirements are not met, the NC code is generated based on the 3D triangular mesh model of the processed part. Then, based on the Modbus and TCP / IP protocols, the data transmission between the visual measurement device and the machine tool numerical control device is completed. Finally, the machine tool executes the program according to the NC code and continues to the next step until the part meets the processing accuracy requirements.

[0085] Compared with the prior art, this application has the following beneficial effects:

[0086] 1. The present application involves an online fast and precise measurement method for complex free-form surface parts. It is based on optical imaging, combined with computer vision technology and geometric measurement principles, and directly measures the processed workpiece on a CNC machine tool. There is no need to remove the workpiece and then install it on the measuring platform for measurement. It can directly achieve comprehensive online real-time control of the manufacturing process of complex free-form surface parts, avoid secondary clamping errors, break through the technical bottleneck of low efficiency of current measurement methods, variable measurement references and insufficient flexibility of adaptive calibration, and is an inevitable development trend of current industrial intelligent measurement.

[0087] 2. This application can establish a complete integrated measurement and processing technology system, realize high-efficiency, high-quality precision processing of complex free-form surface parts, and form corresponding standards, specifications or process procedures to provide technical guidance and normative basis for subsequent applications.

[0088] 3. This application will form theoretical and technological innovations in the field of high-end equipment and new intelligent manufacturing, and will promote the integration of online rapid and precise measurement as an important link into the high-end equipment manufacturing production line, promote the collaborative innovation and industrial upgrading of high-end equipment, drive the formation and rapid development of the high-end equipment manufacturing industry chain, and enhance the core competitiveness of the high-end equipment manufacturing industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0089] Figure 1 Flow chart of online rapid and precise measurement of complex free-form surface parts in CNC machining;

[0090] Figure 2 Schematic diagram of online rapid and precise measurement of complex free-form surface parts during CNC machining;

[0091] Figure 3 Octree network structure;

[0092] Figure 4 Schematic diagram of the calibration board structure. DETAILED DESCRIPTION

[0093] In order to enable those skilled in the art to better understand the present invention, the technical solution of the present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0094] Example 1

[0095] like Figure 1-4 As shown, a device for online fast and precise measurement of complex free-form surface parts, the measuring device includes a six-degree-of-freedom industrial robot, an AGV, a visual measuring device and a multi-axis CNC machine tool, the six-degree-of-freedom industrial robot is installed on the AGV, the visual measuring device is installed at the end of the six-degree-of-freedom industrial robot, the visual measuring device includes a sinusoidal grating projector and a multi-eye camera, the sinusoidal grating projector is located at the center of the end of the six-degree-of-freedom industrial robot, the multi-eye camera is evenly distributed around the projector, and the camera optical axis is tilted toward the projector optical axis; the robot is used to guide the visual measuring device to complete the image information collection of the processed parts from multiple visions. The CNC machining of the complex free-form surface parts is completed by a multi-axis CNC machine tool, and the multi-axis CNC machine tool includes a CNC system, which is wirelessly connected to the six-degree-of-freedom industrial robot and the AGV through a control panel. The multi-eye camera is a four-eye camera.

[0096] The calibration plate is also included, and the calibration plate is installed on the protective door of the multi-axis CNC machine tool. The calibration plate is a circular target calibration plate with a fixed-pitch pattern array and a black background with white circles, such as Figure 4 shown.

[0097] Example 2

[0098] An online fast and precise measurement method for complex free-form surface parts includes the following steps:

[0099] S1. 3D reconstruction of 3D point cloud data of processed parts

[0100] After the four cameras have collected images, the deformed fringe images collected from the processed parts are reconstructed into three-dimensional point clouds to obtain part point cloud data from four perspectives.

[0101] S1.1. Calibration of visual measurement device system

[0102] Each camera in the visual measurement device forms a monocular structured light system with a projector. The entire system is calibrated as four monocular structured light systems to obtain the external parameters of the camera and projector.

[0103] S1.1.1. After the processing of the complex free-form surface parts is completed, the protective door of the CNC machine tool is opened, and the CNC system communicates with the AGV to control the six-degree-of-freedom industrial robot arm to move to the vicinity of the CNC machine tool. The six-degree-of-freedom industrial robot arm drives the visual measurement device to collect images of the calibration plate, and continuously changes the position and posture of the visual measurement device to collect multiple sets of calibration plate images;

[0104] S1.1.2. Take the center of the first circle in the upper left corner of the calibration plate as the origin, horizontally to the right as the x-axis direction, vertically downward as the y-axis direction, and determine the z-axis direction by the right-hand rule to complete the establishment of the world coordinate system, and obtain the relationship between the world coordinates of the key points on the calibration plate and their corresponding points in the pixel coordinate system:

[0105]

[0106] In the formula, (X W , Y W , Z W ) represents an arbitrary point in the world coordinate system, (u, v) represents the corresponding point in the camera pixel coordinate system, dx and dy represent the length of each pixel of the camera in the X direction and Y direction respectively; (u0, v0) is the coordinate of the pixel at the intersection of the camera optical axis and the imaging plane; f is the focal length of the lens, and the matrix R and vector T are the rotation matrix and translation vector of the camera coordinate system transformed to the world coordinate system respectively.

[0107] S1.1.3. Based on multiple sets of data, the monotonic matrix Hc between the two coordinate systems is calculated to obtain the internal and external parameters and distortion parameters of the camera;

[0108] S1.1.4. Use the LM (Levenberg Marquardt) optimization algorithm to update and optimize the internal and external parameters and distortion parameters of the camera.

[0109] S1.1.5. Map the key point coordinates in the camera pixel coordinate system to the projector pixel coordinate system, establish the relationship between the camera pixel coordinate system and the projector pixel coordinate system, and thus obtain the external parameter matrix of the camera, projector and world coordinate system.

[0110]

[0111] In the formula, (u p , v p ) represents the coordinate value in the projector pixel coordinate system, φ C Indicates the conversion factor.

[0112] S1.2. Control the movement of the six-degree-of-freedom industrial robot arm to drive the visual measurement device to collect images of complex free-form surface parts, continuously change the position and posture of the visual measurement device, and collect multiple sets of complex free-form surface part images;

[0113] S1.3, 3D reconstruction point cloud data

[0114] After the industrial robot arm guides the visual measurement device to complete the multi-visual and multi-directional image information acquisition of the processed parts, the external parameter matrix and the absolute phase information of the parts to be measured can be used to complete the conversion from the acquired fringe image (two-dimensional pixel coordinates) to the three-dimensional point cloud of the parts (three-dimensional space coordinates), and complete the three-dimensional reconstruction of the part point cloud data.

[0115] S1.4. Optimize the quality of reconstructed point cloud data

[0116] In the process of 3D point cloud reconstruction, there is a DMA (Discontinuity-induced Measurement Artefact) error in the area where the reflectivity of the part is discontinuous when the projector is projecting, which will cause the final reconstructed point cloud to have a large error. First, the sub-pixel edge detection method is used to determine the sub-pixel boundary of the reflectivity discontinuous area, and the phase interpolation calculation is performed through the phase of the nearby unaffected area to correct the phase error and obtain higher-precision point cloud data.

[0117] S2. Accurate registration of 3D point clouds

[0118] The four point clouds are registered using the coarse and fine registration methods of the point cloud based on the system calibration parameters, and then spliced ​​and fused to obtain point cloud data with the complete morphology of the processed parts to achieve accurate measurement.

[0119] S2.1 Preprocessing and denoising of 3D point clouds

[0120] The part point cloud obtained after 3D reconstruction generates a large number of noise points due to interference such as shadows and reflections. These noise points will have a great impact on subsequent point cloud processing. The specific process of removing these noises is as follows:

[0121] S2.1.1. Read four sinusoidal fringe images of the same frequency and set the shadow threshold;

[0122] S2.1.2, traverse the gray value of each pixel in each fringe image, and determine whether the gray value of each pixel at different times is always less than or equal to the shadow threshold. If this condition is met, it is a shadow area;

[0123] S2.1.3, making the shadow area into a mask image as a shadow mask;

[0124] S2.1.4, during 3D reconstruction, when indexing a pixel in the mask, mark the point as a shadow noise point;

[0125] S2.1.5. After removing the noise points, the point cloud composed of the remaining points is the point cloud without shadow noise.

[0126] S2.2. Rough registration of 3D point cloud

[0127] After obtaining noise-free point cloud data, due to the huge number of point clouds, it is necessary to downsample the point clouds without losing the part morphology, reduce the data volume, speed up the processing speed, and complete the rough registration of the part point clouds under the four camera perspectives, so that the poses of the point clouds under multiple perspectives are basically consistent. The specific process is as follows:

[0128] Use the octree data structure to describe the part point cloud, find the point in the voxel that is closest to the voxel center, and use this point as the sampling point of the current voxel to achieve downsampling;

[0129] Obtain the homogeneous transformation relationship between the projector and the four camera coordinate systems;

[0130]

[0131] Where M c =[R c ,T c ] is the homogeneous transformation matrix from the world coordinate system to the camera coordinate system, M p =[R p ,T p ] is the homogeneous transformation matrix from the world coordinate system to the projector coordinate system, X c , X p and X w They are the points in the camera, projector and world coordinate systems respectively.

[0132] The point clouds in the four camera coordinate systems are transformed into the projector coordinate system to complete the rough registration of the part point cloud, providing a more accurate initial registration value for the subsequent fine registration of the point cloud.

[0133] S2.3. Precise registration of 3D point clouds

[0134] The roughly registered point cloud is further finely registered to obtain a better registration effect and a complete three-dimensional shape of the part. The specific process is as follows:

[0135] S2.3.1. Represent the associated points in the two point clouds to be registered with point sets P and Q, and calculate the homogeneous transformation matrix from set Q to set P:

[0136]

[0137] In the formula, p i =[p xi ,p yi ,p zi ] is the coordinate of the i-th point in the set P, q i =[q xi ,q yi ,q zi ] is the coordinate of the i-th point in the set Q, p iand q i is the associated point, R is the rotation matrix and T is the translation vector.

[0138] S2.3.2 Calculation and The optimal solution can be calculated based on the geometric relationship between the associated point sets P and Q as follows:

[0139] S2.3.3. According to the obtained rotation matrix and translation vectors Perform homogeneous coordinate transformation on the original part point cloud to complete the registration of the original point cloud and the target point cloud;

[0140] S2.3.4. Fuse the point clouds from multiple perspectives, stitch the two frames of point clouds together, compensate for the missing point clouds due to overexposure at each perspective, and achieve precise alignment of the part point clouds.

[0141] S3. Precision measurement of complex free-form surface parts

[0142] The point cloud of the part is segmented by the curvature-constrained part point cloud segmentation method. Then, the shielded Poisson surface reconstruction method based on normal constraints is used to obtain the three-dimensional triangular mesh model of the part. Finally, the model is sliced ​​and the target cross-section data characterization method based on multivariate nonlinear optimal regression is used to obtain the cross-section data of the model and complete the three-dimensional precision measurement of the part.

[0143] S3.1. Point cloud segmentation of complex free-form surface parts

[0144] The effective marking information of the part point cloud is insufficient and the feature scale is single, which makes it difficult to effectively extract the features of the target point cloud with complex overall structure and large local differences, and reduces the accuracy of the part point cloud segmentation. To achieve accurate segmentation of complex free-form surface parts, the specific process is as follows:

[0145] S3.1.1. Input the part point cloud data, and first perform octree voxel coarse division on the point cloud according to its density; then, set the minimum voxel size threshold d min， Fine voxel segmentation of point cloud;

[0146] S3.1.2. In the finely divided octree voxels, according to the idea of ​​principal component analysis, the covariance matrix C is constructed using the point set Pi = [xi yi zi] in each voxel.

[0147]

[0148] In the formula, is the center of mass of the voxel.

[0149] S3.1.3, the matrix C is a symmetric positive definite matrix, and its eigenvalue decomposition is performed.

[0150]

[0151] Where λ1, λ2, λ3 are the three eigenvalues ​​of the matrix C, and e1, e2, e3 are the eigenvectors in the three directions of the corresponding point set.

[0152] S3.1.4, calculate the curvature δ, The curvature parameter value is used to determine the point cloud characteristics of the point cloud target, and the part target point cloud segmentation based on curvature constraints is realized.

[0153] S3.2. Construction of 3D triangular mesh model of parts

[0154] The shielded Poisson reconstruction method based on normal vector constraints can retain the surface details of the parts to the greatest extent and improve the accuracy of the three-dimensional triangular mesh model of the part surface point cloud. The specific process is as follows:

[0155] S3.2.1. Input the target point cloud data, represent the point cloud data with an octree, subdivide it so that each sampling point is a leaf node with a depth of D, and calculate the vector field of node N

[0156]

[0157] In the formula, q is a point in the point cloud set Q, α N,q is the weighting coefficient, is the normal vector information of q, F N (q) is the node function of node N.

[0158] S3.2.2. By function The vector v consisting of the Laplace operator projected on the function space solves the function Complete the Poisson equation The solution of

[0159] Where Δ is the Laplace operator, represents the indicator function corresponding to the surface, is the divergence operator.

[0160] S3.2.3. After solving the Poisson equation, the marching cube method MC is used to extract the isosurface, complete the surface reconstruction of the part point cloud, and construct a three-dimensional triangular mesh model of the part surface point cloud.

[0161] S3.3. Three-dimensional precision measurement of parts

[0162] S3.3.1. Slice the triangular mesh model of the part to obtain the cross-sectional data of the part, and divide the cross-sectional data into different point sets to obtain the cross-sectional point set data of various irregular target parts.

[0163] S3.3.2. Establish a characterization model based on the multivariate nonlinear optimal regression of the Marquardt method to fit the cross-sectional geometry of the component and calculate the measurement parameters of the component such as radius, chord length, thickness, etc.

[0164] S3.3.3. Based on the minimum inclusion area theory, a target cross-section error assessment model is established to complete the entire process of target precision measurement and quality evaluation.

[0165] S4. Online evaluation of machining accuracy of machined parts

[0166] The measured parameters of the machined parts are compared with the standard model template. If the parts machining accuracy requirements are met, the parts are qualified and the next part is processed by CNC.

[0167] If the part processing accuracy requirements are not met, the NC code is generated based on the 3D triangular mesh model of the processed part. Then, based on the Modbus and TCP / IP protocols, the data transmission between the visual measurement device and the machine tool numerical control device is completed. Finally, the machine tool executes the program according to the NC code and continues to the next step until the part meets the processing accuracy requirements.

[0168] The above embodiments are merely illustrative of the principles and effects of the present invention, and are not intended to limit the present invention. Anyone familiar with the technology may modify or change the above embodiments without violating the spirit and scope of the present invention. Therefore, all equivalent modifications or changes made by those with ordinary knowledge in the technical field without departing from the spirit and technical ideas disclosed by the present invention should still be covered by the claims of the present invention.

Claims

1. An online fast and precise measuring device for complex free-form surface parts, characterized in that: The measuring device includes a six-degree-of-freedom industrial robot, an AGV, a visual measuring device and a multi-axis CNC machine tool. The six-degree-of-freedom industrial robot is installed on the AGV, and the visual measuring device is installed at the end of the six-degree-of-freedom industrial robot. The visual measuring device includes a sinusoidal grating projector and a multi-eye camera. The sinusoidal grating projector is located at the center of the end of the six-degree-of-freedom industrial robot, and the multi-eye cameras are evenly distributed around the projector, and the camera optical axis is inclined toward the projector optical axis.

2. The on-line rapid and precise measuring device for complex free-form surface parts according to claim 1, characterized in that: The CNC machining of the complex free-form surface parts is completed by a multi-axis CNC machine tool, which includes a CNC system. The CNC system is wirelessly connected to a six-degree-of-freedom industrial robot arm and an AGV through a control panel.

3. The on-line rapid and precise measuring device for complex free-form surface parts according to claim 1, characterized in that: The multi-eye camera is a four-eye camera.

4. The on-line rapid and precise measuring device for complex free-form surface parts according to claim 1, characterized in that: It also includes a calibration plate, which is installed on the protective door of the multi-axis CNC machine tool.

5. The on-line rapid and precise measuring device for complex free-form surface parts according to claim 1, characterized in that: The calibration plate is a circular target calibration plate with a fixed-spacing pattern array and a black background and white circles.

6. An online fast and precise measurement method for complex free-form surface parts, the measurement method is based on the measurement device according to any one of claims 1 to 5, and specifically comprises the following steps: S1, 3D reconstruction of 3D point cloud data of processed parts; S2, accurate registration of 3D point clouds; S3. Precision measurement of complex free-form surface parts; S4. Online assessment of machining accuracy of machined parts.

Images