A method and system for detecting contour deformation of large workpieces

Abstract

A large workpiece contour deformation detection method and system, belonging to the field of artificial intelligence technology, the detection method comprises: constructing a measurement field: a measurement planning detection field based on a K-means clustering algorithm; establishing a workpiece feature digital model: extracting the workpiece features including shape features and position features; processing multi-sensor fusion data to integrate a laser tracking module and a visual measurement module and unify the spatial coordinates of each module; constructing a three-dimensional model, comparing the three-dimensional model with the feature digital model, calculating the error using the coordinate value, and deformation is considered when the error exceeds a threshold, thereby realizing quantitative detection of the contour deformation of the large workpiece.

Description

Technical Field

[0001] The present invention relates to the field of machine vision technology, and in particular to a large workpiece contour deformation detection method and system. Background Art

[0002] As the development direction of my country's manufacturing industry towards high-end, intelligent and green has gradually become clear. At present, China widely uses digital technology based on full three-dimensional models in the design and development of large and complex workpiece products such as aviation, aerospace, high-speed rail passenger cars, and automobiles. However, there is still immaturity in the digital detection technology based on full three-dimensional models, resulting in relatively low efficiency and high cost in manufacturing product processing and quality inspection. Therefore, solving the problem of high-precision detection of three-dimensional morphology deviations of large and complex workpiece products and realizing the application of product digital measurement technology are of great significance to improving the manufacturing and assembly quality and efficiency of automobiles, rail passenger cars and aircraft, which will help promote the transformation of traditional manufacturing to advanced manufacturing, improve the quality and efficiency of production and manufacturing, and bring significant economic benefits.

[0003] For the assembly process of the manufacturing industry, the precision measurement technology of large workpiece surface can realize the accurate collection of product assembly element information, and is the key technology to realize the interconnection between the physical state of assembly and the information network. As the assembly process develops towards intelligence, the precision measurement technology of large workpieces has become a key factor and an important force in promoting the rapid development of intelligent assembly, and is constantly developing towards the goal of intelligent measurement and feedback control. Large-scale digital detection mainly faces the following problems:

[0004] Problem 1: The purpose of building a measurement field is to meet the measurement needs of large workpieces. Due to the large size of large workpieces, a laser tracker or a binocular vision machine often cannot fully collect comprehensive data. Therefore, it is necessary to configure multiple measuring instruments and reasonably plan their number and positions.

[0005] Question 2: During the inspection of large workpieces, the workpiece itself or the inspection equipment often moves and shifts, which will cause errors in the measurement data and affect the inspection accuracy. If motion information compensation is not performed, the position and posture changes caused by the movement will be mapped to the measurement data, resulting in changes in the measurement value rather than the shape, which will cause false positives in the inspection results. Summary of the invention

[0006] The embodiment of the present invention provides a method and system for detecting contour deformation of a large workpiece. The technical solution is as follows:

[0007] On the one hand, a method for detecting contour deformation of a large workpiece is provided, the method being implemented by an electronic device and comprising:

[0008] Constructing the measurement field: measurement planning and detection field based on K-means clustering algorithm;

[0009] Establishing a workpiece feature digital model: extracting the workpiece features including shape features and position features, and obtaining their coordinate values ​​for unique naming and storage;

[0010] Processing multi-sensor fusion data: the multi-sensor includes a laser tracking module, a visual measurement module and unifies the spatial coordinates of each module;

[0011] Build a 3D model: The binocular vision module obtains 3D point cloud data, performs denoising, splicing, and rapid surface reconstruction on the 3D point cloud data;

[0012] Contour deformation detection: Compare the 3D model with the characteristic digital model, and use the coordinate value to calculate the error. If the error exceeds the threshold, it is considered deformation.

[0013] Preferably, the position, velocity and acceleration motion information of the large workpiece are obtained through a laser tracking module and a visual measurement module; based on the motion information, a motion model of the motion trajectory and posture change of the large workpiece is established; and the measurement data is corrected and calibrated through an interpolation method to eliminate errors caused by motion.

[0014] Preferably, the method for extracting contours from the workpiece shape feature is:

[0015] S201 uses CAI TA to extract component surface features;

[0016] S202 sets the reference plane P0, and moves the reference plane P0 by an offset distance Li through a plane offset operation to obtain an offset plane Pi;

[0017] S203 intersects the offset plane P i with the component outer surface S1 to obtain the component outer contour C i; S204 creates contour control points on the contour C i according to the curvature and specific detection requirements;

[0018] S205 repeats steps S202-S204 to complete the extraction of the shape features of the component.

[0019] Preferably, the fast surface reconstruction method is:

[0020] S301 obtains deviation detection data between the surface and the point cloud;

[0021] S302 selects an optimized NURBS surface, adjusts the direction according to the coordinates of all control points of the surface and the corresponding deviations, and observes the deformation position of the surface of the operation interface;

[0022] S303: selecting a control point to be adjusted and setting adjustment parameters;

[0023] S304 obtains a reorganized surface by using the surface formed by the new set of adjusted control points. Preferably, the K-means clustering algorithm plans the measurement field;

[0024] S401 Input known measurement point data set to synthesize the number of clusters K;

[0025] S402 distributes the measurement point data set to K clusters, and each measurement point belongs to only one cluster;

[0026] S403 calculates the centroid of each cluster as the centroid point representing the cluster;

[0027] S404 repeats the following steps until the centroid position no longer changes: reallocate each measurement point to the centroid with the shortest distance from it, update the measurement points included in each cluster; recalculate the centroid position of each cluster;

[0028] S405 determines two adjacent centroids between clusters, and the two centroids are used to determine the observation station and the placement position of the laser tracker;

[0029] S406 sets the placement area of ​​the laser tracker in the measurement field according to the position of the centroid, the measurement distance range and the pitch range of the laser tracker.

[0030] Preferably, the naming method is:

[0031] S501 prefix segment, used to indicate the part type,

[0032] S502 is an infix segment used to represent the main characteristic values ​​of the part. The main characteristic values ​​are size and shape. The characteristic values ​​are input into a hash function for calculation to generate a unique hash value.

[0033] S503 suffix segment, used to indicate the part number.

[0034] On the other hand, a large workpiece contour deformation detection system is provided, and the device is applied to a large workpiece contour deformation detection system method, and the device comprises:

[0035] Workspace construction module: measurement planning detection field based on K-means clustering algorithm, used to construct a workspace suitable for contour deformation detection;

[0036] Workpiece feature digital model building module: used to extract the shape features and position features of the workpiece, and save its coordinate values ​​with unique names;

[0037] Multi-sensor fusion data processing module: including laser tracking module and visual measurement module, used to unify the spatial coordinates of each module;

[0038] 3D model building module: Use the binocular vision module to obtain 3D point cloud data, and obtain a 3D model through denoising, splicing and fast surface reconstruction;

[0039] Contour deformation detection module: compares the 3D model with the workpiece feature digital model, calculates the error using the coordinate value, and determines that the workpiece contour is deformed when the error exceeds the threshold.

[0040] Preferably, it further comprises a data preprocessing module: preprocessing the data acquired from the multi-sensor fusion data module, including denoising, filtering and data correction, so as to improve the accuracy and reliability of the data.

[0041] Preferably, it also includes an error calculation module: according to the comparison result of the three-dimensional model and the workpiece characteristic digital model, the deformation error of the workpiece contour is calculated, and whether it is deformation is determined according to a preset threshold.

[0042] Preferably, it also includes real-time monitoring of data collection and processing during the measurement process to ensure the stability and accuracy of the system and issue an alarm in a timely manner.

[0043] Preferably, a machine learning module is also included: the contour deformation data of the workpiece is trained and analyzed by a machine learning algorithm to establish a workpiece deformation model to achieve automatic recognition and prediction of unknown workpiece deformation.

[0044] Preferably, it also includes a motion information compensation module, which performs motion compensation on the measurement data based on the motion information of the laser tracking module and the visual measurement module to eliminate errors caused by motion.

[0045] Preferably, the motion information compensation includes: motion information acquisition, motion model establishment, motion compensation algorithm, and real-time motion compensation.

[0046] Preferably, the motion information is acquired by acquiring the motion information of the workpiece or the measuring device, including the position, speed and acceleration, through a laser tracking module and a visual measurement module. The acquisition of the motion information can be achieved by using sensors, encoders or other measuring devices.

[0047] Preferably, the motion model is established by establishing a motion model of the workpiece or the measuring device according to the acquired motion information. The motion model may be a mathematical model based on the kinematic principle of the object, and is used to describe the motion trajectory and posture changes of the workpiece or the measuring device.

[0048] Preferably, a motion compensation algorithm is designed based on the established motion model to perform motion compensation on the measured data. The motion compensation algorithm can correct and rectify the measured data through mathematical calculations or interpolation methods according to the relationship between the motion model and the measured data to eliminate errors caused by motion.

[0049] Preferably, the real-time motion compensation: during the measurement process, motion information is acquired in real time, and real-time motion compensation is performed on the measurement data according to a motion compensation algorithm. Through real-time motion compensation, errors caused by motion can be corrected in time, thereby improving the accuracy and reliability of the measurement.

[0050] On the other hand, an electronic device is provided, which includes a processor and a memory, wherein the memory stores at least one instruction, and the at least one instruction is loaded and executed by the processor to implement the above-mentioned large workpiece contour deformation detection system method.

[0051] On the other hand, a computer-readable storage medium is provided, wherein at least one instruction is stored in the storage medium, and the at least one instruction is loaded and executed by a processor to implement the above-mentioned large workpiece contour deformation detection system method.

[0052] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0053] The measurement field is planned through the K-means algorithm, and the shape control points and position coordinates obtained by CAITA are extracted as the workpiece feature model. Then, the laser tracking and visual measurement data are integrated to obtain the 3D point cloud, and the surface is reconstructed to obtain the 3D model. The coordinates of the 3D model and the workpiece feature model are compared, and the error is calculated to determine whether the contour is deformed, so as to achieve quantitative detection of contour deformation of large workpieces; and data preprocessing, error calculation and real-time monitoring modules are added to improve accuracy; in order to eliminate motion errors, it can also obtain motion information, establish motion models, and design motion compensation algorithms to correct measurement data. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0055] Figure 1 It is a flow chart of a large workpiece contour deformation detection method provided by the present invention;

[0056] Figure 2 It is a schematic diagram of shape extraction in a large workpiece contour deformation detection method provided by the present invention;

[0057] Figure 3 It is a three-dimensional point cloud denoising logic block diagram of a large workpiece contour deformation detection method provided by the present invention;

[0058] Figure 4It is a three-dimensional reconstruction comparison model diagram in a large workpiece contour deformation detection method provided by the present invention; DETAILED DESCRIPTION

[0059] In order to make the technical problems, technical solutions and advantages to be solved by the present invention more clear, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

[0060] like Figure 1 As shown, a large workpiece contour deformation detection method, the method comprising:

[0061] S101 Construction of measurement field: measurement planning and detection field based on K-means clustering algorithm.

[0062] Specifically, the K-means clustering algorithm plans the measurement field; S401 inputs the known measurement point data set to synthesize the number of clusters K; S402 distributes the measurement point data set to K clusters, and each measurement point belongs to only one cluster; S403 calculates the centroid of each cluster as the centroid point representing the cluster; S404 repeats the following steps until the centroid point position no longer changes: a) reallocates each measurement point to the centroid point with the smallest distance from it, and updates the measurement points included in each cluster; b) recalculates the centroid point position of each cluster; S405 determines two adjacent centroid points between clusters, and these two centroid points are used to determine the rotation station position and the placement position of the laser tracker; S406 sets the placement area of ​​the laser tracker in the measurement field according to the position of the centroid point, the measurement distance range and the pitch range of the laser tracker.

[0063] The principle is: a measurement planning model based on the K-means clustering algorithm. The K-means clustering algorithm is an iterative data partitioning algorithm and an unsupervised learning method. If there are n data, the values ​​in the data are standardized as follows:

[0064]

[0065] Assume that the probability of a measurement point being selected as the mth (K≥m>1) cluster center is:

[0066]

[0067] Given the measurement point data set and cluster set, the loss function is defined as:

[0068]

[0069] Given K clusters, the number of measuring stations is K-1. The measuring points can be divided into K sets. In order to determine the location of the transfer station and the instrument, two adjacent point sets are merged to find the new centroid, that is:

[0070]

[0071] Taking the centroid as the intersection point, placing a laser tracker at a position perpendicular to this part of the measurement space can minimize the maximum measurement error of the station. The closest laser tracker is set according to the measurement range of the measurement distance and pitch angle, and the farthest laser tracker is limited by the farthest measurement distance that meets the accuracy. The space between the two extreme distances is the feasible domain space of the laser tracker.

[0072] It is worth mentioning that the purpose of constructing a measurement field is to meet the measurement needs of large workpieces. Due to the large size of large workpieces, a laser tracker or a binocular vision machine often cannot fully collect comprehensive data. Therefore, it is necessary to configure multiple measuring instruments and reasonably plan their number and positions.

[0073] By configuring multiple measuring instruments, the entire surface of the workpiece can be covered, and the data of the workpiece can be fully acquired, avoiding the problem of missing or inaccurate data due to the inability of a single instrument to cover the entire area. By rationally planning the number and position of instruments, the measurement efficiency can be maximized. Multiple instruments working at the same time can collect data in parallel and shorten the measurement time. In summary, by rationally planning the number and position of instruments, the advantages of multiple instruments can be fully utilized to achieve comprehensive and accurate measurement of the shape characteristics of large workpieces.

[0074] S102 establishes a workpiece feature digital model: extracts the workpiece features including shape features and position features, and obtains their coordinate values ​​for unique naming and storage.

[0075] Specifically, Figure 2 The method for contour extraction in the workpiece shape feature shown is: S201 uses CAITA (Computer-Aided Inspection and Test of Assemblies) to extract the surface features of the component; S202 sets the reference plane P0, and through the operation of offsetting the plane, moves the reference plane P0 by the offset distance Li to obtain the offset plane Pi; S203 intersects the offset plane Pi with the outer surface S1 of the component to obtain the component outer contour Ci; S204 creates contour control points on the contour Ci according to the curvature and specific detection requirements; S205 repeats steps S202-S204 to complete the extraction of the shape features of the component, and extracts the shape features of the component by intersecting the offset plane and the contour.

[0076] The S501 prefix segment is used to indicate the part type; the S502 infix segment is used to indicate the main characteristic values ​​of the part, which are size and shape. The characteristic values ​​are input into the hash function for calculation to generate a unique hash value; the S503 suffix segment is used to indicate the part number. The above naming method is convenient for debugging and optimization and has uniqueness and traceability, which facilitates the subsequent rapid detection of deformed parts.

[0077] It is worth mentioning that by setting the reference plane and detection requirements, contour control points that meet the requirements can be obtained, thereby realizing the extraction of the contour features of the workpiece shape. It is suitable for complex workpiece contours and improves the accuracy and efficiency of large workpiece contour extraction.

[0078] S103 processes multi-sensor fusion data: the multi-sensor includes a laser tracking module and a visual measurement module and unifies the spatial coordinates of each module.

[0079] The principle is that the core of multi-sensor data fusion is the unification of multi-sensor coordinates, that is, by building a multi-source system measurement field and observing the same common points to obtain the relationship between different coordinate systems, the coordinate unification between systems is completed. Assume that there are n transfer station common points in the measurement network, and the first transfer station common point p i In M j The set of common point measurement values ​​of the transfer station (j=1,2···,n) is A j (j=1,2,···,n). Let A0 be the set of optimal estimated positions of all transfer points under M0. Since the coordinate transformation is a rigid transformation, we can find M j The orientation estimation of (j=1,2···,m) under M0, that is, finding the common target point set A j (j=1,2,···,m) to A0-aligned coordinate transformation matrix.

[0080] x j 、z j Confirmed A j The translation vector T aligned to A0 j ,θ xj ,θ yj ,θ zj Confirmed A j The rotation matrix R aligned to A0 j , then the optimal orientation matrix δ(M j The solution of )(j=1,2,···,m) can be converted into the solution of A j The rotation matrix R of the optimal coordinate transformation to A0 j , translation vector T j At this time, the least squares principle of unequal precision distances between common points of multiple stations is adopted, which is expressed by the formula.

[0081]

[0082] If the common target point p i In M j can be measured, then its measured value is expressed as ξ ji =(x ji y ji z ji ) T , the standard deviation of each direction measurement is σ jix , σ jiy , σ jiz , then its covariance matrix is ​​V jiξ .

[0083]

[0084] The set of measurement samples at all stations is E i (ξ ji ∈E i ), weights are assigned according to the covariance matrix of the Cartesian coordinate system of each station, ξ ji The weight matrix of (j≠0) can be expressed as W ji

[0085]

[0086] make

[0087]

[0088] but:

[0089]

[0090] Assuming that the sensor units are independent of each other, their measurement variance matrix Q can be expressed as:

[0091]

[0092]

[0093] Where J represents the Jacobian matrix

[0094] According to formula (4), the data fusion weight matrix W can be obtained ij :

[0095]

[0096] In summary, it can be determined that the result of the data fusion of the measured values ​​in the global coordinate system obtained after the observation values ​​at multiple stations are transferred can be expressed by formula (1-10):

[0097] P j =∑W ij P ij

[0098] Among them, P j represents the coordinate value of the observation data fusion of the jth measured point in the global coordinate system; P ij represents the observed value of the jth measured point at the i-th measuring station; W ij Represents the data fusion weight of the jth measured point at the i-th measuring station.

[0099] Therefore, it can be seen that the covariance matrix V of the test point after data fusion can be expressed as:

[0100]

[0101] The covariance value V obtained after multi-sensor data fusion is smaller than the measured covariance value V at any independent station. i It can be seen that the coordinate estimate of the measured point in the global coordinate system after data fusion estimation is more accurate than the coordinate value in the global coordinate system obtained after the observation value obtained at each station is transferred. Therefore, high-precision acquisition and rapid processing of multi-sensor fusion measurement data of large-scale components can be achieved.

[0102] S104 builds a three-dimensional model: the binocular vision module obtains three-dimensional point cloud data, performs denoising, splicing, and fast surface reconstruction on the three-dimensional point cloud data;

[0103] Specifically, the rapid surface reconstruction method is as follows: S301 obtains the deviation detection data between the surface and the point cloud; S302 selects the optimized NURBS surface, and adjusts the direction according to the coordinates of all control points of the surface and the corresponding deviations, and observes the deformation part of the surface of the operation interface; S303 selects the control points that need to be adjusted and sets the adjustment parameters; S304 obtains the reconstructed surface by using the surface composed of a new set of adjusted control points.

[0104] Among them, the binocular vision module obtains three-dimensional point cloud data, such as Figure 3 The three-dimensional point cloud denoising method includes: obtaining noisy original point cloud data, performing K_D tree space division, grouping the point cloud data, judging for each data point P whether the number of spherical neighboring points of the data point P is less than a threshold value, if not, obtaining more accurate surface normal vector information, and performing bilateral filtering on the point cloud data after normal vector calculation to make the point cloud surface smoother. For data points whose number of spherical neighboring points is less than the threshold value, bilateral filtering is directly performed to finally obtain denoised point cloud data, in which noise points are effectively removed while retaining the surface features of the original point cloud.

[0105] The rapid surface analysis and reconstruction in the reconstruction of 3D CAD models is different from the traditional design process. It starts from the known relevant information of the physical object, and performs data collection, data processing and 3D reconstruction on the physical prototype. By constructing a 3D model with the same structural features as the original model, the original model is designed or re-innovated. In surface reconstruction, if the control parameters are too large, the deviation of the reconstructed surface will be too large. If the control parameters are too small, the fitting surface will be wrinkled and cause defects such as discontinuous surface curvature. Taking the example product as an example, a surface error evaluation method is constructed to optimize the surface parameters and control the surface fitting quality. The steps of surface reconstruction based on the "NURBS surface control points" secondary development of CAA / CAT IA are as follows: S301 obtains the deviation detection data between the surface and the point cloud;

[0106] S302 selects the optimized NURBS surface, and adjusts the direction according to the coordinates of all control points of the surface and the corresponding deviations, and observes the deformation of the surface of the operation interface; S303 selects the control points that need to be adjusted and sets the adjustment parameters; S304 obtains the reorganized surface by using the surface composed of a new set of adjusted control points.

[0107] Table 1 Parameter information of fitting surface control points

[0108]

[0109]

[0110] According to the deviation analysis results in Table 1, select the control point numbered 10. The dialog box will show that the deviation direction of the control point is (248.74, -733.95, 632.01), and the deviation analysis results show that the deviation distance is -1.3mm. Activate the response and generate the adjusted surface. Special note here: The accuracy and smoothness of the fitting surface are two conflicting technical indicators. The higher the accuracy of the fitting surface, the worse the smoothness, and vice versa. An ideal fitting surface needs to take into account smoothness as much as possible within the deviation range. Although the deviation between the fitting surface and the point cloud is 1.3mm, the setting of the adjustment parameters is sufficient to ensure that the deviation of the adjusted surface is not greater than the maximum allowable range (±1mm), so the adjustment distance here is 0.5mm. Use the same method to adjust the 6 control points with large deviations on the fitting surface.

[0111] After optimizing and cutting the surface with the "NURBS surface control point" tool, the surface was tested for deviation. The test results showed that the maximum positive deviation between the fitted point cloud and the reconstructed surface was +0.11mm, and the maximum negative deviation was -0.15mm.

[0112] Two methods are used to reconstruct the remaining three-dimensional surface to obtain a complete surface model, such as Figure 4Compared with the surface reconstructed by the optimization method, the model reconstructed by the traditional method has poor smoothness and obvious convex deformation.

[0113] Through systematic calculation, the accuracy of reconstructed models by the rapid surface optimization method is generally improved by 1.5 times, and the reconstruction time is reduced to. Using the equal illumination mapping curve analysis, the equal illumination curves of the reconstructed surface smoothness are more uniform, the surface is smoother and of higher quality.

[0114] The 3D deviation results show that the maximum deviation of the surface model reconstructed by the traditional method exceeds the required range because the allowable deviation of the part is ±1mm. The rapid surface reconstruction method reconstructs the model to meet the accuracy requirements. The local annotation and deviation bar show that the average deviation is less than 0.15mm, and the quality of rapid model reconstruction is significantly better than that of the traditional method.

[0115] S105 Contour deformation detection: Compare the 3D model with the characteristic digital model, calculate the error using the coordinate value, and if the error exceeds the threshold, it is considered deformation.

[0116] The principle is that contour deformation detection is a method for detecting shape changes of 3D models. First, it is compared with the characteristic digital model, which is a reference model defined or established in advance and contains the expected ideal shape and features. During the comparison process, the error between the coordinate values ​​of each point in the 3D model and the coordinate values ​​of the corresponding points in the characteristic digital model is calculated to determine whether the shape has changed. The error can be calculated using Euclidean distance or other appropriate distance measurement methods. A threshold is set. When the calculated error exceeds the threshold, it is considered that deformation has occurred. The threshold is set according to the requirements of the specific application and the actual situation.

[0117] S106 Motion compensation.

[0118] Specifically, the position, velocity and acceleration motion information of large workpieces are obtained through the laser tracking module and the visual measurement module;

[0119] Based on the motion information, a motion model of the motion trajectory and posture change of the large workpiece is established;

[0120] Through interpolation methods, the measured data are modified and corrected to eliminate errors caused by movement.

[0121] It is worth mentioning that by obtaining precise motion information, establishing an accurate motion model, and applying motion compensation algorithms for real-time processing, the system can eliminate errors caused by motion. This will improve the accuracy and reliability of the measurement results.

[0122] Further, a detection system is constructed according to the above method, including the following components:

[0123] Workspace construction module: The measurement planning detection field based on the K-means clustering algorithm is used to construct a workspace suitable for contour deformation detection; the K-means clustering algorithm is used to cluster the data in the measurement planning detection field, and similar data points are classified into one category, thereby constructing a workspace suitable for contour deformation detection, providing a suitable workspace for subsequent contour deformation detection, so as to perform detection more accurately.

[0124] Workpiece feature digital model building module: used to extract the shape features and position features of the workpiece, and save its coordinate values ​​with a unique name; extract the shape features and position features of the workpiece, and save its coordinate values ​​with a unique name, in order to identify and describe the workpiece, and facilitate subsequent contour deformation detection and error calculation.

[0125] Multi-sensor fusion data processing module: including laser tracking module and visual measurement module, used to unify the spatial coordinates of each module; including laser tracking module and visual measurement module, used to unify the spatial coordinates of each module, fuse the data obtained by different sensors, and obtain consistent spatial coordinates to improve the accuracy and reliability of measurement.

[0126] 3D model construction module: Use the binocular vision module to obtain 3D point cloud data, and obtain the 3D model through denoising, splicing and rapid surface reconstruction; Use the binocular vision module to obtain 3D point cloud data, and obtain the 3D model through denoising, splicing and rapid surface reconstruction. The 3D model is used to describe the shape and structure of the workpiece, providing basic data for subsequent contour deformation detection.

[0127] Contour deformation detection module: compare the three-dimensional model with the workpiece feature digital model, calculate the error using the coordinate value, and determine the workpiece contour deformation when the error exceeds the threshold. Compare the three-dimensional model with the workpiece feature digital model, calculate the error using the coordinate value, and determine the workpiece contour deformation when the error exceeds the threshold. By comparing the three-dimensional model and the feature digital model of the workpiece, it is detected whether the workpiece contour is deformed to determine whether the quality and shape of the workpiece meet the requirements. Further, it also includes a data preprocessing module: preprocess the data obtained from the multi-sensor fusion data module, including denoising, filtering and data correction to improve the accuracy and reliability of the data. Compare the three-dimensional model with the workpiece feature digital model, calculate the error using the coordinate value, and determine the workpiece contour deformation when the error exceeds the threshold. By comparing the three-dimensional model and the feature digital model of the workpiece, it is detected whether the workpiece contour is deformed to determine whether the quality and shape of the workpiece meet the requirements.

[0128] Furthermore, it also includes an error calculation module: according to the comparison results of the three-dimensional model and the workpiece feature digital model, the deformation error of the workpiece contour is calculated, and whether it is deformed is determined according to a preset threshold. According to the comparison results of the three-dimensional model and the workpiece feature digital model, the deformation error of the workpiece contour is calculated, and whether it is deformed is determined according to a preset threshold. The significance of this step is to evaluate the degree of contour deformation of the workpiece by calculating the error to determine whether the quality and shape of the workpiece meet the requirements.

[0129] Furthermore, it also includes real-time monitoring of data collection and processing during the measurement process to ensure the stability and accuracy of the system, and timely alarms to detect and solve problems in a timely manner to ensure the accuracy and reliability of the measurement results.

[0130] Furthermore, it also includes a machine learning module: the contour deformation data of the workpiece is trained and analyzed through a machine learning algorithm, and a workpiece deformation model is established to realize automatic recognition and prediction of the deformation of an unknown workpiece. The automatic recognition and prediction of the deformation of an unknown workpiece can be realized, and the contour deformation of the workpiece is modeled and analyzed through a machine learning method to improve the automation capability and accuracy of the system.

[0131] Furthermore, it also includes a motion information compensation module, which performs motion compensation on the measurement data based on the motion information of the laser tracking module and the visual measurement module to eliminate errors caused by motion. According to the motion information of the measuring device or the workpiece, the measurement data is corrected and rectified to eliminate errors caused by motion and improve the accuracy of measurement.

[0132] During the inspection of large workpieces, the workpiece itself or the inspection equipment often moves and shifts, and the movement will cause errors in the measurement data, affecting the inspection accuracy. If motion information compensation is not performed, the position and posture changes caused by the movement will be mapped to the measurement data, resulting in changes in the measurement value rather than changes in shape. This will cause false positives in the inspection results, which will be misjudged as shape changes that are actually caused by movement, resulting in distorted and inaccurate inspection results. By obtaining the motion information of the workpiece, establishing a motion model, and designing a motion compensation algorithm, the impact of the motion component on the measurement data can be analyzed, which achieves high-precision distinction between motion errors and shape changes, effectively improving the reliability of the inspection results. Eliminating motion errors is very important to ensure inspection quality.

[0133] In view of this, the motion information compensation includes: motion information acquisition, motion model establishment, motion compensation algorithm, and real-time motion compensation.

[0134] The motion information acquisition: the motion information of the workpiece or the measuring device, including the position, speed and acceleration, is acquired through the laser tracking module and the visual measurement module. The motion information can be acquired by using sensors, encoders or other measuring devices.

[0135] The motion model is established: a motion model of the workpiece or the measuring device is established according to the acquired motion information. The motion model may be a mathematical model based on the kinematic principle of the object, and is used to describe the motion trajectory and posture change of the workpiece or the measuring device.

[0136] Based on the established motion model, a motion compensation algorithm is designed to perform motion compensation on the measured data. The motion compensation algorithm can correct and rectify the measured data through mathematical calculation or interpolation method according to the relationship between the motion model and the measured data to eliminate the error caused by motion.

[0137] The real-time motion compensation: during the measurement process, motion information is acquired in real time, and real-time motion compensation is performed on the measurement data according to a motion compensation algorithm. Through real-time motion compensation, errors caused by motion can be corrected in time, thereby improving the accuracy and reliability of the measurement.

[0138] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for detecting contour deformation of a large workpiece, characterized in that: The method comprises: Constructing a measurement field: A measurement planning detection field based on a K-means clustering algorithm, wherein the K-means clustering algorithm plans the measurement field: S401 inputs a known measurement point data set to synthesize the number of clusters K; S402 distributes the measurement point data set to K clusters, with each measurement point belonging to only one cluster; S403 calculates the centroid of each cluster as the centroid point representing the cluster; S404 repeats the following steps until the centroid point position no longer changes: reallocates each measurement point to the centroid point with the smallest distance from it, updates the measurement points included in each cluster, and recalculates the centroid point position of each cluster; S405 determines two adjacent centroid points between clusters, and these two centroid points are used to determine the rotation station position and the placement position of the laser tracker; S406 sets the placement area of ​​the laser tracker in the measurement field according to the position of the centroid point, the measurement distance range and the pitch range of the laser tracker; Establish workpiece feature digital model: extract the workpiece features including shape features and position features, and obtain their coordinate values ​​for unique naming and storage. The method for contour extraction in the workpiece shape features is as follows: S201 uses CAITA to extract component surface features; S202 sets the reference plane P0, and through the operation of offsetting the plane, moves the reference plane P0 by the offset distance Li to obtain the offset plane Pi; S203 intersects the offset plane Pi with the component outer surface S1 to obtain the component outer contour Ci; S204 creates contour control points on the contour Ci according to the curvature and specific detection requirements; S205 repeats steps S202-S204 to complete the extraction of the shape features of the component; Processing multi-sensor fusion data: The multi-sensor includes a laser tracking module and a visual measurement module, and unifies the spatial coordinates of each module; Constructing a three-dimensional model: The binocular vision module obtains three-dimensional point cloud data, performs denoising, splicing and fast surface reconstruction on the three-dimensional point cloud data, and the fast surface reconstruction method is as follows: S301 obtains deviation detection data between the surface and the point cloud; S302 selects the optimized NURBS surface, and adjusts the direction according to the coordinates of all control points of the surface and the corresponding deviations, and observes the deformation part of the surface of the operation interface; S303 selects the control points that need to be adjusted and sets the adjustment parameters; S304 obtains the reconstructed surface by using the surface composed of a new set of adjusted control points; Contour deformation detection: Compare the 3D model with the characteristic digital model, and use the coordinate value to calculate the error. If the error exceeds the threshold, it is considered deformation.

2. The method according to claim 1, characterized in that: The following steps are also included: The position, velocity and acceleration motion information of large workpieces can be obtained through the laser tracking module and the visual measurement module; Based on the motion information, a motion model of the motion trajectory and posture change of the large workpiece is established; Through interpolation methods, the measured data are modified and corrected to eliminate errors caused by movement.

3. The method according to claim 1, characterized in that The naming method is: S501 prefix segment, used to indicate the part type, S502 infix segment is used to represent the main characteristic values ​​of the part, which are size and shape. Input the characteristic value into the hash function for calculation to generate a unique hash value; S503 suffix segment, used to indicate the part number.

4. A detection system constructed according to any one of the methods described in claims 1-3, characterized in that: Includes the following components: Workspace construction module: measurement planning detection field based on K-means clustering algorithm, used to construct a workspace suitable for contour deformation detection; Workpiece feature digital model building module: used to extract the shape features and position features of the workpiece, and save its coordinate values ​​with unique names; Multi-sensor fusion data processing module: including laser tracking module and visual measurement module, used to unify the spatial coordinates of each module; 3D model building module: Use the binocular vision module to obtain 3D point cloud data, and obtain a 3D model through denoising, splicing and fast surface reconstruction; Contour deformation detection module: compares the 3D model with the workpiece feature digital model, calculates the error using the coordinate value, and determines that the workpiece contour is deformed when the error exceeds the threshold.

5. The system according to claim 4, characterized in that It also includes a data preprocessing module: preprocessing the data obtained from the multi-sensor fusion data module, including denoising, filtering and data correction, to improve the accuracy and reliability of the data.

6. The system according to claim 4, characterized in that It also includes an error calculation module: based on the comparison results of the three-dimensional model and the workpiece feature digital model, the deformation error of the workpiece contour is calculated, and whether it is deformation is determined according to a preset threshold.

7. The system according to claim 4, characterized in that It also includes a motion information compensation module, which performs motion compensation on the measurement data based on the motion information of the laser tracking module and the visual measurement module to eliminate errors caused by motion.

8. The system according to claim 7, characterized in that The motion information compensation includes: motion information acquisition, motion model establishment, motion compensation algorithm, and real-time motion compensation.

9. The system according to claim 8, characterized in that The motion information acquisition: the motion information of the workpiece or the measuring device, including the position, speed and acceleration, is acquired through the laser tracking module and the visual measurement module. The motion information can be acquired by using sensors, encoders or other measuring devices.

10. The system according to claim 8, characterized in that Based on the established motion model, a motion compensation algorithm is designed to perform motion compensation on the measured data. The motion compensation algorithm can correct and rectify the measured data through mathematical calculation or interpolation method according to the relationship between the motion model and the measured data to eliminate the error caused by motion.

Images