Method and device for measuring thickness of large-area coating of curved-surface workpiece

Abstract

The invention provides a method and device for measuring the thickness of a large-area coating of a curved-surface workpiece, and relates to the technical field of coating thickness measurement, and the method comprises the steps: obtaining the point cloud data of a target coating region of the curved-surface workpiece, and calculating the Gaussian curvature of each point; dividing a curved surface corresponding to the target coating area into a plurality of target curved surface patches with different curvature levels according to Gaussian curvature, and further generating a curved surface model; dividing the initial grid based on the curved surface model to obtain a plurality of initial measurement points; controlling the mechanical arm to measure the thickness of the target coating area according to each preliminary measurement point to obtain a preliminary thickness data set; adjusting grid nodes of the initial grid according to the initial thickness data set to obtain a plurality of target measurement points; controlling the mechanical arm to measure the thickness of the target coating area according to each target measuring point to obtain a target thickness data set; and generating a thickness distribution cloud picture according to the target thickness data set. According to the invention, high-precision measurement of the coating thickness of the large-area curved-surface workpiece is realized.

Description

Technical Field

[0001] This application relates to the field of coating thickness measurement technology, specifically to a method and apparatus for measuring the coating thickness of a large area on a curved workpiece. Background Technology

[0002] In industrial production and quality inspection, accurate measurement of the coating thickness on the workpiece surface plays a crucial role in ensuring product quality and performance. For workpieces with curved surfaces, the surface typically exhibits both convex and concave regions. During the deposition of coating particles, influenced by gravity, the coating thickness is relatively thinner in the convex regions, while the coating thickness is thicker in the central concave region.

[0003] Current techniques for measuring coating thickness on curved workpieces primarily employ point-by-point measurement. This method fails to fully consider the geometric characteristics of the curved surface, resulting in a lack of representativeness in the selection of measurement points and an inability to objectively and comprehensively reflect the overall coating thickness distribution of the workpiece. This problem is particularly pronounced when measuring the coating thickness of large-area curved workpieces. Existing methods struggle to effectively measure the coating thickness over large areas, significantly reducing the accuracy of the measurement results and severely impacting the reliability and effectiveness of workpiece coating quality assessment. Therefore, improving the accuracy of overall coating thickness measurement for large-area curved workpieces is a pressing technical issue that needs to be addressed. Summary of the Invention

[0004] In view of the above-mentioned shortcomings of the prior art, this application provides a method and apparatus for measuring the coating thickness of large-area curved workpieces, which effectively solves the problem of inaccurate measurement of the overall coating thickness of large-area curved workpieces.

[0005] In a first aspect, this application provides a method for measuring the coating thickness over a large area of ​​a curved workpiece, the method comprising: Acquire point cloud data of the target coating area of ​​the curved workpiece, and calculate the Gaussian curvature of each point based on the point cloud data; Based on the Gaussian curvature, the surface corresponding to the target coating region is divided into multiple target surface patches with different curvature levels, and a surface model is generated based on the target surface patches. An initial mesh was generated based on the surface model to obtain multiple preliminary measurement points; Based on each of the preliminary measurement points, the robotic arm is controlled to perform thickness measurement of the target coating area to obtain a preliminary thickness dataset. Adjust the grid nodes of the initial grid based on the preliminary thickness dataset to obtain multiple target measurement points; Based on each of the target measurement points, the robotic arm is controlled to perform thickness measurement of the target coating area to obtain a target thickness dataset; Generate a thickness distribution cloud map based on the target thickness dataset.

[0006] In an optional implementation, calculating the Gaussian curvature of each point based on the point cloud dataset includes: A density-based clustering algorithm is used to filter out background noise from the point cloud data to obtain initial point cloud data; The high-frequency noise of the initial point cloud data is filtered out using the moving least squares method to obtain the target point cloud data; The Gaussian curvature is obtained by calculating the curvature tensor of each point based on the target point cloud data through local neighborhood fitting of the surface.

[0007] In an optional implementation, the surface corresponding to the target coating region is divided into multiple target surface patches with different curvature levels according to the Gaussian curvature, and a surface model is generated based on the target surface patches, including: The corresponding neighboring points of the first target point in the target point cloud data are determined by spatial nearest neighbor search; Calculate the overall curvature similarity based on the Gaussian curvature of the first target point and the Gaussian curvature of the neighboring points; If the overall curvature similarity is less than or equal to the similarity threshold, the first target point and the neighboring points are divided into the same surface patch; otherwise, the second target point is selected to determine the neighboring points and the surface patch is re-divided until all points of the target point cloud data are divided to obtain multiple initial surface patches. Based on the comparison of the Gaussian curvature and graded curvature threshold of all points on each initial surface patch, each initial surface patch is divided into target surface patches with different curvature levels. The target surface patches are merged to obtain the surface model.

[0008] In an optional implementation, an initial mesh is generated based on the surface model to obtain multiple preliminary measurement points, including: The surface boundary of the freeform surface is determined based on the Gaussian curvature and normal vector at each point on the surface model. Based on the surface boundary, a Chebyshev mesh is generated for the free-form surface. The mesh spacing of the Chebyshev mesh varies in the corresponding regions of the target surface patch at different curvature levels, thus obtaining the initial mesh. Each grid node of the initial grid is used as the initial measurement point.

[0009] In an optional implementation, the robotic arm is controlled to perform thickness measurements of the target coating area based on each of the preliminary measurement points to obtain a preliminary thickness dataset, including: A motion path is generated based on each of the preliminary measurement points, and the movement of the robotic arm is controlled according to the motion path. When the robotic arm moves to each of the initial measurement points, the posture of the robotic arm is adjusted according to the point cloud data; The robotic arm, after adjusting its posture, performs thickness measurements on the initial measurement points to obtain the initial thickness dataset.

[0010] In an optional implementation, generating a motion path based on each of the preliminary measurement points includes: Based on the curvature level and spatial distribution of each of the preliminary measurement points, multiple connected regions are divided, and the connected regions are prioritized to obtain the motion sequence; Traverse the initial measurement points of each of the connected regions to generate multiple local paths, and set transition paths in each of the connected regions; The local paths are connected sequentially according to the motion sequence, and the transition paths are connected using a smooth curve to obtain the motion path.

[0011] In an optional implementation, the mesh nodes of the initial mesh are adjusted according to the preliminary thickness dataset to obtain multiple target measurement points, including: Traverse all grid nodes of the initial grid as target grid nodes, and calculate the absolute thickness difference between the two points based on the thickness data of the target grid node and the corresponding neighboring grid nodes. Calculate the dynamic correction threshold between the two points based on the Gaussian curvature of the target grid node and the neighboring grid nodes; If the absolute thickness difference is greater than the dynamic correction threshold, then intermediate measurement points are added based on the coordinates of the target mesh node and the neighboring mesh nodes; If the absolute thickness difference is less than or equal to the dynamic correction threshold, then the spacing between the target mesh node and the neighboring mesh node is maintained. All grid nodes of the initial grid and the supplementary intermediate measurement points are used as the target measurement points.

[0012] In an optional implementation, a dynamic correction threshold between two points is calculated based on the Gaussian curvature of the target mesh node and the neighboring mesh nodes, including: Calculate the geometric mean curvature based on the Gaussian curvature of the target grid node and the Gaussian curvature of the neighboring grid nodes; The relative curvature deviation between the two nodes is calculated based on the Gaussian curvature of the target grid node, the Gaussian curvature of the neighboring grid nodes, and the geometric mean curvature. The exponential decay correction coefficient is determined based on the geometric mean curvature, and the linear decay correction coefficient is determined based on the relative deviation of the curvature. The dynamic correction threshold is obtained by combining the exponential decay correction coefficient and the linear decay correction coefficient between the two points with a preset critical difference.

[0013] In an optional implementation, generating a thickness distribution cloud map based on the target thickness dataset includes: Data interpolation is performed based on the target thickness dataset to fill in the blank areas between each target measurement point, thereby obtaining thickness distribution data. Color mapping is performed based on the thickness distribution data to obtain the thickness distribution cloud map.

[0014] Secondly, this application provides a large-area coating thickness measuring device for curved workpieces, the device comprising: The data processing module is used to acquire point cloud data of the target coating area of ​​the curved workpiece and calculate the Gaussian curvature of each point based on the point cloud data. The model generation module is used to divide the surface corresponding to the target coating area into multiple target surface patches with different curvature levels according to the Gaussian curvature, and generate a surface model based on the target surface patches. The mesh generation module is used to generate an initial mesh based on the surface model to obtain multiple preliminary measurement points; The preliminary measurement module is used to control the robotic arm to perform thickness measurement of the target coating area based on each of the preliminary measurement points, and to obtain a preliminary thickness dataset. The grid adjustment module is used to adjust the grid nodes of the initial grid according to the preliminary thickness dataset to obtain multiple target measurement points; The target measurement module is used to control the robotic arm to perform thickness measurement of the target coating area according to each of the target measurement points, so as to obtain a target thickness dataset; The cloud map generation module is used to generate a thickness distribution cloud map based on the target thickness dataset.

[0015] The method and apparatus for measuring the coating thickness of curved workpieces provided in this application achieve accurate geometric modeling of complex curved workpieces through high-precision 3D scanning and surface fitting. Chebyshev mesh generation is performed using curvature grading and normal vector analysis, improving the rationality of the measurement point layout. Simultaneously, a dynamic threshold mechanism based on the geometric mean of Gaussian curvature and relative deviation is introduced to adaptively adjust the spacing between measurement points. Measurement points are automatically densified in areas with significant coating thickness gradient changes, effectively capturing local abrupt changes while balancing measurement accuracy and efficiency. Finally, a robotic arm is controlled to automatically measure according to an automatically generated motion path, obtaining thickness measurement data to generate a full-coverage thickness distribution cloud map, achieving high-precision measurement and full-area visualization of the coating thickness of large-area curved workpieces. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a schematic flowchart of a method for measuring the coating thickness of a curved workpiece in an embodiment of this application. Figure 2 This is a three-dimensional scanning schematic diagram of a local coating area of ​​a curved workpiece in an embodiment of this application; Figure 3 This is a schematic diagram of a curved surface model of a local coating area of ​​a curved workpiece in an embodiment of this application; Figure 4 This is a schematic diagram of the initial mesh division of a local coating area of ​​a curved workpiece in an embodiment of this application; Figure 5 This is a schematic diagram of the movement path of the robotic arm in a certain local coating area of ​​a curved workpiece in an embodiment of this application; Figure 6 This is a schematic diagram of the thickness distribution cloud map of a certain local coating area of ​​a curved workpiece in an embodiment of this application; Figure 7 This is a schematic diagram of the structure of the large-area coating thickness measuring device for curved workpieces provided in the embodiments of this application; Figure 8 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0018] Key component symbols: 200 - Large-area coating thickness measuring device for curved workpieces; 210 - Data processing module; 220 - Model generation module; 230 - Mesh generation module; 240 - Preliminary measurement module; 250 - Mesh adjustment module; 260 - Target measurement module; 270 - Cloud map generation module; 300 - Electronic equipment; 310 - Processor; 320 - Communication interface; 330 - Memory; 340 - Communication bus. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be further described clearly and completely below with reference to the accompanying drawings of the embodiments. It should be noted that the described embodiments are merely some embodiments of this application, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0020] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0021] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of this application.

[0022] In aircraft manufacturing, the surface coating of skin components is a key factor determining the core performance of the workpiece, such as corrosion resistance and wear resistance. However, most skin components are large-area curved surfaces, and the differences in Gaussian curvature at the surface nodes create convex and concave feature areas. Existing workpiece coating thickness measurement methods generally use a fixed-interval grid distribution, which is difficult to combine with the complex structural morphology of the workpiece and the actual variation law of coating thickness, resulting in an inaccurate match between the measurement points and the actual characteristics of the object being inspected. This problem is particularly prominent in the coating thickness measurement scenario of large-area curved workpieces, making it difficult to objectively and comprehensively reflect the overall coating thickness distribution of the workpiece, seriously affecting the reliability and effectiveness of workpiece coating quality assessment.

[0023] This application provides a method for measuring the coating thickness of a large area of ​​curved workpieces, which effectively solves the problem of inaccurate measurement of the overall coating thickness of large-area curved workpieces. Figure 1 This is a first schematic diagram of the process for measuring the coating thickness of a curved workpiece over a large area, as provided in the embodiments of this application. Figure 1 As shown, the method includes the following steps: S100: Obtain point cloud data of the target coating area of ​​the curved workpiece, and calculate the Gaussian curvature of each point based on the point cloud data.

[0024] In this embodiment of the application, the curved workpiece includes, but is not limited to, curved workpieces such as the skin parts of an aircraft, and the target coating area can be a large-area curved coating area.

[0025] As an optional implementation of this application, a structured light scanner can be used. This scanner projects a coded grating onto the surface of the target coating area of ​​the curved workpiece, and uses a binocular camera to capture the distortion pattern caused by the surface undulations of the grating, thus obtaining a three-dimensional scan of the curved workpiece. Then, based on the three-dimensional scan, the three-dimensional coordinates of each pixel are calculated using the principle of triangulation, thereby obtaining the point cloud data of the target coating area of ​​the curved workpiece.

[0026] For example, taking a large-area curved surface skin part covered with a paint layer as an example, eight 5mm diameter reflective target balls made of highly reflective material are attached to the edges and center of the skin part. The marked points on the reflective target balls serve as coordinate references. During scanning, the reflective target balls are individually identified and their coordinates are recorded, which are subsequently used for alignment of the inner and outer surfaces. The scanning resolution of the structured light scanner is adjusted according to the accuracy requirements, for example, set to 0.1mm / point. The skin part is scanned and images are captured using a structured light scanner and a binocular camera. Figure 2 This is a three-dimensional scanning schematic diagram of a local coating area of ​​a curved workpiece in an embodiment of this application, such as... Figure 2 As shown, the final generated point cloud contains 100,000 to 150,000 pixels per square meter. Based on this 3D scan image, the 3D coordinates of each pixel are calculated using the principle of triangulation to obtain the point cloud data.

[0027] In this embodiment of the application, the Gaussian curvature of each point is calculated based on the point cloud data, specifically including the following steps: S110. Use a density-based clustering algorithm to filter out background noise from the point cloud data to obtain initial point cloud data.

[0028] As an optional implementation of this application, a density-based clustering algorithm can be used to remove background noise and outliers to obtain initial point cloud data. Density-based clustering is an unsupervised learning algorithm that partitions clusters based on density connectivity. A cluster is a continuous set of regions with sufficiently high density, while regions below a density threshold are considered noise points. If the point cloud data in this application embodiment... ε The neighborhood contains at least MinPts samples:

[0029] These samples are the core points, among which Representing point cloud data x of ε The set of all samples within the neighborhood.

[0030] Then, clusters are constructed using density reachability and density accessibility. Ultimately, all density-accessible core points and their ε-neighborhood samples are grouped into the same cluster. Samples not included in any cluster are identified as noise points, thus identifying clusters of arbitrary shapes and automatically filtering outliers. Density reachability is defined as: if a sample chain exists... And if the densities of adjacent samples are directly accessible, then the sample x 1 and sample x n The density can reach [a certain value].

[0031] S120. The moving least squares method is used to filter out high-frequency noise in the initial point cloud data to obtain the target point cloud data.

[0032] As an optional implementation of this application, the moving least squares method can be used to smooth high-frequency noise in the initial point cloud data, thereby effectively suppressing point cloud jitter caused by high-frequency noise such as orange peel texture. The moving least squares method, through local weighted least squares fitting, filters out microscopic irregular disturbances while preserving the macroscopic geometric features of the surface, improving data continuity and smoothness. After least squares fitting processing, the coordinate stability of the initial point cloud data is significantly enhanced, providing a high-precision and low-noise geometric basis for subsequent key point extraction and robotic arm measurement path planning.

[0033] Optionally, after the above noise filtering process, the amount of target point cloud data can be controlled to within 100,000 points per square meter while ensuring the integrity of the features.

[0034] S130. Calculate the curvature tensor of each point by fitting a surface to the local neighborhood based on the target point cloud data, and obtain the Gaussian curvature.

[0035] In this embodiment, an arbitrary point is used as a seed, and a local neighborhood fitting surface is performed based on a region growing algorithm to calculate the Gaussian curvature of each point in the target point cloud data.

[0036] First, taking the target point as the center, select its k nearest neighbors, and use the least squares method to fit the surface to obtain the surface equation. Then, after solving for the coefficients, the surface equation is transformed into the surface parametric equation form as follows:

[0037] In the above formula, , representing the coordinate variables in the two-dimensional parameter domain, used to uniquely identify points on a three-dimensional surface using two-dimensional indices. T Indicates the frequency.

[0038] Then, the Gaussian curvature of the target point is calculated based on the properties of the spatial surface curve. The calculation formula is as follows:

[0039] In the above formula, E , F and G Denotes the coefficients of the first fundamental form, where , , , L , M and N Denotes the coefficients of the second fundamental form, where , , , n Represents the unit normal vector of the surface. r u and rv This represents the first-order partial derivative of the surface parametric equation. r uu , r uv and r vv The second-order partial derivatives of the surface parametric equations are expressed as follows: ,

[0040] , ,

[0041] In this embodiment of the application, after Gaussian curvature calculation, the corresponding normal vector and Gaussian curvature are associated with the corresponding points in the target point cloud data, so that each point in the target point cloud data contains three-dimensional coordinates, normal vector and Gaussian curvature.

[0042] This application embodiment acquires point cloud data of the target area of ​​a curved workpiece through high-precision 3D scanning, and performs scanning noise removal processing on the point cloud data to effectively eliminate outliers and high-frequency interference, thereby improving the surface fitting accuracy. The surface geometric features are quantified by calculating Gaussian curvature, providing data for subsequent region division by curvature and optimization of measurement point distribution.

[0043] S200. Divide the surface corresponding to the target coating area into multiple target surface patches with different curvature levels according to the Gaussian curvature, and generate a surface model based on the target surface patches.

[0044] In this embodiment, the target point cloud data of the target coating region is fitted into a non-uniform rational B-spline NURBS smooth surface. Based on the Gaussian curvature of each point, the surface corresponding to the target coating region is divided into multiple target surface patches, each with a corresponding curvature level. Finally, a surface model is generated based on the target surface patches. The generation of the surface model specifically includes the following steps: S210. Determine the corresponding neighboring points of the first target point in the target point cloud data through spatial nearest neighbor search.

[0045] In this embodiment of the application, any center point in the target point cloud data is taken as the first target point. P 0. Nearest neighbor search is used to filter points whose spatial distance is less than or equal to twice the average spacing of the initial point cloud. k Each point is considered a neighborhood point, and these neighborhood points are uniformly distributed in space. Optional, k We can choose 8, then the 8 neighboring points are denoted as follows: .

[0046] S220. Calculate the overall curvature similarity based on the Gaussian curvature of the first target point and the Gaussian curvature of the neighboring points.

[0047] In this embodiment, the curvature difference of a single point pair can be quantified using relative curvature difference, thereby avoiding interference from the magnitude of the absolute curvature value in the difference judgment. The formula for calculating relative curvature difference is as follows:

[0048] In the above formula, Indicates the first i The relative curvature difference between each neighboring point and the first target point K 0 represents the Gaussian curvature of the first target point. K i Indicates the first i Gaussian curvature of a neighborhood point , represents a local minimum, used to avoid ideal planes. K Division overflow when 0=0.

[0049] The overall curvature similarity between the first target point and its eight neighboring points is obtained by combining the curvature differences of the average relative curvature differences. The calculation formula is as follows:

[0050] In the above formula, This indicates the overall curvature similarity.

[0051] S230. If the overall curvature similarity is less than or equal to the similarity threshold, the first target point and its neighboring points are divided into the same surface patch. Otherwise, the second target point is selected to determine the neighboring points and the surface patch is re-divided until all points of the target point cloud data are divided, and multiple initial surface patches are obtained.

[0052] In this embodiment, the similarity threshold can be set according to the actual situation of the curved workpiece. Optionally, the similarity threshold can be set to 15%. If the overall curvature similarity corresponding to the first target point is less than or equal to 15%, it is determined that the curvature of the 8 neighboring points is similar to that of the first target point, and they can be classified into the same surface patch. Otherwise, it is determined that the curvature of the 8 neighboring points is significantly different from that of the first target point, and the neighboring points may belong to different surface features. They need to be divided into different surface patches. Then, any center point in the target point cloud data is reselected as the second target point, and the neighboring points are re-determined to re-divide the surface patches until all points in the target point cloud data are divided, resulting in multiple initial surface patches.

[0053] As an optional implementation of this application, the included angle between the normal vectors of adjacent initial surface patches can be set. a ≤30° to ensure topological continuity.

[0054] S240. Based on the comparison of the Gaussian curvature and graded curvature threshold of all points on each initial surface patch, divide each initial surface patch into target surface patches with different curvature levels.

[0055] In the embodiments of this application, graded curvature thresholds can be set according to the actual situation of the curved workpiece. For example, two graded curvature thresholds can be set, and the Gaussian curvature of all points on each initial surface piece can be traversed. The Gaussian curvature of each point is compared with the two graded curvature thresholds, and each initial surface piece is further divided into target surface pieces with low curvature, medium curvature, and high curvature.

[0056] For example, the two graded curvature thresholds can be set to 0.1 mm. - ¹ and 0.3mm - ¹, If the Gaussian curvature K of the initial surface patch at a certain point is ≤0.1mm - ¹, then this point is classified as a point in the low-curvature target surface patch; if 0.1mm - ¹ <K≤0.3mm - ¹, then this point is classified as a point in the target surface patch with medium curvature; if K>0.3mm - ¹, then the point is assigned to a point in the high curvature target surface patch. After all the points in the initial surface patch are divided, the points of the corresponding curvature level are integrated to obtain target surface patches of different curvature levels.

[0057] S250. Merge the various target surface patches to obtain the surface model.

[0058] In this embodiment, the fitted target surface patches are merged to form a complete NURBS smooth surface model, and then the overall surface model is fine-tuned to improve the surface quality and smoothness. Figure 3 This is a schematic diagram of a curved surface model of a local coating area of ​​a curved workpiece in an embodiment of this application, such as... Figure 3 As shown, a surface model is obtained by fitting a surface patch to the target point cloud data, thereby restoring the design shape of surface workpieces such as skin parts.

[0059] This application's embodiments divide the workpiece into different surface patches based on Gaussian curvature similarity and classify these patches according to different curvature levels, which facilitates differentiated measurement point layouts. The model generated by fitting the target point cloud data restores the true geometric shape of the workpiece surface, improving the surface continuity and smoothness. Through surface patch division and surface fitting, the subsequent measurement point distribution better matches the actual surface characteristics, enhancing the adaptability and accuracy of the thickness measurement path, especially improving the detection reliability of complex surface areas.

[0060] S300: Based on the surface model, an initial mesh is generated to obtain multiple preliminary measurement points.

[0061] In this embodiment, a boundary identification method combining Gaussian curvature and normal vector analysis is used to determine the boundaries of the freeform surface in the surface model, and a Chebyshev mesh is generated on the freeform surface. Each mesh node is used as a preliminary measurement point. The determination of the preliminary measurement points specifically includes the following steps: S310. Determine the surface boundary of the freeform surface based on the Gaussian curvature and normal vector of each point on the surface model.

[0062] As an optional implementation of this application, a Gaussian curvature threshold and a normal vector deviation threshold can be set based on the 8-neighborhood data of each point on the surface model, according to the actual situation of the surface workpiece. If the relative curvature difference between a target point and its neighboring points is greater than or equal to the Gaussian curvature threshold of 0.5mm... -1 If the deviation of the angle between the normal vectors of the two points is greater than or equal to the normal vector deviation threshold, then the target point is determined to be in a geometric abrupt change region and marked as a boundary candidate point. The Gaussian curvature change reflects abrupt changes in the surface curvature, while the normal vector deviation reflects a transition in surface orientation. Together, they can accurately capture physical boundaries such as edges and corners on freeform surfaces. Optionally, the Gaussian curvature threshold can be set to 0.5 mm. -1 The normal vector deviation threshold is set to 15°.

[0063] By performing joint analysis on each point on the surface model in sequence, marking all candidate boundary points that meet the conditions, eliminating isolated noise points through connected component analysis, retaining a continuous set of boundary points, and using a curve fitting algorithm to generate closed and smooth boundary contours, the automatic identification of complex freeform surface boundaries is achieved.

[0064] S320. Generate Chebyshev meshes for free-form surfaces based on surface boundaries. The mesh spacing of the Chebyshev meshes varies in the regions corresponding to the target surface patches at different curvature levels, thus obtaining the initial mesh.

[0065] In this embodiment, the Chebyshev mesh is a structured mesh with nodes densely distributed in the boundary region and sparsely distributed in the central region, suitable for geometric feature analysis and measurement point planning of freeform surfaces. The coordinate formula of a Chebyshev node is as follows:

[0066] In the above formula, u i Represents the first in the two-dimensional parameter domain i One Chebyshev node, N This represents the total number of Chebyshev nodes.

[0067] Two-dimensional parameter domain ( u,v Chebyshev nodes within the surface are obtained through the surface parametric equations. Mapped onto a 3D curved surface model, it forms a Chebyshev mesh with a rectangular topology.

[0068] Simultaneously, by quantizing the output using a piecewise function, the mesh spacing for the corresponding region of the low-curvature target surface patch is set to 10mm, the mesh spacing for the corresponding region of the medium-curvature target surface patch is set to 5-10mm, and the mesh spacing for the corresponding region of the high-curvature target surface patch is set to 5mm. A smooth transition mesh with a width of 10mm is set at the boundary zone between the corresponding regions of the target surface patches of different curvature levels to obtain the initial mesh.

[0069] S330. Use each grid node of the initial grid as a preliminary measurement point.

[0070] In this embodiment of the application, the grid nodes at the intersection of each grid in the generated initial grid are used as the preliminary measurement points of the target coating area of ​​the curved workpiece.

[0071] For example, if the flattened dimensions of a target coating area of ​​a curved workpiece are 0.5m × 0.5m, then an initial mesh generation will produce approximately 2500 preliminary measurement points. Figure 4 This is a schematic diagram of the initial mesh division of a local coating area of ​​a curved workpiece in an embodiment of this application, as shown below. Figure 4 As shown in the figure, the grid nodes are the initial measurement points.

[0072] This application employs a Chebyshev mesh for initial mesh generation, resulting in a denser distribution of initial measurement points in the surface boundary region and a relatively sparser distribution in the central region, effectively adapting to the geometric characteristics of freeform surfaces. This initial mesh layout improves the detection accuracy in boundary and high-curvature regions while avoiding redundant measurement points in low-variety areas.

[0073] S400: Control the robotic arm to perform thickness measurement of the target coating area based on each preliminary measurement point to obtain a preliminary thickness dataset.

[0074] As an optional implementation of this application, a rotatable thickness gauge probe can be installed on the robotic arm of the collaborative robot, and the robotic arm can be controlled to measure the thickness of the target coating area based on a preliminary measurement point. Specifically, the steps include: S410. Generate a motion path based on each preliminary measurement point, and control the movement of the robotic arm based on the motion path.

[0075] In this embodiment, a motion path is generated based on the coordinates of the initial measurement points using a path planning algorithm according to the shortest path principle. The robotic arm is then controlled to move sequentially to each initial measurement point according to the motion path, avoiding redundant movement and keeping the total measurement time within 30 minutes. The generation of the motion path specifically includes the following steps: S411. Divide the region into multiple connected regions based on the curvature level and spatial distribution of each preliminary measurement point, and prioritize the connected regions to obtain the motion sequence.

[0076] Optionally, based on the curvature properties and spatial distribution of the initial measurement points, each discrete initial measurement point is divided into connected region clusters to reduce long-distance movement across regions. First, using curvature level as the classification standard, initial measurement points corresponding to high, medium, and low curvature regions are assigned to different connected region clusters, ensuring that measurement points in regions with the same curvature level are preferentially assigned to the same connected region cluster. Then, under spatial continuity constraints, connected region sub-clusters are divided, grouping spatially continuous points into the same connected region sub-cluster to avoid forcibly merging physically separate regions, such as two independent folds in a skinned part. Finally, multiple connected regions are divided based on each connected region cluster and connected region sub-cluster.

[0077] Meanwhile, due to the high density of measurement points in high-curvature regions, the robotic arm makes more frequent attitude adjustments in these areas. Therefore, priority can be given to reducing attitude switching losses caused by subsequent cross-regional movements. Thus, the access priority of connected regions is set sequentially according to the movement order of high curvature, medium curvature, and low curvature to obtain the movement sequence, thereby reducing the frequent angle adjustments caused by the robotic arm's cross-regional movements and improving the overall smoothness of the measurement operation.

[0078] S412. Traverse the preliminary measurement points of each connected region, generate multiple local paths, and set transition paths in each connected region.

[0079] In this embodiment, based on the row and column structure of a rectangular Chebyshev mesh, a serpentine scanning traversal strategy is adopted: odd-numbered rows are measured sequentially along the column direction, and even-numbered rows are measured sequentially along the column direction, thereby effectively reducing large-scale idle travel between adjacent rows and improving measurement efficiency. Starting from the starting node of each connected region, after completing the traversal of the current row point by point, the process moves along the row direction to the next row and then reverses, repeating this cycle until all nodes of the connected region are covered, thereby generating local paths for each connected region.

[0080] Simultaneously, transition paths are set at the boundaries of different connected regions to adapt to the robot arm's motion constraints and reduce idle travel. Furthermore, by limiting the angle between the line connecting adjacent measurement points and the robot arm's current posture to ≤30°, the path direction can be optimized to match the posture continuity requirements of the robot arm's end effector, significantly reducing joint adjustment amplitude and dynamic errors during measurement, and improving motion stability and measurement repeatability.

[0081] S413. Connect the local paths sequentially according to the motion sequence, and use smooth curves to connect the transition paths to obtain the motion path.

[0082] In this embodiment, local paths are sequentially connected according to the movement sequence of connected regions. Optimal transition points are selected in adjacent connected regions, taking into account factors such as shortest distance, minimal attitude change, and path continuity to reduce idle travel and abrupt attitude changes caused by cross-region movement. After selecting the optimal transition points, a 5th-order B-spline curve is used to fit and connect the transition paths to obtain the motion path. Because the 5th-order B-spline curve has high-order continuity, it can simultaneously ensure smooth transitions of position, velocity, and acceleration at the starting point, intermediate point, and ending point, thereby eliminating mechanical vibration and impact and improving motion stability and measurement accuracy.

[0083] For example, Figure 5 This is a schematic diagram of the robotic arm's motion path in a specific coated area of ​​a curved workpiece in an embodiment of this application, such as... Figure 5 As shown in the figure, the control robot arm sequentially performs thickness measurements at each preliminary measurement point according to the motion path shown in the figure.

[0084] S420. When the robotic arm moves to each preliminary measurement point, adjust the posture of the robotic arm according to the point cloud data.

[0085] In this embodiment, when the robotic arm reaches each preliminary measurement point, the surface normal vector information contained in the target point cloud data can be used to drive the joints of the robotic arm in real time to adjust the spatial attitude of the thickness gauge probe, ensuring that the probe axis is precisely aligned with the local normal direction of the target coating area surface. Through normal vector-guided attitude adaptive control, the tilt angle deviation of the thickness gauge probe can be controlled within a very small range, effectively suppressing thickness reading drift caused by changes in the incident angle, and reducing the thickness measurement error caused by tilting to ±2. μm Within.

[0086] S430. Based on the adjusted posture of the robotic arm, the thickness of the initial measurement points is measured to obtain a preliminary thickness dataset.

[0087] In this embodiment, the thickness gauge probe on the robotic arm, after its posture is adjusted, contacts the coating surface with a contact force controlled within 0.5N to avoid damaging the coating. The coating thickness is measured using the eddy current method. Each initial measurement point is measured three times, and the average value is taken. Finally, a preliminary thickness dataset containing "measurement point coordinates, thickness value, and Gaussian curvature value" is output. For example, the data for a certain measurement point in the preliminary thickness dataset is: measurement point coordinates X =100mm, Y =50mm, Z =5mm, corresponding coating thickness value T =300 μm This corresponds to a Gaussian curvature K = 0.3 mm. -1 .

[0088] This application embodiment generates an optimal motion path through path planning. The robotic arm moves sequentially to each preliminary measurement point according to the shortest path principle to measure the coating thickness, effectively reducing idle travel and redundant movement, and significantly improving measurement efficiency. At the same time, it combines the point cloud normal vector to adjust the probe posture of the thickness gauge in real time, ensuring that the measurement direction is always perpendicular to the curved surface, significantly reducing the risk of human error and contact damage, effectively achieving high-precision and high-efficiency large-area continuous measurement, and improving the consistency of measurement data and the reliability of detection.

[0089] S500: Adjust the grid nodes of the initial grid based on the preliminary thickness dataset to obtain multiple target measurement points.

[0090] As a further implementation of this application, by analyzing the thickness difference between adjacent measurement points, measurement points can be dynamically supplemented, and the grid spacing can be reduced, thereby improving the detection accuracy in areas of thickness gradient change while avoiding redundant measurements in areas of low change. The adjustment of the initial grid nodes specifically includes the following steps: S510. Traverse all grid nodes of the initial grid as target grid nodes, and calculate the absolute thickness difference between the two points based on the thickness data of the target grid node and the corresponding neighboring grid nodes.

[0091] In this embodiment of the application, each grid node in the initial grid is traversed as the target grid node. P i Includes 3D coordinates from the initial mesh ( X i , Y i , Z i (Unit: mm, corresponding coating thickness value) T i ,unit μ m The corresponding Gaussian curvature K i Unit: mm -1 and initial grid spacing d 0. The initial grid spacing is determined by the curvature level of different regions. For example, the initial grid spacing of low curvature regions is 10mm, while the initial grid spacing of high curvature regions is 5mm.

[0092] For each target grid node in the initial grid P i and its neighboring areas Each neighboring grid node within Calculate the absolute thickness difference between two points using the following formula:

[0093] In the above formula, This represents the absolute thickness difference between two points: the target mesh node and its corresponding neighboring mesh nodes. T i This represents the coating thickness value of the target mesh node. T j This represents the coating thickness value of the neighboring grid nodes.

[0094] The absolute thickness difference directly reflects the thickness gradient change between two points. The larger the difference, the worse the coating thickness uniformity in that area, and the higher the possibility of out-of-tolerance risk.

[0095] S520. Calculate the dynamic correction threshold between the two points based on the Gaussian curvature of the target grid node and the neighboring grid nodes.

[0096] In this embodiment, the magnitude and degree of curvature change affect the coating thickness uniformity. Therefore, the spacing between two points is dynamically adjusted based on the Gaussian curvature changes of the target mesh node and neighboring mesh nodes. A dynamic correction threshold is introduced to improve sensitivity. The calculation steps for the dynamic correction threshold are as follows: S521. Calculate the geometric mean curvature based on the Gaussian curvature of the target grid node and the Gaussian curvature of the neighboring grid nodes.

[0097] In this embodiment, the formula for calculating the geometric mean curvature is as follows:

[0098] In the above formula, K avg Represents the geometric mean curvature. K i Represents the target mesh node P i Gaussian curvature, K j Represents neighboring grid nodes P j Gaussian curvature.

[0099] S522. Calculate the relative curvature deviation between two nodes based on the Gaussian curvature of the target grid node, the Gaussian curvature of the neighboring grid nodes, and the geometric mean curvature.

[0100] In this embodiment, the relative curvature deviation between two nodes reflects the degree of drastic change in the curvature of the local surface. The greater the relative curvature deviation between the two nodes, the more drastic the change in surface curvature between the two points. The formula for calculating the relative curvature deviation is as follows:

[0101] In the above formula, Indicates the relative deviation of curvature, with 10 in the denominator. -6This is to avoid K avg Meaningless calculations when =0.

[0102] S523. Determine the exponential decay correction coefficient based on the geometric mean curvature, and determine the linear decay correction coefficient based on the relative deviation of curvature.

[0103] In this embodiment, since the dynamic correction threshold needs to be reduced in high curvature regions to improve sensitivity, an exponential decay function is used to quantify the dynamic correction intensity based on the geometric mean curvature to obtain the exponential decay correction coefficient. The calculation formula for the exponential decay correction coefficient is as follows:

[0104] In the above formula, This represents the exponential decay correction factor. This represents the sensitivity adjustment coefficient, which can be taken as 0.2 to 0.4.

[0105] If the curvature changes drastically between two nodes, the dynamic correction threshold needs to be further reduced to capture abrupt changes in thickness. Therefore, a linear attenuation function is used to quantify the dynamic correction intensity based on the relative curvature deviation, and the linear attenuation correction coefficient is obtained. The formula for calculating the linear attenuation correction coefficient is as follows:

[0106] In the above formula, This represents the linear decay correction factor. m This represents a constant setting value that ensures the dynamic correction threshold after correction is not lower than the reasonable detection lower limit of the thickness gauge probe.

[0107] In this embodiment, when the relative curvature deviation between two nodes is large, further reduction of the linear attenuation correction coefficient may make the dynamic correction threshold too small. Therefore, the actual resolution of the thickness gauge probe for measuring coating thickness is obtained, and a reasonable detection lower limit of the thickness gauge probe is obtained through multiple sets of coating thickness detection experiments. Based on this reasonable detection lower limit and the curvature of different measurement points of each coating, the constant setting value is determined.

[0108] S524. Based on the exponential decay correction coefficient and the linear decay correction coefficient between the two points, and combined with the preset critical difference, obtain the dynamic correction threshold.

[0109] In this embodiment, the formula for calculating the dynamic correction threshold between the target grid node and its corresponding neighboring grid nodes is as follows:

[0110] In the above formula, This indicates a dynamically adjusted threshold. This represents the preset critical difference value, which can be determined comprehensively based on the initial thickness dataset and the actual coating distribution of the curved workpiece.

[0111] S530. If the absolute thickness difference is greater than the dynamic correction threshold, then intermediate measurement points are added based on the coordinates of the target grid node and the neighboring grid nodes.

[0112] In this embodiment, when the absolute thickness difference between the target mesh node and its corresponding neighboring mesh nodes is greater than the dynamic correction threshold, at the target mesh node... P i and neighboring grid nodes P j Add a new measurement point in the neighborhood of the midpoint of the line connecting them. P mid ,in: If the region containing the two points is a low-curvature surface, linear interpolation is used for calculation. P mid The three-dimensional coordinates are determined to ensure that they lie on the spatial line connecting the target mesh node and the corresponding neighboring mesh nodes. The calculation formula is as follows:

[0113] In the above formula, X mid , Y mid and Z mid These represent the new measurement points. P mid The three coordinate values, X i , Y i and Z i Represents the target mesh node P i The three coordinate values, X j , Y j and Z j Represents neighboring grid nodes P j The three coordinate values.

[0114] If the regions where the two points are located are regions of medium curvature and high curvature, surface interpolation can be performed. Based on the surface parameter equations of the constructed surface model, the midpoint of the shortest path between the two points on the surface can be calculated. P mid The three-dimensional coordinates are used to avoid surface deviation caused by linear interpolation.

[0115] As an optional implementation of this application, if subsequent adjustments to the node mesh are needed based on a new thickness dataset according to actual conditions, a preliminary estimate can be given through linear thickness interpolation to guide the priority of subsequent measurements. The calculation formula is as follows:

[0116] In subsequent actual measurements, the measured values ​​can be used. Replacement estimate And recalculate Compared with measured values , Compared with measured values If any difference is greater than the preset critical difference, a new measurement point needs to be added at the midpoint of the corresponding line segment, with a maximum of 2 iterations.

[0117] S540. If the absolute thickness difference is less than or equal to the dynamic correction threshold, then maintain the spacing between the target mesh node and the neighboring mesh nodes.

[0118] In this embodiment, when the absolute thickness difference between the target mesh node and its corresponding neighboring mesh nodes is less than or equal to the dynamic correction threshold, it indicates that the coating thickness uniformity between the two points meets the requirements, there is no significant risk of exceeding the tolerance, and the original initial mesh spacing is maintained. d 0, no new measurement points are needed.

[0119] S550, take all grid nodes of the initial grid and the supplementary intermediate measurement points as target measurement points.

[0120] In this embodiment of the application, all grid nodes of the original initial grid and the supplementary intermediate measurement points are integrated to obtain the target measurement point.

[0121] For example, the grid nodes are adjusted for the initial 2,500 measurement points generated in the initial grid division, and finally about 3,000 target measurement points are generated.

[0122] This application's embodiments utilize a dynamic threshold correction mechanism combined with curvature variations and thickness gradients to adaptively supplement new measurement points on the initial grid. Measurements are densified in high-curvature or abrupt thickness change regions to improve detection accuracy, while sparse measurement points are maintained in uniform areas to avoid redundancy. This effectively balances measurement efficiency and accuracy, enabling focused monitoring of critical areas and enhancing the reliability and intelligence of coating quality assessment.

[0123] S600: Control the robotic arm to perform thickness measurement of the target coating area according to each target measurement point, and obtain the target thickness dataset.

[0124] In this embodiment, referring to steps S410-S430 of controlling the robotic arm to measure the thickness of the target coating area based on preliminary measurement points, a motion path is generated based on the coordinates of the target measurement points using a path planning algorithm according to the shortest path principle. The generation of the motion path is performed as described in steps S411-S413. The robotic arm is then controlled to move sequentially to each target measurement point according to the motion path. The posture of the robotic arm is adjusted based on the point cloud data, and the thickness is measured at each target measurement point to obtain a target thickness dataset.

[0125] As an optional implementation of this application, the thickness measurement can be re-executed based on the actual measurement requirements of the curved workpiece and the adjusted target measurement points to obtain a new thickness dataset. Steps S500-S600 are then repeated based on the new thickness dataset until the thickness difference between each adjacent target measurement point meets the requirements. Finally, the robotic arm is controlled to perform thickness measurement based on the latest target measurement points to obtain the latest target thickness dataset.

[0126] S700: Generate a thickness distribution cloud map based on the target thickness dataset.

[0127] In this embodiment, the target thickness dataset obtained by controlling the robotic arm to measure based on the target measurement points is subjected to three-dimensional visualization processing to generate a thickness distribution cloud map. The generation of the thickness distribution cloud map specifically includes the following steps: S710. Perform data interpolation based on the target thickness dataset to fill in the blank areas between each target measurement point and obtain thickness distribution data.

[0128] As an optional implementation of this application, discrete target measurement points can be filled in the blank areas between target measurement points using Kriging interpolation. Based on the thickness values ​​and spatial distances of adjacent target measurement points, the thickness of the intermediate point is calculated to obtain thickness distribution data of the cloud map. The correlation of Kriging interpolation is quantified using the Variogram function. Based on the spatial distribution and curvature characteristics of the target measurement points, thickness distribution data that both closely matches the actual measured values ​​and reflects the overall trend of coating thickness on curved workpieces is generated.

[0129] S720. Perform color mapping based on thickness distribution data to obtain a thickness distribution cloud map.

[0130] In this embodiment, a color gradient can be set based on the thickness distribution data to generate a cloud map, thereby obtaining a thickness distribution cloud map and visually distinguishing the thickness state of the target coating area of ​​the curved workpiece. Figure 6 This is a schematic diagram of the thickness distribution cloud map of a certain local coating area of ​​a curved workpiece in an embodiment of this application, such as... Figure 6 As shown, the color gradient is set as follows: thickness value d < 185. μm Then it is mapped to red, 185 μm≤Thickness value d≤200 μm The color is then mapped to green, and the thickness value d > 215. μm The color is then mapped to blue, and a thickness distribution cloud map is finally obtained in which the color changes uniformly with the thickness state, which fully reflects the thickness distribution state of the target coating area of ​​the curved workpiece.

[0131] This application's embodiments accurately reconstruct the geometric features of curved workpieces through 3D scanning and surface fitting. Combined with curvature level zoning and Chebyshev mesh generation, the rationality of the measurement point layout is improved. Simultaneously, the spacing between measurement points is dynamically adjusted, and high repeatability measurement is achieved based on automatic positioning and attitude adjustment by a robotic arm. Finally, a visualized thickness cloud map is generated, significantly improving the automation, intelligence, and reliability of large-area coating thickness measurement for curved workpieces.

[0132] Based on the same technical concept as the above embodiments, this application also provides a large-area coating thickness measuring device for curved workpieces. Figure 7 This is a schematic diagram of the structure of the large-area coating thickness measuring device for curved workpieces provided in the embodiments of this application, as shown below. Figure 7 As shown, the large-area coating thickness measuring device 200 for the curved workpiece includes: The data processing module 210 is used to acquire point cloud data of the target coating area of ​​the curved workpiece and calculate the Gaussian curvature of each point based on the point cloud data.

[0133] The model generation module 220 is used to divide the surface corresponding to the target coating area into multiple target surface patches with different curvature levels according to the Gaussian curvature, and generate a surface model based on the target surface patches.

[0134] The mesh generation module 230 is used to generate an initial mesh based on the surface model to obtain multiple preliminary measurement points.

[0135] The preliminary measurement module 240 is used to control the robotic arm to perform thickness measurement of the target coating area based on each preliminary measurement point, and to obtain a preliminary thickness dataset.

[0136] The mesh adjustment module 250 is used to adjust the mesh nodes of the initial mesh according to the preliminary thickness dataset to obtain multiple target measurement points.

[0137] The target measurement module 260 is used to control the robotic arm to perform thickness measurement of the target coating area according to each target measurement point, and obtain the target thickness dataset.

[0138] The cloud map generation module 270 is used to generate a thickness distribution cloud map based on the target thickness dataset.

[0139] The large-area coating thickness measuring device for curved workpieces provided in this application embodiment can achieve high-precision measurement and full-area visualization of the coating thickness of large-area curved workpieces, thereby improving the reliability and intelligence level of coating quality assessment.

[0140] It is understood that the implementation method of the large-area coating thickness measurement method for curved workpieces in the above embodiment is also applicable to this embodiment and can achieve the same technical effect, so it will not be described again here.

[0141] Based on the same concept, this application also provides an electronic device. Figure 8 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application, such as... Figure 8 As shown, the electronic device 300 may include a processor 310, a communication interface 320, a memory 330, and a communication bus 340, wherein the processor 310, the communication interface 320, and the memory 330 communicate with each other via the communication bus 340. The processor 310 can call logical instructions in the memory 330 to execute the steps of the large-area coating thickness measurement method for curved workpieces as described in the above embodiments. For example, this includes: S100: Obtain point cloud data of the target coating area of ​​the curved workpiece, and calculate the Gaussian curvature of each point based on the point cloud data; S200: Divide the surface corresponding to the target coating area into multiple target surface patches with different curvature levels according to the Gaussian curvature, and generate a surface model based on the target surface patches; S300: Based on the surface model, an initial mesh is generated to obtain multiple preliminary measurement points; S400: Control the robotic arm to perform thickness measurement of the target coating area based on each preliminary measurement point to obtain a preliminary thickness dataset; S500: Adjust the mesh nodes of the initial mesh based on the preliminary thickness dataset to obtain multiple target measurement points; S600: Control the robotic arm to perform thickness measurement of the target coating area according to each target measurement point, and obtain the target thickness dataset; S700: Generate a thickness distribution cloud map based on the target thickness dataset.

[0142] The processor 310 can be a central processing unit (CPU). The processor can also 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, or combinations of the above types of chips.

[0143] Furthermore, the logical instructions in the aforementioned memory 330 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part 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 application. 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.

[0144] The memory 330 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created by the processor, etc. Furthermore, the memory may include high-speed random access memory and non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, the memory may optionally include memory remotely located relative to the processor, which can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0145] Based on the same concept, embodiments of this application also provide a computer-readable storage medium storing a computer program containing at least one piece of code executable by a master control device to control the master control device to implement the steps of the large-area coating thickness measurement method for curved workpieces as described in the above embodiments. For example, it includes: S100: Obtain point cloud data of the target coating area of ​​the curved workpiece, and calculate the Gaussian curvature of each point based on the point cloud data; S200: Divide the surface corresponding to the target coating area into multiple target surface patches with different curvature levels according to the Gaussian curvature, and generate a surface model based on the target surface patches; S300: Based on the surface model, an initial mesh is generated to obtain multiple preliminary measurement points; S400: Control the robotic arm to perform thickness measurement of the target coating area based on each preliminary measurement point to obtain a preliminary thickness dataset; S500: Adjust the mesh nodes of the initial mesh based on the preliminary thickness dataset to obtain multiple target measurement points; S600: Control the robotic arm to perform thickness measurement of the target coating area according to each target measurement point, and obtain the target thickness dataset; S700: Generate a thickness distribution cloud map based on the target thickness dataset.

[0146] Based on the same technical concept, this application also provides a computer program, which, when executed by a main control device, is used to implement the above-described method embodiments.

[0147] The computer program may be stored, in whole or in part, on a computer-readable storage medium packaged with the processor, or in part or in whole on a memory not packaged with the processor.

[0148] Based on the same technical concept, this application also provides a processor for implementing the above-described method embodiments. The processor can be a chip.

[0149] In summary, the method and apparatus for measuring the coating thickness of curved workpieces provided in this application achieve accurate geometric modeling of complex curved workpieces through high-precision 3D scanning and surface fitting. Furthermore, Chebyshev mesh generation, combined with curvature grading and normal vector analysis, improves the rationality of the measurement point layout. Simultaneously, a dynamic threshold mechanism based on the Gaussian curvature geometric mean and relative deviation is introduced to adaptively adjust the measurement point spacing, automatically densifying measurement points in areas with significant coating thickness gradient changes, effectively capturing local abrupt changes, and balancing measurement accuracy and efficiency. Finally, by controlling a robotic arm to automatically measure according to an automatically generated motion path, the thickness measurement data is obtained to generate a full-coverage thickness distribution cloud map, achieving high-precision measurement and full-area visualization of the coating thickness of large-area curved workpieces.

[0150] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0151] The embodiments described above are merely examples of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these modifications and improvements all fall within the protection scope of this application.

[0152] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application 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. Such 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 this application.

Claims

1. A method for large area coating thickness measurement of a curved workpiece, characterized in that, The method comprises: acquiring point cloud data of a target coating area of a curved workpiece, and calculating Gaussian curvature of each point according to the point cloud data; dividing a corresponding curved surface of the target coating area into a plurality of target curved surface patches of different curvature levels according to the Gaussian curvature, and generating a curved surface model based on the target curved surface patches; dividing an initial grid based on the curved surface model to obtain a plurality of preliminary measurement points; controlling a mechanical arm to perform thickness measurement of the target coating area according to each preliminary measurement point to obtain a preliminary thickness data set; adjusting grid nodes of the initial grid according to the preliminary thickness data set to obtain a plurality of target measurement points; controlling the mechanical arm to perform thickness measurement of the target coating area according to each target measurement point to obtain a target thickness data set; generating a thickness distribution cloud chart according to the target thickness data set.

2. The method of large area coating thickness measurement of a curved workpiece of claim 1, wherein, The method comprises: filtering background noise of the point cloud data by using a density-based clustering algorithm to obtain initial point cloud data; filtering high-frequency noise of the initial point cloud data by using a moving least squares method to obtain target point cloud data; calculating a curvature tensor of each point by locally fitting a curved surface according to the target point cloud data to obtain the Gaussian curvature.

3. The method of claim 2, wherein The method comprises: determining neighbor points corresponding to a first target point in the target point cloud data by spatial neighbor searching; calculating overall curvature similarity according to the Gaussian curvature of the first target point and the Gaussian curvature of the neighbor points; if the overall curvature similarity is less than or equal to a similarity threshold, dividing the first target point and the neighbor points into the same curved surface patch, otherwise, selecting a second target point to determine neighbor points to redivide the curved surface patch until all points of the target point cloud data are divided to obtain a plurality of initial curved surface patches; dividing each initial curved surface patch into target curved surface patches of different curvature levels by comparing the Gaussian curvature of all points on each initial curved surface patch with a hierarchical curvature threshold; merging each target curved surface patch to obtain the curved surface model.

4. The method of large area coating thickness measurement of a curved workpiece of claim 3, wherein, The method comprises: determining a curved surface boundary of a free curved surface according to the Gaussian curvature and a normal vector of each point on the curved surface model; generating a Chebyshev grid of the free curved surface based on the curved surface boundary, the grid spacing of the Chebyshev grid being different in regions corresponding to the target curved surface patches of different curvature levels to obtain the initial grid; taking each grid node of the initial grid as the preliminary measurement point.

5. The method of large area coating thickness measurement of a curved workpiece of claim 1, wherein, The method comprises: generating a motion path according to each preliminary measurement point, and controlling the mechanical arm to move according to the motion path; adjusting a posture of the mechanical arm according to the point cloud data when the mechanical arm moves to each preliminary measurement point; Based on the adjusted posture of the mechanical arm, thickness measurement is performed on the preliminary measurement points to obtain the preliminary thickness data set.

6. The method of large area coating thickness measurement of a curved workpiece of claim 5, wherein, A motion path is generated according to each of the preliminary measurement points, including: A plurality of connected regions are divided according to the curvature level and spatial distribution of each of the preliminary measurement points, and the connected regions are prioritized to obtain a motion sequence; The preliminary measurement points of each of the connected regions are traversed to generate a plurality of local paths, and a transition path is set in each of the connected regions; The local paths are connected in sequence according to the motion sequence, and the transition paths are connected using a smooth curve to obtain the motion path.

7. The method of claim 5, wherein According to the preliminary thickness data set, the grid nodes of the initial grid are adjusted to obtain a plurality of target measurement points, including: All grid nodes of the initial grid are traversed as target grid nodes, and the absolute thickness difference between two points is calculated according to the thickness data of the target grid node and the corresponding neighborhood grid node; According to the Gaussian curvature of the target grid node and the neighborhood grid node, a dynamic correction threshold value between the two points is calculated; If the absolute thickness difference is greater than the dynamic correction threshold value, intermediate measurement points are supplemented according to the coordinates of the target grid node and the neighborhood grid node; If the absolute thickness difference is less than or equal to the dynamic correction threshold value, the spacing between the target grid node and the neighborhood grid node is maintained; All grid nodes of the initial grid and the supplemented intermediate measurement points are taken as the target measurement points.

8. The method of claim 7, wherein, According to the Gaussian curvature of the target grid node and the neighborhood grid node, a dynamic correction threshold value between two points is calculated, including: The geometric mean curvature is calculated according to the Gaussian curvature of the target grid node and the Gaussian curvature of the neighborhood grid node; The curvature relative deviation of the two nodes is calculated according to the Gaussian curvature of the target grid node, the Gaussian curvature of the neighborhood grid node, and the geometric mean curvature; The exponential decay correction coefficient is determined according to the geometric mean curvature, and the linear decay correction coefficient is determined according to the curvature relative deviation; According to the exponential decay correction coefficient and the linear decay correction coefficient between two points, the dynamic correction threshold value is obtained in combination with a preset critical difference value.

9. The method of claim 1, wherein A thickness distribution cloud chart is generated according to the target thickness data set, including: Data interpolation is performed according to the target thickness data set to fill in the blank areas between each of the target measurement points to obtain thickness distribution data; Color mapping is performed according to the thickness distribution data to obtain the thickness distribution cloud chart.

10. A large area coating thickness measurement apparatus for a curved workpiece, characterized by, The device includes: A data processing module for obtaining point cloud data of a target coating region of a curved workpiece, and calculating the Gaussian curvature of each point according to the point cloud data; A model generation module for dividing the target coating region corresponding curved surface into a plurality of target surface pieces of different curvature levels according to the Gaussian curvature, and generating a curved surface model based on the target surface pieces; A grid division module for dividing an initial grid based on the curved surface model to obtain a plurality of preliminary measurement points; A preliminary measurement module for controlling a mechanical arm to perform thickness measurement of the target coating region according to each of the preliminary measurement points to obtain a preliminary thickness data set; a grid adjustment module configured to adjust grid nodes of the initial grid according to the preliminary thickness dataset to obtain a plurality of target measurement points; a target measurement module configured to control the robot arm to perform thickness measurement of the target coating region according to each of the target measurement points to obtain a target thickness dataset; and a cloud map generation module configured to generate a thickness distribution cloud map according to the target thickness dataset.

Images