A laser processing planning method and apparatus for convex curved surface workpieces based on mesh distortion

Abstract

This invention discloses a laser processing planning method and apparatus for convex curved workpieces based on mesh distortion. It simultaneously captures distorted mesh images of the laser mesh on the convex curved workpiece using four industrial cameras (positive X, negative X, positive Y, and negative Y). Mesh node information is extracted using an edge detection algorithm, and the local gradient of the workpiece surface is quickly calculated. The normal vector is calculated using the gradient vector cross product method and automatically calculated directly using a Python program, achieving efficient prediction of the surface normal vector and reducing the time and cost of traditional point-based cloud computing methods. By setting a threshold δ, the deviation of the mesh normal vector is sorted and iteratively calculated, automatically identifying areas with small local curvature changes and merging the meshes to form segmented processing areas. This adaptive segmentation method reduces the complexity of the processing path, improves processing efficiency, and ensures that the laser incident angle is aligned with the average normal vector of the processing area, optimizing the laser processing quality.

Description

Technical Field

[0001] This invention relates to the field of laser processing technology, and specifically to a method and apparatus for laser processing planning of convex curved workpieces based on mesh distortion. Background Technology

[0002] Perovskite solar cells, as a third-generation solar cell technology, possess advantages such as high efficiency, low cost, low light performance, and environmental friendliness, demonstrating outstanding application potential in the photovoltaic power generation field. The scribing of perovskite solar cells is one of the key steps in the manufacturing process. It utilizes laser equipment to precisely scribble lines on the perovskite thin film to achieve series connection of the cells and optimize performance.

[0003] The curved surface characteristics of perovskite solar cells typically stem from the flexibility of their substrate material or pre-formed curved structures (such as PET, metal foil, etc.). If the planar material is scribed first, subsequent bending may lead to the following problems: 1. The scribed grooves may deform or break during bending, damaging the electrical isolation structure and causing series connection failure between sub-cells (e.g., misalignment of P2 connection); 2. Different film layers (e.g., FTO, perovskite, metal electrodes) may delaminate or crack due to bending stress, affecting cell stability. Traditional scribing equipment and methods cannot adapt to changes in curved surfaces, making it difficult to guarantee scribing accuracy and straightness. Scibing accuracy and straightness are crucial to cell performance; inaccurate scribing may lead to uneven current distribution within the cell, thus affecting the photoelectric conversion efficiency and stability. Therefore, direct scribing on curved surfaces requires laser equipment with three-dimensional dynamic focusing and curved surface path planning capabilities to ensure scribing accuracy and straightness.

[0004] In many engineering cases, the actual shape of curved perovskite solar cells does not match the CAD model. Therefore, it is necessary to obtain the shape and normal vector of the workpiece surface through reverse engineering and automatically plan the non-programming scribing machining path based on the reverse model.

[0005] For the following five reasons, block machining planning based on surface normal vectors has become a necessary prerequisite for automatic path planning without programming: 1) The travel of the motion mechanism is limited, making it difficult to cover large-sized curved surfaces in one go; 2) The curvature of different regions of the curved workpiece may vary greatly, and uniform machining parameters cannot be adapted globally; 3) When machining curved workpieces continuously, mechanical motion errors gradually amplify; 4) Global path planning for curved workpieces requires processing a large amount of data, resulting in a heavy computational burden, which can be accelerated by parallel computing; 5) Failure of a single global path planning may lead to the scrapping of the entire curved workpiece. However, previous methods for obtaining surface normal vectors are inefficient and require a lot of computational resources, increasing the cost of obtaining normal vectors. For example, the implicit function partial derivative calculation method requires accurate analysis of the surface equation, and the computational complexity is high for high-order NURBS or B-spline surfaces, making it difficult to meet the real-time requirements of block machining planning; the covariance matrix analysis method based on PCA requires processing a large amount of nearest neighbor data, resulting in a large computational load, especially with a significant decrease in efficiency in large-sized curved surfaces; the main purpose of traditional surface reconstruction technology is to obtain point cloud positions, which is time-consuming and costly. Summary of the Invention

[0006] To address the problems existing in the background art, the present invention provides a laser processing planning method and apparatus for convex curved surface workpieces based on mesh distortion, which achieves efficient and low-cost rapid calculation of surface normal vectors and adaptive block processing planning through mesh distortion analysis.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] This invention provides a laser processing planning method for convex curved surface workpieces based on mesh distortion, comprising:

[0009] The convex curved workpiece is placed below the laser grid projector, and the laser grid is projected onto the surface of the convex curved workpiece.

[0010] Using the beam emission direction of the laser grid projector as the Z-axis direction of the spatial coordinate system, four cameras are set in the positive and negative X-axis directions and the positive and negative Y-axis directions, respectively, and the optical axes of the four cameras are perpendicular to the Z-axis direction. Four distorted grid images are captured by the four cameras respectively.

[0011] Extract the mesh node information of the workpiece surface from the captured distorted mesh image, record the grid point coordinates of each mesh, calculate the local gradient of the surface, and calculate the normal vector of the surface using the gradient cross product method.

[0012] By setting a normal vector deviation threshold δ, the deviations of the normal vectors of adjacent meshes are sorted and iteratively calculated, automatically identifying areas with small local curvature changes, and merging the meshes to generate block processing areas.

[0013] Sampling is performed along the grid lines in the curved surface to generate laser processing path points, which are then combined with the average normal vector after segmentation to generate the laser processing path.

[0014] In the above technical solution: the orthogonal arrangement of the four cameras ensures complete decoupling of data in the two directions during gradient calculation, avoiding cross-interference introduced by the tilted viewpoint. The distorted mesh captured by the cameras directly corresponds to the local gradient of the surface along the X and Y directions. In addition, the optical axes of all cameras are perpendicular to the Z-axis of the laser projection, so that image distortion is only caused by changes in the height of the workpiece surface, eliminating the need for additional compensation for perspective distortion caused by camera tilt and improving the accuracy of distortion capture.

[0015] Furthermore, the captured distorted mesh image is denoised and lens distortion compensated, and the mesh node information of the workpiece surface is extracted through an edge detection algorithm.

[0016] Furthermore, the local gradient of the surface The slope represents the rate of change of the height of the surface along the X and Y directions at grid points, and is calculated using the following formula:

[0017] ;

[0018] ;

[0019] ;

[0020] in, Indicates the surface at point The slope along the X direction (positive value indicates an increase, negative value indicates a decrease). Indicates the surface at point The slope along the Y direction (positive value indicates an increase, negative value indicates a decrease). The coordinates are , It is the height at which the lasers intersect on the curved surface.

[0021] Furthermore, the formula for calculating the normal vector N of the surface is as follows:

[0022] ;

[0023] ;

[0024] ;

[0025] ;

[0026] ;

[0027] Then normalize to get

[0028] .

[0029] Furthermore, the normal vector deviations between all adjacent meshes are sorted, and the adjacent mesh with the smallest current normal vector deviation is determined iteratively by setting a normal vector deviation threshold δ. If it does not exceed δ, the mesh is merged into processing blocks α, β, γ, ..., and the coordinates and normal vectors of all grid points in processing blocks α, β, γ, ... are stored respectively.

[0030] Furthermore, the generated laser processing path is smoothed between blocks by using a B-spline curve fitting algorithm to smooth changes in processing posture (changes in normal vector).

[0031] Further, adjust the height of the laser grid projector until the laser grid emitted by the laser grid projector completely covers the surface of the convex curved workpiece.

[0032] Furthermore, the two cameras located in the positive and negative directions of the X-axis are on the same central axis, and the two cameras located in the positive and negative directions of the Y-axis are on the same central axis. The height of the four cameras is adjustable and the height is consistent.

[0033] The present invention also provides a laser processing planning device for convex curved workpieces based on mesh distortion, for implementing the above-mentioned laser processing planning method for convex curved workpieces, including a laser mesh projector, a cross, columns, a working platform and four cameras. A telescopic component movable along the Z-axis is installed in the middle of the cross, and the laser mesh projector is installed at the driving end of the telescopic component. The working platform is located directly below the cross, and four vertical columns are respectively provided on the top of the working platform. The tops of the four columns are respectively connected to the four ends of the cross, and a camera is installed on each column.

[0034] Furthermore, the camera is mounted on the column at an adjustable height, the work platform has four adjustable legs at the bottom, and the column and the cross-shaped end bracket are equipped with leveling components.

[0035] The adjustable laser grid projector in this device allows the projection grid to completely cover the workpiece. The projection range and camera angle can be adjusted according to the size and shape of the convex curved workpiece, and the camera height can be adjusted to adapt to convex curved workpieces of different specifications, thus improving the versatility and adaptability of the system.

[0036] Compared with the prior art, the present invention has the following beneficial effects:

[0037] (1) The laser mesh on the convex curved workpiece is captured simultaneously by four industrial cameras (X positive, X negative, Y positive and Y negative), and the mesh node information is extracted by combining the edge detection algorithm. The local gradient of the workpiece surface is quickly calculated, the gradient vector cross product method is used to calculate the normal vector, and the calculation is automatically performed by the Python program to achieve efficient prediction of the surface normal vector, reducing the time and cost of the traditional acquisition point cloud computing normal vector method.

[0038] (2) By setting a threshold δ, the deviation of the mesh normal vector is sorted and iteratively calculated, the region with small local curvature change is automatically identified, and the mesh is merged to form processing blocks α, β, γ... This adaptive block method reduces the complexity of the processing path, improves the processing efficiency, and at the same time ensures that the laser incident angle is aligned with the average normal vector of the processing area, thus optimizing the laser processing quality. Attached Figure Description

[0039] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0040] Figure 1 This is a schematic diagram of the laser processing planning device for convex curved workpieces in Example 1;

[0041] Figure 2 This is a laser mesh image of the convex curved workpiece surface after distortion in Example 2;

[0042] Figure 3 In Example 2, the laser processing path is generated based on the grid lines and normal vectors;

[0043] Figure 4 The images are the original pictures captured by the camera in Example 2;

[0044] Figure 5 In Example 2, a Gaussian filtering algorithm is used to denoise the original images captured by the camera.

[0045] Figure 6 In Example 2, Harris corner detection was used to obtain grid point information;

[0046] Figure 7 This is a diagram illustrating the traditional point cloud computing normal vector method.

[0047] Figure 8 This is a flowchart illustrating step (8) in Example 2;

[0048] The specific reference numerals in the attached figures are as follows:

[0049] 1. Laser grid projector; 2. Telescopic assembly; 3. Cross; 4. Leveling assembly; 5. Column; 6. Waist-shaped hole; 7. Camera; 8. Work platform; 9. Hydraulic leveling support rod. Detailed Implementation

[0050] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0051] Example 1

[0052] This embodiment provides a laser processing planning device for convex curved surface workpieces based on mesh distortion, such as... Figure 1 As shown, the system includes a laser grid projector 1, a cross 3, columns 5, a work platform 8, and four cameras 7 (using 20-megapixel monochrome industrial cameras). A telescopic assembly 2, movable along the Z-axis, is installed in the center of the cross 3. The laser grid projector 1 is mounted on the drive end of the telescopic assembly 2. The telescopic assembly 2 can be either electrically adjustable (e.g., an electric push rod) or manually adjustable (screw + scale) using a lead screw. The work platform 8 is located directly below the cross 3. Four vertical columns 5 are mounted on the top of the work platform 8. The tops of the four columns 5 are connected to the four ends of the cross 3 via leveling components 4. Each column 5 is equipped with a height-adjustable camera 7. Specifically, vertically oriented oblong holes 6 can be made in the columns 5, and the cameras 7 can be fixed in the oblong holes 6 with bolts, allowing for height adjustment. Alternatively, a linear module + stepper motor structure can be used to automatically adjust the height of the cameras 7, enabling automatic focusing and angle adjustment. Multiple hydraulic leveling support rods 9 are located at the bottom of the work platform 8 to adjust its horizontal angle.

[0053] The hydraulic system provides significant force and precise position control. The hydraulic leveling support rod 9 allows independent height adjustment of each support point to level the work platform 8. Workpieces with varying surface curvatures may require adjustments to the work platform 8 at different positions to ensure accurate projection and capture. When the workpiece curvature varies significantly, the angle of the work platform 8 needs adjustment to better cover the curved surface with the laser projection grid, preventing projection distortion or inaccurate capture by the camera 7 due to platform 8 tilt. For large workpieces, the camera 7 requires different mounting heights to capture the distorted grid across the entire surface. The sliding rail design of the waist-shaped hole 6 allows for flexible position adjustment of the camera 7, enabling vertical movement without disassembly, adapting to convex curved workpieces of different sizes and specifications, thus improving the system's versatility and adaptability.

[0054] Example 2

[0055] This embodiment provides a laser processing planning method for convex curved surface workpieces based on mesh distortion. It is implemented using the apparatus in Embodiment 1 and specifically includes the following steps:

[0056] (1) such as Figure 2 As shown, the length of the convex curved workpiece is 950mm and the height is 370mm. The convex curved workpiece is placed under the laser grid projector 1 with a wavelength of 650nm. A laser grid pattern is projected onto the surface of the convex curved workpiece with a grid spacing of 100mm. The telescopic component 2 is adjusted until the laser grid emitted by the laser grid projector 1 completely covers the surface of the convex curved workpiece.

[0057] (2) Using the beam emission direction of the laser grid projector 1 as the Z-axis direction of the spatial coordinate system, four cameras 7 are set in the positive and negative directions of the X-axis and the positive and negative directions of the Y-axis respectively to establish the workpiece coordinate system. The optical axes of the four cameras 7 are all perpendicular to the Z-axis direction. The two cameras 7 located in the positive and negative directions of the X-axis are on the same central axis, and the two cameras 7 located in the positive and negative directions of the Y-axis are on the same central axis. The height of the four cameras 7 is adjustable and the height is consistent. Four distorted grid images are captured by the four cameras 7 respectively.

[0058] (3) Denoise and compensate for lens distortion in the captured distorted mesh image, and extract mesh node information on the workpiece surface using the Harris corner detection algorithm.

[0059] When capturing images using a 20-megapixel monochrome industrial camera, sensor noise, ambient light interference, or electronic noise may cause random noise in the image. To remove this noise, Gaussian filtering is used to convolve the image, suppressing high-frequency noise and preserving edge information. This method improves the accuracy of subsequent algorithms, reduces noise interference, and avoids misidentifying noise as intersections or edges in subsequent edge detection. It also improves the efficiency of subsequent algorithms, resulting in a cleaner preprocessed image and reducing the complexity of subsequent algorithms.

[0060] The Harris corner detection algorithm employs non-maximum suppression, retaining only pixels with the largest gradient magnitude in the gradient direction, thus refining edges to a single pixel width. It uses dual-threshold detection, setting both a high and a low threshold to preserve strong edges, while weak edges (between the two thresholds) are only retained when connecting strong edges. This algorithm has low time complexity, improving the efficiency of grid point extraction.

[0061] (4) Record the coordinates of each grid point in the positive X-axis, negative X-axis, positive Y-axis, and negative Y-axis directions. These coordinates can be expressed as follows: ,in It is the height at which the lasers intersect on the curved surface.

[0062] (5) Local gradient of the surface The slope represents the rate of change of the height of the surface along the X and Y directions at grid points, and is calculated using the following formula:

[0063] ;

[0064] ;

[0065] ;

[0066] in, Indicates the surface at point The slope along the X direction (positive value indicates an increase, negative value indicates a decrease). Indicates the surface at point The slope along the Y direction (positive values ​​indicate an increase, negative values ​​indicate a decrease).

[0067] (6) The normal vector N of the surface can be calculated by the cross product of the gradient vectors:

[0068] ;

[0069] ;

[0070] ;

[0071] ;

[0072] .

[0073] (7) To obtain the unit normal vector, normalization is required:

[0074] .

[0075] (8) Sort the normal vector deviations between all adjacent meshes. Iterate through the adjacent meshes with the smallest current normal vector deviation by setting a normal vector deviation threshold δ. If it does not exceed δ, merge the meshes into processing blocks α, β, γ... and store the coordinates and normal vectors of all grid points in processing blocks α, β, γ... respectively. The specific process of this step is as follows: Figure 8 .

[0076] In steps (4) to (8) above, the grid node information is input into the calculation program written in Python to obtain the normal vector, and the grid is merged through the region growth algorithm.

[0077] The following is a portion of the Python code:

[0078] def region_growing_clustering(mesh_faces, face_normals, threshold_angle_deg):

[0079] num_faces = len(mesh_faces)

[0080] threshold_cos=np.cos(np.radians(threshold_angle_deg)) # Convert the angle to cosine value

[0081] #Initialize unvisited facets

[0082] unvisited=set(range(num_faces))

[0083] clusters=[]

[0084] face_adjacency=build_face_adjacency(mesh_faces) # Face adjacency relationships need to be built beforehand.

[0085] while unvisited:

[0086] #Randomly select a seed face

[0087] seed = unvisited.pop()

[0088] cluster=[seed]

[0089] queue = deque([seed])

[0090] while queue:

[0091] current_face = queue.popleft()

[0092] # Traverse the adjacent faces of the current face

[0093] for neighbor in face_adjacency[current_face]:

[0094] if the neighbor is unvisited:

[0095] #Calculate the cosine of the angle between the normal vectors

[0096] cos_angle=np.dot(face_normals[current_face], face_normals[neighbor])

[0097] # If the similarity meets the threshold, then merge.

[0098] if cos_angle>=threshold_cos:

[0099] unvisited.remove(neighbor)

[0100] cluster.append(neighbor)

[0101] queue.append(neighbor)

[0102] clusters.append(cluster)

[0103] return clusters.

[0104] (9) Uniformly sample the grid lines in the curved surface to generate discrete laser processing path points, and combine them with the average normal vector after block division to generate the laser processing path, such as... Figure 3 As shown, the changes in processing posture are smoothed between blocks by using a B-spline curve fitting algorithm. First, the chord length is parameterized, and then the accumulated chord length method is used to generate uniform node vectors. After substituting these vectors into the processing point set to form an overdetermined system of equations, the control points are solved.

[0105] ;

[0106] ;

[0107] ;

[0108] ;

[0109] Where Q is the type-value matrix (n×3), and N is the basis function matrix (n×m, element N). i,p (u k P is the control point matrix (m×3).

[0110] The B-spline curve fitting algorithm can effectively smooth the processing path curve and avoid laser head jitter.

[0111] To verify the advantages of the method of this invention in actual processing, a comparative analysis was conducted with traditional methods in terms of normal vector calculation efficiency. This invention only requires the collection of approximately 75 mesh nodes for calculation (e.g., Figures 4-6 As shown), traditional dense point cloud methods require collecting approximately 12,000 point clouds (e.g. Figure 7 As shown in the figure, the local neighborhood is fitted using the PCA (Principal Component Analysis) method, which is time-consuming and has high data redundancy. In contrast, this invention uses laser grid projection and an industrial camera 7 to jointly acquire distorted images, extracts key corner points through image processing to construct a sparse grid, significantly reducing the number of nodes, and directly solves the normal vector based on the grid topology, avoiding neighborhood search and covariance matrix calculation for each point.

[0112] Through experimental comparison, under the same hardware environment (Intel i7 CPU, single-threaded execution), the traditional point cloud method takes about 6.8 seconds to calculate 12,000 point clouds and synthesize the normal vectors of triangular facets within the region, while the present invention only takes 2.3 seconds under sparse nodes, improving the computational efficiency by about 66.2% and significantly shortening the preprocessing time, which is especially suitable for online processing planning needs.

[0113] The laser processing planning method for convex curved surface workpieces based on mesh distortion of the present invention has the following advantages:

[0114] (1) A more efficient way to calculate normal vectors

[0115] The existing technology disclosed in CN115761137A employs a fusion method of point cloud data and normal vector regression. While it can generate high-precision surfaces, its computational efficiency is low due to the involvement of neural network calculations and images from multiple lighting directions, making it unsuitable for time-sensitive scenarios. This invention, however, uses four industrial cameras 7 to simultaneously acquire distorted mesh images and combines edge detection to extract mesh nodes, directly calculating the surface normal vector. This eliminates the need for point cloud data acquisition and neural network regression, thus improving the efficiency of normal vector calculation.

[0116] (2) Low computational load and low hardware requirements reduce usage costs.

[0117] The existing technology with publication number CN115761137A uses point cloud data fusion neural networks for surface reconstruction, which has high computational complexity, requires large memory and high computing power, and is not suitable for resource-constrained environments. CN201410647810.6 relies on binocular vision matching, involving precise camera calibration and stereo matching calculations, which have high computational resource requirements. This invention uses only four industrial cameras and an edge detection algorithm to extract mesh nodes, eliminating the need for large-scale point cloud data storage and reducing memory requirements. It employs simple mathematical operations of gradient calculation and normal vector calculation, avoiding deep learning and complex data fusion, resulting in low computational load and the ability to run on ordinary industrial computers or embedded devices.

[0118] (3) Optimize laser processing efficiency through adaptive mesh merging

[0119] The prior art disclosed in CN115761137A mainly focuses on surface reconstruction and does not use normal vectors for processing efficiency optimization. The prior art disclosed in CN104408772A only achieves mesh projection-assisted 3D reconstruction, also without optimizing processing efficiency. This invention proposes normal vector deviation sorting and iterative calculation, adaptively merging meshes in areas with small local curvature changes, and generating processing paths through mesh line sampling. This achieves integration from measurement to processing planning, reducing complex data processing steps. Furthermore, by reducing frequent attitude adjustments of the processing head, the complexity of the processing path is reduced, thus improving processing efficiency.

[0120] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A laser processing planning method for convex curved surface workpieces based on mesh distortion, characterized in that, include: The convex curved workpiece is placed below the laser grid projector, and the laser grid is projected onto the surface of the convex curved workpiece. Using the beam emission direction of the laser grid projector as the Z-axis direction of the spatial coordinate system, four cameras are set in the positive and negative X-axis directions and the positive and negative Y-axis directions, respectively, and the optical axes of the four cameras are perpendicular to the Z-axis direction. The height of the four cameras is adjustable and the height is consistent. Four distorted grid images are captured by the four cameras respectively. Extract the mesh node information of the workpiece surface from the captured distorted mesh image, record the grid point coordinates of each mesh, calculate the local gradient of the surface, and calculate the normal vector of the surface using the gradient cross product method. The local gradient of a surface represents the rate of change of the surface's height along the X and Y directions at grid points; The normal vector deviations between all adjacent meshes are sorted. The adjacent mesh with the smallest normal vector deviation is determined iteratively by setting a normal vector deviation threshold δ. If it does not exceed δ, the meshes are merged to generate block processing regions. The coordinates of all grid points and normal vectors of each block processing region are stored respectively. Sampling is performed along the grid lines in the curved surface to generate laser processing path points. The average normal vector after block division is combined to generate the laser processing path. The generated laser processing path is then smoothed by a B-spline curve fitting algorithm between blocks to reduce changes in processing posture.

2. The laser processing planning method for convex curved surface workpieces based on mesh distortion according to claim 1, characterized in that, The captured distorted mesh image is denoised and lens distortion compensated, and the mesh node information on the workpiece surface is extracted using an edge detection algorithm.

3. The laser processing planning method for convex curved surface workpieces based on mesh distortion according to claim 2, characterized in that, Local gradient of a surface The formula for calculating the rate of change of the height of the surface along the X and Y directions at grid points is as follows: in, This indicates that the surface is at point P. i,j The slope along the X direction at that point. This indicates that the surface is at point P. i,j The slope along the Y direction at point P i,j The coordinates are (x i,j ,y i,j ,z i,j Z i,j It is the height at which the lasers intersect on the curved surface.

4. The laser processing planning method for convex curved surface workpieces based on mesh distortion according to claim 3, characterized in that, The formula for calculating the normal vector N of the surface is as follows: N=T x ×T y ; Then normalize to get 5. The laser processing planning method for convex curved surface workpieces based on mesh distortion according to claim 1, characterized in that, Adjust the height of the laser grid projector until the laser grid emitted by the laser grid projector completely covers the surface of the convex curved workpiece.

6. The laser processing planning method for convex curved surface workpieces based on mesh distortion according to claim 1, characterized in that, Two cameras located in the positive and negative directions of the X-axis are on the same central axis, and two cameras located in the positive and negative directions of the Y-axis are on the same central axis. The height of all four cameras is adjustable and they maintain a consistent height.

7. A laser processing planning device for convex curved surface workpieces based on mesh distortion, characterized in that, The method for planning laser processing of convex curved workpieces as described in any one of claims 1 to 6 includes a laser grid projector, a cross, columns, a work platform, and four cameras. A telescopic component movable along the Z-axis is installed in the middle of the cross, and the laser grid projector is installed at the drive end of the telescopic component. The work platform is located directly below the cross, and four vertical columns are respectively provided on the top of the work platform. The tops of the four columns are respectively connected to the four ends of the cross, and a camera is installed on each column.

8. The laser processing planning device for convex curved surface workpieces based on mesh distortion according to claim 7, characterized in that, The camera is mounted on the column at an adjustable height. The bottom of the work platform is provided with four adjustable legs. The column and the end bracket of the cross are provided with leveling components.

Images