Computer vision-based planar part laser processing method, medium, and apparatus
By using computer vision and 3D Gaussian registration algorithms, automated registration of planar parts and templates is achieved, solving the problems of poor accuracy and low efficiency in traditional laser processing. This improves the automation and accuracy of laser processing and makes it suitable for a wider variety of parts.
Patent Information
- Application Number
- CN202610509958.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-04-17
- Publication Date
- 2026-08-25
AI Technical Summary
Traditional laser processing methods for planar parts suffer from problems such as poor accuracy and efficiency in manual positioning, limited compatibility with a small variety of parts, and poor generalization, making it difficult to meet the needs of efficient and automated processing.
A computer vision-based approach is used to acquire point clouds of planar parts and templates through non-contact measurement. A three-dimensional Gaussian registration algorithm is used to achieve automated registration of the parts and templates, and laser processing is performed based on the pose transformation relationship.
It has achieved automation and high precision in laser processing of planar parts, improved work efficiency, and can be adapted to more types of parts.
Smart Images
Figure CN122625849A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of laser processing technology for planar parts, and in particular to a method, medium, and equipment for laser processing planar parts based on computer vision. Background Technology
[0002] There are many ways to process parts. In this invention, laser processing is used to refer to general processing fields to illustrate specific application scenarios. Laser processing methods include laser cleaning, laser welding, and laser cutting. Planar parts refer to sheet metal parts with different shapes, characterized by a flat upper surface, sides that are often beveled or non-beveled vertical sides, and an overall outline that is often a polygon, including squares. The upper surface may have polygonal or circular openings. Traditionally, laser processing of planar parts requires manual positioning. Generally, a fixture is used to clamp the part, and then a worker selects several positioning points on it, aligning the template and part positions. Then, a robot or machine tool carrying a laser processing tool is used to process the part. In this method, the part is generally square, requiring manual positioning of corner points, or has a special shape requiring fixture clamping. With the increasing demand for laser processing, the variety of parts is also increasing, and efficiency requirements are becoming higher. Traditional registration methods for planar parts have disadvantages such as high manual intervention, poor accuracy, low efficiency, limited compatibility with various parts, and poor generalization. Summary of the Invention
[0003] The purpose of this invention is to provide a computer vision-based laser processing method, medium, and equipment for planar parts, aiming to improve the efficiency, automation, and precision of laser processing, and to adapt to a wider range of parts. The specific technical solution is as follows:
[0004] A computer vision-based laser processing method for planar parts, the method comprising the following steps:
[0005] S100. Obtain planar part point cloud by non-contact measurement of planar part based on computer vision, and obtain the edge of planar part point cloud by processing planar part point cloud; obtain part template point cloud by sampling part template by sampling algorithm, and obtain part template point cloud edge by processing part template point cloud.
[0006] S200. The three-dimensional Gaussian registration algorithm is used to register the edge of the planar part point cloud and the edge of the part template point cloud to obtain the pose transformation relationship between the planar part point cloud and the part template point cloud.
[0007] S300: Convert the part template according to the pose transformation relationship, and perform laser processing on the planar part according to the processing parameters in the part template.
[0008] Further, in step S100, the planar part is measured using a structured light camera or a profilometer to obtain the point cloud of the planar part.
[0009] Further, in step S100, obtaining the edge of the planar part point cloud by processing the planar part point cloud includes the following steps: converting the planar part point cloud into a first 3D original image, and the first 3D... Figure 2 Value map; for the first three dimensions Figure 2 The first Canny edge is obtained using the Canny edge detection algorithm; the centroid is calculated for each edge point in the first Canny edge to obtain the first contour direction; the first 3D shape is obtained using the first contour direction. Figure 2 The corresponding edge point in the value map is the first actual edge; the data of the point in the original first 3D image is obtained based on the first actual edge; the data of the point in the original first 3D image is projected onto the corresponding projection line to obtain the parameter of the point, and the point with the largest parameter value is selected as the first true edge corresponding to the first actual edge.
[0010] Further, in step S100, sampling the part template to obtain the part template point cloud by sampling algorithm includes: converting the part template into a topological format of triangular facets, establishing a local coordinate system for each triangular facet, calculating the points in each triangular facet according to the sampling interval, and integrating all points to obtain the part template point cloud.
[0011] Furthermore, obtaining the edge of the part template point cloud by processing the part template point cloud includes the following steps: converting the part template point cloud into a second 3D image, the original image, and the second 3D image. Figure 2 Value map; for the second three dimensions Figure 2 The value image is used to obtain the second Canny edge using the Canny edge detection algorithm; the centroid is calculated for each edge point in the second Canny edge to obtain the second contour direction; the second 3D shape is obtained using the second contour direction. Figure 2 The corresponding edge point in the value map is the second actual edge; the data of the points in the original second three-dimensional map is obtained based on the second actual edge; the data of the points in the original second three-dimensional map are projected onto the corresponding projection line to obtain the parameters of the points, and the point with the largest parameter value is selected as the second true edge corresponding to the second actual edge.
[0012] Further, in step S200, a three-dimensional Gaussian registration algorithm is used to register the first real edge and the second real edge to obtain the pose transformation relationship between the planar part point cloud and the part template point cloud.
[0013] Further, step S200 includes the following steps:
[0014] S210. Based on the principle of randomness, select multiple initial angles as candidates;
[0015] S220. For each candidate initial angle, determine the initial pose transformation matrix based on the initial angle;
[0016] S230. For each candidate initial angle, the first true edge is transformed according to the initial pose transformation matrix, and then registered with the second true edge according to the registration objective function. The registration objective function is optimized by the Levenberg-Marquardt algorithm to calculate the registration matrix. Finally, the initial pose transformation matrix is multiplied by the registration matrix to obtain the final registration matrix.
[0017] S240. Select the final registration matrix with the best registration result under all initial angles as the pose transformation relationship between the planar part point cloud and the part template point cloud.
[0018] Furthermore, in step S230, the registration objective function formula is as follows:
[0019]
[0020] in, Let i be a point in the first true edge under the initial pose; i is the index of the point in the first true edge, which is from 1 to n, and n is the number of points in the first true edge. Let j be the index of a point in the second real edge, which is from 1 to m, and m be the number of points in the second real edge. for and Euclidean distance; This is a rotation and translation transformation.
[0021] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the computer vision-based laser processing method for planar parts as described above.
[0022] The present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the computer vision-based laser processing method for planar parts as described above.
[0023] The present invention provides a method, medium, and equipment for laser processing of planar parts based on computer vision, which has the following beneficial effects:
[0024] This invention obtains a planar part point cloud through non-contact measurement of the planar part based on computer vision, and then processes the planar part point cloud to obtain its edges. A sampling algorithm is used to sample the part template to obtain a part template point cloud, and the same algorithm is applied to obtain its edges. A 3D Gaussian registration algorithm is used to register the edges of the planar part point cloud and the part template point cloud, revealing their pose transformation relationship. The part template is then converted based on this pose transformation relationship, and the planar part is laser-processed according to the processing parameters in the part template. This invention automates the registration of the planar part and part template in the laser processing of planar parts, avoiding the manual registration required in traditional planar part laser processing. This improves the efficiency, automation, and precision of laser processing of planar parts, and allows for the adaptation to a wider range of planar parts. Attached Figure Description
[0025] Figure 1 A schematic flowchart of a computer vision-based laser processing method for planar parts is provided for an embodiment of the present invention;
[0026] Figure 2 This is a schematic diagram of a planar part structure;
[0027] Figure 3 This is a schematic diagram of the sampling method for a part template;
[0028] Figure 4 This is a binary image of the point cloud;
[0029] Figure 5 This is a schematic diagram of the refined algorithm from the canny edge to the actual edge;
[0030] Figure 6 This is a schematic diagram of the refined algorithm from the actual edge to the true edge;
[0031] Figure 7 A 3D Gaussian registration model diagram;
[0032] Figure 8 This is a structural block diagram of a computer device according to an embodiment of the present invention. Detailed Implementation
[0033] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. The advantages and features of the present invention will become clearer from the following description. It should be noted that the drawings are all in a very simplified form and use non-precise proportions, and are only used to facilitate and clearly illustrate the purpose of the embodiments of the present invention.
[0034] Example 1
[0035] This embodiment provides a computer vision-based laser processing method for planar parts. (See attached document.) Figure 1 As shown, the method includes the following steps:
[0036] S100. Obtain planar part point cloud by non-contact measurement of planar part based on computer vision, and obtain the edge of planar part point cloud by processing planar part point cloud; obtain part template point cloud by sampling part template by sampling algorithm, and obtain part template point cloud edge by processing part template point cloud.
[0037] In one embodiment, a structured light camera or profilometer is used to measure a planar part to obtain a point cloud of the planar part. Specifically, devices capable of extracting 3D point clouds, such as structured light cameras or profilometers, are mounted on the motion mechanism of a robot or machine tool, moving above the planar part to collect its point cloud. For example, a laser processing device may be a robot, and the measuring device may be a 3D structured light camera. This camera is mounted on the robot via a platform and other necessary connecting devices. The robot carries the 3D structured light camera to the planar part to collect its point cloud.
[0038] In one embodiment, obtaining the edge of the planar part point cloud by processing the planar part point cloud includes the following steps: converting the planar part point cloud into a first three-dimensional original image, and the first three-dimensional... Figure 2 Value map; for the first three dimensions Figure 2 The first Canny edge is obtained using the Canny edge detection algorithm; the centroid is calculated for each edge point in the first Canny edge to obtain the first contour direction; the first 3D shape is obtained using the first contour direction. Figure 2 The corresponding edge point in the value map is the first actual edge; the data of the point in the original first 3D image is obtained based on the first actual edge; the data of the point in the original first 3D image is projected onto the corresponding projection line to obtain the parameter of the point, and the point with the largest parameter value is selected as the first true edge corresponding to the first actual edge.
[0039] In an exemplary embodiment, see Figure 2-7As shown, the cube bounding box of the part's point cloud is calculated, including the maximum and minimum z values zMax and zMin, the maximum and minimum x values xMax and xMin, and the maximum and minimum y values yMax and yMin. Then, the length L of the 3D image is calculated based on xMax - xMin, and the width W of the 3D image is calculated based on yMax and yMin. The dimensions of the 3D image are determined based on L and W. Generally, the original dimensions of the 3D image are length: L + 100, width: W + 100. Then, the sampling interval Interval is set. Since the actual point cloud interval is not exactly 1mm, the actual point cloud interval is set as the sampling interval. The dimensions of the sampled 3D image are length: (L + 100) / Interval, width: (W + 100) / Interval. Further 3D modeling is required. Figure 2 Value map, 3D Figure 2 The dimensions of the value map are the same as the dimensions of the 3D sample map.
[0040] We need to fill the original 3D image, the sampled 3D image, and the 3D image with point clouds. Figure 2 Value Map. Let the point cloud have a point P(x,y,z). Then the coordinates of point P in the original 3D graph are (round(Px-xMin), round(Py-yMin)), where the round() function rounds the input number to the nearest integer. The data of point P is stored in the original 3D graph. The coordinates of point P in the sampled 3D graph are (round((Px-xMin) / Interval), round((Py-yMin) / Interval)), and the data of point P is stored in the sampled 3D graph. Point P in 3D... Figure 2 The coordinates in the value map are consistent with their coordinates in the 3D sample map, and the corresponding coordinates in the 3D map are... Figure 2 Fill in 1 in the value graph, because it is three-dimensional. Figure 2 The initial value of the value in the value graph is 0, so in three dimensions... Figure 2 In the value graph, the part without data is 0, and the part with data is 1.
[0041] Figure 4 In the diagram, the outer black frame represents the border of the 3D image, the inner white area represents the data area, the red lines represent Canny edges, the green lines represent actual edges, and the blue lines represent ground truth edges. Canny edge detection methods are needed to process the 3D image. Figure 2 The value map is used to obtain the Canny edges. Since the location corresponding to the Canny edge may not be the actual location, it is necessary to map the Canny edges to the actual edges. The Canny edge detection results are traversed, and the 8-neighborhood of each Canny edge is extracted. If it contains a 3D... Figure 2For the values in the data map, extract the corresponding data from the original 3D image. Calculate the centroid of these points, and obtain the vector Vcc by connecting the current point's position cu and the centroid ca of the data points. Let Vcc point in the direction of the part, and -Vcc point outwards from the contour. Following the direction of -Vcc, extract the coordinates of points with a value of 1 in the binary image along the line passing through point cu with the direction of -Vcc. The coordinates of the point closest to point cu are then the actual edge coordinates. In this way, the Canny edge is mapped to the actual edge.
[0042] It is necessary to map from actual edges to true edges. Since the edges of the actual part's point cloud sometimes have bevels, the edges obtained by the above method are actually the lower edges of the bevels. Due to the shooting angle of the 3D camera, the acquired point cloud may not capture all bevel edges, so the actual edges are inaccurate. A more accurate extraction of the part's upper surface edges is needed. Using the same method, traverse all edge points, let the current point be ct, and calculate the contour direction Vcc. Extract the data of points in the original 3D image on all lines passing through the ct point with the direction -Vcc. Select the closest projection line based on the direction of Vcc. If Vcc approaches the direction (1, 0, 0), the projection line is chosen as (cos(45), 0, sin(45)); if Vcc approaches the direction (-1, 0, 0), the projection line is chosen as (-cos(45), 0, sin(45)); if Vcc approaches the direction (0, 1, 0), the projection line is chosen as (0, cos(45), sin(45)); if Vcc approaches the direction (0, -1, 0), the projection line is chosen as (0, -cos(45), sin(45)). The projection line passes through the point (0, 0, 0). Projecting all points onto the projection line, the point with the largest parameter value is the true edge we are looking for. The parameter value refers to the value t that corresponds to each point projected onto the projection line when the projection line is written as a parametric equation.
[0043] In one embodiment, obtaining the part template point cloud by sampling the part template through a sampling algorithm includes: converting the part template into a topological format of triangular facets, establishing a local coordinate system for each triangular facet, calculating the points in each triangular facet according to the sampling interval, and integrating all the points to obtain the part template point cloud.
[0044] In one embodiment, obtaining the edge of the part template point cloud by processing the part template point cloud includes the following steps: converting the part template point cloud into a second 3D original image, and the second 3D... Figure 2 Value map; for the second three dimensions Figure 2The value image is used to obtain the second Canny edge using the Canny edge detection algorithm; the centroid is calculated for each edge point in the second Canny edge to obtain the second contour direction; the second 3D shape is obtained using the second contour direction. Figure 2 The corresponding edge point in the value map is the second actual edge; the data of the points in the original second three-dimensional map is obtained based on the second actual edge; the data of the points in the original second three-dimensional map are projected onto the corresponding projection line to obtain the parameters of the points, and the point with the largest parameter value is selected as the second true edge corresponding to the second actual edge.
[0045] S200. The three-dimensional Gaussian registration algorithm is used to register the edge of the planar part point cloud and the edge of the part template point cloud to obtain the pose transformation relationship between the planar part point cloud and the part template point cloud.
[0046] In one embodiment, a three-dimensional Gaussian registration algorithm is used to register the first real edge and the second real edge to obtain the pose transformation relationship between the planar part point cloud and the part template point cloud.
[0047] In a preferred embodiment, step S200 includes the following steps:
[0048] S210. Based on the principle of randomness, select multiple initial angles as candidates.
[0049] Specifically, the true edges of the part point cloud need to be rotated according to the initial angle. Since the registration algorithm has a certain degree of randomness, the registration results of the true edges of the part point cloud and the true edges of the part template point cloud with different initial angles will be different. Therefore, it is necessary to calculate multiple sets of registration results for the two edges under different initial angles, and then select the best result as the final result. Since this invention registers the edges of two planes, the initial angle is the angle of rotation around the z-axis in Euler angles. Generally, based on the principle of randomness, 100 sets of random numbers are calculated, and these random numbers are normalized to between 0 and 360 degrees as candidates for the initial angle. Then, in one iteration, a candidate angle is selected as the initial angle.
[0050] S220. For each candidate initial angle, determine the initial pose transformation matrix based on the initial angle.
[0051] Specifically, the initial offset value is the difference between the centroids of their respective point clouds. Then, the rotation angle around the z-axis and the offset values of the x-axis and y-axis together constitute the initial pose transformation matrix.
[0052] S230. For each candidate initial angle, the first true edge is transformed according to the initial pose transformation matrix, and then registered with the second true edge according to the registration objective function. The registration objective function is optimized by the Levenberg-Marquardt algorithm to calculate the registration matrix. Finally, the initial pose transformation matrix is multiplied by the registration matrix to obtain the final registration matrix.
[0053] Specifically, the registration objective function formula is as follows:
[0054]
[0055] in, Let i be a point in the first true edge under the initial pose; i is the index of the point in the first true edge, which is from 1 to n, and n is the number of points in the first true edge. Let j be the index of a point in the second real edge, which is from 1 to m, and m be the number of points in the second real edge. for and Euclidean distance; This is a rotation and translation transformation.
[0056] The physical meaning of the above registration objective function formula is that through rotation and translation transformation... After and weighted average The weighted average rule for overlapping points is that the closer the points are, the greater the weight, and the farther the points are, the lower the weight. The objective function is minimized when the two point clouds completely overlap.
[0057] Specifically, the registration matrix is a 6-dimensional vector containing 3 Euclidean rotation angles and 3 translations to determine the pose relationship between the point clouds of the planar part and the point cloud of the part template in space.
[0058] S240. Select the final registration matrix with the best registration result under all initial angles as the pose transformation relationship between the planar part point cloud and the part template point cloud.
[0059] S300: Convert the part template according to the pose transformation relationship, and perform laser processing on the planar part according to the processing parameters in the part template.
[0060] Specifically, the robot or machine tool transforms the part template according to the pose transformation relationship, and processes the part according to the processing parameters in the part template.
[0061] This invention provides an embodiment of a planar part that obtains a point cloud through non-contact measurement of the planar part using computer vision. The point cloud is then processed to obtain its edges. A sampling algorithm is used to sample a part template, resulting in a part template point cloud. The part template point cloud is then processed to obtain its edges. A 3D Gaussian registration algorithm is employed to register the edges of the planar part point cloud and the part template point cloud, revealing their pose transformation relationship. The part template is then converted based on this pose transformation relationship, and the planar part is laser-processed according to the processing parameters in the part template. This automation of the planar part-template registration process in the laser processing of planar parts avoids the manual registration required in traditional planar part laser processing. This improves the efficiency, automation, and precision of laser processing of planar parts, and allows for the adaptation to a wider range of planar parts.
[0062] Example 2
[0063] This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the computer vision-based laser processing method for planar parts described above.
[0064] The storage medium may be a magnetic disk, optical disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk drive (HDD), or solid-state drive (SSD), etc.; the storage medium may also include combinations of the above types of memory.
[0065] Example 3
[0066] This embodiment provides a computer device, which includes: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps of the computer vision-based laser processing method for planar parts described above.
[0067] like Figure 8As shown, the computer device 70 may include: at least one processor 71, such as a CPU (Central Processing Unit), at least one communication interface 73, a memory 74, and at least one communication bus 72. The communication bus 72 is used to enable communication between these components. The communication interface 73 may include a display screen and a keyboard; optionally, the communication interface 73 may also include a standard wired interface or a wireless interface. The memory 74 may be high-speed RAM (Random Access Memory) or non-volatile memory, such as at least one disk storage device. Optionally, the memory 74 may also be at least one storage device located remotely from the aforementioned processor 71. The memory 74 stores application programs, and the processor 71 calls the program code stored in the memory 74 to execute any of the above-described method steps.
[0068] The communication bus 72 can be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The communication bus 72 can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 8 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.
[0069] The memory 74 may include volatile memory, such as random-access memory (RAM); the memory may also include non-volatile memory, such as flash memory, hard disk drive (HDD) or solid-state drive (SSD); the memory 74 may also include a combination of the above types of memory.
[0070] The processor 71 can be a central processing unit (CPU), a network processor (NP), or a combination of CPU and NP.
[0071] The processor 71 may further include a hardware chip. This hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The PLD may be a complex programmable logic device (CPLD), a field-programmable gate array (FPGA), a generic array logic (GAL), or any combination thereof.
[0072] Optionally, the memory 74 is also used to store program instructions. The processor 71 can invoke the program instructions to implement the computer vision-based laser processing method for planar parts as described in this invention.
[0073] Those skilled in the art should understand that the present invention can be implemented in many other specific forms without departing from the spirit and scope of the invention. Any changes or modifications made by those skilled in the art based on the embodiments of the present invention and the above disclosure shall fall within the protection scope of the claims.
Claims
1. A method for laser processing of planar parts based on computer vision, characterized in that, The method includes the following steps: S100. Obtain planar part point cloud by non-contact measurement of planar part based on computer vision, and obtain the edge of planar part point cloud by processing planar part point cloud; obtain part template point cloud by sampling part template by sampling algorithm, and obtain part template point cloud edge by processing part template point cloud. S200. The three-dimensional Gaussian registration algorithm is used to register the edge of the planar part point cloud and the edge of the part template point cloud to obtain the pose transformation relationship between the planar part point cloud and the part template point cloud. S300: Convert the part template according to the pose transformation relationship, and perform laser processing on the planar part according to the processing parameters in the part template.
2. The laser processing method for planar parts based on computer vision according to claim 1, characterized in that, In step S100, the planar part is measured using a structured light camera or a profilometer to obtain the point cloud of the planar part.
3. The laser processing method for planar parts based on computer vision according to claim 1, characterized in that, In step S100, obtaining the edge of the planar part point cloud by processing the planar part point cloud includes the following steps: converting the planar part point cloud into a first three-dimensional original image and a first three-dimensional binary image; and using the Canny edge detection algorithm to obtain the first Canny edge from the first three-dimensional binary image. Calculate the centroid of each edge point in the first canny edge to obtain the first contour direction; use the first contour direction to obtain the corresponding edge point in the first three-dimensional binary image as the first actual edge; obtain the data of the points in the original image of the first three-dimensional image based on the first actual edge. The data of points in the original 3D image are projected onto the corresponding projection lines to obtain the parameters of the points. The point with the largest parameter value is selected as the first real edge corresponding to the first actual edge.
4. The laser processing method for planar parts based on computer vision according to claim 3, characterized in that, In step S100, sampling the part template to obtain the part template point cloud by sampling algorithm includes: converting the part template into a topological format of triangular facets, establishing a local coordinate system for each triangular facet, calculating the points in each triangular facet according to the sampling interval, and integrating all the points to obtain the part template point cloud.
5. The computer vision-based laser processing method for planar parts according to claim 4, characterized in that, The process of obtaining the edge of the part template point cloud by processing the part template point cloud includes the following steps: converting the part template point cloud into the original image of the second 3D image and the binary image of the second 3D image; and using the Canny edge detection algorithm to obtain the second Canny edge of the binary image of the second 3D image. Calculate the centroid of each edge point in the second canny edge to obtain the second contour direction; use the second contour direction to obtain the corresponding edge point in the second three-dimensional binary image as the second actual edge. Data of points in the original image of the second three-dimensional image are obtained based on the second actual edge; The data of points in the original second 3D image are projected onto the corresponding projection lines to obtain the parameters of the points. The point with the largest parameter value is selected as the second real edge corresponding to the second actual edge.
6. The computer vision-based laser processing method for planar parts according to claim 5, characterized in that, In step S200, the first real edge and the second real edge are registered using a three-dimensional Gaussian registration algorithm to obtain the pose transformation relationship between the planar part point cloud and the part template point cloud.
7. The laser processing method for planar parts based on computer vision according to claim 6, characterized in that, Step S200 includes the following steps: S210. Based on the principle of randomness, select multiple initial angles as candidates; S220. For each candidate initial angle, determine the initial pose transformation matrix based on the initial angle; S230. For each candidate initial angle, the first true edge is transformed according to the initial pose transformation matrix, and then registered with the second true edge according to the registration objective function. The registration objective function is optimized by the Levenberg-Marquardt algorithm to calculate the registration matrix. Finally, the initial pose transformation matrix is multiplied by the registration matrix to obtain the final registration matrix. S240. Select the final registration matrix with the best registration result under all initial angles as the pose transformation relationship between the planar part point cloud and the part template point cloud.
8. The laser processing method for planar parts based on computer vision according to claim 7, characterized in that, In step S230, the registration objective function formula is as follows: in, Let i be a point in the first true edge under the initial pose; i is the index of the point in the first true edge, which is from 1 to n, and n is the number of points in the first true edge. Let j be the index of a point in the second real edge, which is from 1 to m, and m be the number of points in the second real edge. for and Euclidean distance; This is a rotation and translation transformation.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the computer vision-based laser processing method for planar parts as described in any one of claims 1-8.
10. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the computer vision-based laser processing method for planar parts as described in any one of claims 1-8.