Robot vision automatic scanning path planning method based on three-dimensional grid model

Abstract

The invention relates to a robot vision automatic scanning path planning method based on a three-dimensional grid model, and belongs to the related field of machining and manufacturing. According to the method, a scanning view cone model of a scanner is obtained according to a series of constraints of a surface structured light scanner, and through Gaussian mapping and clustering analysis, through segmentation and combination of triangular patches, a three-dimensional grid model is obtained; the method comprises the steps of obtaining a plurality of areas convenient to scan of a workpiece, generating candidate viewpoints according to a scanning view cone model, introducing viewpoint measurement quality parameters, selecting optimal viewpoints, and carrying out path planning on all the obtained optimal viewpoints according to a greedy algorithm, so that automatic viewpoint arrangement and automatic path generation are realized; the problems of repeated measurement, measurement omission, low efficiency and the like of manual teaching viewpoint workpiece measurement are solved, and the method has the advantages of high precision, high efficiency, high integrity, automation and the like.

Description

TECHNICAL FIELD

[0001] The application relates to a robot vision automatic scanning path planning method based on a three-dimensional grid model and belongs to the related field of mechanical processing and manufacturing. BACKGROUND

[0002] With the development of industrial production and manufacturing, three-dimensional point cloud measurement is gradually utilized due to its non-contact, rapidness and precision. Integrating a handheld surface structure light scanner with an industrial robot not only utilizes the advantages of the handheld scanner such as simplicity, high efficiency, replaceability and low cost, but also fully utilizes the flexibility and repeatability of the industrial robot, which is beneficial to reducing the labor intensity of workers and promoting the construction of a digital intelligent production line. Due to the limitations of the scanner and the robot, only part of the appearance of a workpiece can be obtained at a single viewpoint, and multiple viewpoints must be planned to measure the three-dimensional model of the workpiece. However, most of the current robot scanning viewpoint planning adopts the mode of manually teaching the viewpoint to measure the workpiece. This mode completely relies on the experience and technical level of the technicians to adjust, and the operation behavior of the technicians usually lacks globality and accuracy, which is easy to cause problems such as repeated measurement or missed measurement, and the work efficiency of the workers is low, so that the whole measurement process takes a long time. Therefore, it is urgent to propose a measurement trajectory planning method to replace manual teaching of the viewpoint, automatically plan the viewpoint and the path, realize scanning of the workpiece by the industrial robot with the surface structure light scanner, and realize complete, high-precision and high-efficiency measurement. SUMMARY

[0003] The application aims to overcome the above-mentioned deficiencies in the prior art and provide a robot vision automatic scanning path planning method based on a three-dimensional grid model. The scanning cone model of the scanner is obtained according to the series of constraints of the surface structure light scanner. Through Gaussian mapping and cluster analysis, the multiple regions of the workpiece that are convenient for scanning are obtained through segmentation and merging of triangular facets. The candidate viewpoints are generated according to the scanning cone model, the optimal viewpoint is selected by introducing the viewpoint measurement quality parameter, and the path planning is performed on all the optimal viewpoints obtained according to the greedy algorithm. Thus, the automatic arrangement of the viewpoint and the automatic generation of the path are realized, and the problems of repeated measurement, missed measurement and low efficiency in the manual teaching of the viewpoint for the workpiece measurement are solved.

[0004] The above-mentioned purpose of the application is mainly achieved by the following technical scheme:

[0005] The robot vision automatic scanning path planning method based on a three-dimensional grid model comprises the following steps:

[0006] A scanning cone model of the scanner is obtained according to the measurement space constraint condition of the surface structure light scanner.

[0007] The workpiece 3D model is obtained by scanning with a surface structured light scanner. Through Gaussian mapping and cluster analysis, the preliminary region division of the workpiece 3D model is obtained. Then, through the segmentation and merging of triangular facets, the workpiece 3D model is divided into multiple scanning regions.

[0008] For each scanned region, candidate viewpoints are generated based on the scanned view cone model;

[0009] Calculate the measurement quality parameters for each candidate viewpoint, evaluate the measurement quality of each candidate viewpoint, and select the candidate viewpoint with the highest measurement quality parameters as the optimal viewpoint.

[0010] Path planning is performed on all the optimal viewpoints obtained using a greedy algorithm.

[0011] In the above-mentioned robot vision automatic scanning path planning method based on a three-dimensional mesh model, the measurement space constraints include visibility constraints, field of view constraints, depth of field constraints, projector constraints, and occlusion constraints.

[0012] In the above-mentioned robot vision automatic scanning path planning method based on a three-dimensional mesh model, the visibility constraint refers to the angle between the scanner's shooting direction and the measurement point being less than a certain value. Suppose that the scanner's shooting direction v along the main axis is tangent to any point on the curved surface at p, and the product of the normal vector n of that point and the scanner's shooting direction v is less than zero, that is: n·v<0.

[0013] In the above-mentioned robot vision automatic scanning path planning method based on a 3D mesh model, the scanning cone model is a cone with a height of H. opt For optimal ranging of a structured light scanner, the diameter of the base circle is L. opt The cone angle is the field of view angle, and the measurement distance of the surface structured light scanner is... D represents the scanner's depth of field.

[0014] In the above-mentioned robot vision automatic scanning path planning method based on a 3D mesh model, the 3D model of the workpiece obtained by scanning with a surface structured light scanner is a simplified STL model.

[0015] In the above-mentioned robot vision automatic scanning path planning method based on a 3D mesh model, the 3D model of the workpiece is obtained by scanning with a surface structured light scanner. Through Gaussian mapping and cluster analysis, a preliminary region division of the workpiece's 3D model is obtained. Then, through the segmentation and merging of triangular facets, the workpiece's 3D model is divided into multiple scanning regions, including:

[0016] The normal vectors of all points on the workpiece surface are mapped onto a unit sphere. Then, cluster analysis is used to divide triangular facets with similar normal vector angles into a single class, resulting in a preliminary region division. Subsequently, unconnected triangular facets within the same class are separated into independent regions. For triangular facet clusters with areas smaller than a set threshold, they are merged into the adjacent cluster with the smallest average normal vector angle. For triangular facet clusters with areas larger than the measurement space range, they are divided in half, ultimately resulting in multiple scanning regions.

[0017] In the above-mentioned robot vision automatic scanning path planning method based on a three-dimensional mesh model, the triangular facets with similar normal angles refer to triangular facets with a difference in normal angle of less than 20°.

[0018] In the above-mentioned robot vision automatic scanning path planning method based on a 3D mesh model, for each scanning region, candidate viewpoints are generated according to the scanning frustum model, including:

[0019] For each scanned region, the particle and the average normal vector are calculated, and the normal vector is drawn from the particle along the normal vector H. opt Set a candidate viewpoint at a distance, then use the point mass as the vertex of a cone with a cone angle of 30° and the normal vector as the perpendicular line, H opt On the circle at the base of the cone with height, six candidate viewpoints are selected at 60° intervals, with the viewpoints pointing towards the point mass. The pose of the viewpoint in three-dimensional space is represented by the candidate viewpoint coordinate system. The Y-axis points from the candidate viewpoint to the center of the sphere, the X-axis is the tangent to the circular plane where the candidate viewpoint is located, and this circular plane is the tangent plane for generating the candidate viewpoint. The Z-axis is determined based on the right-hand rule.

[0020] In the above-mentioned robot vision automatic scanning path planning method based on a 3D mesh model, the measurement quality parameters of each candidate viewpoint are calculated, including:

[0021] Cf p =a·Cf vis +b·Cf θ

[0022] Among them: Cf p Cf is the measurement quality parameter for candidate viewpoints. vis Cf is a visibility parameter. θ Here, 'a' is the image quality parameter, and 'a' is the visibility parameter Cf. vis The weight value of b, where b is the imaging quality parameter Cf. θ The weight value.

[0023] In the above-mentioned robot vision automatic scanning path planning method based on a 3D mesh model, a is set to 3 and b is set to 2.

[0024] In the above-mentioned robot vision automatic scanning path planning method based on a 3D mesh model, the visibility parameter Cf vis The number of local feature triangles corresponding to this viewpoint is represented by the following formula:

[0025]

[0026] Among them, S p 'The sum of the areas of all triangular faces within this partition that satisfy the visualization constraints of this viewpoint.' This represents the sum of the areas of all triangular faces within the partition.

[0027] In the above-mentioned robot vision automatic scanning path planning method based on a 3D mesh model, the imaging quality parameter Cf θ This indicates the image quality of all triangles at that viewpoint, and the calculation formula is as follows:

[0028]

[0029] Cf θ =S n ·N n

[0030] Among them, S n The normalized triangle area coefficient represents the proportion of the area of ​​this triangle in the total area of ​​all triangles in this partition, and θ is the angle between the vector from the center of the triangle to the candidate viewpoint and the normal vector of the triangle.

[0031] In the above-mentioned robot vision automatic scanning path planning method based on a 3D mesh model, path planning is performed on all the obtained optimal viewpoints using a greedy algorithm, including:

[0032] Select the robot's starting position as the starting point of the scanning path. Find the point closest to the starting point in the optimal viewpoint array composed of all optimal viewpoints and mark it as the second point of the scanning path. Then find the unmarked point closest to the second point in the optimal viewpoint array and mark it as the third point of the scanning path. Continue in this manner until all viewpoints in the optimal viewpoint array are marked. Connect the points at the robot's initial position with the reordered viewpoints in sequence to generate the shortest scanning path for the surface scanning measurement robot.

[0033] A robot vision-based automatic scanning path planning system based on a 3D mesh model includes:

[0034] The view frustum model generation module generates the scanning view frustum model of the scanner based on the measurement space constraints of the surface structured light scanner.

[0035] The scanning area division module obtains the three-dimensional model of the workpiece by scanning with a surface structured light scanner. Through Gaussian mapping and cluster analysis, it obtains the preliminary area division of the three-dimensional model of the workpiece. Then, through the segmentation and merging of triangular facets, the three-dimensional model of the workpiece is divided into multiple scanning areas.

[0036] The candidate viewpoint generation module generates candidate viewpoints for each scanned region based on the scanned viewpoint model.

[0037] The optimal viewpoint acquisition module calculates the measurement quality parameters of each candidate viewpoint, evaluates the measurement quality of each candidate viewpoint, and selects the candidate viewpoint with the highest measurement quality parameters as the optimal viewpoint.

[0038] The path planning module performs path planning for all the optimal viewpoints obtained using a greedy algorithm.

[0039] A computer device includes a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method described above.

[0040] Compared with the prior art, the present invention has at least the following beneficial effects:

[0041] (1) This invention provides a robot vision automatic scanning path planning method based on a three-dimensional mesh model. The scanning cone model of the scanner is obtained according to a series of constraints of the surface structured light scanner. Through Gaussian mapping and cluster analysis, and after the segmentation and merging of triangular facets, multiple areas of the workpiece that are easy to scan are obtained. Candidate viewpoints are generated according to the scanning cone model. The viewpoint measurement quality parameters are introduced to select the optimal viewpoint. The path planning is performed on all the optimal viewpoints obtained according to the greedy algorithm. Thus, the automatic placement of viewpoints and the automatic generation of paths are realized, which solves the problems of repeated measurement, missed measurement and low efficiency in workpiece measurement by manually teaching viewpoints.

[0042] (2) This invention proposes an automatic viewpoint arrangement and path generation method to control the robot to clamp the surface structure light scanner to measure the workpiece, which has the advantages of high precision, high efficiency, strong integrity and automation.

[0043] (3) The present invention realizes the automatic arrangement of the viewpoint for scanning the workpiece, and automatically plans the path according to the viewpoint to facilitate the robot to read the movement; by selecting the optimal viewpoint through the viewpoint quality parameters, the scanning accuracy and integrity are improved, the calculation process is automated, and compared with manual hand measurement, the labor cost is reduced and the measurement efficiency is improved; the improvement of measurement accuracy avoids the batch scrapping of products in the subsequent processing process, reducing the loss cost and raw material consumption.

[0044] (4) The technical solution of the present invention fills the technical gap in the industry at home and abroad: When a workpiece is scanned in three dimensions, it is usually done by a technician holding a scanner and scanning by manual teaching. However, this method relies entirely on the experience and technical level of the technician, which often lacks globality and accuracy, and is prone to problems such as repeated measurement or omission of measurement. Moreover, the efficiency of manual work is low, which makes the whole measurement process take a long time. Through the present invention, the automatic arrangement of viewpoints and path planning are realized, and the robot is used to execute the set path for three-dimensional measurement, so that the scanning has the advantages of high integrity, high precision, high efficiency and low cost. Attached Figure Description

[0045] Figure 1 A flowchart of a robot vision automatic scanning path planning method based on a three-dimensional mesh model provided in an embodiment of the present invention;

[0046] Figure 2 The scanning cone model provided in the embodiments of the present invention;

[0047] Figure 3 This is a schematic diagram of the candidate viewpoint distribution provided in an embodiment of the present invention. Detailed Implementation

[0048] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments:

[0049] like Figure 1 As shown in the figure, this embodiment of the invention provides a robot vision automatic scanning path planning method based on a three-dimensional mesh model, which specifically includes the following steps:

[0050] I. Based on the constraints of the measurement space of the surface structured light scanner (a series of constraints such as depth of field and field of view), the scanning cone model of the scanner is obtained, such as... Figure 2 The image shown is a scanning cone model provided in an embodiment of the present invention.

[0051] The constraints of the measurement space include visibility constraints, field of view constraints, depth of field constraints, projector constraints, and occlusion constraints.

[0052] The scanning cone model is a cone with a height of H. opt For optimal ranging of a structured light scanner, the diameter of the base circle is L. opt The cone angle is the field of view angle, and the measurement distance of the surface structured light scanner is... D represents the scanner's depth of field.

[0053] Visibility constraints refer to the requirement that the angle between the scanner's shooting direction and the measurement point is less than a certain value. Suppose that the scanner's shooting direction v along the main axis is tangent to any point on the surface at p. The product of the normal vector n of that point and the scanner's shooting direction v must be less than zero, i.e., n·v<0.

[0054] 2. Based on the 3D model of the workpiece obtained by scanning with a surface structured light scanner, in this embodiment, a rough STL model of the workpiece is obtained. Through Gaussian mapping and cluster analysis, a preliminary region division of the workpiece model is obtained. Then, through the segmentation and merging of triangular facets, the workpiece model is finally divided into multiple regions that are easy to scan. The specific methods include:

[0055] The normal vectors of all points on the workpiece surface are mapped onto a unit sphere. Cluster analysis is then used to group triangular facets with similar normal vector angles into a single class, resulting in a preliminary region division. Next, unconnected triangular facets within the same class are separated into independent regions. Clusters of triangular facets with areas smaller than a set threshold are merged into adjacent clusters with the smallest average normal vector angle. Clusters of triangular facets with areas larger than the measurement space are further divided in half, resulting in two regions. This ultimately yields multiple regions suitable for scanning.

[0056] 3. For each part of the scanned area, candidate viewpoints are generated based on the scanned cone model.

[0057] For each region, the particle and the average normal vector are calculated, and the normal vector is calculated from the particle along the normal vector H. opt Set a candidate viewpoint at a distance, then use the point mass as the vertex of a cone with a cone angle of 30° and the normal vector as the perpendicular line, H opt Six candidate viewpoints are selected at 60° intervals on the circle at the base of the cone, with the viewpoints pointing towards the point mass. The pose of each viewpoint in three-dimensional space is represented by the candidate viewpoint coordinate system. The Y-axis points from the candidate viewpoint to the center of the sphere, the X-axis is the tangent line to the circular plane containing the candidate viewpoint (which is the tangent plane used to generate the candidate viewpoints), and the Z-axis is determined based on the right-hand rule. Figure 3 The diagram shown is a schematic diagram of the candidate viewpoint distribution provided in an embodiment of the present invention.

[0058] Fourth, introduce viewpoint measurement quality parameters to evaluate the measurement quality of each candidate viewpoint, and select the candidate viewpoint with the highest measurement quality as the optimal viewpoint.

[0059] Introducing viewpoint measurement quality parameters:

[0060] Cf p =a·Cf vis +b·Cf θ

[0061] Among them, Cf p Cf is the measurement quality parameter for the viewpoint. vis Cf is a visibility parameter. θ Here are the imaging quality parameters. a and b are the weight values ​​for the two parameters mentioned above; a is set to 3 and b to 2.

[0062] Furthermore, the visibility parameter Cfvis This represents the number of local feature triangles corresponding to this viewpoint, and its calculation formula is as follows:

[0063]

[0064] Among them, S p 'The sum of the areas of all triangular faces within this partition that satisfy the visualization constraints of this viewpoint.' This represents the sum of the areas of all triangular faces within the partition.

[0065] Furthermore, the imaging quality parameter Cf θ This indicates the image quality of all triangles corresponding to this viewpoint, and its calculation formula is as follows:

[0066]

[0067] Cf θ =S n ·N n

[0068] Among them, S n The normalized triangle area coefficient represents the proportion of the area of ​​this triangle in the total area of ​​all triangles in this partition, and θ is the angle between the vector from the center of the triangle to the candidate viewpoint and the normal vector of the triangle.

[0069] 5. Perform path planning for all the optimal viewpoints obtained using a greedy algorithm.

[0070] First, the robot's starting position is selected as the starting point of the scanning path. The point closest to the starting point in the optimal viewpoint array is found and marked as the second point on the scanning path. Then, the unmarked point closest to the second point is found in the viewpoint array and marked as the third point on the scanning path, and so on, until all viewpoints in the viewpoint array are marked. Connecting the robot's initial position point with the reordered viewpoints sequentially generates the shortest scanning path for the area scanning measurement robot.

[0071] The present invention also provides a robot vision automatic scanning path planning system based on a three-dimensional mesh model, comprising:

[0072] The view frustum model generation module generates the scanning view frustum model of the scanner based on the measurement space constraints of the surface structured light scanner.

[0073] The scanning area division module obtains the three-dimensional model of the workpiece by scanning with a surface structured light scanner. Through Gaussian mapping and cluster analysis, it obtains the preliminary area division of the three-dimensional model of the workpiece. Then, through the segmentation and merging of triangular facets, the three-dimensional model of the workpiece is divided into multiple scanning areas.

[0074] The candidate viewpoint generation module generates candidate viewpoints for each scanned region based on the scanned viewpoint model.

[0075] The optimal viewpoint acquisition module calculates the measurement quality parameters of each candidate viewpoint, evaluates the measurement quality of each candidate viewpoint, and selects the candidate viewpoint with the highest measurement quality parameters as the optimal viewpoint.

[0076] The path planning module performs path planning for all the optimal viewpoints obtained using a greedy algorithm.

[0077] The present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the above-described method.

[0078] This invention improves the measurement accuracy, completeness, and efficiency of the workpiece model by dividing a rough workpiece model into regions, generating candidate viewpoints based on the scanning cone, selecting the optimal viewpoint by introducing viewpoint measurement quality parameters, and finally planning a path for the obtained optimal viewpoint. Based on the obtained path, the workpiece is 3D scanned by a robot clamping scanner.

[0079] The above description is only the best specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the protection scope of the present invention.

[0080] The contents not described in detail in this specification are common knowledge to those skilled in the art.

Claims

1. A robot vision-based automatic scanning path planning method based on a 3D mesh model, characterized in that, include: The scanning cone model of the scanner is obtained based on the measurement space constraints of the surface structured light scanner; The workpiece 3D model is obtained by scanning with a surface structured light scanner. Through Gaussian mapping and cluster analysis, the preliminary region division of the workpiece 3D model is obtained. Then, through the segmentation and merging of triangular facets, the workpiece 3D model is divided into multiple scanning regions. For each scanned region, candidate viewpoints are generated based on the scanned view cone model; Calculate the measurement quality parameters for each candidate viewpoint, evaluate the measurement quality of each candidate viewpoint, and select the candidate viewpoint with the highest measurement quality parameters as the optimal viewpoint. Path planning is performed on all the optimal viewpoints obtained using a greedy algorithm.

2. The robot vision automatic scanning path planning method based on a three-dimensional mesh model according to claim 1, characterized in that, The measurement space constraints include visibility constraints, field of view constraints, depth of field constraints, projector constraints, and occlusion constraints.

3. The robot vision automatic scanning path planning method based on a three-dimensional mesh model according to claim 2, characterized in that, The visibility constraint refers to the angle between the scanner's shooting direction and the measurement point being less than a certain value. Suppose that the scanner's shooting direction v along the main axis is tangent to any point on the curved surface at p, and the product of the normal vector n at that point and the scanner's shooting direction v is less than zero, that is: n·v<0.

4. The robot vision automatic scanning path planning method based on a three-dimensional mesh model according to claim 1, characterized in that, The scanning cone model is a cone with a height of H. opt For optimal ranging of a structured light scanner, the diameter of the base circle is L. opt The cone angle is the field of view angle, and the measurement distance of the surface structured light scanner is... D represents the scanner's depth of field.

5. The robot vision automatic scanning path planning method based on a three-dimensional mesh model according to claim 1, characterized in that, The 3D model of the workpiece obtained by the surface structured light scanner is a simplified STL model.

6. The robot vision automatic scanning path planning method based on a three-dimensional mesh model according to claim 1, characterized in that, The 3D model of the workpiece is obtained by scanning with a structured light scanner. Through Gaussian mapping and cluster analysis, a preliminary region division of the 3D model is obtained. Then, through triangular facet segmentation and merging, the 3D model of the workpiece is divided into multiple scanning regions, including: The normal vectors of all points on the workpiece surface are mapped onto a unit sphere. Then, cluster analysis is used to divide triangular facets with similar normal vector angles into a single class, resulting in a preliminary region division. Subsequently, unconnected triangular facets within the same class are separated into independent regions. For triangular facet clusters with areas smaller than a set threshold, they are merged into the adjacent cluster with the smallest average normal vector angle. For triangular facet clusters with areas larger than the measurement space range, they are divided in half, ultimately resulting in multiple scanning regions.

7. The robot vision automatic scanning path planning method based on a three-dimensional mesh model according to claim 6, characterized in that, The triangular facets with similar angles between their normal vectors refer to triangular facets with an angle difference of less than 20° between their normal vectors.

8. The robot vision automatic scanning path planning method based on a three-dimensional mesh model according to claim 1, characterized in that, For each scanned region, candidate viewpoints are generated based on the scanned view frustum model, including: For each scanned region, the particle and the average normal vector are calculated, and the normal vector is drawn from the particle along the normal vector H. opt Set a candidate viewpoint at a distance, then use the point mass as the vertex of a cone with a cone angle of 30° and the normal vector as the perpendicular line, H opt On the circle at the base of the cone with height, six candidate viewpoints are selected at 60° intervals, with the viewpoints pointing towards the point mass. The pose of the viewpoint in three-dimensional space is represented by the candidate viewpoint coordinate system. The Y-axis points from the candidate viewpoint to the center of the sphere, the X-axis is the tangent to the circular plane where the candidate viewpoint is located, and this circular plane is the tangent plane for generating the candidate viewpoint. The Z-axis is determined based on the right-hand rule.

9. The robot vision automatic scanning path planning method based on a three-dimensional mesh model according to claim 1, characterized in that, Calculate the measurement quality parameters for each candidate viewpoint, including: Cf p =a·Cf vis +b·Cf θ Among them: Cf p Cf is the measurement quality parameter for candidate viewpoints. vis Cf is a visibility parameter. θ Here, 'a' is the image quality parameter, and 'a' is the visibility parameter Cf. vis The weight value of b, where b is the imaging quality parameter Cf. θ The weight value.

10. The robot vision automatic scanning path planning method based on a three-dimensional mesh model according to claim 9, characterized in that, Set a to 3 and b to 2.

11. The robot vision automatic scanning path planning method based on a three-dimensional mesh model according to claim 9, characterized in that, The visibility parameter Cf vis The number of local feature triangles corresponding to this viewpoint is represented by the following formula: Among them, S p 'The sum of the areas of all triangular faces within this partition that satisfy the visualization constraints of this viewpoint.' This represents the sum of the areas of all triangular faces within the partition.

12. The robot vision automatic scanning path planning method based on a three-dimensional mesh model according to claim 9, characterized in that, The imaging quality parameter Cf θ This indicates the image quality of all triangles at that viewpoint, and the calculation formula is as follows: Cf θ =S n ·N n Among them, S n The normalized triangle area coefficient represents the proportion of the area of ​​this triangle in the total area of ​​all triangles in this partition, and θ is the angle between the vector from the center of the triangle to the candidate viewpoint and the normal vector of the triangle.

13. The robot vision automatic scanning path planning method based on a three-dimensional mesh model according to claim 1, characterized in that, Path planning is performed on all the optimal viewpoints obtained using a greedy algorithm, including: Select the robot's starting position as the starting point of the scanning path. Find the point closest to the starting point in the optimal viewpoint array composed of all optimal viewpoints and mark it as the second point of the scanning path. Then find the unmarked point closest to the second point in the optimal viewpoint array and mark it as the third point of the scanning path. Continue in this manner until all viewpoints in the optimal viewpoint array are marked. Connect the points at the robot's initial position with the reordered viewpoints in sequence to generate the shortest scanning path for the surface scanning measurement robot.

14. A robot vision-based automatic scanning path planning system based on a three-dimensional mesh model, characterized in that, include: The view frustum model generation module generates the scanning view frustum model of the scanner based on the measurement space constraints of the surface structured light scanner. The scanning area division module obtains the three-dimensional model of the workpiece by scanning with a surface structured light scanner. Through Gaussian mapping and cluster analysis, it obtains the preliminary area division of the three-dimensional model of the workpiece. Then, through the segmentation and merging of triangular facets, the three-dimensional model of the workpiece is divided into multiple scanning areas. The candidate viewpoint generation module generates candidate viewpoints for each scanned region based on the scanned viewpoint model. The optimal viewpoint acquisition module calculates the measurement quality parameters of each candidate viewpoint, evaluates the measurement quality of each candidate viewpoint, and selects the candidate viewpoint with the highest measurement quality parameters as the optimal viewpoint. The path planning module performs path planning for all the optimal viewpoints obtained using a greedy algorithm.

15. A computer device comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the steps of the method of claim 1.

Images