Wheelset painting equipment and painting method based on vision guidance

Abstract

This invention discloses a wheelset painting system based on vision guidance. The system comprises a rotary drive device, a vision guidance module, a robotic arm, a painting actuator, and a control device. Based on image information from the vision guidance module, the system extracts a set of wheelset surface coordinate points, generates a painting path, and controls the robotic arm and the painting actuator to paint the wheelset along the painting path. This system utilizes a vision guidance control system to accurately identify and track train wheelsets, automatically plan the painting path, and adjust painting parameters, achieving efficient and uniform painting results.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent industrial equipment, and in particular to a wheelset painting device and a painting method based on vision guidance. Background Art

[0002] This invention relates to a vision-guided wheelset painting device and method, designed to improve the automation level of train wheelset painting, ensure painting quality, reduce labor costs, and minimize environmental pollution. The device utilizes a vision-guided control system to accurately identify and track train wheelsets, automatically plan the painting path, and automatically adjust painting parameters, achieving efficient and uniform painting results.

[0003] The utility model patent with application number CN216094496U discloses a wheelset painting device, which includes a manipulator, a brush body, a control component and a rotary drive mechanism. The device paints according to a preset path corresponding to the wheelset model, and then drives the wheelset to rotate and contact the brush body to achieve painting.

[0004] However, since the structures of the hub, spokes, etc. on the wheelset surface are different and the surface is undulating, setting a preset path for each model of wheelset is inefficient. Painting only by the brush body abutting against the wheelset and the wheelset rotating cannot control the paint thickness and uniformity on the surfaces of different structures. Therefore, there is an urgent need for a wheelset painting device and a painting method based on vision guidance. Summary of the Invention

[0005] In order to solve the above problems, the present invention provides a wheelset painting device and a painting method based on vision guidance, which aims to solve the problems of uneven painting and low efficiency on complex surfaces of wheelsets during the painting process.

[0006] The technical solution adopted in the present invention is:

[0007] A wheelset painting device based on vision guidance comprises: a rotation drive device, a vision guidance module, a robotic arm, a painting actuator, and a control device.

[0008] Rotation drive device: used to support and drive the wheelset to rotate so as to paint from different angles;

[0009] Vision guidance module: used to capture image information of wheelsets;

[0010] Painting actuator: equipped with painting tools, performs painting operations according to the instructions of the control device;

[0011] Robotic arm: connected to the paint actuator, driving the paint actuator to move along the paint path;

[0012] Control device: receiving image information from the visual guidance module, extracting a set of coordinate points on the wheelset surface, generating a painting path, and controlling the robotic arm and the painting actuator to paint the wheelset along the painting path;

[0013] Each robotic arm is provided with a visual guidance module and a paint actuator. The visual guidance module can be hidden in the groove 10 at the end of the robotic arm through a folding mechanism or a mechanical rotation mechanism. When the paint actuator is performing a painting operation, the visual guidance module rotates and is hidden in the groove at the end of the robotic arm.

[0014] Preferably, the visual guidance module uses a binocular 3D camera installed on the robotic arm, and the robotic arm brings it to a designated position to collect images and obtain sampling points;

[0015] Preferably, the designated position includes a first designated position and a second designated position;

[0016] Preferably, the rotary drive device drives the wheelset to rotate, so that the wheelset captures image information of the wheelset at 0° and 180° respectively and paints it, thereby ensuring that the entire surface of the wheelset is covered with paint while saving the number of rotations.

[0017] Preferably, the painting tool equipped with the painting actuator is a brush;

[0018] Preferably, the control device extracts a set of wheelset surface coordinate points by filtering, reducing density, and removing discrete noise points and background from the image information;

[0019] Preferably, the control device generates a painting path by generating a bounding box, and generates a painting path by intersecting the parallel sections generated according to the vertices of the bounding box with the set of coordinate points on the wheelset surface; when the depth of the bounding box is greater than or equal to a first threshold, the painting path adopts an "O" type path, and when the depth of the bounding box is less than the first threshold, the painting path adopts a "Z" type path.

[0020] Preferably, the control device controls the robotic arm by setting a cylindrical obstacle model and an improved gravitational potential energy algorithm to perform robotic arm path control.

[0021] A wheelset painting method based on vision guidance, using any of the wheelset painting equipment described above, is characterized by comprising:

[0022] S1: The robot arm moves the visual guidance module at its output end to the first specified position and the second specified position;

[0023] S2: Use the vision guidance module to capture the image information of the wheelset;

[0024] S3: The control device receives the image information of the visual guidance module, extracts the set of coordinate points on the wheelset surface, generates a painting path, and controls the robot arm and the painting actuator to paint the wheelset along the painting path;

[0025] S4: The rotation drive device drives the wheelset to rotate a certain angle, and steps S1, S2, and S3 are repeated until all positions of the wheelset to be painted are painted;

[0026] S5: After painting is completed, the vision guidance module inspects the wheelset surface again to ensure the painting quality.

[0027] Compared with the prior art, the present invention has the following beneficial effects:

[0028] High degree of automation: greatly reduce manual operations and improve production efficiency;

[0029] Strong adaptability: suitable for wheelsets of different sizes and shapes, with good versatility;

[0030] High precision: Visual guidance eliminates the need to enter the wheelset model, size, and appearance, ensuring the accuracy of the painting position and range;

[0031] Stable quality: Automated control ensures the consistency of coating quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0033] Figure 1 Schematic diagram of a wheelset painting device based on vision guidance provided by an embodiment of the present invention;

[0034] Figure 2 This is a schematic diagram of painting path planning for different surfaces of a wheelset provided by an embodiment of the present invention;

[0035] Figure 3 This is a collision detection model for a wheelset painting device provided by an embodiment of the present invention;

[0036] Figure 4 It is a schematic diagram of the resultant force of the gravitational potential energy method provided by an embodiment of the present invention.

[0037] In the figure, 1. Frame, 2. Control device, 3. Paint can, 4. Robotic arm, 5. Vision guidance module, 6. Paint actuator, 7. Rotation drive device, 8. Second designated position, 9. First designated position, 10. Groove, 11. Wheel set, YZT, cylinder, JN, capsule body. DETAILED DESCRIPTION

[0038] The following will be combined with the accompanying drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0039] Example 1

[0040] Since the structures of the hub, spokes, etc. on the wheelset surface are different and the surface is undulating and changeable, it is impossible to paint the wheelset surface through a simple preset painting path. This embodiment provides a wheelset painting device based on vision guidance.

[0041] like Figure 1 As shown, the wheelset painting equipment based on vision guidance includes a rotary drive device 7, a vision guidance module 5, a robotic arm 4, a painting actuator 6, a control device 2, and also includes a frame 1 and a paint tank 3.

[0042] Rotation drive device 7: used to support and drive the wheelset 11 to rotate so as to paint from different angles;

[0043] Vision guidance module 5: used to capture image information of wheelset 11;

[0044] Painting actuator 6: equipped with painting tools, performs painting operations according to the instructions of the control device;

[0045] Robotic arm 4: connected to the paint actuator, drives the paint actuator to move along the paint path according to the instructions of the control device;

[0046] Control device 2: receives image information from the visual guidance module 5, extracts a set of coordinate points on the wheelset surface, generates a painting path, and controls the robotic arm and the painting actuator to paint the wheelset along the painting path.

[0047] Preferably, the visual guidance module 5 uses a binocular 3D camera, which is installed on the robotic arm 4, and is brought to a designated position by the robotic arm 4 to collect images and obtain sampling points;

[0048] Specifically, the visual guidance module 5 uses a binocular 3D camera or a structured light camera, with an accuracy requirement of ≤2% within 2m and a working range of 0.1m-10m.

[0049] Specifically, a robot arm 4 is provided on both sides of the wheelset 11, and each robot arm 4 is provided with a visual guidance module 5 and a paint actuator 6. The visual guidance module 5 can be hidden in the groove 10 at the end of the robot arm through a folding mechanism or a mechanical rotation mechanism. When the paint actuator 6 is performing the painting operation, the visual guidance module 5 rotates and is hidden in the groove 10 at the end of the robot arm 4. See the attached figure. Figure 1 Left robotic arm; when the visual guidance module 5 is performing image acquisition operation, it rotates 90 degrees from the groove 10 to be parallel to the wheel axle (x-axis direction), and the paint actuator 6 rotates 90 degrees to be perpendicular to the wheel axle (y-axis direction), see the attached Figure 1 The right robotic arm has the effect of ensuring that the visual guidance module 5 does not interfere with the paint actuator during use.

[0050] Preferably, the designated position includes a first designated position 9 and a second designated position 8, the first designated position 9 is above the outer side of the wheel, and the second designated position 8 is above the inner side of the wheel; the first designated position 9 and the second designated position 8 are in the same z-axis plane as the center of the axle; their specific positions can be determined by the wheel size, camera parameters, etc., and it is necessary to ensure that the visual guidance module 5 can clearly and completely capture the wheelset surface image at the designated position.

[0051] Specifically, first, the initial position of the wheelset is 0°, and the visual guidance module 5 obtains the image information of the wheelset at the first designated position 9 and the second designated position 8 respectively, and then performs the painting operation; then, the rotation drive device drives the wheelset to rotate, and after the wheelset rotates 180°, the visual guidance module 5 captures the image information of the wheelset at the first designated position 9 and the second designated position 8 again and performs the painting operation. The rotation operation is used to deal with the influence of the wheel axle on the image below being blocked, and ensure that the full surface of the wheelset is covered with paint on the basis of saving the number of rotations.

[0052] Preferably, the painting tool equipped with the painting actuator 6 is a brush;

[0053] Preferably, the robotic arm 4 is a six-axis robotic arm.

[0054] Preferably, the control device 2 extracts a set of wheelset surface coordinate points by performing through-filtering processing, cube density reduction processing, discrete noise points and background removal on the image information;

[0055] Specifically, the control device receives image information from the visual guidance module 5. To reduce the computational complexity of subsequent processing, the initial sampling point data is preprocessed. This preprocessing process reduces the density of sampling points while preserving existing features to speed up subsequent processing. The preprocessed standardized sampling points are then refined to optimize their surface information.

[0056] Specifically, the initial sampling point data may contain noise and outliers due to uncertain background and production conditions. This redundant information can affect the subsequent calculation speed and path planning process. Furthermore, by combining the through-filter algorithm with the rotary drive device, selecting an appropriate range in the z direction ensures that the paint can cover the entire wheelset. This redundant information is removed using the through-filter algorithm. The specific steps are as follows:

[0057] By setting the value range of x, y, and z directions, the required sampling point information can be filtered out. Suppose the sampling point data is (x i ,y i ,z i )∈P,x i ,y i ,z i Represents the coordinates of the i-th point. The ranges of the sampling points in the three directions are [x min ,x max ],[y min ,y max ],[z min ,z max ], the sampling point through-filter calculation is as follows:

[0058] P′={(x i ,y i ,z i )∈P|x min ≤x i ≤x max ,y min ≤y i ≤y max ,z min ≤z i ≤z max}

[0059] After completing the straight-through filtering, due to the huge amount of sampling point data and the fact that the paint actuators for painting operations often have a certain width, overly dense sampling points actually increase the computational complexity for path planning. Properly reducing the density of sampling points can improve the efficiency of the painting path and streamline the data that needs to be processed for subsequent calculations, speeding up the operation. However, reducing the density of sampling points can also lead to the loss of some contour information. Existing technologies often use curvature calculation methods to reduce density. This method estimates the curvature of the sampling point by calculating the angle between the normal line of the sampling point and the adjacent points. Although the operation is simple, the boundary between the background and the target and some relatively isolated edge information will be overly de-densified.

[0060] This application proposes to use cube density reduction processing. First, the sampling points are divided into a group of small cubes, and then the center point formula is used to calculate the position of the center point of the small cube, as shown in the following formula:

[0061]

[0062] Among them, zx max ,zx min ,zy max ,zy min ,zz max ,zz min Respectively represent the minimum and maximum values ​​of the x, y, and z coordinates of all sampling points in a small cube.

[0063] Finally, the calculated center point of the small cube replaces all sampling points within the group of small cubes to achieve density reduction. Cube density reduction is good at preserving edge information and also has a certain smoothing effect on curvature, which can improve the smoothness of the painting process.

[0064] After de-densification, the sampling point data collected may contain some discrete noise points due to equipment and environmental factors. These discrete noise points will have a significant impact on subsequent sampling point processing tasks. If left unprocessed, they may be included in trajectory planning during the painting process.

[0065] Specifically, this application adopts Gaussian filtering to form a Gaussian distribution graph by analyzing the average distance value between each sampling point and the adjacent k points in the sampling point set, calculating its mean and standard deviation, and analyzing whether the sampling point is a discrete noise point outside the standard.

[0066] Let the Nth point in the sampling point correspond to the x, y, z coordinates as X n , Y n , Z n The distance P from this point to any point m , as shown below:

[0067]

[0068] Calculate the mean μ and standard deviation σ of the distance between each point and the adjacent k points,

[0069]

[0070] A threshold M is set based on the standard deviation of the average distance of the sampling points, and sampling points with a standard deviation greater than the threshold M are deleted.

[0071] After the filtering, density reduction, and point removal processes described above, only the surface of the workbench remains as useless information. The prior art typically uses the least squares method to fit a plane. However, since the distribution of sampling points is not completely random, this significantly affects the accuracy of the least squares method. Therefore, this application uses a ground filtering algorithm for background removal. Flipping the sampling point, assuming a piece of cloth covers the bottom of the sampling point, the area covered by the cloth is the ground background area.

[0072] The basic formula for simulating cloth covering is as follows:

[0073]

[0074] Where m is the mass parameter of the cloth particle, X is the simulated position of the cloth particle at time t, and F ext (X, t) represents external driving factors (gravity, collision, etc.), F int (X,t) represents the internal driving factors (internal connections between particles).

[0075] The position of a cloth particle is determined by both external and internal driving factors. Assuming only external factors exist and the internal factors are 0, the position of the next cloth particle movement can be calculated by giving a value of Δt:

[0076]

[0077] Where m is the mass parameter of the cloth particle, G is a constant, and Δt is the time step.

[0078] The inversion phenomenon of the vacant part in the sampling point plane is controlled by controlling the movement relationship of two adjacent particles. That is, if both particles can move, they will move the same distance in opposite directions at the same time. The movement distance is as follows:

[0079]

[0080] Where d is the displacement of the particle; when the particle is movable, b is equal to 1; when it is immovable, b is equal to 0; p i For p o The adjacent particles of n are normalized to the unit vector in the vertical direction of the sampling point.

[0081] The specific steps of the ground filtering algorithm are as follows:

[0082] 1) Reverse the sampling point data, set the mesh size of the simulated cloth as required and set it above the highest point;

[0083] 2) Project the inverted sampling points and cloth particles to the same horizontal plane, find the adjacent points of the cloth particles, and record the intersection height value.

[0084] 3) By comparing the height value of the nearest neighbor point of the cloth particle with the recorded intersection height value, determine whether the particle needs to be moved. If the height value of the nearest neighbor point is less than or equal to the intersection height value, the height value of the particle is set to the relative height value and it is set to be immovable.

[0085] 4) Calculate the displacement imposed on each cloth particle by its surroundings based on the internal driving factors in the cloth simulation.

[0086] 5) Repeat (3) and (4) until the maximum height change of all cloth particles is small enough or reaches a preset iteration threshold to ensure that the position of the particles is stable.

[0087] 6) Calculate the height difference between the sampling point and the cloth particle. Based on a pre-set threshold, determine the height difference between the sampling point and the cloth particle. If the height difference is less than the threshold, the point is marked as a ground point; otherwise, it is marked as a non-ground point. After deleting the ground point, extract the set of wheel set surface coordinate points.

[0088] Preferably, the control device generates the painting path by generating a bounding box, and generates a painting path by intersecting the parallel sections generated by the vertices of the bounding box with the set of coordinate points on the wheelset surface. When the depth of the bounding box is greater than or equal to a first threshold, the painting path adopts an "O" type path, and when the depth of the bounding box is less than the first threshold, the painting path adopts a "Z" type path, see the attached Figure 2 Planning of painting paths for different wheelset surfaces.

[0089] Specifically, the bounding box method generates a minimum bounding box for a model. Commonly used methods include AABB and OBB. The OBB method, also known as an oriented bounding box, provides good enclosed space for various objects and offers higher real-time performance than the AABB method, making it more suitable for wheelsets with uncertain shapes.

[0090] Once the bounding box type is determined, a series of parallel and equidistant slices can be generated based on the vertices of the bounding box. The resulting intersections are the desired trajectory lines, which can be obtained using projection or intersection methods.

[0091] This application takes into account the characteristics of the wheelset having convex and flat parts. According to the characteristics of different parts, two different path trajectories are used when generating the path trajectory of the trajectory line. A "Z"-type path trajectory is used for the flat part, and an "O"-type path trajectory is used for the convex part. Because the "Z"-type path trajectory consists of straight lines and broken line segments, it has a simple structure and is easy to calculate and generate. When the surface changes of the part are relatively gentle, it can reduce the travel distance and time of the paint actuator; the "O"-type path trajectory has better adaptability, the speed change of the robot arm 4 is small, it can reduce movement vibration and loss, and is more suitable for convex parts.

[0092] Specifically, when the bounding box depth is greater than or equal to a first threshold, the painting path adopts an "O"-shaped path. When the bounding box depth is less than the first threshold, the painting path adopts a "Z"-shaped path. By selecting the appropriate painting trajectory plan based on the different surface characteristics of the part, the efficiency and painting effect of the painting operation can be improved.

[0093] Preferably, the control device controls the robotic arm 4 by setting a cylindrical obstacle model and an improved gravitational potential energy algorithm to perform path control of the robotic arm 4.

[0094] Specifically, as the working environment and motion trajectory of the robot arm 4 increase in complexity during operation, the likelihood of interference and collision between the robot arm 4 and its surroundings, as well as between the robot arm 4 and its own connecting rods, increases. Such collisions can damage the robot arm 4, causing property damage and even threatening the operator's personal safety. During the wheelset painting process, since the wheelset primarily consists of wheels and axles and has a relatively fixed structure, the robot arm 4 and obstacles are simplified to reduce the complexity of the collision detection algorithm.

[0095] Since the wheels and axles are both cylindrical, they are wrapped in a cylindrical shape as obstacles. The capsule JN commonly used in the field is used to wrap the four-link robot arm, and the collision detection problem between the two is converted into a collision problem between the capsule JN and the cylinder YZT. Figure 3 The collision detection model for the wheelset painting equipment shown.

[0096] Preferably, the specific coordinate position of the cylinder YZT can be obtained through the set of coordinate points on the aforementioned wheelset surface, and the size of the cylinder YZT can reserve a certain margin compared to the size of the wheelset surface to improve the reliability of the collision calculation; collision detection is performed by calculating the shortest distance between the capsule body JN surface and the cylinder YZT surface, which has the effect of simplifying the collision calculation model.

[0097] Gravitational potential energy algorithm is a classic algorithm in robot arm path planning. It realizes path planning by constructing a virtual potential field. Figure 4 The diagram of the resultant force of the gravitational potential energy method is shown. The motion space of the robot arm is regarded as a virtual potential field composed of a repulsive potential field and an attractive potential field. The overall repulsive potential field is composed of the repulsive field of each obstacle in the workspace. The direction of the repulsive potential field is away from the obstacle, while the attractive potential field is generated by the target position and is directed toward attracting the robot arm 4 to the target position. Therefore, by setting the target position as the lowest potential field, theoretically, the movement of the robot arm 4 will eventually reach this position. According to the potential energy gap between the current position of the robot arm 4 and the target position, the movement direction and speed of the robot arm 4 are calculated, and the path planning of the robot arm 4 can be finally realized.

[0098] This application improves the traditional gravitational potential energy algorithm. To solve the problem that the robot arm 4 is far away from the target position and generates a large attraction, which may cause a collision, a maximum distance limit d is set between the robot arm 4 and the target position. max ,δ is the attraction factor, d(p,p g ) is the position p of the robot 4 to the target position p g The distance between them. The modified attractive potential function is:

[0099]

[0100] The attraction function is:

[0101]

[0102] In order to solve the problem that the target position and the obstacle position are close to each other, the repulsive force potential field function is modified according to the relative distance between the robot arm 4 and the target position. The improved repulsive force potential field function is:

[0103]

[0104] Where μ is the repulsive force factor, d s represents the exclusion range of each obstacle, d(p,p o ) is the minimum distance between the robot arm 4 and the obstacle, d(p,p g ) is the distance between the robot arm 4 and the target position, and n is a positive coefficient. The repulsive force function is to find the partial derivative along the obstacle and the target position respectively:

[0105]

[0106]

[0107] i o Represents the unit direction vector from the obstacle to the robotic arm 4, i g Represents the unit direction vector of the robot arm 4 to the target position.

[0108] Therefore, the total force applied to the robot arm 4 is:

[0109] F to (p) = F at (p)+F re (p)

[0110] This application completes the obstacle avoidance path planning of the robot arm 4 in the task space by setting a cylindrical obstacle model and an improved gravitational potential energy algorithm to control the path of the robot arm 4.

[0111] A wheelset painting method based on vision guidance, using any of the wheelset painting equipment described above, is characterized by comprising:

[0112] S1: The robot arm moves the visual guidance module at its output end to the first specified position and the second specified position;

[0113] S2: Use the vision guidance module to capture the image information of the wheelset;

[0114] S3: The control device receives the image information of the visual guidance module, extracts the set of coordinate points on the wheelset surface, generates a painting path, and controls the robot arm and the painting actuator to paint the wheelset along the painting path;

[0115] S4: The rotation drive device drives the wheelset to rotate a certain angle, and steps S1, S2, and S3 are repeated until all positions of the wheelset to be painted are painted;

[0116] S5: After painting is completed, the vision guidance module inspects the wheelset surface again to ensure the painting quality.

[0117] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A wheelset painting device based on vision guidance, comprising: Rotary drive device, visual guidance module, robotic arm, paint actuator, control device, among which, The rotary drive device is used to support and drive the wheelset to rotate so as to paint from different angles; The visual guidance module is used to capture image information of the wheelset; The robotic arm is connected to the paint actuator to drive the paint actuator to move along the paint path; The painting actuator is equipped with a painting tool and performs painting operations according to the instructions of the control device; The control device receives image information from the visual guidance module, extracts a set of coordinate points on the wheelset surface, generates a painting path, and controls the robotic arm and the painting actuator to paint the wheelset along the painting path; The robotic arm is provided with the visual guidance module and the paint execution mechanism. The visual guidance module is connected to the groove at the end of the robotic arm in a concealed manner through a folding mechanism or a mechanical rotation mechanism. When the paint execution mechanism is performing a painting operation, the visual guidance module is rotated and hidden in the groove at the end of the robotic arm. The control device generates a painting path by generating a bounding box, and generates a painting path by intersecting the parallel sections generated by the vertices of the bounding box with the set of coordinate points on the wheelset surface; when the depth of the bounding box is greater than or equal to a first threshold, the painting path adopts an "O" type path, and when the depth of the bounding box is less than the first threshold, the painting path adopts a "Z" type path.

2. The wheelset painting device based on vision guidance according to claim 1, characterized in that: The visual guidance module uses a binocular 3D camera installed on the robotic arm, and the robotic arm brings it to a designated position for image acquisition to obtain sampling points.

3. The wheelset painting device based on vision guidance according to claim 2, characterized in that: The designated positions include a first designated position and a second designated position; the rotation drive device drives the wheelset to rotate, so that the wheelset captures image information of the wheelset at 0° and 180° respectively and paints it, corresponding to the visual guidance module at the first designated position and the second designated position respectively.

4. The wheelset painting device based on vision guidance according to claim 1, characterized in that: The control device extracts a set of wheelset surface coordinate points by performing filtering processing, density reduction processing, discrete noise point and background removal on the image information.

5. The wheelset painting device based on vision guidance according to claim 4, characterized in that: The filtering process adopts straight-through filtering, and its calculation formula is as follows: P′={(x i ,y i ,z i )∈P|x min ≤x i ≤x max ,y min ≤y i ≤y max ,z min ≤z i ≤z max } Among them, the sampling point data is (x i ,y i ,z i )∈P,x i ,y i ,z i They represent the coordinates of the i-th point, and the ranges of the sampling points in the three directions are [x min ,x max ],[y min ,y max ],[z min ,z max ].

6. The wheelset painting device based on vision guidance according to claim 5, characterized in that: The density processing adopts cube density reduction processing. First, the processed sampling points are divided into a group of small cubes. Then, the center point formula is used to calculate the position of the center point of the small cube, as shown in the following formula: Among them, zx max ,zx min ,zy max ,zy min ,zz max ,zz min They represent the minimum and maximum values ​​of the x, y, and z coordinates of all sampling points in a small cube, respectively. Finally, the calculated center point of the small cube replaces all sampling points in the group of small cubes.

7. The wheelset painting device based on vision guidance according to claim 1, characterized in that: The control device controls the robotic arm by setting a cylindrical obstacle model and an improved gravitational potential energy algorithm to control the path of the robotic arm.

8. The wheelset painting device based on vision guidance according to claim 7, characterized in that: The improved gravitational potential energy algorithm is specifically as follows: Sets the maximum distance limit d between the robot arm and the target position max ,δ is the attraction factor, d(p,p g ) is the robot position p to the target position p g The distance between The attraction function is: The repulsive force function is: Where μ is the repulsive force factor, d s represents the exclusion range of each obstacle, d(p,p o ) is the minimum distance between the robot arm and the obstacle, d(p,p g ) is the distance between the robot arm and the target position, n is a positive coefficient, i o Represents the unit direction vector from the obstacle to the robotic arm, i g The unit direction vector representing the robot arm to the target position; The total force applied to the robot arm is: F to (p)=F at (p)+F re (p)。 9. A wheelset painting method based on vision guidance, using the wheelset painting device according to any one of claims 1 to 8, characterized in that: include: S1: The robot arm moves the visual guidance module at its output end to the first specified position and the second specified position; S2: Use the vision guidance module to capture the image information of the wheelset; S3: The control device receives the image information of the visual guidance module, extracts the set of coordinate points on the wheelset surface, generates a painting path, and controls the robot arm and the painting actuator to paint the wheelset along the painting path; S4: The rotation drive device drives the wheelset to rotate a certain angle, and steps S1, S2, and S3 are repeated until all positions of the wheelset to be painted are painted; S5: After painting is completed, the vision guidance module inspects the wheelset surface again to ensure the painting quality.

Images