Automatic coating trajectory planning method for composite material part

Abstract

The invention particularly relates to an automatic coating trajectory planning method for a composite material workpiece. The method comprises the following steps that 1, robot zero point calibration, 3D camera and robot pose relation calibration, multi-robot pose relation calibration, workpiece displacement mechanism and robot pose calibration and spray gun and robot tail end pose relation calibration are conducted; 2, a paint spraying robot carrying the 3D camera is controlled to carry out iterative shooting and point cloud fusion on the workpiece, a complete working three-dimensional point cloud model is obtained, and the real edge of the complete working three-dimensional point cloud model is recognized; and 3, trajectory planning is carried out, the trajectory planning is carried out in two parts, namely complex curved surface path planning and boundary trajectory planning, and the two parts of planning are fused to finally form a profiling migration trajectory of the robot driving the spray gun. And a track planning mode is adopted, so that track planning is more complete and fine, and the spraying quality of the material edge is improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of automatic spraying, and particularly relates to a composite part automatic coating track planning method. BACKGROUND

[0002] In recent years, with the rapid development of machine vision and artificial intelligence technologies, robots are increasingly widely applied in the field of industrial spraying. At present, the main track planning methods are manual drag teaching method and offline programming method. The drag teaching method refers to that a skilled operator completes a spraying process by manually dragging the end of the robot, and records the teaching track, so that the robot automatically repeats the teaching track to complete the spraying operation. The drag teaching method is simple in process, but the spraying quality is highly dependent on the operator's technology and experience, and is easily affected by human factors. Moreover, in the spraying teaching process, the teacher is exposed to the spraying environment, and the health is easily damaged. The offline programming method is based on the CAD model of the workpiece to be sprayed, and combines specific spraying material deposition model, constraint conditions and optimization criteria to complete the spraying track planning. This method can effectively avoid the influence of operator's subjective factors on the spraying quality, and effectively ensure the spraying quality. However, this method has problems such as complicated operation of making the CAD model of the workpiece, long time consumption, and difficulty in establishing the workpiece model, which has a great influence on the working efficiency of the spraying system. The method of using a visual sensor to obtain the surface information of the workpiece to be sprayed to plan the spraying track of the workpiece has only been developed in the past 20 years. FlexPaint of European Spraying Organization first began to study the use of a visual sensor to obtain the surface information of the workpiece to be sprayed to plan the spraying track of the workpiece, in which a laser sensor is used to obtain data. Later, Hoegg et al. used three TOF cameras to realize three-dimensional reconstruction, providing data for path planning. Ye Xiao et al. used three-dimensional laser scanning technology and least square method to realize three-dimensional workpiece recognition and track planning. At present, the spraying track planning based on visual recognition has been applied to the fields of furniture board and civil aircraft sheet metal part painting.

[0003] However, the helicopter composite component has more complex shape, more free-form surfaces and more part styles, and so far, there has been no report on automatic recognition and automatic painting of composite parts. SUMMARY

[0004] The purpose of the present application is: The aviation composite component is mainly made by manual operation, which depends on the spraying experience of the operator. The manual spraying of complex components has poor uniformity of paint film, and is prone to process defects such as pinholes, vertical flow, hanging particles and orange peel, which is difficult to meet the requirements of the current aviation manufacturing "blowout period" for the coating quality and efficiency of the composite component. The present application breaks through the key technologies of three-dimensional intelligent recognition, track flexible planning, contour uniform speed tracking and operation space calibration, and forms a composite part automatic coating track planning method.

[0005] The technical scheme of the present application is: According to the first aspect of the present application, a composite material part automatic coating track planning method is proposed, which is applied to the layout of "double spray painting robot + workpiece displacement mechanism". The workpiece displacement mechanism is arranged between the two spray painting robots, and is used to clamp the workpiece to be sprayed and has multi-dimensional translation freedom in the plane. The spray gun is connected to the end of the spray painting robot through a spray gun displacement mechanism, and the 3D camera is rigidly connected to the robot. The automatic coating track planning method comprises the following steps: Step 1: robot zero point calibration, 3D camera and robot pose relationship calibration, multi-robot pose relationship calibration, workpiece displacement mechanism and robot pose calibration, and spray gun and robot end pose relationship calibration. Step 2: control the spray painting robot carrying the 3D camera to iteratively shoot and fuse the point cloud of the workpiece, obtain the complete working three-dimensional point cloud model and identify the real edge thereof. Step 3: trajectory planning, which is divided into two parts, namely complex curved surface path planning and boundary trajectory planning, and the final profiled walking trajectory of the robot driving the spray gun is formed by fusion of the two parts.

[0006] In one possible embodiment, in the step 1, the 3D camera and robot pose relationship calibration specifically comprises the following process: The camera is carried by the robot to shoot a fixed calibration board from multiple different poses, and based on the consistency of the feature points of the calibration board described in the robot base coordinate system, the transformation matrix of the camera coordinate system to the robot end coordinate system is iteratively optimized and solved.

[0007] In one possible embodiment, in the step 1, the pose relationship between the multiple robots specifically comprises the following process: taking the base coordinate system of the first robot as the main coordinate system, shooting the same fixed calibration board through the camera at the end of each robot, and using the camera pose estimation result, the pose relationship of the base coordinate system of the second robot relative to the main coordinate system is calculated through coordinate transformation.

[0008] In one possible embodiment, in the step 1, the pose relationship between the spray gun and the robot end specifically comprises the following process: the spray gun head is made to contact a fixed point in space in different poses, the robot end pose is recorded, and the position parameters of the spray gun coordinate system relative to the robot end coordinate system are solved by least squares fitting.

[0009] In one possible embodiment, in the step 2, it specifically comprises the following process: S21, initial shooting and edge extraction: control the robot to a preset initial pose to shoot the workpiece, extract the initial point cloud and edge point set P of the workpiece; S22, edge segmentation and re-shooting pose calculation: sequentially segment the edge point set P into m sub-edges, calculate the point set mean point s' of each sub-edge, and fit a plane in the neighborhood of s' to obtain a normal vector η, and calculate the re-shooting pose according to s', η and a preset optimal shooting distance D; S23, iterative re-shooting and model fusion: control the robot to shoot according to the calculated re-shooting pose, fuse the newly obtained point cloud with the existing model, and obtain an updated edge point set P'; S24, edge verification and iteration termination judgment: compare the edge point set P with P', if there is a point in P' whose distance from the point in P is less than a set threshold β, then eliminate the point from P'; if P' is not empty after elimination, then take P' as new edge information, and return to step S22 for the next round of re-shooting; if P' is empty or other termination conditions are met, then obtain a complete workpiece three-dimensional point cloud model.

[0010] In one possible embodiment, in the step 3, the complex curved surface path planning comprises: Surface segmentation based on feature vectors: calculate the feature vectors of each point in the point cloud model, and based on the region growing method, classify the points with a feature vector included angle less than a threshold θ0 into the same surface sheet to achieve the segmentation of the complex curved surface; Adaptive solving of TCP trajectory points: perform main slicing on the point cloud model according to the spraying gun spacing d, and perform region slicing according to the spraying characteristics; for each main slicing region, fit a plane using the adjacent point cloud data thereof; calculate the intersection of the main slicing plane, the region slicing plane and the fitted plane as the position of the TCP trajectory point, and take the normal vector of the fitted plane as the pointing direction of the spraying gun at the trajectory point.

[0011] In one possible embodiment, in the step 3, the boundary trajectory planning in step 3 comprises: Boundary point identification: for any point P0 in the point cloud, obtain K-neighbor points thereof and project them to the fitted plane; analyze the distribution included angles of the neighbor points with respect to P0 in the projection plane; if the sum θsum of the included angles between the adjacent vectors is less than a set threshold θa, then determine that P0 is a boundary point; Boundary point spraying gun posture determination: take the normal vector η of the fitted plane of the neighbor points of the boundary point as the reference, rotate η by a fixed angle θ around the rotation axis K determined by the direction of the line connecting the front and rear boundary points, and take the resulting direction vector η' as the pointing direction of the spraying gun at the boundary point.

[0012] According to the second aspect of the present application, an automatic painting system is provided for implementing the above method, comprising: At least two paint spraying robots; A displacement mechanism arranged between the paint spraying robots for clamping and translating the workpiece; 3D camera, rigidly fixed on the end effector of the paint spraying robot; Spray gun, rigidly fixed on the end effector of the paint spraying robot; Control system, in communication connection with the paint spraying robot, the displacement mechanism, the 3D camera, for executing the system calibration, the three-dimensional intelligent perception and the trajectory planning algorithm, and controlling the paint spraying robot and the displacement mechanism to move cooperatively.

[0013] In one possible embodiment, the displacement mechanism has at least one translational degree of freedom, for expanding the effective working space of the paint spraying robot, so as to realize full-coverage spraying of large workpieces.

[0014] In one possible embodiment, the 3D camera is coaxially or adjacently fixed on the same rigid connection plate with the spray gun, and the connection plate is installed on the robot end flange.

[0015] According to a third aspect of the present application, a computer readable storage medium is provided, which stores a computer program, and the computer program is executed by a processor to implement the steps of the above-mentioned automatic coating trajectory planning method.

[0016] According to a fourth aspect of the present application, an electronic device is provided, which comprises a memory, a processor and a computer program stored on the memory and executable on the processor, and the processor implements the steps of the above-mentioned automatic coating trajectory planning method when executing the computer program.

[0017] The present application has the following advantages: (1) Multi-dimensional precision guarantee: 3D vision + laser tracking composite calibration is adopted to realize the absolute positioning precision of the robot end; (2) Precise dynamic compensation mechanism: through the space monitoring of more than 12 points, multi-robot cooperative calibration is adopted, and a cross double-sided vision system is used, so that the pose precision is ≤0.03mm.

[0018] (3) More complete trajectory planning: the trajectory planning mode is adopted, so that the trajectory planning is more complete and delicate, and the spraying quality of the material edge is improved. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1 Robot working station schematic diagram; Among them: 1-6-axis robot 2- spray gun displacement mechanism 3- workpiece to be sprayed 4- workpiece displacement mechanism 5- spray gun 6- 3D camera Figure 2 Working space calibration process; Figure 3 Boundary comparison diagram; Figure 4 Segmentation diagram; Figure 5 Schematic diagram for obtaining trajectory points; Figure 6 Point cloud slice diagram; Figure 7 Schematic diagram of boundary points and interior points; Figure 8 Diagram of the included angle; Figure 9 Schematic diagram of boundary extraction for flat panel components. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] The features and illustrative embodiments of various aspects of the present invention will now be described in detail. Numerous specific details are set forth in the following detailed description to provide a thorough understanding of the invention. However, it will be apparent to those skilled in the art that the invention may be practiced without requiring some of these specific details. The following description of embodiments is merely intended to provide a better understanding of the invention by illustrating examples of the invention. The invention is by no means limited to any specific setups and methods set forth below, but covers any improvements, substitutions, and modifications to structures, methods, and devices without departing from the spirit of the invention. Well-known structures and techniques are not shown in the drawings and the following description to avoid unnecessarily obscuring the invention.

[0022] It should be noted that, unless otherwise specified, the embodiments of the present invention and the features thereof can be combined with each other, and the various embodiments can be referenced and cited in each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0023] To achieve precise and automated coating trajectory planning, it is necessary to combine the layout of spraying equipment, the calibration of the work space, three-dimensional intelligent sensing technology, and flexible planning technology for the spraying trajectory of composite material components.

[0024] 1. Spraying Equipment Layout: This spraying equipment includes a 3D camera, a robot, a positioning mechanism, and a spray gun. To achieve complete perception and spraying of large composite components (4m×3m×2m) and ensure full coverage of the spraying area for large parts, a "dual robot + positioning mechanism" layout is adopted. The positioning mechanism is placed between two spraying robots and can move horizontally. The 3D camera and spray gun are fixed to the end effector of the spraying robot, and the camera and robot are rigidly connected. For details, please see [link / details]. Figure 1 Schematic diagram of the robot's workstation.

[0025] 2. Workspace calibration process (e.g.) Figure 2 As shown): 1) Robot zero-point calibration: Record the relative position of the robot's initial position to the chessboard grid, and calibrate the robot's zero point through iterative solution of small neighborhoods. 2) Calibration of the pose relationship between the 3D camera and the robot: Using the principle that the object-image space transformation matrix equals the robot's object-image coordinates without changing their coordinates, the transformation matrix between the robot space and the camera space is calculated to construct the pose link of the painting system coordinate system. The specific calculation method is as follows: The camera coordinate system and the robot coordinate system are transformed using the following formula:

[0026] Wherein, rotation matrix There are three angular parameters, translation matrix There are three positional parameters.

[0027] The robot takes photos of a stationary calibration plate from at least 12 different positions and orientations, recording the robot's end effector data and joint angle data for each photo. By calibrating the camera, the camera's extrinsic parameters are obtained, from which the description of a point on the calibration plate relative to the normal vector in the camera coordinate system can be extracted. and ,Give A rough initial measurement transforms the description of a point on the calibration plate and its normal vector in the camera coordinate system into a description of the same point on the calibration plate and its normal vector in the robot base coordinate system, as given by the following two formulas:

[0028]

[0029] Calculate separately , The standard deviation, based on an iterative method, is... , standard deviation The partial derivatives of the three angle parameters and the three position parameters are close to zero as a convergence condition, and the relationship between the pose of the 3D camera and the robot is obtained.

[0030] 3) Position relationship calibration between multiple robots: The position relationship between multiple robots is established by using a 3D camera fixed to the end of the robot to photograph a stationary calibration plate.

[0031] Let robot 1's base coordinate system be the principal coordinate system. The goal is to find the pose relationships between the other robots' base coordinate systems and the principal coordinate system. There are two robots, with robot 1's base coordinate system being the principal coordinate system. The pose relationship between the camera coordinate system and the robot's end effector coordinate system is known as follows: , The calibration board is fixed in a position where both robots can take pictures, and the two robots photograph the calibration board. The pose relationship between the calibration board coordinate system and the two camera coordinate systems is obtained based on the camera pose estimation method. , Therefore, the pose relationship between the base coordinate system of robot 2 and the base coordinate system of robot 1 is as follows.

[0032]

[0033] 4) Workpiece Positioning Mechanism and Robot Pose Calibration: The workpiece positioning mechanism has one translation axis that moves the workpiece. The calibration plate is rigidly connected to the positioning mechanism. The robot end effector is fixed at a certain position and photographs the calibration plate. The calibration plate is then moved a certain distance along the translation axis and stopped, and the end effector's values ​​are recorded. This process is repeated multiple times. The pose of the calibration plate relative to the 3D camera is obtained using a camera pose estimation method. A corner point on the calibration plate obtained from multiple photographs is then extracted. And pass the obtained corner coordinates through By transforming to the robot's base coordinate system, straight line fitting is performed on these corner points to obtain the relationship between the translation axis and the robot's posture.

[0034] 5) Spray gun and robot end-effector pose calibration: The painting and robot end-effector pose localization consists of two parts—position calibration and attitude calibration. Position calibration involves making the spray gun head contact a fixed spatial point from different directions to ensure that the spray gun head position remains unchanged. Record the end-effector data for all calibration points. The pose relationship between the spray gun coordinate system and the robot base coordinate system is as follows:

[0035]

[0036] The least squares fit was used to obtain the results. , that is, the position parameters of the spray gun coordinate system relative to the robot end coordinate system. The pose calibration can directly use the default value or be obtained through the end pose. That is, the X-axis, Y-axis, and Z-axis directions of the spray gun coordinate system are the Z-axis, X-axis, and Y-axis directions of the robot end coordinate system respectively.

[0037] 3. Three-dimensional intelligent perception technology: Preset an initial pose for the robot that is directly facing the center of the workpiece surface and has a spacing of 1 m, and guide the robot to take an initial shot of the workpiece. Rely on the region growing algorithm based on eight neighborhoods to extract the point cloud and edges of the workpiece. After sorting the edge points extracted from the shot, store them in the set , where .

[0038] From the edge point set , sequentially intercept a sub-point set with a length of , , where b < n. In this way, the workpiece edge is divided into m parts, and m is obtained by the following formula

[0039] where, is the total length of the workpiece edge.

[0040] Then, taking the m sub-edges after segmentation as a reference, conduct a round of m supplementary shots of the workpiece. The shooting pose is determined by the following method: For the mth supplementary shot, use the sub-point set as auxiliary information, and obtain the mean coordinate point of the point set through the following formula, and obtain the point cloud of the points on the workpiece within a certain neighborhood of the point. Then fit a plane to obtain the plane normal vector η. Translate the point a certain distance D (D is determined by the best shooting distance of the camera) along η in the direction of the robot body. Obtain the pre-shooting position s of the camera through the formula, and use η as the pointing direction of the camera optical axis. The default roller angle is zero degrees. Thus, determine the supplementary shooting pose of the camera.

[0041]

[0042]

[0043] Fuse the workpiece point cloud extracted after each supplementary shot with the previous workpiece point cloud. Finally, obtain a new workpiece point cloud model after this round of supplementary shooting is completed, and extract the edge point set of the new point cloud model.

[0044] The edge point set The workpiece edge point set extracted after reshooting Compare and calculate sets Each point in the set The distance to each point in the set is such that when the distance is less than a threshold β (β is determined by camera accuracy and registration accuracy), the point is selected from the set. The point is removed from the set and considered a true edge point, still existing after re-evaluation. After one round of retrieval and filtering, the set is... The set of points that were removed in the initial image represents the true edge line of the workpiece. For example... Figure 3 As shown. If the point set All points were removed, meaning the point cloud model stopped growing after the reshoot, indicating that the workpiece point cloud model was complete. If, after filtering and removal, the set... If not empty, then according to the set after removal. The edge point information in the image is then used to perform a new round of re-photographing according to the method described above, until the workpiece geometric information is completely acquired.

[0045] 4. Trajectory Planning: Currently, large aerospace components can be classified into four categories based on their shape: box-shaped, free-form surface, tube-shaped, and flat plate-shaped.

[0046] In trajectory planning technology, the essence is to obtain the pose of a series of path points on the surface of the workpiece model, and then drive the robot to move the spray gun to follow the shape of the workpiece surface. In this process, the uniformity and gloss of the paint film sprayed on the workpiece surface are ensured.

[0047] Trajectory planning is divided into two parts: path planning for complex surfaces and boundary trajectory planning.

[0048] 1) Path planning for complex surfaces: For complex surfaces such as box-shaped or freeform workpieces, the normal vector of the tangent plane at each discrete point on the surface points to different directions. Therefore, the normal vector of the fitted plane is obtained by fitting a plane with each discrete point and the points in its neighborhood. This is used as the feature vector of each discrete point, and the segmentation of complex surfaces is achieved based on this feature vector.

[0049] First, calculate the feature vector of each point in the workpiece model. Take any point in the workpiece model. ,Depend on The nearest neighbor algorithm finds the neighbors that are closest to the target point. Points , by point set The quadratic surface can be fitted using the least squares method, therefore the point The normal vector of the tangent plane in this quadratic surface It can be obtained from the following formula:

[0050] in The equation for the quadratic surface is given, and by traversing the point cloud model, the feature vector of each point can be obtained.

[0051] Given the feature vectors of each point, specify any point. Let be the seed point, and its eigenvector be... The seed point is calculated using the following formula. eigenvectors With judgment point eigenvectors The included angle :

[0052] assumed , When two points lie on the same surface, the angle between their normal vectors is less than 1. ,when When the seed point and the decision point are on the same surface, they are included in a surface. By doing so, all points adjacent to the seed point can be searched to obtain a simple surface.

[0053] Next, find the points that have not yet been successfully connected as new seed points, and repeat the above steps until all points have been connected into a patch, such as... Figure 4 The segmentation diagram is shown.

[0054] 1) Technical solutions for obtaining trajectory points Considering the different paint film flow properties on different workpiece surfaces, such as the fact that pipe parts are more prone to paint flow problems than flat parts, and that paint flow is directly related to the pressure gun spacing d in trajectory planning, the algorithm for solving trajectory points was implemented to automatically solve reasonable TCP trajectory points based on different pressure gun spacings d.

[0055] To ensure the accuracy of the planned trajectory points, the workpiece is cut using horizontal and vertical tangent planes, dividing it into a series of small, approximately planar surfaces. The intersection points of the plane fitted by these small surfaces with the two tangent planes are then used to obtain the TCP trajectory points close to the workpiece surface. Figure 5 As shown.

[0056] Construct bounding boxes for point cloud data, based on the gun spacing in the self-tuning spraying parameters. Assuming the maximum span of the point cloud model in the slice direction is The number of slice layers can be determined by... To obtain.

[0057] Determine the number of slice layers and slice the point cloud. Separate the point cloud data and determine the slice region for each coordinate point, such as... Figure 6As shown. Along the slicing direction, odd-numbered slices are selected sequentially as the plane containing the path. Two adjacent slices of the odd-numbered slices and two point cloud data regions between these three slices are selected. If one side of an odd-numbered slice has no adjacent slices, only the slices and point cloud data of the other side are selected. The selected regions are sliced ​​according to the trajectory direction. The spacing between slices is determined based on the minor axis length of the elliptical cone torch, determining the number of slice layers. The point cloud data in the slice region is then separated. The fitting plane equation is obtained by sequentially selecting region slices and their adjacent point clouds. The intersection of the main slice plane, the region slice plane, and the fitting plane, obtained through the following equation, is the location of the TCP trajectory point:

[0058] The normal vector of the fitted plane is the direction of the spray gun, and the intersection of the main slice and the fitted plane is the direction of the spraying path.

[0059] 2) In boundary trajectory planning, the workpiece boundary is extracted as the location of the path points of the boundary trajectory, and its orientation is realized using the surfaces adjacent to the boundary points as auxiliary information. Therefore, boundary trajectory planning can be divided into two parts: determining the location of the path points of the boundary trajectory and determining their orientation.

[0060] a) Determining the location of boundary path points Based on the characteristics of the uniformly distributed point cloud workpiece model, it can be seen that the K-nearest neighbors of boundary points are mainly distributed on one side of the boundary points, while the K-nearest neighbors of interior points are distributed around the interior points, such as... Figure 7 As shown.

[0061] Take any point on the workpiece model ,Depend on The nearest neighbor algorithm finds the neighbors that are closest to the target point. points and Form a set , by set Fitting plane equations , point set Project onto the plane using the following formula Forming a point set :

[0062] by In plane Projection points in With the origin of the coordinate system, at Neutral The farthest point and constitute axis, The normal vector of a plane Use axes to establish a coordinate system coordinate system The fundamental vector is In the coordinate system The representation in is When these three vectors are written as a single... When using a matrix, this matrix represents the rotation matrix that indicates the transformation relationship between the two coordinate systems, denoted as: Through homogeneous transformation matrix From its original base coordinate system Switch to Therefore, dimensionality reduction can be achieved.

[0063] pass A The tan2 function can be used to obtain The points in the middle are respectively with In coordinate system The included angle is shown below:

[0064] The obtained angles are sorted in ascending order. After sorting, the angles between the points adjacent to the target point are calculated using equation (3-36). ,like Figure 8 As shown, The sum of all included angles is calculated using the following formula. : (3-36) Depend on Figure 7 It can be seen that when At that time, the target point Let be the boundary point, where It is a threshold, generally Boundary extraction of flat panel components, such as... Figure 9 As shown.

[0065] b) Determination of the attitude of boundary path points

[0066]

[0067] in, , , ,and .

[0068] Fit the normal vector of the plane to the set of nearest plane points of the boundary points. For reference, and using the two path points as the axis of rotation. Let the normal vector of the fitting plane about the axis of rotation Rotate by a fixed angle A new direction vector is obtained ,by As the spray gun is pointed, and with the rotation axis The direction is used as the spraying path direction.

[0069] Example 1 – Trajectory Planning for a 600mm×600mm Flat Plate Specimen (60μm Film Thickness) 1. The process layout of this embodiment is as follows: a "dual robot + positioning mechanism" layout is adopted, with the positioning mechanism placed between the two painting robots. The positioning mechanism can be translated. The 3D camera and spray gun are fixed to the end of the painting robot, and the camera and robot are rigidly connected.

[0070] 2. Spatial calibration: 1) Robot zero-point calibration: Record the robot's initial pose relative to the chessboard grid, and solve the robot's zero point using a small neighborhood iterative algorithm (iterations ≥ 10 times) to ensure that the zero-point positioning error is ≤ 0.01 mm.

[0071] 2) 3D Camera and Robot Pose Calibration: a. Fix the calibration plate (same as the zero-point calibration) on the test piece platform. The robot drives the camera to take pictures in 12 different poses (the poses cover the top of the test piece and 4 sides, with 2-3 angles in each direction); b. Record the robot's end effector data (joint angles, end effector coordinates) for each shot. Extract the X coordinates of the calibration plate corner points in the camera system using the camera's extrinsic parameters. C With normal vector V C c. with Based on the initial value, iteratively optimize to... X B (Robot system corner coordinates) V B The standard deviation of the (robot system normal vector) is <0.005mm. After convergence, the pose relationship is determined.

[0072] 3) Due to the limited selection of parts, only one painting robot is needed, eliminating the need for multi-robot pose calibration and calibration of the displacement mechanism and robot.

[0073] 3. Three-dimensional intelligent perception: The robot moves a 3D camera to a position 1m directly above the center of the test specimen (the camera's optical axis is perpendicular to the test specimen plane) to capture the initial point cloud of the specimen. A region growing algorithm is used to obtain the discrete points corresponding to the boundary discrete points (0,0,0), (600,0,0), (600,600,0), and (0,600,0). Subset segmentation: The subset length d = 100mm is selected (adapting to the 600mm side length of the test specimen, ensuring each side is divided into 6 segments), resulting in the subset Q. m ={P i | P i ∈R 3Let Q be a subset of points i = 1, 2, ..., 6 (m = 4 edges × 6 segments = 24 subsets). Calculate each Q. m The mean coordinates are used to fuse the point cloud from 24 re-enhanced shots with the initial point cloud to obtain a new edge point set P'. Verification logic: Calculate the distance between each point in P and its corresponding point in P', setting a threshold β = 0.02 mm (determined by camera accuracy + registration accuracy); Integrity judgment: If the edges of the flat sample are regular, and all points in P' after re-enhanced shots coincide with P (distance < β), it indicates that the point cloud model is complete (no missed areas), and re-enhanced shots are stopped.

[0074] 4. Flat plate test piece trajectory planning: 1) Surface principal trajectory planning: Randomly select a point in the point cloud of the sample sample. P 0( x 0, y Take K=6 neighboring points (0,0) and fit a quadratic surface (actually a plane z=0). From the plane equation f(x,y,z)=z=0, the partial derivative yields the normal vector η= (0,0,1), and the normal vectors of all points are consistent. Take any point as a seed point and calculate the angle θ=0° between its normal vector and the normal vectors of other points. Set a threshold θ=5° (for flat surfaces, the angle is much smaller than the threshold). All points are grouped into a single planar region, eliminating the need for multi-region segmentation. (Pressure gun spacing) d Based on the required film thickness (60μm) and coating characteristics (epoxy primer, 50% solids content, 80% spraying efficiency), the following calculations were made: d =100 mm (Ensure a 50% overlap in coating to avoid uneven film thickness); Number of slices n Maximum span of the specimen along the X-axis Dmax =600 mm n=6 layers.

[0075] 2) Boundary trajectory planning: The boundary of the flat specimen is a regular rectangle, and the boundary trajectory needs to be planned separately to ensure that the edge film thickness meets the standard. Through calculation, the boundary path attitude parameters are obtained: Reference normal vector η: the normal vector of the plane adjacent to the boundary point is taken as η=(0,0,1) (perpendicular to the specimen plane); Rotation axis K: the line connecting the front and rear path points of the boundary is taken as the rotation axis; Rotation angle θ: to avoid edge leakage, θ=5° is set (so that the spray gun is slightly tilted to the outside of the specimen to cover the edge of the boundary).

Claims

1. A method for automated coating trajectory planning of composite material parts, characterized in that, Includes the following steps: Step 1: Perform robot zero-point calibration, 3D camera and robot pose relationship calibration, pose relationship calibration among multiple robots, workpiece displacement mechanism and robot pose calibration, and spray gun and robot end effector pose relationship calibration. Step 2: Control the painting robot equipped with the 3D camera to iteratively photograph and fuse the workpiece with point clouds to obtain a complete working 3D point cloud model and identify its real edges; Step 3: Perform trajectory planning. Trajectory planning is divided into two parts: complex surface path planning and boundary trajectory planning. The two parts are integrated to form the contour-following trajectory of the robot driving the spray gun.

2. The automated coating trajectory planning method for composite material parts according to claim 1, characterized in that, In step 1, the calibration of the pose relationship between the 3D camera and the robot specifically includes the following process: By using a robot carrying a camera to capture images of a fixed calibration plate from multiple different poses, and based on the consistency of the description of the feature points of the calibration plate in the robot's base coordinate system, the transformation matrix from the camera coordinate system to the robot's end effector coordinate system is iteratively optimized.

3. The automated coating trajectory planning method for composite material parts according to claim 1, characterized in that, In step 1, the pose relationship between multiple robots specifically includes the following process: using the base coordinate system of the first robot as the main coordinate system, each robot's end-effector camera takes pictures of the same fixed calibration plate, and using the camera pose estimation results, the pose relationship of the base coordinate system of the second robot relative to the main coordinate system is calculated through coordinate transformation.

4. The automated coating trajectory planning method for composite material parts according to claim 1, characterized in that, In step 1, the calibration of the pose relationship between the spray gun and the robot end effector specifically includes the following process: by making the spray gun head contact a fixed point in space with different postures, the pose of the robot end effector is recorded, and the position parameters of the spray gun coordinate system relative to the robot end effector coordinate system are solved by least squares fitting.

5. The automated coating trajectory planning method for composite material parts according to claim 1, characterized in that, Step 2 specifically includes the following process: S21. Initial Shooting and Edge Extraction: Control the robot to a preset initial pose to shoot the workpiece, extracting the initial point cloud and edge point set P of the workpiece; S22. Edge Segmentation and Supplementary Shooting Pose Calculation: Sequentially segment the edge point set P into m sub-edges, calculate the mean point s' of each sub-edge, and fit a plane in the neighborhood of s' to obtain the normal vector η. Calculate the supplementary shooting pose based on s', η, and the preset optimal shooting distance D; S23. Iterative Supplementary Shooting and Model Fusion: Control the robot to shoot according to the calculated supplementary shooting pose, and fuse the newly acquired point cloud with the existing model to obtain the updated edge point set P'; S24. Edge Verification and Iteration Termination Judgment: Compare the edge point sets P and P'. If there is a point in P' whose distance to a point in P is less than a set threshold β, remove that point from P'; if P' is not empty after removal, use it as new edge information and return to step S22 for the next round of supplementary shooting; if P' is empty or meets other termination conditions, a complete 3D point cloud model of the workpiece is obtained.

6. The automated coating trajectory planning method for composite material parts according to claim 1, characterized in that, In step 3, the complex surface path planning includes: Feature vector-based surface segmentation: Calculate the feature vectors of each point in the point cloud model, and based on the region growing method, classify the points whose feature vectors have an angle less than the threshold θ0 into the same surface patch to achieve the segmentation of complex surfaces; TCP trajectory point adaptive solution: The point cloud model is sliced ​​into main slices according to the spray gun spacing d, and regional slices are made according to the spraying characteristics; for each main slice region, a plane is fitted using its adjacent point cloud data; the intersection of the main slice plane, the regional slice plane and the fitted plane is calculated as the position of the TCP trajectory point, and the normal vector of the fitted plane is used as the direction of the spray gun at the trajectory point.

7. The automated coating trajectory planning method for composite material parts according to claim 1, characterized in that, In step 3, the boundary trajectory planning includes: Boundary point identification: For any point P0 in the point cloud, obtain its K nearest neighbors and project them onto the fitting plane; analyze the distribution angles of each nearest neighbor relative to P0 in the projection plane; if the sum of the angles between adjacent vectors θsum is less than the set threshold θa, then P0 is determined to be a boundary point. Boundary point spray gun attitude determination: Taking the normal vector η of the plane fitted by the neighboring points of the boundary point as the reference, rotate η around the rotation axis K determined by the direction of the line connecting the front and rear boundary points by a fixed angle θ, and the resulting direction vector η' is used as the direction of the spray gun at that boundary point.

8. An automated coating system, characterized in that, A method for implementing an automated coating trajectory planning method for composite material parts according to any one of claims 1-7 includes: At least two painting robots; A positioning mechanism, located between the painting robots, is used to clamp and translate the workpiece; A 3D camera is rigidly fixed to the end effector of the painting robot; The spray gun is rigidly fixed to the end effector of the painting robot; The control system is communicatively connected to the painting robot, the positioning mechanism, and the 3D camera. It is used to execute the system calibration, three-dimensional intelligent perception and trajectory planning algorithms, and control the coordinated movement of the painting robot and the positioning mechanism.

9. A computer-readable storage medium, characterized in that, It stores a computer program, which, when executed by a processor, implements the steps of the automated coating trajectory planning method according to any one of claims 1-7.

10. An electronic device, characterized in that, The method includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the automated coating trajectory planning method according to any one of claims 1-7.

Images