Method for extracting glue spraying track of complex curved surface of aircraft panel

Abstract

The invention provides an aircraft panel complex curved surface glue spraying track extraction method, which comprises the following steps: acquiring point cloud data of an aircraft panel, and constructing an actual three-dimensional point cloud model by using a point cloud registration algorithm; performing model matching on the designed three-dimensional point cloud model of the aircraft panel and the actual three-dimensional point cloud model, and extracting a designed glue spraying track; the designed glue spraying track in the three-dimensional space is mapped to a two-dimensional plane through manifold, Kalman filtering processing is conducted on the point clouded wallboard 2D model or the collected three-dimensional point cloud and the glue spraying track point cloud on the two-dimensional plane, and a two-dimensional plane glue spraying track is obtained; and converting the two-dimensional plane glue spraying track into a predicted three-dimensional point cloud track in the wallboard glue coating curved surface through spatial transformation, and outputting the predicted three-dimensional point cloud track as a predicted glue spraying track. The glue spraying path of complex curved surfaces such as aircraft panels and outer skins can be accurately extracted, and the path is executed through the robot, so that the aircraft manufacturing and assembling efficiency and quality are improved, and meanwhile, the labor cost is reduced.

Description

Technical Field

[0001] This invention belongs to the field of aerospace manufacturing engineering and aircraft assembly technology, and relates to a method for extracting the adhesive spraying trajectory of complex curved surfaces of aircraft panels, which is used for automated surface spraying during the assembly process of aircraft. Background Technology

[0002] In the aerospace manufacturing sector, the introduction of intelligent adhesive coating systems has significantly improved production efficiency and quality. Automated adhesive coating equipment not only effectively controls production costs but also avoids the health threats posed to workers by volatile adhesives in traditional manual operations. These intelligent equipment, with their superior motion precision and adaptability, can operate stably in complex industrial environments, leading to their widespread adoption in the aerospace industry. Nevertheless, the application of intelligent equipment in aircraft panel adhesive coating processes still faces numerous technical bottlenecks, primarily due to the complex curved surface characteristics of the panel structure and the difficulty in accurately planning the spraying path. These technical challenges restrict the comprehensive promotion and in-depth application of intelligent adhesive coating systems in this field.

[0003] Currently, research on adhesive spraying trajectory extraction for complex curved surfaces has made some progress, with the main methods based on 2D image processing and 3D vision technology. Compared with 2D image processing, 3D vision technology can acquire richer object pose information, showing a significant advantage in adhesive spraying trajectory extraction for complex curved surfaces such as aircraft panels. Existing adhesive spraying trajectory extraction techniques typically rely on the edge contours of the panel captured by a camera. However, in complex curved surface scenarios such as aircraft panels, due to the complexity of the surface geometry and optical features, traditional methods are prone to misjudging the adhesive spraying trajectory, making it difficult to meet the high-precision adhesive spraying requirements of aircraft manufacturing. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention proposes a method for extracting the adhesive spraying trajectory of complex curved surfaces on aircraft panels.

[0005] The technical solution adopted in this invention is as follows:

[0006] A method for extracting adhesive spraying trajectories on complex curved surfaces of aircraft panels includes the following steps:

[0007] S1: Collect point cloud data from different angles of the wall panel, and reconstruct the actual three-dimensional point cloud model of the wall panel through a point cloud registration algorithm.

[0008] Adjust the pose of the wall panel and collect point cloud data of the wall panel at different angles; input the collected point cloud data into the pre-designed point cloud segmentation model for segmentation, retain only the point cloud data of the wall panel area, and use the segmented data to perform point cloud registration and reconstruction through the ICP (Iterative Closest Point) point cloud registration algorithm. Finally, after preprocessing, the complete actual 3D point cloud model of the wall panel is obtained.

[0009] The ICP point cloud registration algorithm specifically involves: searching for feature points in each segmented point cloud dataset, matching and registering feature points from different point cloud datasets. The registration is based on minimizing the Euclidean distance between identical feature points in different point cloud datasets, resulting in the registration rotation matrix R between the different point cloud datasets. * And the registration translation matrix T * Its objective function is:

[0010]

[0011] Where R * This is the registration rotation matrix, representing the rotation transformation after registration of the aircraft panel point cloud; T * It is the registration translation matrix, representing the translation transformation after registration of the aircraft panel point cloud; P i It is an aircraft panel point cloud P S The i-th feature point in; q i It is a point cloud P S China and P i The corresponding nearest feature point, N is the number of feature points, R is the rotation matrix of the aircraft panel point cloud, and T is the translation matrix of the aircraft panel point cloud.

[0012] The steps for preprocessing the reconstructed 3D point cloud model of the wall panel are as follows: First, the point cloud data is downsampled to reduce the amount of data while retaining the feature information of the wall panel; then, noise points are removed by radius filtering to obtain the final actual 3D point cloud model of the wall panel.

[0013] S2: Match the design 3D point cloud model of the wall panel with the actual 3D point cloud model of the wall panel to obtain the design adhesive spraying trajectory.

[0014] The 3D design model of the wall panel is converted into a point cloud to obtain the design 3D point cloud model. The actual 3D point cloud model of the wall panel is then transformed into the coordinate system of the design 3D point cloud model, and the overlap between the two is optimized to achieve the best result. The preset glue spraying trajectory is extracted from the design 3D point cloud model, and the corresponding path is matched in the actual 3D point cloud model of the wall panel using the KNN nearest neighbor search algorithm. Abnormal points in the trajectory are detected and removed, and after sequential processing, the design glue spraying trajectory points are output.

[0015] The KNN nearest neighbor search algorithm is based on the following: In the design of the 3D point cloud model, the search origin P... a (x a ,y a ,z a ) and the search point P in the point cloud of the aircraft panel b (x b ,y b ,z b The Euclidean distance d between the two points is minimized, as shown in the following formula, where the search origin P is the origin.a To design sampling points in a 3D point cloud model, search point P b In the point cloud of the aircraft panel, the search origin P is intersected. a The nearest point:

[0016]

[0017] S3: Spatial registration area to be bonded.

[0018] Point cloud data of the area to be bonded (such as reinforcements, structural components, etc.) is collected and mapped to the reference model space of the wall panel to obtain the bonding surface. Then, ICP or other point cloud registration algorithms are used to spatially align the bonding surface with the actual 3D point cloud model of the wall panel, optimize the positional relationship after registration, maximize the overlap between the bonding surface and the actual 3D point cloud model, and obtain the surface mapping relationship and the geometric constraints of the wall panel point cloud to improve the accuracy of glue spray trajectory mapping and subsequent spatial fitting, serving as the geometric basis for subsequent glue spray trajectory mapping and prediction.

[0019] S4: Generate a predicted adhesive spray trajectory.

[0020] The designed adhesive spraying trajectory obtained in S2 is mapped onto a 2D plane using a Riemannian manifold to obtain the 2D designed adhesive spraying trajectory. The designed 3D point cloud model and the actual 3D point cloud model are mapped onto the 2D plane to obtain the 2D designed point cloud model and the 2D actual point cloud model. The 2D designed adhesive spraying trajectory, along with the 2D designed point cloud model and the 2D actual point cloud model, are then subjected to Kalman filtering to obtain the predicted 2D adhesive spraying trajectory. The predicted 2D adhesive spraying trajectory is then spatially transformed and projected onto the actual 3D point cloud model of the wall panel to output the predicted adhesive spraying trajectory. Specifically:

[0021] S4.1: Treat the point cloud of the designed adhesive spraying trajectory as discrete points on the manifold surface. Using the Riemannian manifold mapping method, while maintaining the geometric characteristics of the designed adhesive spraying trajectory curve, project the three-dimensional trajectory points onto the local two-dimensional parameter plane where the wall panel is located to form the initial two-dimensional adhesive spraying trajectory.

[0022] S4.2: Map the design 3D point cloud model and the actual 3D point cloud model onto a 2D plane to obtain a 2D design point cloud model and a 2D actual point cloud model.

[0023] S4.3: Align the two-dimensional glue spraying trajectory with the two-dimensional projection model of the wall panel or its contour data obtained by projecting from the actual point cloud. Construct a cost function that minimizes the distance between the glue spraying trajectory and the boundary points of the wall panel, and use the Kalman filter algorithm to generate the predicted two-dimensional glue spraying trajectory. Finally, perform noise reduction and trajectory smoothing on the predicted two-dimensional glue spraying trajectory.

[0024] S4.4: Based on the surface mapping relationship in S3 and the geometric constraints of the panel point cloud, the predicted two-dimensional planar adhesive spraying trajectory is mapped back to three-dimensional space through spatial inverse transformation, and then reprojected onto the actual three-dimensional point cloud model of the panel to generate the predicted adhesive spraying trajectory.

[0025] S5: Perform optimal quadratic Kalman filtering optimization on the designed glue spraying trajectory and the predicted glue spraying trajectory to obtain the optimal glue spraying trajectory, and use the optimal glue spraying trajectory as the glue spraying trajectory followed by the robot glue spraying process.

[0026] Furthermore, when it is impossible to extract the complete adhesive spraying trajectory directly from the 3D design model, or when there are problems such as deformation or defects in the boundary shape of the bonding area, an offset processing method based on the boundary contour point cloud is used to generate the adhesive spraying trajectory.

[0027] The offset calculation method for the bias processing is as follows: For a point P0(x0,y0) in the boundary contour point cloud after the wall panel and the bonding surface are aligned, its corresponding offset point to be determined is Q(x0,y0). q ,y q ),but The offset distance is d(u). First, in the boundary contour point cloud, using P0 as the search point, search for the nearest point P1(x1,y1) on the bonding surface. straight line The normal vector is (x1≠x0, y1≠y0), thus obtaining the bias point Q. q (x q ,y q ):

[0028]

[0029] The beneficial effects of this invention are:

[0030] The proposed method for extracting adhesive spraying trajectories on complex curved surfaces of aircraft panels can accurately extract the adhesive spraying paths of complex curved surfaces such as aircraft panels and outer skins. These paths are then executed by robots, thereby improving the efficiency and quality of aircraft manufacturing and assembly while reducing labor costs. Furthermore, this method employs visual guidance technology to ensure that the accuracy of the adhesive spraying trajectory is not affected by factors such as panel deformation, demonstrating broad application prospects. Attached Figure Description

[0031] Figure 1 This is a flowchart of the implementation method. Detailed Implementation

[0032] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. 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.

[0033] This embodiment, in conjunction with the accompanying drawings, further illustrates a method for extracting adhesive spraying trajectories from complex curved surfaces of aircraft panels.

[0034] The adhesive spraying trajectory extraction method described herein uses an aircraft panel trajectory extraction and adhesive spraying device to extract the adhesive spraying trajectory and execute the adhesive spraying operation. The extraction device includes a structured light camera, a multi-degree-of-freedom manipulator, an external axis, an industrial control computer, and tooling fixtures. The fixture is used to hold the aircraft panel, ensuring its stable position during point cloud acquisition, model reconstruction, and adhesive spraying operations, thus guaranteeing data accuracy and adhesive spraying precision. The industrial control computer controls the overall operation of the device, processes point cloud data, constructs the model, extracts and optimizes the adhesive spraying trajectory, and generates commands to precisely control the equipment for adhesive spraying. The fixture is connected to the industrial control computer via an external axis, driving the aircraft panel to rotate and move in multiple angles and postures. This, combined with a structured light camera, allows for the acquisition of point cloud data from different perspectives of the panel, ensuring comprehensive point cloud information of the complex curved surface of the panel, providing a data foundation for the subsequent accurate reconstruction of the 3D point cloud model. A multi-degree-of-freedom robotic arm, based on the final generated optimal adhesive spraying trajectory, drives the adhesive spraying device in 3D space to perform adhesive spraying operations on the complex curved surface of the aircraft panel according to preset paths and parameters (such as spraying speed and adhesive amount). Its multi-axis linkage capability allows it to flexibly adapt to various curvature and angle changes on the panel surface, ensuring that the adhesive spraying device always acts on the adhesive application area with the appropriate posture and distance, thereby guaranteeing the uniformity and accuracy of adhesive spraying. The structured light camera includes a color camera module, a structured light projection module, a processor module, a power supply module, and mechanical structural components. It is used to collect point cloud data of the aircraft panel from different angles during the aircraft panel pose adjustment process. After transmission and processing, it provides a basis for the reconstruction of the 3D point cloud model. The color camera module can assist in acquiring the color and texture information of the panel surface, while the structured light projection module helps improve the acquisition accuracy of the point cloud data by projecting specific patterns, so as to more accurately capture the geometric features of the complex curved surface of the panel. The multi-degree-of-freedom manipulator and the structured light camera achieve hand-eye calibration through calibration plates and calibration blocks to determine the coordinate relationship.

[0035] like Figure 1 The method for extracting the adhesive spray trajectory includes the following steps:

[0036] S1: Collect point cloud data from different angles of the wall panel, and reconstruct the actual three-dimensional point cloud model of the wall panel through a point cloud registration algorithm;

[0037] The external axis is rotated and the pose of the wall panel is adjusted by controlling the terminal device. The point cloud data of the wall panel at different angles is collected by the structured light camera. The collected point cloud data is then input into a pre-designed point cloud segmentation model for segmentation. Only the point cloud data of the wall panel area is retained. The segmented data is then used to perform point cloud registration and reconstruction through the ICP (Iterative Closest Point) point cloud registration algorithm. Finally, after preprocessing, the actual three-dimensional point cloud model of the complete wall panel is obtained.

[0038] The ICP point cloud registration algorithm specifically involves: searching for feature points in each segmented point cloud dataset, matching and registering feature points from different point cloud datasets. The registration is based on minimizing the Euclidean distance between identical feature points in different point cloud datasets, resulting in the registration rotation matrix R between the different point cloud datasets. * And the registration translation matrix T * Its objective function is:

[0039]

[0040] Where R * This is the registration rotation matrix, representing the rotation transformation after registration of the aircraft panel point cloud; T * It is the registration translation matrix, representing the translation transformation after registration of the aircraft panel point cloud; P i It is an aircraft panel point cloud P S The i-th feature point in; q i It is a point cloud P S China and P i The corresponding nearest feature point, N is the number of feature points, R is the rotation matrix of the aircraft panel point cloud, and T is the translation matrix of the aircraft panel point cloud.

[0041] The steps for preprocessing the reconstructed 3D point cloud model of the wall panel are as follows: First, the point cloud data is downsampled to reduce the amount of data while retaining the feature information of the wall panel; then, noise points are removed by radius filtering to obtain the final actual 3D point cloud model of the wall panel.

[0042] S2: Match the design 3D point cloud model of the wall panel with the actual 3D point cloud model of the wall panel to obtain the design adhesive spraying trajectory.

[0043] The 3D design model of the wall panel is converted into a point cloud to obtain the design 3D point cloud model; the actual 3D point cloud model of the wall panel is transformed into the coordinate system of the design 3D point cloud model, and the overlap between the two is optimized to achieve the best result.

[0044] The pre-set glue spraying trajectory is extracted from the design 3D point cloud model. The corresponding path is matched in the actual 3D point cloud model of the wall panel using the KNN nearest neighbor search algorithm. Abnormal points in the trajectory are detected and removed. After sequential processing, the designed glue spraying trajectory points are output.

[0045] The basis for KNN nearest neighbor search is: the search origin P in the design of the 3D point cloud model. a (x a ,y a ,z a ) and the search point P in the point cloud of the aircraft panel b (x b ,y b ,z b The Euclidean distance d between the two points is minimized, as shown in the following formula, where the search origin P is the origin. a To design sampling points in a 3D point cloud model, search point P b In the point cloud of the aircraft panel, the search origin P is intersected. a The nearest point:

[0046]

[0047] S3: Spatial registration area to be bonded.

[0048] Point cloud data of the area to be bonded (such as reinforcements, structural components, etc.) is collected and mapped to the reference model space of the wall panel to obtain the bonding surface. Then, ICP or other point cloud registration algorithms are used to spatially align the bonding surface with the actual 3D point cloud model of the wall panel, optimize the positional relationship after registration, maximize the overlap between the bonding surface and the actual 3D point cloud model, and obtain the surface mapping relationship and the geometric constraints of the wall panel point cloud to improve the accuracy of glue spray trajectory mapping and subsequent spatial fitting, serving as the geometric basis for subsequent glue spray trajectory mapping and prediction.

[0049] S4: Generate a predicted adhesive spray trajectory.

[0050] The designed adhesive spraying trajectory obtained in S2 is mapped onto a 2D plane using a Riemannian manifold to obtain the 2D designed adhesive spraying trajectory. The designed 3D point cloud model and the actual 3D point cloud model are mapped onto the 2D plane to obtain the 2D designed point cloud model and the 2D actual point cloud model. The 2D designed adhesive spraying trajectory, along with the 2D designed point cloud model and the 2D actual point cloud model, are then subjected to Kalman filtering to obtain the predicted 2D adhesive spraying trajectory. The predicted 2D adhesive spraying trajectory is then spatially transformed and projected onto the actual 3D point cloud model of the wall panel to output the predicted adhesive spraying trajectory. Specifically:

[0051] S4.1: Treat the point cloud of the designed adhesive spraying trajectory as discrete points on the manifold surface. Using the Riemannian manifold mapping method, while maintaining the geometric characteristics of the designed adhesive spraying trajectory curve, project the three-dimensional trajectory points onto the local two-dimensional parameter plane where the wall panel is located to form the initial two-dimensional adhesive spraying trajectory.

[0052] S4.2: Map the design 3D point cloud model and the actual 3D point cloud model onto a 2D plane to obtain a 2D design point cloud model and a 2D actual point cloud model.

[0053] S4.3: Align the two-dimensional glue spraying trajectory with the two-dimensional projection model of the wall panel or its contour data obtained by projecting from the actual point cloud. Construct a cost function that minimizes the distance between the glue spraying trajectory and the boundary points of the wall panel, and use the Kalman filter algorithm to generate the predicted two-dimensional glue spraying trajectory. Finally, perform noise reduction and trajectory smoothing on the predicted two-dimensional glue spraying trajectory.

[0054] S4.4: Based on the surface mapping relationship in S3 and the geometric constraints of the panel point cloud, the predicted 2D planar adhesive spraying trajectory is mapped back to 3D space through inverse spatial transformation and then reprojected onto the actual 3D point cloud model of the panel to generate the predicted adhesive spraying trajectory. This process, based on the surface mapping relationship in S3 and the geometric constraints of the panel point cloud, ensures that the reconstructed 3D point set accurately fits the actual panel surface.

[0055] The core of this step lies in using a "3D→2D→3D" mapping strategy to process trajectories in a 2D plane, which can effectively improve path smoothness and surface adaptability, laying the foundation for subsequent trajectory optimization and robot execution.

[0056] In other embodiments, when it is not possible to directly extract the complete adhesive spraying trajectory from the 3D design model in certain assembly scenarios, or when there are problems such as deformation or defects in the boundary shape of the bonding area, an offset processing method based on the boundary contour point cloud is used to generate the adhesive spraying trajectory. This method does not rely on the design trajectory; it directly derives the spraying path that can be executed by the robot by analyzing the boundary point cloud of the panel and the bonding area.

[0057] The offset calculation method for the bias processing is as follows: For a point P0(x0,y0) in the boundary contour point cloud after the wall panel and the bonding surface are aligned, its corresponding offset point to be determined is Q(x0,y0). q ,y q ),but The offset distance is d(u). First, in the boundary contour point cloud, using P0 as the search point, search for the nearest point P1(x1,y1) on the bonding surface. straight line The normal vector is (x1≠x0, y1≠y0), thus obtaining the bias point Q. q (x q,y q ):

[0058]

[0059] S5: Perform optimal quadratic Kalman filtering optimization on the designed glue spraying trajectory and the predicted glue spraying trajectory to obtain the optimal glue spraying trajectory, and use the optimal glue spraying trajectory as the glue spraying trajectory followed by the robot glue spraying process.

[0060] First, a second-order Kalman filter optimization is performed on the designed and predicted glue spraying trajectories to obtain the optimized glue spraying trajectory. Next, outliers in the glue spraying trajectory path are identified and removed. The Bessel Catmull-Rom curve interpolation technique is used to insert additional transition points between trajectory points at different intervals to achieve a smooth transition of the glue spraying trajectory path. Finally, according to the specific requirements of the glue spraying process, robot glue spraying instructions are formulated, and the robot and its glue spraying device are controlled according to these instructions to perform precise glue spraying operations on the wall panel.

[0061] The specific steps of quadratic Kalman filter optimization are as follows: First, define the object and the objective. Let the curve consist of N discrete points, and the state vector of each discrete point be x. i =[x i ,y i ,z i ] T The design curve is: p d (i)=[x d (i),y d (i),z d (i)] T The measurement curve is: p m (i)=[x m (i),y m (i),z m (i)] T Then the filtered target estimation curve is: p e (i) = [x(i), y(i), z(i)] T Each point i is considered as an estimation step.

[0062] Then, a second-order Kalman filter model is performed, defining the assumption that the curve is smooth and the changes between points are stable, and establishing a state transition model:

[0063] x i =x i-1 +v i +w i

[0064] Among them, the difference v estimated using the design curve i =p d (i)-p d(i-1), w i ~N(0,Q) represents the process noise characterizing the design model error, using the state transition Jacobian matrix. The identity matrix locally linearizes the model, enabling subsequent Kalman filtering prediction and update steps to be effectively applied to nonlinear scenarios.

[0065] After completing the modeling, the points in the measurement point cloud are then observed. i =x i +v i , Using the observation-transition Jacobian matrix (Unit matrix) locally linearizes the model, enabling the observation information to be effectively utilized by the Kalman filter update step.

[0066] After completing the observations, perform model initialization and determine the initial values. There are two methods for state initialization: Method 1: Directly use the coordinates of the 0th point of the design curve as the initial value, i.e., x0 = p. d (0). Method 2: Take the average of the design curve and the measurement curve at point 0 as the initial value, i.e. Covariance initialization (For example, the standard deviation σ0 = 1 mm) is used to correct the initial value.

[0067] After completing model initialization, perform Kalman filter iterations to optimize discrete points. First, predict the state of the current point: p is the estimated value of the previous point. d (i)-p d (i-1) represents the difference of the design curve, and then the covariance is predicted. P i-1|i-1 Let F be the covariance of the previous point. i Let Q be the state transition identity matrix, and let Q be the process noise.

[0068] Complete the prediction before updating. Kalman gain increases the confidence level of the observations.

[0069] P i|i-1 The prior covariance of the output during the prediction phase. Let R be the transpose of the observation transition Jacobian matrix, and R be the observation noise covariance. Update the state, correct the prior state, and obtain the "posterior state estimate": For the prior state estimate of the output in the prediction phase, K i For Kalman gain, z i These are actual observed values. For the predicted observations. Update the covariance formula: P i|i =(IK i Hi )P i|i-1 I is a 3×3 identity matrix, K i H i P is the correction factor. i|i-1 The prior covariance of the output during the prediction phase. This completes the second-order Kalman filter optimization.

[0070] Subsequently, Bézier curves are used to interpolate the determined optimal glue spraying path to add transition nodes between path points, thus obtaining a smooth glue spraying path. This smooth path is considered an estimate of the optimal glue spraying trajectory and serves as a guiding path for the robot to perform the glue spraying operation.

[0071] More specifically, given points P0, P1, ..., P n Its Catmull-Rom curve is as follows:

[0072]

[0073] In the formula: t is the sampling parameter, which can be set according to the sparsity requirements, and the value range is [0,1], representing that in P i and P i+1 The interpolation position between them. After sampling, the position planning of the glue application trajectory is completed. By combining the corresponding projection point vector of the trajectory point on the curved surface, the normal pose can be obtained, and finally the robot glue application trajectory is extracted.

[0074] The above description is only a partial embodiment of the present invention and does not limit the implementation and protection scope of the present invention. Any improvements and modifications made by those skilled in the art based on the disclosure of the present invention without departing from the scope of the present invention should be within the protection scope of the present invention.

Claims

1. A method for extracting the adhesive spray trajectory of complex curved surfaces of aircraft panels, characterized in that, Includes the following steps: S1: Collect point cloud data from different angles of the wall panel, and reconstruct the actual three-dimensional point cloud model of the wall panel through a point cloud registration algorithm; S2: Match the design 3D point cloud model of the wall panel with the actual 3D point cloud model of the wall panel to obtain the design adhesive spraying trajectory; S3: Spatial registration area to be bonded; S4: Generate the predicted adhesive spray trajectory; S5: Obtain the optimal adhesive spraying trajectory.

2. The method for extracting the adhesive spray trajectory of complex curved surfaces of aircraft panels according to claim 1, characterized in that, S1 specifically involves: adjusting the pose of the wall panel and collecting point cloud data of the wall panel at different angles; inputting the collected point cloud data into a pre-designed point cloud segmentation model for segmentation, retaining only the point cloud data of the wall panel area, and using the segmented data to perform point cloud registration and reconstruction through the ICP point cloud registration algorithm, and finally obtaining the complete actual three-dimensional point cloud model of the wall panel after preprocessing.

3. The method for extracting the adhesive spray trajectory of complex curved surfaces of aircraft panels according to claim 2, characterized in that, The ICP point cloud registration algorithm works as follows: It searches for feature points in each segmented point cloud dataset, matches and registers feature points from different point cloud datasets, and minimizes the Euclidean distance between identical feature points in different point cloud datasets. This results in the registration rotation matrix R between the different point cloud datasets. * And the registration translation matrix T * Its objective function is: Where R * This is the registration rotation matrix, representing the rotation transformation after registration of the aircraft panel point cloud; T * This is the registration translation matrix, representing the translation transformation after registration of the aircraft panel point cloud; P i It is an aircraft panel point cloud P S The i-th feature point in; q i It is a point cloud P S China and P i The corresponding nearest feature point, N is the number of feature points, R is the rotation matrix of the aircraft panel point cloud, and T is the translation matrix of the aircraft panel point cloud.

4. The method for extracting the adhesive spray trajectory of complex curved surfaces of aircraft panels according to claim 2, characterized in that, Preprocessing includes downsampling and noise reduction.

5. The method for extracting the adhesive spray trajectory of complex curved surfaces of aircraft panels according to claim 2, characterized in that, Specifically, S2 involves: converting the three-dimensional design model of the wall panel into a point cloud to obtain a design three-dimensional point cloud model; converting the actual three-dimensional point cloud model of the wall panel into the coordinate system of the design three-dimensional point cloud model; and optimizing the overlap between the two to achieve the best result. The pre-set glue spraying trajectory is extracted from the design 3D point cloud model. The corresponding path is matched in the actual 3D point cloud model of the wall panel using the KNN nearest neighbor search algorithm. Abnormal points in the trajectory are detected and removed. After sequential processing, the designed glue spraying trajectory points are output.

6. The method for extracting the adhesive spray trajectory of complex curved surfaces of aircraft panels according to claim 5, characterized in that, The KNN nearest neighbor search algorithm is based on the following: In the design of a 3D point cloud model, the search origin P... a (x a ,y a ,z a ) and the search point P in the point cloud of the aircraft panel b (x b ,y b ,z b The Euclidean distance d between the two points is minimized, as shown in the following formula, where the search origin P is the origin. a To design sampling points in a 3D point cloud model, search point P b In the point cloud of the aircraft panel, the search origin P is intersected. a The nearest point:

7. The method for extracting the adhesive spray trajectory of complex curved surfaces of aircraft panels according to claim 5, characterized in that, S3 specifically involves: collecting point cloud data of the area to be bonded, mapping it to the reference model space of the wall panel to obtain the bonding surface, and then using a point cloud registration algorithm to spatially align the bonding surface with the actual three-dimensional point cloud model of the wall panel, optimizing the positional relationship after registration, maximizing the overlap between the bonding surface and the actual three-dimensional point cloud model, and obtaining the surface mapping relationship and the geometric constraints of the wall panel point cloud.

8. The method for extracting the adhesive spray trajectory of complex curved surfaces of aircraft panels according to claim 7, characterized in that, Specifically, S4 is: S4.1: Using the Riemannian manifold mapping method, the point cloud of the designed adhesive spraying trajectory is projected onto the local two-dimensional parameter plane where the wall panel is located, while maintaining the geometric features of the designed adhesive spraying trajectory curve, to form the initial two-dimensional adhesive spraying trajectory. S4.2: Map the design 3D point cloud model and the actual 3D point cloud model onto a 2D plane to obtain a 2D design point cloud model and a 2D actual point cloud model; S4.3: Align the two-dimensional adhesive spraying trajectory with the two-dimensional projection model of the wall panel or the contour data obtained by projecting it from the actual point cloud, generate the predicted two-dimensional adhesive spraying trajectory using the Kalman filter algorithm, and perform noise reduction and trajectory smoothing processing. S4.4: Based on the surface mapping relationship in S3 and the geometric constraints of the panel point cloud, the predicted two-dimensional planar adhesive spraying trajectory is mapped back to three-dimensional space through spatial inverse transformation, and then reprojected onto the actual three-dimensional point cloud model of the panel to generate the predicted adhesive spraying trajectory.

9. The method for extracting the adhesive spray trajectory of complex curved surfaces of aircraft panels according to claim 8, characterized in that, Specifically, S5 involves performing optimal quadratic Kalman filtering optimization on the designed glue spraying trajectory and the predicted glue spraying trajectory to obtain the optimal glue spraying trajectory.

10. A method for extracting the adhesive spray trajectory of a complex curved surface of an aircraft panel according to any one of claims 1-9, characterized in that, When it is not possible to extract the complete adhesive spraying trajectory directly from the 3D design model, or when there are problems with the boundary shape of the bonding area, an offset processing method based on the boundary contour point cloud is used to generate the adhesive spraying trajectory.

Images