Aircraft skin feature recognition and edge milling path planning method and system

Abstract

The present invention relates to a method for aircraft skin feature recognition and milling path planning. The method includes preprocessing of original point cloud images, point cloud image feature extraction, time-series-based point cloud posture calculation, feature point-based path planning, processing trajectory fitting and re-segmentation. The method targets the skin frame structure of an aircraft, uses an industrial robot as a carrier, and utilizes a line laser scanner to obtain skin trimming reference information. It provides two data processing modes, online and offline, and plans the processing path according to the process, guiding the robot to complete the shape-following motion task, thereby improving the adaptability of the processing process. The method utilizes the continuity of the changes in the feature points during the processing process, and optimizes the consistency of the feature points based on the time sequence; during online trimming, the feature structure of the processing reference surface is used for fitting, and the original point cloud images at adjacent moments are compared, so that the relative posture of the two is mapped in the base coordinate system, and the position state of the point cloud at the target moment is quickly confirmed by the relative posture relationship.

Description

Technical Field

[0001] The present invention belongs to the technical field of aircraft skin edge milling, and in particular relates to a method and system for aircraft skin feature recognition and edge milling path planning based on traditional algorithms. Background Art

[0002] During aircraft assembly, the skin gap requirement is generally less than 0.8-1mm. With the fifth-generation aircraft's pursuit of stealth performance, the gap requirement has also been increased to 0.3-0.5mm, or even higher.

[0003] During assembly, a trial assembly is performed inside the frame, and the seam clearance requirements are manually marked to determine the size of the skin trimming allowance, and then the processing allowance is cut off. Since the trimming is done inside the frame, ordinary CNC milling machines are difficult to use, and manual trimming is the only option. Currently, the method for manual trimming allowance is to mark the composite skin along the matching datum on the assembly frame, then remove the composite skin from the frame for rough trimming, and then put it back on the frame to fit the comparison datum, mark it again, and then remove it from the frame for trimming. This process is repeated many times to ensure that the fitting clearance is around 1mm. Manual trimming is labor-intensive, time-consuming, and inefficient. Moreover, because the trimming surfaces are mostly curved, the above-mentioned manual trimming method can no longer meet the trimming accuracy requirements of the skin. With the widespread use of carbon fiber composite skins, manual trimming also produces a large amount of toxic dust, which can enter the human body through the respiratory system or skin contact, causing irreversible and serious harm to the workers' health.

[0004] In order to ensure the quality of skin seam trimming, improve trimming efficiency, and meet the design and manufacturing standards of the new generation of aircraft, developing an automated and intelligent trimming method for the inner skin has become an important issue that technicians need to face. Summary of the Invention

[0005] In order to solve the problems that manual trimming methods are labor-intensive, time-consuming, inefficient, and hazardous to operate, and cannot meet the trimming accuracy requirements of the new generation of aircraft skins, the present invention proposes a new aircraft skin feature recognition and milling path planning method.

[0006] This method targets the skin frame structure of an aircraft. Using a KUKA industrial robot as a carrier, a line laser scanner is used to obtain the skin trimming benchmark information. Both online and offline data processing modes are provided, and a reasonable and smooth processing path is formed according to the process. This guides the robot to complete the conformal motion task and improves the adaptability of the processing process.

[0007] The implementation system for this method is primarily software-based, with a hardware platform established for verification. The software system primarily comprises a data analysis and processing system located on a host computer; the hardware platform primarily consists of a 3D line laser scanner, a KUKA robot, a milling module, a visual calibration plate, and the necessary components for system assembly and functional implementation. Through the integration of the hardware platform and software system, this system enables precise machining benchmark extraction and real-time machining path planning in both online and offline modes.

[0008] In general, the aircraft skin feature recognition and milling path planning method includes the following processing steps:

[0009] 1) Point cloud image preprocessing

[0010] Because surface feature point cloud images acquired by line laser scanning have the characteristics of high resolution, small data volume, and low anti-interference ability, and considering factors such as the small scale of the target features of the object to be measured, the relatively complex structure, and the high real-time requirements of the processing process, this method (system) requires preprocessing of the original point cloud image to reduce errors in the subsequent feature extraction and path planning processes. By analyzing the results of the point cloud obtained from the simulation process experiment, this method uses a method including threshold filtering and Gaussian filtering to process the acquired point cloud. On the hardware side, the time-based mean filtering function of the scanner is utilized to achieve optimization of the acquired point cloud.

[0011] 2) Point cloud image feature extraction

[0012] According to the processing technology, the trajectory path of trimming (milling) processing is mainly based on some characteristic structures of the reference frame, such as the chamfered edges in the reference frame. Therefore, in the acquired point cloud image, the system needs to first extract and segment the processing object to distinguish the skin from the reference frame and determine the location of the features. Then the reference frame point cloud image is derived, and the surface change law of the reference frame is used to achieve coarse extraction of feature points, and on this basis, the characteristic structure of the reference frame is further segmented. Since the coarsely extracted feature points are easily affected by factors including point cloud filtering and oscillation, in order to accurately obtain the position of the feature points, the system uses the feature structure point cloud fitting method to perform fine feature point extraction.

[0013] 3) Point cloud pose calculation based on time series

[0014] During the trimming process, the milling processing module needs to adjust its posture according to the reference processing surface of the frame. According to the process information, the reference surface has a pitch angle change. Therefore, the tool coordinate system established on the processing tool spindle needs to rotate with the change of the pitch angle under the control of the system. After adjusting the base coordinate system and the tool coordinate system to a reasonable posture, the system will adjust the posture with the main motion direction of the milling process as the axis. In the online trimming mode, since the processing process has high requirements for real-time performance, the system will directly use the original image obtained by the line laser scanning to perform relative posture calculation based on time series.

[0015] 4) Path planning based on feature points

[0016] After acquiring the feature points, the system will perform path planning based on this series of feature points. The system will re-segment the path points based on the feature points and use the interpolation method to fully utilize the changing trends of each feature point within the interpolation segment to calculate the processing path.

[0017] 5) Processing trajectory fitting and re-segmentation

[0018] The above-mentioned path planning function based on feature points is mainly to meet the high real-time system processing requirements, and is mostly used in online processing mode. On this basis, the offline processing mode has added the function of processing trajectory fitting and re-segmentation to further improve the accuracy and rationality of path planning. Through the path planning function, the system obtains a series of path points calculated based on the feature points, and fits the path points with high-order curves to form a high-order curve expression in space, thereby unifying the path points. Then, based on the high-order curve expression, the system will re-segment the path according to the required speed, calculate the coordinates of each path point at the corresponding speed, and guide the movement of the milling processing module at the end of the robot arm.

[0019] Specifically, the present invention provides a method for aircraft skin feature recognition and edge milling path planning, the method comprising:

[0020] S1. Preprocess the original point cloud image;

[0021] S2. Point cloud image feature extraction;

[0022] S3. Point cloud pose calculation based on time series;

[0023] S4. Path planning based on feature points;

[0024] S5. Processing trajectory fitting and re-segmentation.

[0025] Furthermore, the pre-processing of the original point cloud image in step S1 of the aircraft skin feature recognition and milling path planning method of the present invention includes:

[0026] S11. Analyze the point cloud results obtained from the simulation process experiment;

[0027] S12. Preprocess the original point cloud image using threshold filtering and Gaussian filtering based on the analysis results of the previous step;

[0028] S13. Use the time-series-based adjacent interval mean filtering method of the line laser scanner to optimize the original point cloud image.

[0029] Furthermore, the point cloud image feature extraction described in step S2 of the aircraft skin feature recognition and milling path planning method of the present invention includes:

[0030] S21. Extract and segment the object from the preprocessed point cloud image to distinguish the skin from the reference frame and determine the location of the features;

[0031] S22. Derivative the reference frame point cloud image, roughly extract feature points using the reference frame surface variation pattern, and segment the feature structure of the reference frame based on this.

[0032] S23. Use the feature structure point cloud fitting method to extract feature points accurately and obtain the accurate feature point positions.

[0033] Furthermore, the point cloud image feature extraction described in step S2 of the aircraft skin feature recognition and milling path planning method of the present invention includes: firstly performing a first-order discrete derivative calculation on the preprocessed point cloud image to obtain a first-order discrete derivative point cloud graphic of the feature points corresponding to the point cloud, and performing segmentation of the processing area through extreme point screening; on this basis, performing a discrete derivative calculation on the first-order discrete derivative point cloud graphic again to obtain a second-order discrete derivative point cloud graphic, and performing segmentation of the feature structure through extreme point screening to complete the picking and segmentation of the features.

[0034] Furthermore, the method described in step S23 of the aircraft skin feature recognition and milling path planning method employs feature structure point cloud fitting to extract feature points accurately and obtain accurate feature point locations, including: utilizing different feature regions and line point cloud imaging characteristics to recalculate the locations of feature points; employing the RANSAC fitting algorithm to perform line fitting, calculating actual feature points through intersection, randomly sampling point cloud data to form a sample set, and iteratively calculating corresponding line parameter results within a target confidence range:

[0035]

[0036] Among them, M is the number of sampling times, N is the minimum number of sampling points required to solve the model, and P is the probability that the corresponding point is an interior point;

[0037] The parameters of the straight line equation obtained by the above algorithm are expressed as a point-wise equation:

[0038]

[0039] Among them, the direction vector is s(m, n, p), and a coordinate point on the straight line is (x0, y0, z0).

[0040] Furthermore, the time-series-based point cloud pose calculation described in step S3 of the aircraft skin feature recognition and milling path planning method of the present invention utilizes point cloud features at adjacent moments and obtains the relative pose relationship of the tool coordinate system at corresponding moments by comparing the pose changes between the two, including:

[0041] S31. Setting the movement direction of the tool coordinate system to coincide with the base coordinate system and the main movement direction of the milling process;

[0042] S32. The tool coordinate system established on the machining tool spindle rotates as the pitch angle of the reference machining surface of the reference frame changes, and the base coordinate system and the tool coordinate system are adjusted to a reasonable position;

[0043] S33. The main motion direction of the milling process is used as the axis to adjust the posture;

[0044] S34. In the online trimming mode, the original point cloud image obtained by line laser scanning is directly used to perform time-series-based relative posture calculation.

[0045] Furthermore, the path planning based on feature points described in step S4 of the aircraft skin feature recognition and milling path planning method of the present invention includes: after obtaining the feature points, re-segmenting the path points based on the feature points, using the interpolation method, and utilizing the change trend of each feature point in the interpolation segment to calculate and obtain the processing path.

[0046] Furthermore, the machining trajectory fitting and re-segmentation described in step S5 of the aircraft skin feature recognition and milling path planning method of the present invention includes:

[0047] S51. For the offline processing mode, through the path planning based on the feature points in step S4, a series of path points are obtained based on the feature point calculation, and a high-order curve fitting is performed on the obtained path points to form a high-order curve expression of the space;

[0048] S52. Based on the high-order curve expression obtained in the previous step, the path is re-divided according to the required speed, the coordinates of each path point at the corresponding speed are calculated, and the movement of the milling processing module is guided.

[0049] On the other hand, the present invention also provides an aircraft skin feature recognition and milling path planning system. This system targets the aircraft skin frame structure, uses a KUKA industrial robot as a carrier, utilizes a line laser scanner to obtain skin trimming reference information, plans the processing path in real time according to the processing technology, and guides the robot to complete the shape-conforming motion task;

[0050] The system provides two data processing modes: online and offline;

[0051] The system consists of a software system and a hardware platform, wherein the software system is a data analysis and processing system located on the host computer; the hardware platform includes a 3D line laser scanner, a KUKA industrial robot, a milling processing module, a visual calibration plate, and system assembly structural parts.

[0052] The system operates according to the above-mentioned aircraft skin feature recognition and milling path planning method.

[0053] In addition, the present invention also relates to the application of the above-mentioned aircraft skin feature recognition and milling path planning method and system in the aircraft assembly industry.

[0054] In summary, the aircraft skin feature recognition and milling path planning method and system of the present invention have the following advantages:

[0055] 1) During the feature point extraction process, there may be reference surface defects, measurement errors, etc., which may lead to feature point recognition errors and deviations. This system uses the continuity of feature point changes in the processing process to optimize the feature points based on time series consistency to avoid the occurrence of sudden feature points that affect the generation of the processing path.

[0056] 2) This system uses the machining reference surface feature structure for fitting, comparing the original point cloud images at adjacent moments, mapping the relative poses of the two in the base coordinate system, and quickly confirming the pose state of the point cloud at the target moment based on the relative pose relationship. The above is the method for processing the point cloud pose during online trimming. In offline mode, the system uses an absolute pose calculation method, using the actual pose state of the feature structure in the visual measurement coordinate system and the base coordinate system to obtain the spatial shape and position of the feature structure used for machining through fitting. The mapping angle between the feature structure and the base coordinate system is then directly calculated to obtain the spatial pose of the target tool coordinate system at the specified moment.

[0057] 3) In order to ensure the continuity of the processing and the real-time performance of the system path planning, this system uses the mean filtering method to pre-process the feature points for consistency, reduce the oscillation error caused by line laser scanning, avoid excessive path jitter, and at the same time ensure the stability of parameters such as speed during the processing, thereby improving the smoothness of the processing process. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments of the present invention. Obviously, the following drawings are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0059] Figure 1 The figure is a flow chart of the overall implementation of the method of the present invention.

[0060] Figure 2 This is the effect diagram of point cloud region segmentation in the method of the present invention.

[0061] Figure 3 This is the effect diagram of point cloud derivation in the method of the present invention.

[0062] Figure 4 Schematic diagram of precise extraction of feature points in point cloud in the method of the present invention. DETAILED DESCRIPTION

[0063] To make the objectives, technical solutions, and advantages of the present invention more clearly apparent, the technical solutions of the present invention will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The present invention may also be implemented or applied through different specific implementation methods, and the details in this specification may be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention.

[0064] At the same time, it should be understood that the scope of protection of the present invention is not limited to the specific embodiments described below; it should also be understood that the terms used in the embodiments of the present invention are for describing specific embodiments rather than for limiting the scope of protection of the present invention.

[0065] Example: A method for aircraft skin feature recognition and milling path planning

[0066] This method targets the skin frame structure of an aircraft. Using a KUKA industrial robot as a carrier, a line laser scanner is used to obtain the skin trimming benchmark information. Both online and offline data processing modes are provided, and a reasonable and smooth processing path is formed according to the process. This guides the robot to complete the conformal motion task and improves the adaptability of the processing process.

[0067] The implementation system for this method is primarily software-based, with a hardware platform established for verification. The software system primarily comprises a data analysis and processing system located on a host computer; the hardware platform primarily consists of a 3D line laser scanner, a KUKA robot, a milling module, a visual calibration plate, and the necessary components for system assembly and functional implementation. Through the integration of the hardware platform and software system, this system enables precise machining benchmark extraction and real-time machining path planning in both online and offline modes.

[0068] Overall, if Figure 1 As shown, the method includes the following steps:

[0069] (1) Preprocessing the original point cloud image;

[0070] (2) point cloud image feature extraction;

[0071] (3) Point cloud pose calculation based on time series;

[0072] (4) Path planning based on feature points;

[0073] (5) Processing trajectory fitting and re-segmentation.

[0074] The system measurement benchmark used in this method is composed of a frame and a skin. The frame and its characteristic structure serve as the reference for conformal processing during the trimming process. Therefore, it is necessary to perform regional segmentation and structural segmentation on the collected point cloud.

[0075] After filtering the original point cloud image, we can get Figure 2 The skin and frame structure shown in the figure clearly distinguishes different feature points and the characteristic structures demarcated by these feature points, such as chamfers and planes. Based on the characteristics of point cloud imaging, its image is primarily composed of discrete point cloud data, while the single sampling data obtained by the line laser scanner is an ordered point cloud line. Therefore, information analysis can be performed on specific points in the point cloud to determine their corresponding significance and complete the segmentation of the point cloud structure.

[0076] From the point cloud data collection results, it can be seen that whether it is the division of the processing object or the division of the characteristic structure, the characteristic points have obvious turning characteristics, that is, through this characteristic point, the surface characteristics of the processing base change to a certain extent. Therefore, according to the different degrees of change and by analogy with the relevant theories of characteristic points in continuous functions, the above characteristic points can be classified. The classification method is shown in the following table:

[0077] Table 1 Feature point classification table

[0078]

[0079]

[0080] As can be seen from the table above, when a step mutation occurs at the discontinuity point, the point cloud image will have a more obvious discontinuity feature. The boundary point can be expressed as the second type of feature point, such as Figure 3 The segmentation feature points between the middle skin structure and the reference frame structure. In the reference frame, the chamfer boundary points change the direction of the structures on both sides, causing them to change continuously. Therefore, they can be considered to meet the characteristics of the first type of feature points. That is, through this feature point, the point cloud representation structure still shows a continuous change trend and forms structures such as chamfers and planes.

[0081] According to the theory of discontinuity, its derivative is mainly applied to continuous functions. While point cloud images are mostly composed of discrete points, this point cloud data can be presented in an orderly and dense manner. Therefore, the slope of the discrete points can be used to approximate the derivative function at the corresponding position, thereby achieving feature point classification of the target point. As shown in the following formula, the slope of the discrete points can be calculated as:

[0082]

[0083] Where k ij is the slope of the discrete point, (x i ,y i )、(x j ,y j ) are the coordinates of discrete points.

[0084] Due to the characteristics of point cloud data such as discretization and weak correlation, the data of each point has a certain degree of randomness, which has a certain impact on the feature point screening and recognition process. In order to improve the stability and accuracy of the process, this method uses the method of adjacent interval mean filtering to optimize the target points, as shown below:

[0085]

[0086] The above calculation can obtain the first-order discrete derivative point cloud graphics corresponding to the feature points of the point cloud. By screening the extreme points, the second type of discontinuity points can be obtained to segment the processing area. Then, the discrete derivative calculation is performed again on the first-order discrete derivative point cloud graphics to obtain the second-order discrete derivative point cloud graphics. By screening the extreme points, the first type of discontinuity points can be obtained to segment the feature structure, completing the feature picking and segmentation.

[0087] The above-mentioned feature structure segmentation method based on point cloud images can obtain different areas including processing areas and feature structures, and can classify feature points based on the first-order and second-order discrete derivative point clouds of the point cloud image. However, during the feature point extraction process of the point cloud, due to the filtering operation in the preprocessing process, there is a certain error in the feature point extraction. Therefore, this system proposes a feature point extraction method based on point cloud images.

[0088] This method consists of two main steps: coarse and fine feature point extraction. Coarse feature point extraction relies primarily on the processed point cloud image, directly extracting it from the discrete derivative point cloud. In the first-order discrete derivative point cloud, extreme values ​​are extracted to obtain segmentation points for the image's processing region. In the second-order discrete derivative point cloud, due to the characteristics of the chamfered boundary, both edges must be extracted, which manifest as two extreme points. Using region screening, the system can extract extreme points in the extreme region and map them to the original point cloud, resulting in a coarse extraction of feature points.

[0089] After completing the rough extraction of feature points, the reference frame area can be divided into three different areas by feature points, including two slopes and one chamfer. In order to accurately calculate the positional relationship corresponding to the chamfer boundary, this system proposes a method for fine extraction of feature points, that is, using different feature areas and line point cloud imaging characteristics to recalculate the location of feature points. In the point cloud coordinate system, two feature points can be described in the form of feature intersection, and the corresponding features are mostly straight lines. Therefore, the system uses the RANSAC fitting algorithm for straight line fitting and calculates the actual feature points through intersection. Figure 4 As shown in the figure, the algorithm will use point cloud data for random sampling to form a sample set, and obtain the corresponding straight line parameter results with confidence within the target range through a large number of iterative calculations:

[0090]

[0091] Where M is the number of sampling times, N is the minimum number of sampling points required to solve the model, and P is the probability that the corresponding point is an inlier.

[0092] The parameters of the straight line equation obtained by this algorithm are expressed as a point-wise equation:

[0093]

[0094] Among them, the direction vector is s(m, n, p), and a coordinate point on the straight line is (x0, y0, z0).

[0095] During the actual machining process, the system needs to adjust its posture according to the inclination characteristics of the reference frame to control the tool coordinate system set in the milling processing module and then drive the tool to adjust to achieve more accurate conformal machining. The traditional posture solution method is to calculate the inclination angle in the visual coordinate system through the line point cloud plane structure obtained by the line laser scanner, and then convert it to the tool coordinate system and the robot base coordinate system. After that, its posture state in space is calculated, and the machining spindle corresponding to the tool coordinate system is coincided with it. The tool coordinate system solution and the corresponding robot arm posture state are inversely solved at each moment. However, this process is often accompanied by more coordinate transformation processes, longer coordinate transformation time and lower-precision solution results. Therefore, while ensuring the machining process, this system combines the actual machining process and control method to propose a relatively efficient and high-precision relative posture calculation method.

[0096] The system sets the tool coordinate system's motion direction to coincide with the base coordinate system and the principal motion direction of the actual machining process, minimizing the number of degrees of freedom required for posture transformations and limiting them to pitch angle adjustments. The system then leverages the consistency and continuity between the machined skin and the reference frame structure during machining to transform complex posture changes into a feature-based continuous process. This translates absolute posture calculations at individual moments into relative posture calculations over continuous time.

[0097] The time-series-based point cloud relative posture calculation method mainly utilizes the point cloud features of adjacent moments, and obtains the relative posture relationship of the tool coordinate system at the corresponding moment by comparing the posture changes of the two. At adjacent measurement moments, the visual measurement coordinate system will produce posture changes in the base coordinate system as the posture of the tool coordinate system at the current end of the robot arm changes. Therefore, for the measurement datum at adjacent moments, although the datum to be measured remains basically unchanged due to its structural consistency and continuity, its plane posture in the visual measurement coordinate plane will change significantly. Due to the processing process of the system, the direction of the milling tool is determined by the plane outside the chamfer, so the features corresponding to the plane structure will directly determine the direction of the tool spindle. In the above method, after feature point extraction and feature structure segmentation, the system fits the feature using the RANSAC method to obtain the point-direction formula of the straight line where it is located, and obtains the unit plane vector of the visual coordinate system in that direction. Similarly, based on the point cloud image at the previous moment, the system can also extract the unit vector of the feature structure at that moment. As shown in the following formula:

[0098]

[0099] in, Indicates T i-1 The characteristic vector a at time Indicates T i-1 Always T iThe tool coordinate system posture transformation at the moment can be obtained by calculating the posture transformation angle of the robot end.

[0100] From the angle calculation formula we can know:

[0101]

[0102] The direction of rotation can be determined by the cross product of the vectors:

[0103]

[0104] After extracting and optimizing feature points based on point cloud images, the feature points can more accurately describe the path, but there is still a certain degree of fluctuation. Therefore, the system will optimize and fit the feature points used to generate the path to ensure the accuracy and smoothness of the final path points in both online and offline modes.

[0105] Based on the original path points, the system uses a point cloud smoothing filtering method and a mean filtering method to iteratively process the path points to obtain a point cloud path with stronger correlation and consistency, as shown in the following formula:

[0106]

[0107] That is, the system takes the target point as the center and performs averaging processing on a certain range of points in the neighborhood, where n=3 is initially set.

[0108] The above method can greatly optimize the processing of path points and can more accurately describe the processing path. Since the sampling and point cloud processing method uses interval processing to increase the system's adaptability and to avoid response delays caused by unstable processing time by means of redundant cycles, the path point cannot be directly used for processing path guidance and the system needs to re-segment the path point. When the interval period is known, the system will use the interpolation method to process the path point. Under the above premise, the following formula is obtained:

[0109]

[0110] Wherein, i=0, 1, 2, ..., 7, 8.

[0111] In offline processing mode, the system will perform spatial high-order curve fitting based on the path points, that is, using the least squares method to obtain the final curve parameters by solving the overdetermined equation. As shown in the following formula, for an N-order curve polynomial:

[0112]

[0113] That is, the above formula can be constructed into a system of equations:

[0114] Y = XK;

[0115] Where Y is the [M×1]-dimensional matrix composed of parameter points, X is the [M×6]-dimensional matrix composed of parameter points, M represents the number of coordinate points involved in the fitting, and K is the [6×1]-dimensional coefficient matrix to be calculated. Using the least squares solution theorem for overdetermined equations, we can obtain:

[0116] K=(X T X) -1 X T Y;

[0117] Then, the coefficient matrix K is obtained through this matrix operation.

[0118] Through the above method, the system will obtain the high-order fitting curve corresponding to the set of feature points, and proportionally divide the acquired path through the quantitative interception method, thereby realizing the re-segmentation of the path points and providing the system with accurate processing path guidance.

[0119] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Any technician familiar with the profession can make some changes or modifications to the technical content disclosed above without departing from the scope of the technical solution of the present invention to obtain equivalent embodiments with equivalent changes. However, any simple modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention should be included in the scope of protection of the claims of the present invention.

Claims

1. A method for aircraft skin feature recognition and edge milling path planning, characterized in that: The method comprises: S1. Preprocess the original point cloud image; S2. Point cloud image feature extraction, including: S21. Extract and segment the object from the preprocessed point cloud image to distinguish the skin from the reference frame and determine the location of features. S22. Derivative the reference frame point cloud image, roughly extract feature points based on the reference frame surface variation pattern, and then segment the reference frame's characteristic structure. S23. Use feature structure point cloud fitting to extract feature points and obtain accurate feature point locations; S3. Time-series-based point cloud pose calculation uses point cloud features at adjacent moments and compares their pose changes to obtain the relative pose relationship of the tool coordinate system at corresponding moments. This includes: S31. Setting the movement direction of the tool coordinate system to coincide with the base coordinate system and the main movement direction of milling; S32. The tool coordinate system established on the machining tool spindle rotates as the pitch angle of the reference machining surface of the reference frame changes, adjusting the base coordinate system and the tool coordinate system to a reasonable position; S33. The main motion direction of the milling process is used as the axis to adjust the posture; S34. In the online trimming mode, the original point cloud image obtained by line laser scanning is directly used to perform time-series-based relative pose calculation. S4. Feature point-based path planning, including: after acquiring feature points, re-segmenting the path points based on the feature points, using interpolation to calculate the machining path based on the changing trends of each feature point within the interpolated segment; S5. Processing trajectory fitting and re-segmentation, including: S51 for offline processing mode, through step S4 based on the path planning of the feature points, based on the calculation of the feature points to obtain a series of path points, the path points obtained by high-order curve fitting, forming a high-order curve expression of the space; S52. Based on the high-order curve expression obtained in the previous step, the path is re-divided according to the required speed, the coordinates of each path point at the corresponding speed are calculated, and the movement of the milling processing module is guided.

2. The method for aircraft skin feature recognition and edge milling path planning according to claim 1, characterized in that: The pre-processing of the original point cloud image described in step S1 includes: S11. Analyze the point cloud results obtained from the simulated machining process experiment; S12. Preprocess the original point cloud image using threshold filtering and Gaussian filtering based on the analysis results in the previous step; S13. Optimize the original point cloud image using a time-series-based neighborhood mean filtering method using a line laser scanner.

3. The method for aircraft skin feature recognition and edge milling path planning according to claim 1, characterized in that: The point cloud image feature extraction described in step S2 includes: firstly performing a first-order discrete derivative calculation on the preprocessed point cloud image to obtain a first-order discrete derivative point cloud graphic corresponding to the feature points of the point cloud, and then performing segmentation of the processing area by screening the extreme points; on this basis, performing a discrete derivative calculation on the first-order discrete derivative point cloud graphic again to obtain a second-order discrete derivative point cloud graphic, and then performing segmentation of the feature structure by screening the extreme points to complete the picking and segmentation of the features.

4. The method for aircraft skin feature recognition and edge milling path planning according to claim 3, characterized in that: The feature point extraction method described in step S23 is used to accurately extract feature points and obtain accurate feature point locations, including: using different feature regions and line point cloud imaging characteristics to recalculate the location of feature points; using the RANSAC fitting algorithm to perform line fitting, and calculating the actual feature points through intersection, using point cloud data for random sampling to form a sample set, and obtaining the corresponding line parameter results with confidence within the target range through iterative calculation: ； Among them, M is the number of sampling times, N is the minimum number of sampling points required to solve the model, and P is the probability that the corresponding point is an interior point; The parameters of the straight line equation obtained by the above algorithm are expressed as a point-wise equation: ； The direction vector is , a coordinate point on the straight line is .

5. An aircraft skin feature recognition and milling path planning system, characterized in that: The system targets aircraft skin frame structures, uses a KUKA industrial robot as a carrier, utilizes a line laser scanner to obtain skin trimming benchmark information, plans the processing path in real time according to the processing technology, and guides the robot to complete the contour motion task; The system provides two data processing modes: online and offline; The system consists of a software system and a hardware platform. The software system is a data analysis and processing system located on the host computer; the hardware platform includes a 3D line laser scanner, a KUKA industrial robot, a milling processing module, a visual calibration plate, and system assembly components. The system operates according to the aircraft skin feature recognition and edge milling path planning method according to any one of claims 1-4.

Images