Three-dimensional reconstruction method and system for machining curved workpiece based on point cloud data

Abstract

The application discloses a curved surface workpiece machining three-dimensional reconstruction method and system based on point cloud data, relates to the field of data processing, and comprises the following steps: a collecting module is used for adaptively adjusting the scanning path, sampling density and scanning distance parameters of a collecting device, and driving the collecting device to acquire workpiece surface point cloud data; a preprocessing module is used for dividing levels according to the curvature variation of the workpiece surface, removing abnormal points in each level and completing missing point clouds; the application flexibly adjusts the scanning parameters in combination with the workpiece machining area parameters, ensures that the point cloud data collection is comprehensive and accurate, can also optimize point cloud processing according to the curvature, accurately extracts features and constructs a model in layers, guarantees the integrity of the model through splicing constraints, and ensures that the model meets the machining requirements through accuracy verification and correction.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to a method and system for three-dimensional reconstruction of curved workpieces based on point cloud data. Background Technology

[0002] Curved surface workpieces are components with curved surfaces, lacking obvious flat surfaces or edges. They are commonly found in automotive bodies, aircraft engine blades, and mold cavities. They require machining processes such as CNC milling and 3D printing, and have high requirements for precision and surface roughness.

[0003] Patent application number 202410096929.2 discloses a method for three-dimensional reconstruction of tool surfaces based on non-contact scanning measurement, including the following steps: Step 1: Tool point cloud data acquisition and preprocessing; Acquiring tool point cloud data: Using a machine tool servo system to control a line laser scanner for measurement motion, acquiring tool point cloud data from multiple positions and angles; Preprocessing tool point cloud data: Treating the three-dimensional point cloud data as two-dimensional data, sequentially processing it for missing values, simplification, and outliers to optimize point cloud quality and improve data usability; Step 2: Tool point cloud data rotation stitching based on tool machining characteristics and rotation measurement motion, Multiple sets of point cloud data are stitched together to obtain complete and accurate point cloud data: Step 3: 3D reconstruction of the tool surface. This application aims to solve the problem that "traditional measurement methods include non-contact offline measurement and contact on-machine measurement. Among them, offline measurement requires removing the machining tool from the machine tool for inspection, which easily causes problems such as secondary clamping errors and changes in machining datum, reducing the accuracy and efficiency of tool measurement; contact on-machine measurement mostly requires point-by-point measurement with probes, which has low measurement efficiency and is often limited by the size of the probe, making it difficult to meet the high-precision measurement requirements of the tool's microstructure. Overall, traditional measurement methods have low measurement efficiency and limited measurement accuracy."

[0004] However, in scenarios where there are existing curved workpiece samples, the surface parameters of the curved workpiece samples are collected and then the model is constructed and the curved workpiece is manufactured with reference to the model, the existing technology has poor accuracy in collecting and modeling the structural parameters of the curved workpiece samples. This results in the need for repeated calibration and improvement of the curved workpieces produced in the future, which is not ideal in terms of efficiency and cost.

[0005] To address this, we propose a method and system for 3D reconstruction of curved workpieces based on point cloud data. Summary of the Invention

[0006] In view of the above-mentioned shortcomings of the existing technology, the present invention provides a method and system for three-dimensional reconstruction of curved workpieces based on point cloud data, which can effectively solve the problems of the existing technology.

[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions;

[0008] This invention discloses a three-dimensional reconstruction system for machining curved workpieces based on point cloud data, comprising:

[0009] The acquisition module adaptively adjusts the scanning path, sampling density, and scanning distance parameters of the acquisition device to drive the acquisition device to acquire point cloud data of the workpiece surface. The preprocessing module divides the acquired point cloud data into layers according to the curvature changes of the workpiece surface, removes outliers within each layer, and completes the missing point cloud. The extraction module receives the layered point cloud data after the missing point cloud is completed, identifies the feature regions of the workpiece surface, and extracts the spatial coordinates and topological relationships of the feature points. The construction module receives the spatial coordinates and topological relationships of the feature points, applies them, constructs local 3D models of each region of the workpiece according to the layered order of the point cloud, establishes splicing constraints between the local 3D models, and splices the local 3D models to obtain the overall 3D model. The verification module compares the constructed overall 3D model with the preset processing accuracy requirements of the workpiece, automatically verifies the accuracy deviations of each region of the model and at splicing points, and marks the out-of-tolerance areas. The output module obtains the verification results, adjusts the parameter format of the overall 3D model based on the verification results, and outputs model data adapted to the surface workpiece processing equipment.

[0010] During the acquisition module's operation phase, the acquisition equipment is adjusted based on the curvature range, dimensional boundaries, and accuracy level parameters of the preset processing area of ​​the curved workpiece.

[0011] Furthermore, when the acquisition module adaptively adjusts the acquisition device parameters, it first calculates the curvature fluctuation coefficient based on the curvature range of the preset processing area of ​​the curved workpiece:

[0012] ;

[0013] In the formula: These are the maximum and minimum curvature values ​​within the preset processing area; The average curvature value within the preset processing area;

[0014] Then, determine the sampling density adjustment coefficient based on the curvature fluctuation coefficient:

[0015] when When ≤0.3, λ=1.0, and the sampling density remains the initial reference density. ;

[0016] When 0.3 < When ≤0.6, The sampling density was adjusted to ;

[0017] When >0.6, The sampling density was adjusted to ;

[0018] Based on the preset dimensions, including length, width, height, and accuracy level of the processing area, the scanning distance is calculated. , Indicates the length, width, and height of the dimension boundaries. The accuracy level is indicated, and the final scanning path is planned according to the sampling density D and the scanning distance d: a spiral scanning path is adopted, with the geometric center of the preset processing area as the scanning starting point. The spiral radius of the scanning path gradually increases from 0 to the maximum circumscribed circle radius of the processing area, and the distance between adjacent spiral loops is strictly equal to d. At the same time, during the scanning path planning process, it is determined in real time whether the path covers all boundary points of the preset processing area. If there are uncovered boundary points, a supplementary scanning path segment parallel to the boundary is added in the corresponding boundary area. The connection between the supplementary path segment and the main spiral path is a rounded transition with a transition arc radius of d / 2, so that the scanning path completely covers the preset processing area and there is no duplicate scanning area.

[0019] in, , Indicates the surface roughness of the workpiece. This indicates the form and position tolerances of the workpiece surface.

[0020] Furthermore, when the preprocessing module divides the collected point cloud data into layers according to the curvature variation of the workpiece surface, it first divides the curvature value of the workpiece surface into n curvature intervals. ,in, These represent the minimum curvature and maximum curvature of the workpiece surface, respectively, and the difference between adjacent curvature intervals. equal;

[0021] Iterate through all the collected point cloud data and calculate the surface curvature corresponding to each point cloud data. ,Will Point cloud data falling within the same curvature interval are grouped into the same level, resulting in n point cloud levels. When removing outliers within each level, for any point p in the point cloud data of each level, the number of point clouds in its neighborhood of radius r is searched. Calculate the average number of neighborhood points for all point clouds within this level. with standard deviation ,when When this happens, point p is determined to be an outlier and is removed.

[0022] When completing a missing point cloud, for the missing regions formed after removing outliers, the point cloud data of the boundaries of the missing regions is obtained, and the local surface equations of the boundary point cloud are established. By generating virtual points at intervals of 0.5×D within the missing region, and substituting the virtual points into the local surface equation to calculate their z-coordinates, the missing point cloud is completed.

[0023] Where D represents the sampling density of the corresponding level.

[0024] Furthermore, when the extraction module identifies the feature regions of the workpiece surface, it simultaneously calculates the normal vector of each point in the layered point cloud data that has been completed to fill in the missing point cloud. and the angle between the normal vectors of adjacent points. , Represents the normal vector between two adjacent points, when When the angle exceeds 30 degrees, the region containing the corresponding adjacent points is marked as a candidate region for the feature region.

[0025] Next, calculate the rate of curvature change of the point cloud within the candidate feature region. , This represents the curvature of points i and j within the candidate region. Represents the Euclidean distance between two points, when > When this happens, the candidate region is determined to be a characteristic region of the workpiece surface;

[0026] In the stage of extracting the spatial coordinates and topological relationships of feature points, the spatial coordinates of all points within the feature region are obtained. And construct the adjacency matrix M of the feature points, which follows the following:

[0027] ;

[0028] In the formula: The threshold for adjacency determination is , and The value is set to 1.2 times the sampling density of the corresponding layer of the feature region, and the topological relationship of the feature points is represented by the adjacency matrix.

[0029] Furthermore, when the construction module constructs local 3D models of various regions of the workpiece using the spatial coordinates and topological relationships of feature points, it first fits the local surface of each region corresponding to each point cloud level using the spatial coordinates of feature points within that region, employing the moving least squares method. The fitting function is:

[0030] ;

[0031] In the formula: This is the final output of a two-dimensional to three-dimensional surface mapping relationship, used to characterize the surface morphology of local regions; The number of feature points involved in the fitting; i is the feature point index; For weighting functions; are basis functions;

[0032] , The distance from the feature point to the point to be fitted. The radius of the influence domain;

[0033] Based on the adjacency matrix M of the topological relationships of the feature points, the connection order of each point in the local surface is determined, and the mesh structure of the local 3D model is generated.

[0034] In the stage of establishing splicing constraints between local 3D models, the coordinate deviation of feature points at the splicing point of two adjacent local 3D models is calculated. , , Let k be the coordinates of the feature point k at the junction of the first local model. Let the coordinates of feature point k at the stitching point of the second local model be set, and set the stitching constraint as follows: , This indicates the splicing precision threshold, and The value is set to 0.8 times the preset machining accuracy requirement of the workpiece;

[0035] In the stage of stitching together the local 3D models, the spatial pose of the local 3D models is adjusted so that the coordinate deviation of the feature points at the stitching point meets the stitching constraint conditions, and finally the overall 3D model is constructed.

[0036] Furthermore, when the verification module compares the constructed overall 3D model with the preset machining accuracy requirements of the workpiece, it first selects m verification points on the overall 3D model according to the principle of uniform distribution, and obtains the model coordinates of each verification point. Simultaneously, the coordinates of the workpiece's actual position are obtained through measuring equipment. ;

[0037] Calculate the accuracy deviation at each verification point ;

[0038] Calculate the accuracy deviation of all verification points, and then calculate the average and maximum deviations.

[0039] The average deviation and the maximum deviation are respectively compared with the workpiece's preset machining accuracy requirements. In comparison, when the average deviation is less than or equal to And the maximum deviation is less than or equal to 1.2 times. When the model is deemed to have acceptable accuracy in that region, the average deviation is considered acceptable. Or the maximum deviation is greater than 1.2 times. When this happens, the model is deemed to have out-of-tolerance accuracy in that region;

[0040] For the model stitching area, additional k verification points are selected on the stitching boundary, and their accuracy deviation is calculated. If there is an accuracy deviation at any of the verification points at the stitching area... Greater than In the case of such a situation, the splicing point is marked as an out-of-tolerance area, and the coordinate range and out-of-tolerance value of the out-of-tolerance area are recorded.

[0041] Furthermore, when the output module adjusts the parameter format of the overall 3D model based on the verification results, if there are no out-of-tolerance parts in the verification results, it directly reads the model data format supported by the surface workpiece processing equipment, converts the overall 3D model into the corresponding format, and adjusts the unit parameters of the model to make the model units consistent with the unit settings of the processing equipment.

[0042] If the verification results show out-of-tolerance areas, first, based on the coordinate range and out-of-tolerance value of the out-of-tolerance areas, perform local correction on the out-of-tolerance areas of the overall 3D model. When correcting, use the point cloud data of the qualified area around the out-of-tolerance area as a reference, and refit the surface of the out-of-tolerance area. The fitting method is the same as the local surface fitting method in the construction module.

[0043] After the correction is completed, the accuracy verification process of the verification module is executed again until there are no out-of-tolerance parts in the model. Then, the corrected overall 3D model is converted into a model data format that is compatible with the processing equipment. When outputting the model data, the accuracy verification report of the model is also output. The report includes the number of verification points, the deviation value of each verification point, the correction status of out-of-tolerance parts, and the final accuracy index.

[0044] Among them, the model data formats supported by the surface workpiece processing equipment include STL, IGES, and STEP.

[0045] Furthermore, the output module also includes a database, which stores the raw point cloud data acquired by the acquisition module, the layered point cloud data processed by the preprocessing module, the feature point spatial coordinates and topological relationship data obtained by the extraction module, the local 3D model data and the overall 3D model data generated by the construction module, and the accuracy verification data of the verification module in the order of timestamps.

[0046] The database contains a data association index, which includes workpiece number, acquisition time, processing area parameters, accuracy requirements, and processing time of each module. All 3D reconstruction-related data for a workpiece can be quickly retrieved by workpiece number.

[0047] The database adopts a distributed storage architecture, storing different types of data on different storage nodes. The original point cloud data and the processed point cloud data are stored on high-speed read and write storage nodes, while the 3D model data and verification data are stored on large-capacity storage nodes. The stored data is stored in an encrypted manner, and the encryption key includes the workpiece number.

[0048] Furthermore, the acquisition module is interconnected with a preprocessing module and an extraction module via a wireless network. The extraction module is interconnected with a construction module via a wireless network. The construction module is interconnected with a verification module and an output module via a wireless network. The verification module is interconnected with the output module via a wireless network.

[0049] Secondly, a three-dimensional reconstruction method for machining curved workpieces based on point cloud data includes:

[0050] Referring to the preset processing area parameters of the workpiece, the scanning path, sampling density, and distance of the acquisition device are adaptively adjusted to drive the device to acquire surface point cloud data. The sampling density is determined according to the curvature fluctuation coefficient, and the path adopts a spiral pattern with supplementary boundary scanning segments. The point cloud is divided into layers according to the curvature of the surface, and points with abnormal numbers of neighboring points in each layer are removed. Virtual points are generated through the boundary local surface equation to complete the missing point cloud. The point cloud normal vector and the angle between adjacent points are calculated, and candidate areas with angles exceeding 30 degrees are marked. The feature region is determined by combining the curvature change rate, and the coordinates of the feature points are acquired simultaneously to construct an adjacency matrix to represent the topological relationship. The local surface is fitted with the feature point coordinates to generate a local 3D model, and splicing constraints are established. The posture of the local model is adjusted synchronously to ensure that the deviation meets the constraints, and the overall 3D model is obtained by splicing. Verification points are selected to compare the model with the actual coordinates of the workpiece to calculate the deviation. The regional accuracy is determined by combining the preset accuracy, and additional verification is performed at the splicing point. The out-of-tolerance parts and related information are marked. If there is no out-of-tolerance, the model format is converted and the units are unified. If there is an out-of-tolerance, it is corrected and re-verified. After passing the verification, the model and an accuracy report containing verification and correction information are output.

[0051] Compared with the known prior art, the technical solution provided by this invention has the following beneficial effects:

[0052] This invention provides a method and system for 3D reconstruction of curved workpieces based on point cloud data. During execution, the method and system adaptively adjust scanning parameters to acquire point cloud data of the workpiece surface. Combined with optimized acquisition based on preset processing area parameters, the data acquisition is improved in terms of targeting and efficiency. Point cloud data is processed layer by layer according to surface curvature changes, accurately removing outliers and filling in missing parts to ensure data quality. Feature regions are identified by calculating point cloud normal vectors and curvature change rates, and an adjacency matrix is ​​constructed to clarify the topological relationships of feature points, laying a precise foundation for model construction. Local 3D models are constructed in a layered order, and splicing constraints are established and the posture is adjusted to complete the overall splicing, improving model integrity and accuracy. Simultaneously, by comparing the model with the actual workpiece coordinates at multiple verification points, the accuracy is accurately determined, and out-of-tolerance areas are marked. Finally, the model parameter format is adjusted based on the verification results, and local corrections are made to meet the requirements when out-of-tolerance conditions exist. The system can also systematically store various types of data and support rapid retrieval, providing high-quality models and data support for curved workpiece processing. Attached Figure Description

[0053] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0054] Figure 1 This is a schematic diagram of a 3D reconstruction system for machining curved workpieces based on point cloud data.

[0055] Figure 2 This is a flowchart illustrating a three-dimensional reconstruction method for machining curved workpieces based on point cloud data. Detailed Implementation

[0056] 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, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0057] The present invention will be further described below with reference to embodiments.

[0058] Example 1:

[0059] This embodiment presents a 3D reconstruction system for machining curved workpieces based on point cloud data, such as... Figure 1 As shown, it includes:

[0060] The acquisition module is used to adaptively adjust the scanning path, sampling density and scanning distance parameters of the acquisition device to drive the acquisition device to acquire point cloud data of the workpiece surface;

[0061] When the acquisition module adaptively adjusts the acquisition device parameters, it first calculates the curvature fluctuation coefficient based on the curvature range of the preset processing area of ​​the curved workpiece:

[0062] ;

[0063] In the formula: These are the maximum and minimum curvature values ​​within the preset processing area; The average curvature value within the preset processing area;

[0064] The above formula uses the maximum curvature value, minimum curvature value and average curvature value within the preset processing area as core parameters. By calculating the difference between the maximum and minimum curvature values ​​and comparing it with the average curvature value, the degree of change in the curvature of the surface in the processing area is quantified. This formula can dynamically identify areas where the curvature of the surface changes gently or dramatically, avoiding data redundancy caused by oversampling in areas with gentle curvature, and also preventing feature loss caused by insufficient sampling in areas with dramatic curvature.

[0065] Then, determine the sampling density adjustment coefficient based on the curvature fluctuation coefficient:

[0066] when When ≤0.3, λ=1.0, and the sampling density remains the initial reference density. ;

[0067] When 0.3 < When ≤0.6, The sampling density was adjusted to ;

[0068] When >0.6, The sampling density was adjusted to ;

[0069] Based on the preset dimensions, including length, width, height, and accuracy level of the processing area, the scanning distance is calculated. , Indicates the length, width, and height of the dimension boundaries. The accuracy level is indicated, and the final scanning path is planned according to the sampling density D and the scanning distance d: a spiral scanning path is adopted, with the geometric center of the preset processing area as the scanning starting point. The spiral radius of the scanning path gradually increases from 0 to the maximum circumscribed circle radius of the processing area, and the distance between adjacent spiral loops is strictly equal to d. At the same time, during the scanning path planning process, it is determined in real time whether the path covers all boundary points of the preset processing area. If there are uncovered boundary points, a supplementary scanning path segment parallel to the boundary is added in the corresponding boundary area. The connection between the supplementary path segment and the main spiral path is a rounded transition with a transition arc radius of d / 2, so that the scanning path completely covers the preset processing area and there is no duplicate scanning area.

[0070] in, , Indicates the surface roughness of the workpiece. Indicates the form and position tolerances of the workpiece surface;

[0071] It should be noted that:

[0072] The surface roughness of the workpiece is determined by collecting the micro-profile curve data of the workpiece surface within a specified sampling length using a surface roughness meter. The deviation distance (offset) of each point on the profile curve relative to the profile centerline is read. The arithmetic mean of all offsets is calculated after taking the absolute value of each offset. The arithmetic mean deviation of the profile is then used as the specific value of the surface roughness.

[0073] Workpiece form and position tolerances (including form tolerances and position tolerances) are calculated by comparing the actual geometric parameters of the workpiece with the ideal geometric parameters and combining the tolerance zone requirements. For example, straightness is determined by measuring the maximum deviation between the actual straight line and the ideal straight line, and position tolerance is determined by calculating the degree of conformity between the coordinate deviation of the actual feature and the datum feature and the tolerance value.

[0074] The formula for calculating the scanning distance comprehensively considers two key indicators affecting machining accuracy: workpiece surface roughness and surface form and position tolerances. These two are integrated through a specific mathematical relationship to calculate the scanning distance. The design philosophy of this formula is to ensure that the scanning distance reflects the workpiece surface quality requirements, guaranteeing that the point cloud data acquired during scanning meets the accuracy requirements of subsequent processing. It avoids situations where inappropriate scanning distance leads to point cloud data failing to accurately represent workpiece surface features, making the determination of the scanning distance more scientific and targeted. It can automatically calculate the appropriate scanning distance based on workpieces with different accuracy requirements, effectively improving the matching degree between point cloud data and actual workpiece surface features, thus providing a guarantee for high-precision 3D reconstruction.

[0075] The preprocessing module is used to divide the collected point cloud data into layers according to the curvature changes of the workpiece surface, remove abnormal points in each layer, and fill in the missing point cloud.

[0076] When the preprocessing module divides the collected point cloud data into layers according to the curvature variation of the workpiece surface, it first divides the curvature values ​​of the workpiece surface into n curvature intervals. ,in, These represent the minimum curvature and maximum curvature of the workpiece surface, respectively, and the difference between adjacent curvature intervals. equal;

[0077] Iterate through all the collected point cloud data and calculate the surface curvature corresponding to each point cloud data. ,Will Point cloud data falling within the same curvature interval are grouped into the same level, resulting in n point cloud levels. When removing outliers within each level, for any point p in the point cloud data of each level, the number of point clouds in its neighborhood of radius r is searched. Calculate the average number of neighborhood points for all point clouds within this level. with standard deviation ,when When this happens, point p is determined to be an outlier and is removed.

[0078] When completing a missing point cloud, for the missing regions formed after removing outliers, the point cloud data of the boundaries of the missing regions is obtained, and the local surface equations of the boundary point cloud are established. By generating virtual points at intervals of 0.5×D within the missing region, and substituting the virtual points into the local surface equation to calculate their z-coordinates, the missing point cloud is completed.

[0079] Where D represents the sampling density of the corresponding level;

[0080] It should be noted that, This represents a functional relationship describing the local surface morphology of the missing region. The function is generated by fitting point cloud data of the missing region's boundary. Its function type is determined based on the actual curvature characteristics of the workpiece surface (e.g., a quadratic polynomial function can be used for low-curvature, gently curvature regions, while a spline function can be used for high-curvature, complex regions). The aim is to represent the two-dimensional coordinates (…). ) participate in function operations to obtain This ultimately yields the three-dimensional coordinates (x, y, z).

[0081] The extraction module is used to receive the layered point cloud data after the missing point cloud is completed, identify the feature regions of the workpiece surface, and extract the spatial coordinates and topological relationships of the feature points.

[0082] When the extraction module identifies the feature regions of the workpiece surface, it simultaneously calculates the normal vector of each point in the layered point cloud data that has been completed to fill in the missing point cloud. and the angle between the normal vectors of adjacent points. , Represents the normal vector between two adjacent points, when When the angle exceeds 30 degrees, the region containing the corresponding adjacent points is marked as a candidate region for the feature region.

[0083] Next, calculate the rate of curvature change of the point cloud within the candidate feature region. , This represents the curvature of points i and j within the candidate region. Represents the Euclidean distance between two points, when > When this happens, the candidate region is determined to be a characteristic region of the workpiece surface;

[0084] In the stage of extracting the spatial coordinates and topological relationships of feature points, the spatial coordinates of all points within the feature region are obtained. And construct the adjacency matrix M of the feature points, which follows the following:

[0085] ;

[0086] In the formula: The threshold for adjacency determination is , and The value is taken as 1.2 times the sampling density of the corresponding layer of the feature region, and the topological relationship of the feature points is represented by the adjacency matrix;

[0087] The module is used to receive the spatial coordinates and topological relationships of feature points, apply the spatial coordinates and topological relationships of feature points, construct local 3D models of each region of the workpiece according to the layering order of the point cloud, establish splicing constraints between local 3D models, and splice the local 3D models to obtain the overall 3D model.

[0088] When constructing local 3D models of various regions of a workpiece using the spatial coordinates and topological relationships of feature points, the module first fits the local surface of each region corresponding to each point cloud level using the spatial coordinates of feature points within that region. The fitting function is as follows:

[0089] ;

[0090] In the formula: This is the final output of a two-dimensional to three-dimensional surface mapping relationship, used to characterize the surface morphology of local regions; The number of feature points involved in the fitting; i is the feature point index; For weighting functions; are basis functions;

[0091] , The distance from the feature point to the point to be fitted. The radius of the influence domain;

[0092] The above formula is based on the spatial coordinates of feature points. The moving least squares method is used to construct the fitting function. The contribution of different feature points is weighted by a weight function in the function. Combined with the basis function, the two-dimensional to three-dimensional surface mapping relationship representing the surface morphology of the local region is finally obtained. This makes the fitted surface closer to the feature points with larger weights. At the same time, it takes into account the information of all feature points in the local region, realizes accurate fitting of the local surface, ensures the locality and smoothness of the fitting, and avoids the distortion of the local surface caused by global fitting. It provides core fitting method support for building high-precision local three-dimensional models.

[0093] Based on the adjacency matrix M of the topological relationships of the feature points, the connection order of each point in the local surface is determined, and the mesh structure of the local 3D model is generated.

[0094] In the stage of establishing splicing constraints between local 3D models, the coordinate deviation of feature points at the splicing point of two adjacent local 3D models is calculated. , , Let k be the coordinates of the feature point k at the junction of the first local model. Let the coordinates of feature point k at the stitching point of the second local model be set, and set the stitching constraint as follows: , This indicates the splicing precision threshold, and The value is set to 0.8 times the preset machining accuracy requirement of the workpiece;

[0095] In the stage of stitching together the local 3D models, the spatial pose of the local 3D models is adjusted so that the coordinate deviation of the feature points at the stitching point meets the stitching constraint conditions, and the overall 3D model is finally constructed.

[0096] The verification module is used to compare the constructed overall 3D model with the preset machining accuracy requirements of the workpiece, automatically verify the accuracy deviation of each area and splicing point of the model, and mark the out-of-tolerance parts.

[0097] When the verification module compares the constructed overall 3D model with the preset machining accuracy requirements of the workpiece, it first selects m verification points on the overall 3D model according to the principle of uniform distribution, and obtains the model coordinates of each verification point. Simultaneously, the coordinates of the workpiece's actual position are obtained through measuring equipment. ;

[0098] Calculate the accuracy deviation at each verification point ;

[0099] Calculate the accuracy deviation of all verification points, and then calculate the average and maximum deviations.

[0100] The average deviation and the maximum deviation are respectively compared with the workpiece's preset machining accuracy requirements. In comparison, when the average deviation is less than or equal to And the maximum deviation is less than or equal to 1.2 times. When the model is deemed to have acceptable accuracy in that region, the average deviation is considered acceptable. Or the maximum deviation is greater than 1.2 times. When this happens, the model is deemed to have out-of-tolerance accuracy in that region;

[0101] For the model stitching area, additional k verification points are selected on the stitching boundary, and their accuracy deviation is calculated. If there is an accuracy deviation at any of the verification points at the stitching area... Greater than In the case of this, the splicing point is marked as an out-of-tolerance area, and the coordinate range and out-of-tolerance value of the out-of-tolerance area are recorded;

[0102] The output module is used to obtain the verification results, adjust the parameter format of the overall 3D model based on the verification results, and output model data adapted to the surface workpiece processing equipment.

[0103] When the output module adjusts the parameter format of the overall 3D model based on the verification results, if there are no out-of-tolerance parts in the verification results, it directly reads the model data format supported by the surface workpiece processing equipment, converts the overall 3D model into the corresponding format, and adjusts the unit parameters of the model to make the model units consistent with the unit settings of the processing equipment.

[0104] If the verification results show out-of-tolerance areas, first, based on the coordinate range and out-of-tolerance value of the out-of-tolerance areas, perform local correction on the out-of-tolerance areas of the overall 3D model. When correcting, use the point cloud data of the qualified area around the out-of-tolerance area as a reference, and refit the surface of the out-of-tolerance area. The fitting method is the same as the local surface fitting method in the construction module.

[0105] After the correction is completed, the accuracy verification process of the verification module is executed again until there are no out-of-tolerance parts in the model. Then, the corrected overall 3D model is converted into a model data format that is compatible with the processing equipment. When outputting the model data, the accuracy verification report of the model is also output. The report includes the number of verification points, the deviation value of each verification point, the correction status of out-of-tolerance parts, and the final accuracy index.

[0106] Among them, the model data formats supported by the surface workpiece processing equipment include STL, IGES, and STEP;

[0107] The output module also contains a database, which stores the raw point cloud data acquired by the acquisition module, the layered point cloud data processed by the preprocessing module, the feature point spatial coordinates and topological relationship data obtained by the extraction module, the local 3D model data and the overall 3D model data generated by the construction module, and the accuracy verification data of the verification module in the order of timestamps.

[0108] The database contains a data association index, which includes workpiece number, acquisition time, processing area parameters, accuracy requirements, and processing time of each module. All 3D reconstruction-related data for a workpiece can be quickly retrieved by workpiece number.

[0109] The database adopts a distributed storage architecture, storing different types of data on different storage nodes. The original point cloud data and the processed point cloud data are stored on high-speed read and write storage nodes, while the 3D model data and the verification data are stored on large-capacity storage nodes. The stored data is stored in an encrypted manner, and the encryption key includes the workpiece number.

[0110] During the data acquisition module operation phase, the data acquisition equipment is adjusted by referring to the curvature range, size boundary, and accuracy level parameters of the preset processing area of ​​the curved workpiece.

[0111] The acquisition module interacts with the preprocessing module and the extraction module via a wireless network. The extraction module interacts with the construction module via a wireless network. The construction module interacts with the verification module and the output module via a wireless network. The verification module interacts with the output module via a wireless network.

[0112] In this embodiment, the acquisition module adaptively adjusts the scanning path, sampling density, and scanning distance parameters of the acquisition device to drive the acquisition device to acquire point cloud data of the workpiece surface. The preprocessing module runs afterward to divide the acquired point cloud data into layers according to the curvature changes of the workpiece surface, removes abnormal points in each layer, and fills in missing point clouds. Then, the extraction module receives the layered point cloud data with the missing point clouds filled in, identifies the feature regions of the workpiece surface, and extracts the spatial coordinates and topological relationships of the feature points. The construction module further receives the spatial coordinates and topological relationships of the feature points, applies the spatial coordinates and topological relationships of the feature points, and constructs local three-dimensional models of each region of the workpiece according to the layered order of the point cloud. It establishes splicing constraints between the local three-dimensional models, splices the local three-dimensional models to obtain the overall three-dimensional model, and compares the constructed overall three-dimensional model with the preset processing accuracy requirements of the workpiece through the verification module. It automatically verifies the accuracy deviation of each region of the model and the splicing points and marks the out-of-tolerance parts. Finally, the output module obtains the verification results, adjusts the parameter format of the overall three-dimensional model based on the verification results, and outputs model data adapted to the surface workpiece processing equipment.

[0113] In the above embodiments, the system flexibly adjusts scanning parameters in conjunction with workpiece processing area parameters during use to ensure comprehensive and accurate point cloud data acquisition. It can also optimize point cloud processing according to curvature, accurately extract features and construct models in layers, ensure model integrity through splicing constraints, and ensure that the model meets processing requirements through accuracy verification and correction. It is compatible with the output format of processing equipment and stores the entire process data for easy traceability. This effectively improves the efficiency and accuracy of 3D reconstruction of curved workpieces, provides reliable data support for subsequent processing, and reduces the risk of processing errors.

[0114] Example 2:

[0115] At the implementation level, based on Example 1, this example refers to... Figure 2 The three-dimensional reconstruction system for machining curved workpieces based on point cloud data in Example 1 will be further described in detail below:

[0116] A 3D reconstruction method for machining curved workpieces based on point cloud data includes:

[0117] Based on the preset processing area parameters of the workpiece, the scanning path, sampling density and distance of the acquisition device are adaptively adjusted to drive the device to acquire surface point cloud data. The sampling density is determined according to the curvature fluctuation coefficient, and the path adopts a spiral shape and is supplemented with boundary scanning segments.

[0118] The point cloud is divided into layers according to the curvature of the surface. Points with abnormal number of neighboring points in each layer are removed. Virtual points are generated by the boundary local surface equation to complete the missing point cloud.

[0119] Calculate the point cloud normal vector and the angle between adjacent points, mark candidate regions with angles exceeding 30 degrees, determine feature regions by combining the rate of curvature change, simultaneously obtain the coordinates of feature points, and construct an adjacency matrix to represent topological relationships.

[0120] A local 3D model is generated by fitting the coordinates of feature points to a local surface, a splicing constraint is established, the posture of the local model is adjusted synchronously to make the deviation meet the constraint, and the overall 3D model is obtained by splicing.

[0121] Select verification points to compare the model with the actual coordinates of the workpiece to calculate the deviation. Combine the preset accuracy to determine the accuracy of the area. Perform additional verification at the splicing point and mark the out-of-tolerance parts and related information.

[0122] If there are no out-of-tolerance errors, convert the model format and unify the units. If there are out-of-tolerance errors, correct them and re-verify. After passing the verification, output the model and an accuracy report containing verification and correction information.

[0123] In summary, the methods and systems described in the above embodiments, during execution, adaptively adjust scanning parameters to obtain point cloud data of the workpiece surface, optimize data acquisition by combining preset processing area parameters of the workpiece, improve the targeting and efficiency of data acquisition, process point cloud data in layers according to surface curvature changes, accurately remove abnormal points and fill in missing parts to ensure data quality, identify feature regions by calculating point cloud normal vectors and curvature change rates, construct adjacency matrices to clarify the topological relationships of feature points, lay an accurate foundation for model construction, construct local 3D models in layered order, establish splicing constraints and adjust posture to complete overall splicing, improve model integrity and accuracy, and accurately determine whether the accuracy is qualified by comparing the model with the actual coordinates of the workpiece through multiple verification points and mark out-of-tolerance parts. Finally, adjust the model parameter format according to the verification results, locally correct to qualified when out-of-tolerance exists, and can also store various types of data in an orderly manner and support fast retrieval, providing high-quality model and data support for curved workpiece processing.

[0124] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A three-dimensional reconstruction system for machining curved workpieces based on point cloud data, characterized in that, include: The acquisition module is used to adaptively adjust the scanning path, sampling density and scanning distance parameters of the acquisition device to drive the acquisition device to acquire point cloud data of the workpiece surface; When the acquisition module adaptively adjusts the parameters of the acquisition device, it first calculates the curvature fluctuation coefficient based on the curvature range of the preset processing area of ​​the curved workpiece: ； In the formula: These are the maximum and minimum curvature values ​​within the preset processing area; The average curvature value within the preset processing area; Then, determine the sampling density adjustment coefficient based on the curvature fluctuation coefficient: when When ≤0.3, λ=1.0, and the sampling density remains the initial reference density. ; When 0.3 < When ≤0.6, The sampling density was adjusted to ; When >0.6, The sampling density was adjusted to ; Based on the preset dimensions, including length, width, height, and accuracy level of the processing area, the scanning distance is calculated. , Indicates the length, width, and height of the dimension boundaries. The accuracy level is indicated, and the final scanning path is planned according to the sampling density D and the scanning distance d: a spiral scanning path is adopted, with the geometric center of the preset processing area as the scanning starting point. The spiral radius of the scanning path gradually increases from 0 to the maximum circumscribed circle radius of the processing area, and the distance between adjacent spiral loops is strictly equal to d. At the same time, during the scanning path planning process, it is determined in real time whether the path covers all boundary points of the preset processing area. If there are uncovered boundary points, a supplementary scanning path segment parallel to the boundary is added in the corresponding boundary area. The connection between the supplementary path segment and the main spiral path is a rounded transition with a transition arc radius of d / 2, so that the scanning path completely covers the preset processing area and there is no duplicate scanning area. in, , Indicates the surface roughness of the workpiece. Indicates the form and position tolerances of the workpiece surface; The preprocessing module is used to divide the collected point cloud data into layers according to the curvature changes of the workpiece surface, remove abnormal points in each layer, and fill in the missing point cloud. The extraction module is used to receive the layered point cloud data after the missing point cloud is completed, identify the feature regions of the workpiece surface, and extract the spatial coordinates and topological relationships of the feature points. The module is used to receive the spatial coordinates and topological relationships of feature points, apply the spatial coordinates and topological relationships of feature points, construct local 3D models of each region of the workpiece according to the layering order of the point cloud, establish splicing constraints between local 3D models, and splice the local 3D models to obtain the overall 3D model. The verification module is used to compare the constructed overall 3D model with the preset machining accuracy requirements of the workpiece, automatically verify the accuracy deviation of each area and splicing point of the model, and mark the out-of-tolerance parts. The output module is used to obtain the verification results, adjust the parameter format of the overall 3D model based on the verification results, and output model data adapted to the surface workpiece processing equipment. During the acquisition module's operation phase, the acquisition equipment is adjusted based on the curvature range, dimensional boundaries, and accuracy level parameters of the preset processing area of ​​the curved workpiece.

2. The three-dimensional reconstruction system for machining curved workpieces based on point cloud data according to claim 1, characterized in that, When the preprocessing module divides the collected point cloud data into levels according to the curvature changes of the workpiece surface, it first divides the curvature values ​​of the workpiece surface into n curvature intervals. ,in, These represent the minimum curvature and maximum curvature of the workpiece surface, respectively, and the difference between adjacent curvature intervals. equal; Iterate through all the collected point cloud data and calculate the surface curvature corresponding to each point cloud data. ,Will Point cloud data falling within the same curvature interval are grouped into the same level, resulting in n point cloud levels. When removing outliers within each level, for any point p in the point cloud data of each level, the number of point clouds in its neighborhood of radius r is searched. Calculate the average number of neighborhood points for all point clouds within this level. with standard deviation ,when When this happens, point p is determined to be an outlier and is removed. When completing a missing point cloud, for the missing regions formed after removing outliers, the point cloud data of the boundaries of the missing regions is obtained, and the local surface equations of the boundary point cloud are established. By generating virtual points at intervals of 0.5×D within the missing region, and substituting the virtual points into the local surface equation to calculate their z-coordinates, the missing point cloud is completed. Where D represents the sampling density of the corresponding level.

3. The three-dimensional reconstruction system for machining curved workpieces based on point cloud data according to claim 1, characterized in that, When the extraction module identifies the feature regions of the workpiece surface, it simultaneously calculates the normal vector of each point in the layered point cloud data that has been completed to fill in the missing point cloud. and the angle between the normal vectors of adjacent points. , Represents the normal vector between two adjacent points, when When the angle exceeds 30 degrees, the region containing the corresponding adjacent points is marked as a candidate region for the feature region. Next, calculate the rate of curvature change of the point cloud within the candidate feature region. , This represents the curvature of points i and j within the candidate region. Represents the Euclidean distance between two points, when > When this happens, the candidate region is determined to be a characteristic region of the workpiece surface; In the stage of extracting the spatial coordinates and topological relationships of feature points, the spatial coordinates of all points within the feature region are obtained. And construct the adjacency matrix M of the feature points, which follows the following: ； In the formula: The threshold for adjacency determination is , and The value is set to 1.2 times the sampling density of the corresponding layer of the feature region, and the topological relationship of the feature points is represented by the adjacency matrix.

4. The three-dimensional reconstruction system for machining curved workpieces based on point cloud data according to claim 1, characterized in that, When the construction module constructs local 3D models of various regions of the workpiece using the spatial coordinates and topological relationships of feature points, it first fits the local surface of each region corresponding to each point cloud level using the spatial coordinates of feature points within that region. The fitting function is as follows: ； In the formula: This is the final output of a two-dimensional to three-dimensional surface mapping relationship, used to characterize the surface morphology of local regions; The number of feature points involved in the fitting; i is the feature point index; For weighting functions; are basis functions; , The distance from the feature point to the point to be fitted. The radius of the influence domain; Based on the adjacency matrix M of the topological relationships of the feature points, the connection order of each point in the local surface is determined, and the mesh structure of the local 3D model is generated. In the stage of establishing splicing constraints between local 3D models, the coordinate deviation of feature points at the splicing point of two adjacent local 3D models is calculated. , , Let k be the coordinates of the feature point k at the junction of the first local model. Let the coordinates of feature point k at the stitching point of the second local model be set, and set the stitching constraint as follows: , This indicates the splicing precision threshold, and The value is set to 0.8 times the preset machining accuracy requirement of the workpiece; In the stage of stitching together the local 3D models, the spatial pose of the local 3D models is adjusted so that the coordinate deviation of the feature points at the stitching point meets the stitching constraint conditions, and finally the overall 3D model is constructed.

5. The three-dimensional reconstruction system for machining curved workpieces based on point cloud data according to claim 1, characterized in that, When the verification module compares the constructed overall 3D model with the preset machining accuracy requirements of the workpiece, it first selects m verification points on the overall 3D model according to the principle of uniform distribution, and obtains the model coordinates of each verification point. Simultaneously, the coordinates of the workpiece's actual position are obtained through measuring equipment. ; Calculate the accuracy deviation at each verification point ; Calculate the accuracy deviation of all verification points, and then calculate the average and maximum deviations. The average deviation and the maximum deviation are respectively compared with the workpiece's preset machining accuracy requirements. In comparison, when the average deviation is less than or equal to And the maximum deviation is less than or equal to 1.2 times. When the model region accuracy is deemed acceptable, the average deviation is greater than 1. Or the maximum deviation is greater than 1.2 times. When this happens, the model is deemed to have out-of-tolerance accuracy in that region; For the model stitching area, additional k verification points are selected on the stitching boundary, and their accuracy deviation is calculated. If there is an accuracy deviation at any of the verification points at the stitching area... Greater than In the case of such a situation, the splicing point is marked as an out-of-tolerance area, and the coordinate range and out-of-tolerance value of the out-of-tolerance area are recorded.

6. The three-dimensional reconstruction system for machining curved workpieces based on point cloud data according to claim 1, characterized in that, When the output module adjusts the parameter format of the overall 3D model based on the verification results, if there are no out-of-tolerance parts in the verification results, it directly reads the model data format supported by the surface workpiece processing equipment, converts the overall 3D model into the corresponding format, and adjusts the unit parameters of the model to make the model units consistent with the unit settings of the processing equipment. If the verification results show out-of-tolerance areas, first, based on the coordinate range and out-of-tolerance value of the out-of-tolerance areas, perform local correction on the out-of-tolerance areas of the overall 3D model. When correcting, use the point cloud data of the qualified area around the out-of-tolerance area as a reference, and refit the surface of the out-of-tolerance area. The fitting method is the same as the local surface fitting method in the construction module. After the correction is completed, the accuracy verification process of the verification module is executed again until there are no out-of-tolerance parts in the model. Then, the corrected overall 3D model is converted into a model data format that is compatible with the processing equipment. When outputting the model data, the accuracy verification report of the model is also output. The report includes the number of verification points, the deviation value of each verification point, the correction status of out-of-tolerance parts, and the final accuracy index. Among them, the model data formats supported by the surface workpiece processing equipment include STL, IGES, and STEP.

7. The three-dimensional reconstruction system for machining curved workpieces based on point cloud data according to claim 1, characterized in that, The output module also contains a database, which is used to store the raw point cloud data acquired by the acquisition module, the layered point cloud data processed by the preprocessing module, the feature point spatial coordinates and topological relationship data obtained by the extraction module, the local three-dimensional model data and the overall three-dimensional model data generated by the construction module, and the accuracy verification data of the verification module in the order of timestamps. The database contains a data association index, which includes workpiece number, acquisition time, processing area parameters, accuracy requirements, and processing time of each module. All 3D reconstruction-related data for a workpiece can be quickly retrieved by workpiece number. The database adopts a distributed storage architecture, storing different types of data on different storage nodes. The original point cloud data and the processed point cloud data are stored on high-speed read and write storage nodes, while the 3D model data and verification data are stored on large-capacity storage nodes. The stored data is stored in an encrypted manner, and the encryption key includes the workpiece number.

8. The three-dimensional reconstruction system for machining curved workpieces based on point cloud data according to claim 1, characterized in that, The acquisition module is interconnected with the preprocessing module and the extraction module via a wireless network. The extraction module is interconnected with the construction module via a wireless network. The construction module is interconnected with the verification module and the output module via a wireless network. The verification module is interconnected with the output module via a wireless network.

9. A method for three-dimensional reconstruction of curved workpiece machining based on point cloud data, wherein the method is an implementation method of the three-dimensional reconstruction system for curved workpiece machining based on point cloud data as described in any one of claims 1-8, characterized in that, include: Based on the preset processing area parameters of the workpiece, the scanning path, sampling density and distance of the acquisition device are adaptively adjusted to drive the device to acquire surface point cloud data. The sampling density is determined according to the curvature fluctuation coefficient, and the path adopts a spiral shape and is supplemented with boundary scanning segments. The point cloud is divided into layers according to the curvature of the surface. Points with abnormal number of neighboring points in each layer are removed. Virtual points are generated by the boundary local surface equation to complete the missing point cloud. Calculate the point cloud normal vector and the angle between adjacent points, mark candidate regions with angles exceeding 30 degrees, determine feature regions by combining the rate of curvature change, simultaneously obtain the coordinates of feature points, and construct an adjacency matrix to represent topological relationships. A local 3D model is generated by fitting the coordinates of feature points to a local surface, a splicing constraint is established, the posture of the local model is adjusted synchronously to make the deviation meet the constraint, and the overall 3D model is obtained by splicing. Select verification points to compare the model with the actual coordinates of the workpiece to calculate the deviation. Combine the preset accuracy to determine the accuracy of the area. Perform additional verification at the splicing point and mark the out-of-tolerance parts and related information. If there are no out-of-tolerance errors, convert the model format and unify the units. If there are out-of-tolerance errors, correct them and re-verify. After passing the verification, output the model and an accuracy report containing verification and correction information.

Images