Blade three-dimensional scanning measurement comparison method and system

Abstract

The invention relates to the technical field of blade scanning, solves the technical problem that processing of precise point cloud data in the prior art consumes too long time and is easily affected by environmental vibration, and particularly relates to a three-dimensional scanning measurement comparison method and system of a blade, and the method comprises the following steps: S1, obtaining three-dimensional point cloud data of a to-be-measured blade to generate a point cloud data set, determining a neighborhood point set based on the point cloud data set and obtaining a plurality of scale features through feature decomposition; s2, calculating a self-adaptive threshold value used for removing noise points according to the scale features, and obtaining a preprocessing point set; according to the method, the robustness in the main shaft direction is improved through eigenvalue weighted aggregation, super-resolution maintenance in the thickness direction of the blade is achieved through anisotropic nuclear width, noise is removed while sharp features are reserved through curvature-guided mixed smoothing, the sawtooth error of the reconstructed blade is greatly reduced, the thin-wall resolution is greatly improved, the calculation efficiency is greatly improved, and the method is suitable for large-scale popularization and application. The method is especially suitable for accurate reconstruction of thin leaves of rice, ferns and the like.

Description

Technical Field

[0001] The present invention relates to the field of blade scanning technology, and in particular to a three-dimensional scanning, measurement and comparison method and system for blades. Background Art

[0002] Three-dimensional scanning of blades generally projects specific structured light onto the blade surface. The structured light will be deformed due to the height change of the blade surface. Two high-speed cameras are used to capture the deformed structured light images from different angles. The three-dimensional coordinate information of each point on the blade surface is calculated using the parallax principle and camera calibration parameters, thereby constructing a three-dimensional point cloud model of the blade. Data comparison is to match and compare the scanned three-dimensional point cloud data of the blade with the design model or standard data, and calculate the deviation value between the two. In the existing technology, accurate blade models can be obtained through high-precision laser scanning, but high-precision laser scanning requires dense point clouds and multiple positioning, which is time-consuming and easily affected by environmental vibrations. Although fast structured light scanning improves efficiency, it is difficult to stably capture micron-level aerodynamic surface deviations, the accuracy is unsatisfactory, and the efficiency is not high. Summary of the Invention

[0003] In response to the shortcomings of the existing technology, the present invention provides a three-dimensional scanning measurement and comparison method and system for blades, which solves the technical problem that the existing technology takes too long to process precise point cloud data and is easily affected by environmental vibrations. It achieves the purpose of significantly shortening the time taken to process precise point cloud data, reducing the results affected by environmental vibrations, and improving processing efficiency.

[0004] To solve the above technical problems, the present invention provides the following technical solution: a three-dimensional scanning measurement and comparison method for a blade, the method comprising the following steps: S1. Obtain the three-dimensional point cloud data of the blade to be tested to generate a point cloud dataset , based on point cloud dataset Determine the neighborhood point set And obtain multiple scale features through feature decomposition ; S2. According to scale characteristics Calculate the adaptive threshold for removing noise points , and obtain the preprocessing point set; S3. Select any point in the preprocessing point set as the origin to construct a three-dimensional rectangular coordinate system, and obtain the surface equation for calculating the partial derivative value in the three-dimensional rectangular coordinate system by the surface fitting method ; S4. Calculate the normal vector based on the partial derivative , based on the normal vector Calculate the Gaussian threshold for reducing point cloud data and average threshold , and obtain a simplified point set; S5. Calculate the characteristic sequence for determining the axis vector based on the simplified point set ; S6. According to the characteristic sequence Calculate the principal axis matrix values ​​used to construct the optimized blade model ; S7, according to the main axis matrix value Build a blade model and calculate vertex curvature for alignment of blade features , based on the Gaussian threshold , average threshold and vertex curvature Obtain the comparison results and send them to the measurement comparison center.

[0005] Furthermore, in S1, the specific implementation steps are as follows: S11, the three-dimensional point cloud data of the blade is based on the level radius Divide into a cube grid with multiple levels and build an octree structure of leaves according to the level radius; S12. Define point cloud dataset , Represents the i-th data point, and selects a point in the point cloud dataset based on the leaf octree structure As the center point, get the point Neighborhood point set ,in, ; S13, according to the neighborhood point set Calculate the covariance matrix value ; S14, the covariance matrix value Perform eigenvalue decomposition to obtain multiple scale features ,in, Represents the features of the nth scale in the sth level.

[0006] Furthermore, in S2, the specific implementation steps are as follows: S21. Based on multiple scale features Calculate the multi-scale curvature ratio used to measure the proportion of scale features ; S22, according to the multi-scale curvature ratio Calculate the fusion feature value used to evaluate the multi-scale feature weights ; S23, according to the neighborhood point set Calculation is used to obtain the adaptive threshold The first characteristic value of ; S24, based on the first feature quantity Calculation is used to obtain the adaptive threshold The second characteristic value of ; S25, based on the first feature value , the second feature quantity and fusion eigenvalues Calculation for following scale features Changes in the adaptive threshold for automatic range adjustment ; S26, according to the adaptive threshold Eliminate noise points; like , then point Remove noise points and mark the removed points ; like , then point is a normal data point; S27, repeat S26 until the point cloud data set is detected and a preliminary pre-processed point set is obtained; S28, according to the elimination point Calculate fill points to fill in empty or missing positions ; S29, fill point Add to the preliminary preprocessing point set and obtain the preprocessing point set.

[0007] Furthermore, in S3, the specific implementation steps are as follows: S31. Select a point in the preprocessing point set Establish a three-dimensional rectangular coordinate system for the origin and define any point in the preprocessing point set The coordinates are , and construct the surface equation of the preprocessed point set based on the three-dimensional rectangular coordinate system ; S32. According to the surface equation Calculate the least squares value ; S33, based on the least squares value Calculate the functions used to construct the surface The coefficient of 、 and The value of S34, the surface equation Convert to surface function ; S35, according to the surface function Calculate the partial derivatives used to obtain surface features, including the first value , second value , the third value , the fourth value and the fifth value .

[0008] Furthermore, in S4, the specific implementation steps are as follows: S41. According to the first value in the partial derivative and the second value Calculating the normal vector ; S42, according to the first value and the second value Calculating the bending value 、 and ; S43, according to the third value , the fourth value , the fifth value and normal vector Calculate the deep curve value 、 and ; S44. Calculate Gaussian curvature for screening surface features , mean curvature and global curvature ; S45. Repeat S41-S44 to traverse all points in the preprocessed point set to obtain the Gaussian curvature of the neighborhood of each point. , mean curvature and global curvature ; S46. Calculate the Gaussian threshold for streamlining the preprocessing point set and average threshold ; S47, according to Gaussian threshold and average threshold Complete the screening of the preprocessing point set; like and , it is marked as a reserved point; like and , then it is a point that can be simplified and eliminated; like or , it is marked as a reserved point; S48, repeat S47 to simplify all pre-processed point sets to obtain a simplified point set and end.

[0009] Furthermore, in S5, the specific implementation steps are as follows: S51. Define the streamlined points in the streamlined point set The coordinates are , the number is B, calculate all the streamlined points The center of mass coordinates; S52, select any streamlined point As a reference point , calculate reference points and other streamlined points The mean distance between ; S53, according to the distance mean Calculating vector weights ; S54, based on vector weight Calculate the covariance matrix used to determine the axial quantities of the blade model ; S55. According to the covariance matrix Calculate the characteristic equation and obtain multiple eigenvalues ; S56, multiple eigenvalues Arrange in ascending order to obtain the eigenvalue sequence .

[0010] Furthermore, in S6, the specific implementation steps are as follows: S61, according to the characteristic sequence Calculate Euclidean distance value ; S62, according to the characteristic sequence Calculates the aggregate weights used to represent the weights of the feature values ; S63, according to the aggregation weight Obtain weighted eigenvalues ​​through weighted aggregation ; S64, according to the weighted eigenvalue The optimized spindle matrix value is obtained through decoupling optimization .

[0011] Furthermore, in S7, the specific implementation steps are as follows: S71, according to the axis vector value Calculates the adaptive kernel width used to maintain thin-walled features of the model ; S72, according to the adaptive kernel width Computing implicit field function values ​​through spatial Gaussian convolution ; S73, calculate each simplified point in the blade model Vertex curvature ; S74, scanning a standard leaf according to S11-S73 and obtaining the standard Gaussian threshold of the standard leaf respectively , standard average threshold and vertex curvature ; S75. Determine the Gaussian deviation threshold of the blade model and the standard blade by empirical method , average deviation threshold and curvature deviation threshold ; S76, according to Gaussian deviation threshold , average deviation threshold and curvature deviation threshold Generate comparison results of leaf models, including high consistency, medium similarity and inconsistency; like and and , the comparison results are highly consistent; like or or , the comparison result is medium similarity; like and and , the comparison result is inconsistent.

[0012] Furthermore, the vertex curvature The calculation formula is: ; in, Indicates a simplified point The gradient, Indicates a simplified point The Hessian matrix value of .

[0013] The present invention also provides a system for the three-dimensional scanning measurement and comparison method of the blade, the system comprising: Scale feature module, used to obtain the three-dimensional point cloud data of the blade to be measured to generate a point cloud dataset , based on point cloud dataset Determine the neighborhood point set And obtain multiple scale features through feature decomposition ; Preprocessing module, used to Calculate the adaptive threshold for removing noise points , and obtain the preprocessing point set; The surface fitting module is used to select any point in the preprocessing point set as the origin to construct a three-dimensional rectangular coordinate system. The surface equation used to calculate the partial derivative value is obtained in the three-dimensional rectangular coordinate system by the surface fitting method. ; A streamlined module for calculating normal vectors based on partial derivatives , based on the normal vector Calculate the Gaussian threshold for reducing point cloud data and average threshold , and obtain a simplified point set; Axis vector module, used to calculate the characteristic sequence for determining the axis vector based on the reduced point set ; Model building module for Calculate the principal axis matrix values ​​used to construct the optimized blade model ; Blade comparison module, used to compare the main axis matrix value Build a blade model and calculate vertex curvature for alignment of blade features , based on the Gaussian threshold , average threshold and vertex curvature Obtain the comparison results and send them to the measurement comparison center.

[0014] By means of the above technical solution, the present invention provides a method and system for three-dimensional scanning, measurement and comparison of blades, which has at least the following beneficial effects: 1. The present invention can make point cloud data more accurate through multi-scale feature processing and adaptive filtering of three-dimensional point cloud data. The octree structure can accelerate three-dimensional data search, quantify the degree of surface curvature, and reduce the accidental deletion of feature points through the curvature threshold method. The use of adaptive threshold can also improve the robustness of the algorithm.

[0015] 2. The present invention constructs a surface fitting equation to extract features of the surface based on the fitted surface, and determines the simplification of the point cloud data based on the eigenvalues, thereby completing the simplification of the point cloud data and constructing a directed bounding box model to prepare for the construction of a three-dimensional model in the subsequent steps. The simplification steps of the point cloud data are fast and efficient. The simplified model can reduce the number of calculation steps and data point operations, greatly improving the algorithm's computational efficiency while ensuring the accuracy of the vertices.

[0016] 3. The present invention can effectively reduce the complexity of algorithm operations through global implicit values, effectively reduce the error rate of point cloud data through axial alignment of point cloud data, reduce processing time by utilizing isosurfaces, and finally, through smoothing processing using a three-dimensional smoothing operator, better preserve the carving details of cultural relics, make the three-dimensional model more accurate, and improve computing efficiency while achieving higher accuracy.

[0017] 4. The present invention improves the robustness in the main axis direction through eigenvalue weighted aggregation, achieves super-resolution maintenance in the thickness direction of the leaf through anisotropic kernel width calculation, and removes noise while retaining sharp features at the veins through curvature-guided hybrid smoothing. Compared with the existing technology, the geometric error of reconstructed leaf serrations is greatly reduced, the resolution of thin-walled areas is greatly improved, and the computational efficiency is greatly improved. It is particularly suitable for the precise reconstruction of thin leaves such as rice and ferns. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings: Figure 1 This is a flow chart of the three-dimensional scanning measurement and comparison method of a blade according to the present invention; Figure 2 This is a structural block diagram of the three-dimensional scanning measurement and comparison system for blades of the present invention.

[0019] In the figure: 1. Scale feature module; 2. Preprocessing module; 3. Surface fitting module; 4. Simplification module; 5. Axis vector module; 6. Model building module; 7. Blade comparison module. DETAILED DESCRIPTION

[0020] To make the above-mentioned objectives, features, and advantages of the present invention more clearly understood, the present invention is further described below in detail with reference to the accompanying drawings and specific embodiments. This will enable a full understanding of how this application uses technical means to solve technical problems and achieve technical effects, and to implement the invention accordingly.

[0021] Due to the technical problems that the existing technology takes too long to process precise point cloud data and is easily affected by environmental vibration, this embodiment proposes a three-dimensional scanning measurement and comparison method for blades, such as Figure 1 As shown, the time consumption of precise point cloud data can be greatly shortened, the results affected by environmental vibration are reduced, and processing efficiency is improved. The method includes the following steps: S1. Obtain the three-dimensional point cloud data of the blade to generate a point cloud dataset , based on point cloud dataset Determine the neighborhood point set And obtain multiple scale features through feature decomposition In the process of acquiring point cloud data, due to the accuracy of the equipment and the influence of dim ambient light, the point cloud data is easily inaccurate. To solve this problem, this embodiment proposes a detailed implementation method as follows: S11, the three-dimensional point cloud data of the blade is based on the level radius Divide into a cube grid with multiple levels, and build an octree structure based on the level radius. The expression is: ; in, represents the radius of the sth level, and , represents the construction coefficient; S12. Define point cloud dataset , Represents the i-th data point, select any point in the point cloud dataset As the center point, get the point Neighborhood point set ; S13, according to the neighborhood point set Calculate the covariance matrix value , the calculation formula is: ; Among them, K represents the number of represents the mean value of the neighborhood point coordinates, Represents the coordinate value of the lth neighborhood point, T represents The transpose of S14, the covariance matrix value Perform eigenvalue decomposition to obtain multiple scale features ,in, Represents the feature of the nth scale in the sth level, and the expression is:

[0022] Among them, eig represents the covariance matrix value Perform feature decomposition. The algorithm processes multi-scale features and adaptive filtering on 3D point cloud data, making the point cloud data more accurate. The octree structure accelerates 3D data search, quantifies the degree of surface curvature, and reduces the chance of feature points being accidentally deleted through the curvature threshold method. The use of adaptive thresholds also improves the robustness of the algorithm.

[0023] S2. According to scale characteristics Calculate the adaptive threshold for removing noise points , and obtain the preprocessed point set; based on S1, perform adaptive filtering on the point cloud data. In order to solve this problem, the detailed implementation steps are as follows: S21. Based on multiple scale features Calculating multiscale curvature ratios , the calculation formula is: ; in, Representing features at multiple scales The kth scale feature in, N represents the scale feature the number of S22, according to the multi-scale curvature ratio Calculate fusion eigenvalues , the calculation formula is: ; in, represents the cth fusion eigenvalue; S23, according to the neighborhood point set Calculate the first feature value , the calculation formula is: ; in, represents the sth first feature quantity; S24, based on the first feature quantity Calculate the second feature value , the calculation formula is:

[0024] in, represents the t-th second feature quantity; S25, based on the first feature value and the second feature quantity Calculating adaptive thresholds , the calculation formula is: ; in, represents the yth adaptive threshold; S26, according to the adaptive threshold Remove noise; like , then point Remove the noise and mark the removal points ; like , then point is a normal data point; S27, repeat S26 until the point cloud data set is detected and a preliminary pre-processed point set is obtained; S28, according to the elimination point Calculate fill points , the calculation formula is: ; in, and Represent and eliminate points respectively Two adjacent points; S29, fill point It is added to the preliminary preprocessing point set and the preprocessing point set is obtained. Through multi-scale feature processing and adaptive filtering of three-dimensional point cloud data, the point cloud data can be made more accurate. The octree structure can accelerate the three-dimensional data search, quantify the degree of surface curvature, and reduce the accidental deletion of feature points through the curvature threshold method. The use of adaptive threshold can improve the robustness of the algorithm.

[0025] S3. Select any point in the preprocessing point set as the origin to construct a three-dimensional rectangular coordinate system. In the three-dimensional rectangular coordinate system, the surface equation qm for calculating the partial derivative is obtained by the surface fitting method. h In order to solve this problem, the specific methods adopted are as follows: S31. Select a point p in the preprocessing point set. d Establish a three-dimensional rectangular coordinate system for the origin and define any point p in the preprocessing point set i The coordinates of (x i ,y i ,z i ), and construct the surface equation qm of the preprocessed point set based on the three-dimensional rectangular coordinate system h , the expression is: qm h =ax 2 +bxy+cy 2 -z Where a, b and c represent the surface equation qm h The coefficient of S32, according to the surface equation qm h Calculate the least squares value Z f , the calculation formula is: Where H represents the number of preprocessing points, x i 、y i and z i Respectively represent the x-, y-, and z-axis coordinate values ​​of the preprocessing point; S33, based on the least squares value Z f Calculate the values ​​of coefficients a, b and c respectively, and the calculation formula is: Among them, Z f represents the fth least squares value; S34, the surface equation qm h Convert to surface function Qm e (x,y,z), the expression is: Among them, Qm e (x,y,z) represents the e-th surface function, X(x,y,z), Y(x,y,z) and Z(x,y,z) represent the function expressions on the x, y and z axes respectively; S35, according to the surface function Qm e (x, y, z) calculates the partial derivative, which includes the first value f x1 , the second value f y2 , the third value f xy3 , the fourth value f xx4 and the fifth value f yy5 , the calculation formula is: in, Represents the surface function Qm e The partial derivatives of (x,y,z), represents the partial derivative of the function in the x direction, It represents the partial derivative of the function in the y direction. By constructing the surface fitting equation, it is possible to extract the features of the surface based on the fitting surface, and determine the simplification of the point cloud data based on the eigenvalues, complete the simplification of the point cloud data, and build a blade model at the same time, preparing for the construction of the three-dimensional model in the subsequent steps. The simplification steps of the point cloud data are fast and efficient. The simplified model can reduce the number of calculation steps and data points, while ensuring the accuracy of the vertices, greatly improving the computational efficiency of the algorithm.

[0026] S4. Calculate the normal vector based on the partial derivative , based on the normal vector Calculate the Gaussian threshold for reducing point cloud data and average threshold , and obtain a simplified point set; based on S3, it is necessary to further obtain the features of the three-dimensional surface through partial derivatives, and simplify the point cloud data based on this feature. To solve this problem, this embodiment proposes a more detailed implementation method as follows: S41. According to the first value in the partial derivative and the second value Calculating the normal vector , the calculation formula is: ; in, Representing surface functions The normal vector of S42, according to the first value and the second value Calculating the bending value 、 and , the calculation formula is: ; ; ; in, 、 and denote the g-th, h-th and j-th bending values ​​respectively; S43, according to the third value , the fourth value and the fifth value Calculate the deep curve value 、 and , the calculation formula is: ; ; ; in, 、 and denote the kth, lth and mth deep curvature values ​​respectively; S44. Calculate Gaussian curvature for screening surface features , mean curvature and global curvature ; S45. Repeat S41-S44 to traverse all points in the preprocessed point set to obtain the Gaussian curvature of the neighborhood of each point. , mean curvature and global curvature ; S46. Calculate Gaussian threshold and average threshold , the calculation formula is: ; ; in, represents the qth Gaussian threshold, represents the wth average threshold, H represents the number of preprocessing points; S47, according to Gaussian threshold and average threshold Complete the screening of the preprocessing point set; like and , it is marked as a reserved point; like and , then it is a point that can be simplified and eliminated; like or , then it is marked as a retained point; the retained points generally indicate that the surface features are very obvious, and they are points on the boundary or arc surface. The simplified points are generally points on the plane or relatively flat arc surface. These points do not need to be fully calculated, and only a very small amount is needed to complete them, so they can be eliminated; S48. Repeat S47 to streamline all pre-processed point sets to obtain a streamlined point set and end. By constructing the surface fitting equation, it is possible to extract features from the surface based on the fitted surface, and determine the simplification of the point cloud data based on the eigenvalues, completing the simplification of the point cloud data. At the same time, the blade model is constructed to prepare for the subsequent three-dimensional model construction step. The point cloud data simplification step is fast and efficient. The streamlined model can reduce the number of calculation steps and data point operations, while ensuring the accuracy of the vertices, greatly improving the algorithm's operational efficiency.

[0027] S5. Calculate the characteristic sequence for determining the axis vector based on the simplified point set After the point cloud data is streamlined, a blade model needs to be constructed on this basis to help build a three-dimensional model. To solve this problem, the detailed implementation steps are as follows: S51. Define the streamlined points in the streamlined point set The coordinates are , the number is B, calculate all the streamlined points The center of mass The coordinates are calculated as follows: ; in, represents the xth centroid coordinate, Indicates a simplified point The x-coordinate value of Indicates a simplified point The y-coordinate value of Indicates a simplified point The z-direction coordinate value of S52, select any streamlined point As a reference point , calculate reference points and other streamlined points The mean distance between , the calculation formula is: ; in, represents the dth distance mean, and D represents the reference point and other streamlining points Number of groups; S53, according to the distance mean Calculating vector weights , the calculation formula is: ; in, represents the hth vector weight; S54, based on vector weight Calculate the covariance matrix used to determine the axial quantities of the blade model , the calculation formula is: ; ; ; Among them, s represents the sum of vector weights, and m represents the simplified point The mean of the weighted sum of The number of; S55. According to the covariance matrix Calculate the characteristic equation and obtain multiple eigenvalues , the calculation formula is: ; ; Among them, det represents The determinant of , I represents the identity matrix, represents the eigenvector; S56, multiple eigenvalues Arrange in ascending order to obtain the characteristic sequence , where E is less than or equal to B.

[0028] S6. According to the characteristic sequence Calculate the principal axis matrix values ​​used to construct the optimized blade model ; Because there are errors in the direction of the eigenvalues ​​in the feature sequence when constructing the blade model, it is necessary to adjust the principal axis matrix value of the blade model before constructing the blade model. To solve this problem, this embodiment proposes a detailed implementation method as follows: S61, according to the characteristic sequence Calculate Euclidean distance value The calculation formula is: ; Where E represents the number of eigenvalues ​​in the feature sequence, represents the e-th feature sequence; S62, according to the characteristic sequence Calculates the aggregate weights used to represent the weights of the feature values , the calculation formula is: ; in, Represents the e-th Euclidean distance value in the feature sequence; S63, according to the aggregation weight Obtain weighted eigenvalues ​​through weighted aggregation , the calculation formula is: ; in, represents the w-th weighted eigenvalue, represents the jth aggregation weight; S64, according to the weighted eigenvalue The optimized spindle matrix value is obtained through decoupling optimization , the calculation formula is: ; ; ; ; in, represents the unit vector of the z-axis, 、 and The axis vector values ​​that represent the principal axis matrix values. This is equivalent to determining the positions of the three main axes. Through the method of eigenvalue weighted aggregation, the robustness of the main axis direction can be improved and the accuracy of model construction can be improved.

[0029] S7, according to the main axis matrix value Construct a blade model and calculate the vertex curvature for comparing the blade model with the standard blade features , and obtain the comparison result, and send the comparison result to the measurement comparison end; after generating the blade model based on S6, the blade model is compared with the standard blade. In order to solve this problem, the specific method adopted in this embodiment is as follows: S71, according to the axis vector value Calculates the adaptive kernel width used to maintain thin-walled features of the model , the calculation formula is: ; in, Represents the principal axis matrix value Any axis vector value in represents the curvature sensitivity factor, Represents the jth kernel width coefficient; the curvature sensitivity factor is often set to 0.5 in practical applications. The kernel width can indicate that the model shrinks in the curvature direction and maintains thin-wall features.

[0030] S72, according to the adaptive kernel width Computing implicit field function values ​​through spatial Gaussian convolution , the calculation formula is: ; Among them, m represents the processing streamlined point The number of, T represents the threshold parameter; the threshold parameter T can be obtained through a random generator and is usually set to a positive real number.

[0031] S73, calculate each simplified point in the blade model Vertex curvature , the calculation formula is: ; in, Indicates a simplified point The gradient, Indicates a simplified point The Hessian matrix value of ; S74, scanning a standard leaf according to S11-S73 and obtaining the standard Gaussian threshold of the standard leaf respectively , standard average threshold and vertex curvature ; S75. Determine the Gaussian deviation threshold of the blade model and the standard blade by empirical method , average deviation threshold and curvature deviation threshold ; S76, according to Gaussian deviation threshold , average deviation threshold and curvature deviation threshold Generate comparison results of leaf models, including high consistency, medium similarity and inconsistency; like and and , the comparison results are highly consistent; like or or , the comparison result is medium similarity; like and and , the comparison result is inconsistent. Eigenvalue weighted aggregation improves robustness in the principal axis direction, anisotropic kernel width calculation achieves super-resolution preservation in the leaf thickness direction, and curvature-guided blending smoothing removes noise while preserving sharp features in leaf veins. Compared to existing technologies, the geometric error of reconstructed leaf serrations is significantly reduced, the resolution of thin-walled regions is significantly improved, and computational efficiency is greatly enhanced. This method is particularly suitable for the accurate reconstruction of thin leaves such as rice and ferns.

[0032] Due to the technical problem that the existing technology takes too long to process precise point cloud data and is easily affected by environmental vibration, this embodiment also proposes a system applied to the scanning measurement comparison method, which can significantly shorten the time taken to process precise point cloud data, reduce the results affected by environmental vibration, and improve processing efficiency. Figure 2 As shown, the system includes a scale feature module 1, a preprocessing module 2, a surface fitting module 3, a simplification module 4, an axis vector module 5, a model building module 6 and a blade comparison module 7.

[0033] Scale feature module 1, used to obtain the three-dimensional point cloud data of the blade to be measured to generate a point cloud dataset , based on point cloud dataset Determine the neighborhood point set And obtain multiple scale features through feature decomposition ; Preprocessing module 2 is used to process Calculate the adaptive threshold for removing noise points , and obtain the preprocessing point set; Surface fitting module 3 is used to select any point in the preprocessing point set as the origin to construct a three-dimensional rectangular coordinate system, and obtain the surface equation for calculating the partial derivative value in the three-dimensional rectangular coordinate system through the surface fitting method ; Simplified module 4, used to calculate the normal vector based on the partial derivative , based on the normal vector Calculate the Gaussian threshold for reducing point cloud data and average threshold , and obtain a simplified point set; Axis vector module 5, used to calculate the characteristic sequence for determining the axis vector based on the reduced point set ; Model building module 6 is used to build Calculate the principal axis matrix values ​​used to construct the optimized blade model ; Blade comparison module 7, used to compare the main axis matrix value Build a blade model and calculate vertex curvature for alignment of blade features , based on the Gaussian threshold , average threshold and vertex curvature Obtain the comparison results and send them to the measurement comparison center.

[0034] Those skilled in the art will appreciate that all or part of the steps in the above-mentioned embodiment methods can be accomplished by instructing the relevant hardware through a program. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0035] Each embodiment in this specification is described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Reference can be made to the same or similar parts between the embodiments. For the above embodiments, since they are basically similar to the method embodiments, the description is relatively simple. For relevant parts, refer to the partial description of the method embodiments.

[0036] The above embodiments provide a detailed introduction to the present invention. Specific examples are used herein to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core ideas. At the same time, for those skilled in the art, according to the ideas of the present invention, there may be changes in the specific implementation methods and application scopes. In summary, the contents of this specification should not be understood as limiting the present invention.

Claims

1. A three-dimensional scanning measurement and comparison method for blades, characterized in that: The method comprises the following steps: S1. Obtain the three-dimensional point cloud data of the blade to be tested to generate a point cloud dataset , based on point cloud dataset Determine the neighborhood point set And obtain multiple scale features through feature decomposition ; S2. According to scale characteristics Calculate the adaptive threshold for removing noise points , and obtain the preprocessing point set; S3. Select any point in the preprocessing point set as the origin to construct a three-dimensional rectangular coordinate system, and obtain the surface equation for calculating the partial derivative value in the three-dimensional rectangular coordinate system by the surface fitting method ; S4. Calculate the normal vector based on the partial derivative , based on the normal vector Calculate the Gaussian threshold for reducing point cloud data and average threshold , and obtain a simplified point set; S5. Calculate the characteristic sequence for determining the axis vector based on the simplified point set ; S6. According to the characteristic sequence Calculate the principal axis matrix values ​​used to construct the optimized blade model ; S7, according to the main axis matrix value Build a blade model and calculate vertex curvature for alignment of blade features , based on the Gaussian threshold , average threshold and vertex curvature Obtain the comparison results and send them to the measurement comparison center.

2. The scanning measurement comparison method according to claim 1, characterized in that: In S1, the specific implementation steps are as follows: S11, the three-dimensional point cloud data of the blade is based on the level radius Divide into a cube grid with multiple levels and build an octree structure of leaves according to the level radius; S12. Define point cloud dataset , Represents the i-th data point, and selects a point in the point cloud dataset based on the leaf octree structure As the center point, get the point Neighborhood point set ,in, ; S13, according to the neighborhood point set Calculate the covariance matrix value ; S14, the covariance matrix value Perform eigenvalue decomposition to obtain multiple scale features ,in, Represents the features of the nth scale in the sth level.

3. The scanning measurement comparison method according to claim 1, characterized in that: In S2, the specific implementation steps are as follows: S21. Based on multiple scale features Calculate the multi-scale curvature ratio used to measure the proportion of scale features ; S22, according to the multi-scale curvature ratio Calculate the fusion feature value used to evaluate the multi-scale feature weights ; S23, according to the neighborhood point set Calculation is used to obtain the adaptive threshold The first characteristic value of ; S24, based on the first feature quantity Calculation is used to obtain the adaptive threshold The second characteristic value of ; S25, based on the first feature value , the second feature quantity and fusion eigenvalues Calculation for following scale features Changes in the adaptive threshold for automatic range adjustment ; S26, according to the adaptive threshold Eliminate noise points; like , then point Remove noise points and mark the removed points ; like , then point is a normal data point; S27, repeat S26 until the point cloud data set is detected and a preliminary pre-processed point set is obtained; S28, according to the elimination point Calculate fill points to fill in empty or missing positions ; S29, fill point Add to the preliminary preprocessing point set and obtain the preprocessing point set.

4. The scanning measurement comparison method according to claim 1, characterized in that: In S3, the specific implementation steps are as follows: S31. Select a point in the preprocessing point set Establish a three-dimensional rectangular coordinate system for the origin and define any point in the preprocessing point set The coordinates are , and construct the surface equation of the preprocessed point set based on the three-dimensional rectangular coordinate system ; S32. According to the surface equation Calculate the least squares value ; S33, based on the least squares value Calculate the functions used to construct the surface The coefficient of 、 and The value of S34, the surface equation Convert to surface function ; S35, according to the surface function Calculate the partial derivatives used to obtain surface features, including the first value , second value , the third value , the fourth value and the fifth value .

5. The scanning measurement comparison method according to claim 1, characterized in that: In S4, the specific implementation steps are as follows: S41. According to the first value in the partial derivative and the second value Calculating the normal vector ; S42, according to the first value and the second value Calculating the bending value 、 and ; S43, according to the third value , the fourth value , the fifth value and normal vector Calculate the deep curve value 、 and ; S44. Calculate Gaussian curvature for screening surface features , mean curvature and global curvature ; S45. Repeat S41-S44 to traverse all points in the preprocessed point set to obtain the Gaussian curvature of the neighborhood of each point. , mean curvature and global curvature ; S46. Calculate the Gaussian threshold for streamlining the preprocessing point set and average threshold ; S47, according to Gaussian threshold and average threshold Complete the screening of the preprocessing point set; like and , it is marked as a reserved point; like and , then it is a point that can be simplified and eliminated; like or , it is marked as a reserved point; S48, repeat S47 to simplify all pre-processed point sets to obtain a simplified point set and end.

6. The scanning measurement comparison method according to claim 1, characterized in that: In S5, the specific implementation steps are as follows: S51. Define the streamlined points in the streamlined point set The coordinates are , the number is B, calculate all the streamlined points The center of mass coordinates; S52, select any streamlined point As a reference point , calculate reference points and other streamlined points The mean distance between ; S53, according to the distance mean Calculating vector weights ; S54, based on vector weight Calculate the covariance matrix used to determine the axial quantities of the blade model ; S55. According to the covariance matrix Calculate the characteristic equation and obtain multiple eigenvalues ; S56, multiple eigenvalues Arrange in ascending order to obtain the eigenvalue sequence .

7. The scanning measurement comparison method according to claim 1, characterized in that: In S6, the specific implementation steps are as follows: S61, according to the characteristic sequence Calculate Euclidean distance value ; S62, according to the characteristic sequence Calculates the aggregate weights used to represent the weights of the feature values ; S63, according to the aggregation weight Obtain weighted eigenvalues ​​through weighted aggregation ; S64, according to the weighted eigenvalue The optimized spindle matrix value is obtained through decoupling optimization .

8. The scanning measurement and comparison method according to claim 1, characterized in that: In S7, the specific implementation steps are as follows: S71, according to the axis vector value Calculates the adaptive kernel width used to maintain thin-walled features of the model ; S72, according to the adaptive kernel width Computing implicit field function values ​​through spatial Gaussian convolution ; S73, calculate each simplified point in the blade model Vertex curvature ; S74, scanning a standard leaf according to S11-S73 and obtaining the standard Gaussian threshold of the standard leaf respectively , standard average threshold and vertex curvature ; S75. Determine the Gaussian deviation threshold of the blade model and the standard blade by empirical method , average deviation threshold and curvature deviation threshold ; S76, according to Gaussian deviation threshold , average deviation threshold and curvature deviation threshold Generate the comparison results of the leaf models, which include three levels: high consistency, medium similarity and inconsistency; like and and , the comparison results are highly consistent; like or or , the comparison result is medium similarity; like and and , the comparison result is inconsistent.

9. The scanning measurement and comparison method according to claim 8, characterized in that: The vertex curvature The calculation formula is: ； in, Indicates a simplified point The gradient, Indicates a simplified point The Hessian matrix value of .

10. A system for use in the scanning measurement and comparison method according to any one of claims 1 to 9, characterized in that: The system includes: Scale feature module (1) is used to obtain the three-dimensional point cloud data of the blade to be measured to generate a point cloud data set , based on point cloud dataset Determine the neighborhood point set And obtain multiple scale features through feature decomposition ; Preprocessing module (2) is used to Calculate the adaptive threshold for removing noise points , and obtain the preprocessing point set; The surface fitting module (3) is used to select any point in the preprocessing point set as the origin to construct a three-dimensional rectangular coordinate system, and obtain the surface equation for calculating the partial derivative value in the three-dimensional rectangular coordinate system through the surface fitting method. ; Simplified module (4) for calculating the normal vector based on the partial derivative , based on the normal vector Calculate the Gaussian threshold for reducing point cloud data and average threshold , and obtain a simplified point set; Axis vector module (5), used to calculate the characteristic sequence for determining the axis vector based on the reduced point set ; Model building module (6) is used to Calculate the principal axis matrix values ​​used to construct the optimized blade model ; The blade comparison module (7) is used to compare the main axis matrix value Build a blade model and calculate vertex curvature for alignment of blade features , based on the Gaussian threshold , average threshold and vertex curvature Obtain the comparison results and send them to the measurement comparison center.

Images