Waviness evaluation method based on sensitive direction recognition and computer equipment

Abstract

The invention discloses a waviness evaluation method based on sensitive direction recognition and computer equipment, and the method comprises the steps: obtaining actual measurement point cloud data of the surface of a measured object, and carrying out the block compression storage; executing principal component analysis and curvature flow analysis in parallel, and establishing a direction weighting function to obtain a sensitive direction; performing theoretical model data registration and NURBS parameterized curved surface reconstruction; and extracting cross section data, carrying out bidirectional center line alternating calculation on the cross section data, and taking the maximum value of the extreme value of the bidirectional center line calculation as a final waviness parameter. The hybrid direction recognition model is constructed through parallel principal component analysis and curvature flow analysis, the improved NURBS curved surface reconstruction technology and the bidirectional midline alternation algorithm are combined, the problem that the measurement precision is insufficient due to the fact that a traditional method depends on manual direction designation is solved, the block compression storage technology is adopted, the processing efficiency is greatly improved compared with the traditional method, and the method is suitable for large-scale popularization and application. The method can effectively identify missing defect areas in traditional course and vertical direction measurement.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of surface quality detection of parts, and particularly relates to a corrugation evaluation method based on sensitive direction identification and a computer device. BACKGROUND

[0002] The corrugation detection technology is applied to the surface quality detection of complex curved surfaces such as aircraft fuselages, wings and engine nacelles, and parts such as high-speed trains and ships, and has important significance for improving the surface manufacturing quality, but the existing corrugation detection technology has the following defects:

[0003] Direction dependence defect: when measuring the corrugation of an aircraft surface, the traditional method measures only the preset heading and vertical heading sections according to the GB / T 1031-2009 standard and related enterprise standards, and cannot capture the true corrugation characteristics of the sensitive direction of curvature change, and has no measurement requirements for the corrugation in the remaining directions;

[0004] Low efficiency of data processing: when processing more than 100 million point cloud data, the memory occupation of the conventional algorithm is more than 64 GB, and the memory requirement is large;

[0005] Theoretical model dependence: accurate CAD model support is required, but in actual measurement, the model is often missing or the version is inconsistent, and the registration error is large;

[0006] Parameter calculation error: the single-direction calculation method of the traditional centerline method leads to a large overestimation of the maximum wave depth, and the parameter calculation error is large. SUMMARY

[0007] The purpose of the application is to provide a corrugation evaluation method based on sensitive direction identification, a computer device, a computer readable storage medium and a computer program product, which can solve the problem of insufficient direction sensitivity, improve the data processing efficiency, eliminate the dependence on theoretical models, and effectively reduce the parameter calculation error.

[0008] In order to achieve the above purpose, one aspect of the application provides a corrugation evaluation method based on sensitive direction identification, comprising:

[0009] Step S1: obtaining the measured point cloud data of the measured object surface, performing block compression storage, and when there is a theoretical model of the object surface, loading the theoretical model and converting it into theoretical point cloud data;

[0010] Step S2: performing principal component analysis and curvature flow analysis in parallel, in the principal component analysis, taking the normal vector set of the measured point cloud data as the input, calculating the eigenvalues of the covariance matrix, and extracting the eigenvector with the largest eigenvalue as the principal direction, in the curvature flow analysis, calculating the curvature change gradient along the parameterization direction of the curved surface, and taking the direction with a curvature change gradient greater than a set threshold as a sensitive candidate direction; and establishing a direction weight function W(D):

[0011] W(D) = a · V pca + b · V curv

[0012] Wherein, V pca and V curv are the principal component contribution and the curvature sensitivity, respectively, and a and b are the principal component contribution V pca and the curvature sensitivity V curv normalized coefficient of the feature, satisfying a+b = 1, the direction with the maximum value of the direction weight function is selected as the final sensitive direction;

[0013] Step S3: data registration and surface reconstruction, when there is a theoretical model, the theoretical model is registered with the measured point cloud data by ICP registration algorithm; when there is no theoretical model, the NURBS parameterized surface is generated based on the measured point cloud by adaptive surface reconstruction;

[0014] Step S4: extracting cross-section data, bidirectional midline alternating calculation is performed on the cross-section data, in the forward calculation, taking the starting point as the reference, the moving average window is used to smooth the profile, and the forward peak and valley are detected; in the reverse calculation, taking the end point as the reference, the Savitzky-Golay filter is used to smooth the profile, and the reverse peak and valley are detected; the maximum value of the extreme value calculated by the bidirectional midline is taken as the final waviness parameter.

[0015] Preferably, in step S3, the NURBS parameterized surface based on the measured point cloud comprises: extracting control points from the measured point cloud by Poisson disk sampling method, establishing UV parameterized control grid of NURBS surface, generating node vector by cumulative chord length parameterization method, and distributing weights to the control points by RBF kernel function.

[0016] Preferably, in step S2, the eigenvalues of the covariance matrix are calculated and sorted to obtain the first three eigenvalues λ1≥λ2≥λ3, and the principal component contribution V pca and the curvature sensitivity V curv are calculated as follows:

[0017] V pca = λ1 / (λ1+λ2+λ3)

[0018]

[0019] Wherein, is the maximum value of the absolute value of the curvature change gradient , and is the curvature change gradient corresponding to the principal direction.

[0020] Preferably, in step S4, the extraction section data comprises: generating a grid slice along the sensitive direction of the whole surface point cloud, performing memory optimization by using an octree space partition method, marking non-connected regions, and establishing a region correlation matrix;

[0021] After extracting the section data, abnormal point filtering and coordinate alignment are performed, the abnormal point filtering comprises calculating a sensitive direction deviation Δz of the section point set, and removing data points with |Δz|>3σ, wherein σ is a standard deviation of the deviation; and the coordinate alignment comprises projecting the section data to a normal plane of the sensitive direction, and performing coordinate system rotation alignment through singular value decomposition.

[0022] Preferably, in step S1, a curvature rapid calculation is further performed: an octree space index structure is constructed, a near-neighbor topological relationship of each point is established, and a Gaussian curvature and an average curvature are calculated through multi-threading.

[0023] Preferably, in step S1, the block compression storage comprises: dividing the point cloud according to a grid block, and using an overlap segmentation method for a boundary region; performing spatial sorting on the point cloud in each block, and encoding and compressing coordinate increment values; constructing a cache pool, retaining only decompression data of a current processing block, and using a memory mapping file technology to reduce peak memory occupation.

[0024] Preferably, the method further comprises:

[0025] A three-dimensional color cloud chart is generated, a vector arrow of the sensitive direction is superimposed and displayed in the cloud chart, the arrow length is proportional to the direction weight function value, a waviness parameter is output, and the waviness parameter comprises a maximum value of a waviness wave depth, a maximum value of a wave depth to wavelength ratio, a maximum value of a wave depth to wave distance ratio, and the wave depth, the wave depth to wavelength ratio and the wave depth to wave distance ratio of the waviness in each wave band.

[0026] Another aspect of the present application provides a computer device comprising a memory, a processor and a computer program stored in the memory, wherein the processor executes the computer program to realize the steps of the above method.

[0027] Still another aspect of the present application provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to realize the steps of the above method.

[0028] Still another aspect of the present application provides a computer program product, which comprises a computer program, and the computer program is executed by a processor to realize the steps of the above method.

[0029] The sensitivity direction recognition-based waviness evaluation method, the computer device, the computer readable storage medium and the computer program product according to the above aspect of the present application can solve the problem of insufficient direction sensitivity, improve data processing efficiency, eliminate dependence on a theoretical model, and effectively reduce parameter calculation errors. BRIEF DESCRIPTION OF DRAWINGS

[0030] In order to more clearly illustrate the technical solutions of the present application, the following will briefly introduce the drawings used in the description of the embodiments of the present application. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor on the basis of these drawings:

[0031] Figure 1 is a flowchart of the sensitivity direction recognition-based waviness evaluation method of an embodiment of the present application;

[0032] Figure 2 is a structural diagram of the computer device of an embodiment of the present application. DETAILED DESCRIPTION

[0033] In order to make the purpose, technical solutions and advantages of the present application more clear, the technical solutions of the present application will be described clearly and completely in the following with reference to the drawings. Obviously, the described embodiments are only some embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the protection scope of the present application.

[0034] An embodiment of the present application provides a sensitivity direction recognition-based waviness evaluation method, as shown in Figure 1 The sensitivity direction recognition-based waviness evaluation method of the present application embodiment includes steps S1-S4.

[0035] Step S1: data loading and preprocessing

[0036] In this step, data loading refers to loading measured point clouds of an object surface and a theoretical model. The measured point cloud loading refers to obtaining measured point cloud data of the surface by a laser tracker or a three-dimensional scanner, and realizing block compression storage for large-scale measured point clouds. The theoretical model loading refers to loading theoretical CAD model data in the format of STEP, STP, IGES and the like, and the theoretical CAD model can be parsed and read in by a public model library or a third-party API.

[0037] In one embodiment, the object surface is measured by a three-dimensional scanner to obtain a large-scale measured point cloud, which is stored by using a block compression strategy, including space division selection, compression encoding and memory management optimization, so as to perform low-memory optimization processing.

[0038] The space division selection refers to dividing the point cloud by 500mm*500mm grid blocks, and using an overlapping segmentation method (overlapping width of 10mm) for the boundary region. The compression encoding refers to performing spatial sorting on the point cloud in each block and encoding and compressing the coordinate increment value. The memory management optimization refers to constructing a cache pool, retaining only the decompression data of the current processing block, and using a memory mapping file technology to reduce the peak memory occupancy.

[0039] If there is a theoretical model of the object surface, a STEP format theoretical file is taken as an example, the theoretical model is read in through an OpenCascade library, and is converted into a theoretical point cloud of the object surface.

[0040] In this step, the data preprocessing refers to performing rapid curvature calculation. In order to reduce the curvature calculation time of the large-scale point cloud, firstly, the near-neighbor topological relationship of each point is established by constructing an octree spatial index structure, and at the same time, the Gaussian curvature kernel function involved in the calculation of the Gaussian curvature K and the average curvature H is calculated by using multi-threading, so as to complete the rapid curvature calculation.

[0041] The curvature calculation formula is as follows, and the neighborhood search radius is 3 times the average distance of the point cloud:

[0042] K = (LN-M 2 ) / (EG-F 2 )

[0043] H = (EN+GL-2FM) / (2(EG-F2))

[0044] Wherein, E, F, G are first-order basic quantities, and L, M, N are second-order basic quantities.

[0045] Step S2: sensitive direction identification

[0046] In this step, the sensitive direction identification includes PCA (Principal Component Analysis, principal component analysis) direction analysis, curvature flow analysis and direction weight evaluation model construction. The PCA direction analysis includes two steps of covariance matrix calculation and eigenvalue solution. The curvature flow analysis includes two steps of curvature gradient calculation and sensitive direction (threshold) screening. The direction weight evaluation model construction includes two steps of feature normalization processing and weight function design.

[0047] The covariance matrix calculation in the PCA direction analysis refers to using the set of normal vectors of the loaded measured points as input, quickly retrieving the k-nearest neighbors using an octree index, and employing a block accumulation algorithm to avoid storing the full covariance matrix. The eigenvalue solving in the PCA direction analysis involves applying the Jacobi rotation method to the 3×3 covariance matrix, with the iteration termination condition being that the absolute value of the off-diagonal elements is less than e. -6 Finally, the eigenvalues ​​are obtained and sorted (λ1≥λ2≥λ3), and the eigenvectors corresponding to λ1 are extracted as the principal direction D1.

[0048] The curvature gradient calculation in the curvature flow analysis refers to calculating the curvature change gradient along the parameterized U and V directions of the surface using an improved Frechet derivative calculation method. Numerical solutions are obtained using the central difference method, with adaptive adjustment of the step size h, as shown below:

[0049]

[0050] h = min(0.1R) c 0.5d avg )

[0051] in, and R represents the gradient of curvature along the U and V directions of the surface, respectively. c Let d be the local radius of curvature. avg The average spacing of the point cloud.

[0052] The sensitive direction screening in the curvature flow analysis refers to constructing a curvature gradient histogram, selecting the directions corresponding to the top 5% of gradient values ​​as sensitive candidate directions, and determining the curvature change threshold θ through iterative calculation, satisfying θ = μ + kσ, where μ is the average rate of curvature change, σ is the standard deviation, and k ∈ [2,3]. Then, cluster analysis is performed on the candidate directions to eliminate discrete noise directions.

[0053] The feature normalization process in the constructed directional weight evaluation model involves calculating the PCA contribution and curvature sensitivity, with the corresponding calculation formulas as follows. The weight function design formula in the constructed directional weight evaluation model is shown below.

[0054] V pca =λ1 / (λ1+λ2+λ3)

[0055]

[0056] W(D)=α·V pca +β·V curv

[0057] Where α and β are the PCA contribution rates V, respectively. pca and curvature sensitivity Vcurv The coefficients after feature normalization satisfy α + β = 1. In one embodiment, coefficient optimization is performed through orthogonal experiments to determine α = 0.55 and β = 0.45. Based on the curvature change gradient corresponding to the main direction, direction optimization is performed, and the direction with the largest W(D) value is selected as the final sensitive direction D. s .

[0058] Step S3: Data Registration and Surface Reconstruction

[0059] In this step, the data registration and surface reconstruction include two cases: theoretical model registration when a theoretical model exists, and model-free surface generation when no theoretical model exists. The theoretical model registration uses an improved ICP algorithm. In the coarse registration stage, the FPFH algorithm is used to extract feature descriptors, and then the RANSAC algorithm is used for registration. In the fine registration stage, the LM optimization algorithm is used, with the objective function shown below, and a regularization coefficient λ = 0.01 is applied to prevent matrix singularities.

[0060] min∑||T(P i )-Q i || 2 +λ·tr(R T R)

[0061] Where tr refers to matrix R T The trace of R, T represents the transformation, P i It is the input data point, Q i λ represents the corresponding target data point, and λ is the regularization parameter.

[0062] The model-free surface generation refers to generating a theoretical surface as a reference surface for waviness evaluation based on the measured data of the surface being measured when the theoretical model of the surface being measured cannot be obtained. This includes control point optimization layout and NURBS surface generation.

[0063] The optimization of control point layout in the model-free surface generation includes initial sampling using Poisson disk sampling, extraction of control points from the measured point cloud, where the number of control points n satisfies n = log₂N (N is the number of sampling points), and the spacing is in the sensitive direction D. s The cosine component multiples are then used to iteratively optimize the control point positions by adjusting the control point positions based on the K-means clustering algorithm.

[0064] The NURBS surface generation in the model-free surface generation involves establishing a UV-parameterized control mesh for the NURBS surface. This includes generating node vectors using the cumulative chord length parameterization method, calculating weights, assigning weights to control points using the RBF (radial basis function) kernel function (as shown in the formula below), solving the equations, and using Tikhonov regularization to handle ill-conditioned problems.

[0065]

[0066] where ω i represents the weight, Q i is the target point, P j is the observation point, the kernel radius r = 3d, and d is the average distance between control points.

[0067] Step S4: cross-section data extraction, cross-section processing and parameter calculation

[0068] In this step, the cross-section data extraction includes two types of input: whole surface point cloud and slice data. When the input is whole surface point cloud, first, generate grid slices along the sensitive direction Ds, then use octree space division method for memory optimization, and finally mark the non-connected regions and establish the region correlation matrix. When the input is slice data, directly enter the cross-section processing and parameter calculation.

[0069] The cross-section processing and parameter calculation include abnormal point filtering, coordinate alignment, bidirectional centerline calculation, and output parameters. The abnormal point filtering refers to calculating the sensitive direction deviation Δz of the cross-section point set, and then removing the data points with |Δz| > 3σ, where σ is the standard deviation of the deviation. The coordinate alignment refers to projecting the cross-section data onto the normal plane of the Ds direction, and performing coordinate system rotation alignment through singular value decomposition (SVD).

[0070] The bidirectional centerline calculation includes forward calculation, reverse calculation, and extreme value fusion. The forward calculation in the bidirectional centerline calculation refers to taking the starting point as the reference, using a moving average window (window length L = 5% of the cross-section length) to smooth the profile, and using the derivative zero-crossing detection method to identify the peaks and valleys. The first-order derivative zero-crossing point is used to detect the peaks and valleys, and the derivative changes from positive to negative is a peak, and vice versa is a valley. The reverse calculation in the bidirectional centerline calculation takes the end point as the reference, uses the Savitzky-Golay filter (polynomial order 3, window length 11 points) for smoothing processing, detects the reverse peak and valley sequence, and performs boundary processing. The mirror extension method is used to avoid end effect. The extreme value fusion in the bidirectional centerline calculation refers to taking the union of the bidirectional detection results, removing the small fluctuations with an amplitude less than 0.02 mm, and taking the extreme maximum value of the bidirectional detection results as the final waviness parameter. The output parameter refers to the output of the evaluation parameters of the waviness, including D / L_max, D / X_max, D_max, and D / L, D / X, D of each wave band.

[0071] D / L_max = Max(D i / L i )

[0072] D / X_max = Max(D i / X i )

[0073] D_max = Max(D i )

[0074] Wherein, D i is the wave depth (amplitude) of the i-th wave band, L i is the corresponding wavelength, X i is the wave distance, D / L is the ratio of wave depth to wavelength, and D / X is the ratio of wave depth to wave distance.

[0075] After obtaining the waviness parameter, the result visualization can be performed.The result visualization refers to generating a three-dimensional color cloud chart, and the color scale range of the three-dimensional color cloud chart is set to [D_max*0.2, D_max]; the vector arrow of the sensitive direction Ds is superimposed and displayed in the cloud chart, and the arrow length is proportional to the value of W(Ds); and an evaluation report is output, including: a waviness parameter statistical table (mean value, extreme value, standard deviation), a profile comparison chart of a key section, a measurement uncertainty evaluation result (k=2), and the like.

[0076] To sum up, the waviness evaluation method based on sensitive direction recognition of the embodiment of the application proposes a "multimodal direction recognition-adaptive surface reconstruction-bidirectional parameter calculation" three-in-one waviness evaluation solution, establishes a PCA and curvature flow hybrid analysis model, solves the problem of insufficient direction sensitivity, develops a parameterized surface self-generation algorithm based on NURBS, eliminates the dependence on a theoretical model, designs a bidirectional centerline alternating calculation strategy, breaks through the system error limitation of single-direction detection, and can effectively reduce the parameter calculation error.

[0077] Wherein, the multimodal direction recognition technology includes fusion of global direction analysis of PCA and local feature response of curvature flow, and a weight function is designed to balance the principal component contribution and the curvature change sensitivity; the adaptive surface reconstruction includes development of a control point optimized distribution method based on Poisson sampling, and adoption of regularized NURBS fitting to solve the problem of ill-conditioned equation; and the bidirectional extreme value detection mechanism includes adoption of moving average for forward calculation to suppress noise, and use of a Savitzky-Golay filter for reverse calculation to maintain features.

[0078] Compared with the traditional method, the waviness evaluation method based on sensitive direction recognition of the embodiment of the application has the following beneficial effects:

[0079] 1. Improve the comprehensiveness and accuracy of complex curved surface waviness detection

[0080] The traditional method relies on manual preset detection direction (such as heading and vertical heading), and it is difficult to adapt to the geometric characteristics of complex surfaces, especially in the double curvature or free surface area, and the direction mismatch problem is prone to occur. The multi-modal sensitive direction recognition technology proposed in the application realizes the adaptive optimization of the detection direction by fusing the global geometric features (PCA main direction analysis) and the local curvature variation features (curvature flow gradient analysis). The application breaks through the limitation of artificial experience and can automatically capture the sensitive propagation direction of the surface ripple defects, for example, in the wing leading edge, fairing transition area and other curvature mutation areas, and can accurately identify the oblique or nonlinear ripple defects that are easily missed by traditional methods. Through the direction weight evaluation model, the global geometric feature contribution and the local curvature sensitivity are systematically balanced to ensure that the selected direction conforms to the overall surface shape and reflects the ripple features of the key area. The application significantly improves the detection coverage and avoids the defect missed detection or misjudgment caused by the deviation of the direction selection, and provides more comprehensive data support for the evaluation of the aerodynamic performance of the aircraft.

[0081] 2. Enhance the robustness and engineering applicability of the detection process

[0082] The application proposes an innovative solution to the problems of lack of theoretical model and large data scale in actual engineering. In the absence of a theoretical model, the application generates a high-precision NURBS parameterized surface based on the measured point cloud through adaptive surface reconstruction technology, solving the strong dependence of traditional registration methods on CAD models. The technology optimizes the layout of control points and uses regularization fitting algorithm to effectively suppress noise interference while ensuring the smoothness of the surface, ensuring that the reconstructed surface is highly consistent with the real geometric shape. In addition, for large-scale point cloud processing needs, the technology uses block compression storage and octree spatial indexing technology to realize stable and efficient processing of large-scale point cloud. The bidirectional centerline alternating calculation strategy eliminates the systematic bias introduced by single-direction calculation by fusing forward and reverse waveform features, significantly improving the calculation reliability of the waviness parameter. These technical improvements enable the method of the application to adapt to the detection needs of different equipment and different working conditions, and still maintain stable output under the conditions of model missing, data missing or complex geometry, greatly expanding the industrial application scenarios of the technology, and the registration error under the condition of model missing can be controlled within 0.1mm.

[0083] The embodiment of the application also provides a computer device, which can be a server, and the internal structure diagram of the computer device can be as shown in Figure 2As shown in the figure. The computer device includes a processor, a memory and a network interface connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium, an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The database of the computer device is used to store the running parameter data of each framework. The network interface of the computer device is used to communicate with the external terminal through the network connection. The computer program is executed by the processor to realize the steps of the method of the embodiment of the application.

[0084] Those skilled in the art can understand that, Figure 2 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. The specific computer device can include more or less components than those shown in the figure, or combine certain components, or have a different component arrangement.

[0085] The embodiment of the application also provides a computer readable storage medium, which stores a computer program, and the computer program is executed by the processor to realize the steps of the method of the embodiment of the application.

[0086] The embodiment of the application also provides a computer program product, which includes a computer program, and the computer program is executed by the processor to realize the steps of the method of the embodiment of the application.

[0087] The above only describes certain exemplary embodiments of the application by way of illustration, and it is self-evident that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the application. Therefore, the above figures and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the application.

Claims

1. A method for evaluating waviness based on sensitive direction recognition, characterized by, The method comprises the following steps: Step S1: obtaining measured point cloud data of the measured object surface, performing block compression storage, when a theoretical model of the object surface exists, loading the theoretical model and converting it into theoretical point cloud data; Step S2: performing principal component analysis and curvature flow analysis in parallel, in the principal component analysis, taking the normal vector set of the measured point cloud data as input, calculating the eigenvalue of the covariance matrix, and extracting the eigenvector with the largest eigenvalue as the principal direction, in the curvature flow analysis, calculating the curvature variation gradient along the parameterization direction of the curved surface, and taking the direction with the curvature variation gradient greater than a set threshold as a sensitive candidate direction; and establishing a direction weight function W(D): W(D) = a · V pca + β · V curv wherein V pca and V curv are the principal component contribution and the curvature sensitivity, respectively, and a and β are the principal component contribution V pca and the curvature sensitivity V curv normalized coefficients of the features, with a + β = 1, and the direction with the largest directional weight function value is selected as the final sensitive direction; Step S3: performing data registration and curved surface reconstruction, when the theoretical model exists, performing registration of the theoretical model and the measured point cloud data through an ICP registration algorithm; when the theoretical model does not exist, generating a NURBS parameterized curved surface based on the measured point cloud through adaptive curved surface reconstruction; Step S4: extracting cross section data, performing bidirectional centerline alternating calculation on the cross section data, in the forward calculation, taking the starting point as the reference, smoothing the profile by using a moving average window, and detecting the forward peak and valley; in the reverse calculation, taking the terminal point as the reference, smoothing the profile by using a Savitzky-Golay filter, and detecting the reverse peak and valley; and taking the maximum value of the extreme values calculated by the bidirectional centerline as the final waviness parameter.

2. The method of claim 1, wherein, In step S3, the generating of the NURBS parameterized curved surface based on the measured point cloud comprises: extracting control points from the measured point cloud by using a Poisson disk sampling method, establishing a UV parameterized control grid of the NURBS curved surface, generating a node vector by using a cumulative chord length parameterization method, and distributing weights to the control points by using an RBF kernel function.

3. The method according to claim 1 or 2, characterized in that, In step S2, the eigenvalues of the covariance matrix are calculated and sorted in size to obtain the first three eigenvalues λ1≥λ2≥λ3, and the principal component contribution V is calculated as follows pca and the curvature sensitivity V curv : V pca = λ1 / (λ1+λ2+λ3) wherein is the curvature variation gradient is the maximum value of the absolute value is the curvature variation gradient corresponding to the main direction 4. The method according to claim 1 or 2, characterized in that, In step S4, the extracting of the cross section data comprises: generating a grid slice along the sensitive direction from the whole surface point cloud, performing memory optimization by using an octree space division method, marking non-connected regions, and establishing a region correlation matrix; After the cross section data is extracted, abnormal point filtering and coordinate alignment are performed, the abnormal point filtering comprises: calculating the sensitive direction deviation Δz of the cross section point set, and removing the data points with |Δz|>3σ, wherein σ is the standard deviation of the deviation; and the coordinate alignment comprises: projecting the cross section data to the normal plane of the sensitive direction, and performing coordinate system rotation alignment through singular value decomposition.

5. The method according to claim 1 or 2, characterized in that, In step S1, fast curvature calculation is further performed: constructing an octree space index structure, establishing the near-neighbor topological relationship of each point, and calculating the Gaussian curvature and the average curvature through multi-threading.

6. The method of claim 1 or 2, wherein, In step S1, the block compression storage comprises: dividing the point cloud by grid blocks, and using an overlapping division method for the boundary regions; performing spatial sorting on the point cloud in each block, and encoding and compressing the coordinate increment value; constructing a cache pool, retaining only the decompression data of the current processing block, and reducing the peak memory occupancy by using a memory mapping file technology.

7. The method according to claim 1 or 2, characterized in that, The method further comprises the following steps: A three-dimensional color cloud chart is generated, and a vector arrow of a sensitive direction is superimposed and displayed on the cloud chart, and the arrow length is proportional to the direction weight function value; and a waviness parameter is output, the waviness parameter including a maximum value of a waviness wave depth, a maximum value of a wave depth to wave length ratio, a maximum value of a wave depth to wave distance ratio, and the wave depth, the wave depth to wave length ratio, and the wave depth to wave distance ratio of the waviness in each wave band.

8. A computer device comprising a memory, a processor, and a computer program stored on the memory, wherein the computer program comprises instructions that, when executed by the processor, cause the processor to perform the method of any one of claims 1-7. The processor executes the computer program to implement the steps of the method of any one of claims 1-7.

9. A computer readable storage medium having stored thereon a computer program, characterized in that, The computer program, when executed by the processor, implements the steps of the method of any one of claims 1-7.

10. A computer program product comprising a computer program, characterized in that, The computer program, when executed by the processor, implements the steps of the method of any one of claims 1-7. The computer program, when executed by the processor, implements the steps of the method of any one of claims 1-7.

Images