Three-dimensional model feature compensation method, device and equipment and computer storage medium

Abstract

The invention relates to a three-dimensional model feature compensation method, device and equipment and a computer storage medium, and belongs to the technical field of three-dimensional model scanning, and the three-dimensional model feature compensation method comprises the following steps: carrying out tensor voting of a normal vector on a target point based on a neighborhood point of the target point in three-dimensional point cloud data, determining the type of the target point based on a tensor voting result; denoising the normal vector of the neighborhood point of the target point based on the type of the target point to obtain the normal vector of the denoised neighborhood point, and calculating the denoising curvature value of the target point based on the normal vector of the denoised neighborhood point; and determining a feature compensation direction of the target point based on the denoising neighborhood point normal vector of the target point, calculating a feature compensation value of the target point based on the denoising curvature value, and performing feature compensation on the target point based on the feature compensation direction and the feature compensation value. According to the invention, accurate feature compensation can be carried out on the three-dimensional model.

Description

Technical Field

[0001] The present invention relates to the technical field of three-dimensional model scanning, and particularly to a method, device, equipment and computer storage medium for compensating features of a three-dimensional model. Background Art

[0002] Structured light three-dimensional scanning is a reconstruction method that projects a laser line onto the surface of an object, synchronously acquires images through two cameras, extracts laser feature stripes, obtains the three-dimensional spatial coordinates of feature points on the object surface through binocular matching, and generates a spatial voxel with direction and distance information by fusing the point cloud on the object surface to quickly construct a high-precision three-dimensional mesh model. Structured light three-dimensional scanning can achieve non-contact measurement and has the advantages of high speed and high precision, and has been widely used in the fields of industrial measurement, reverse engineering, cultural relic digitization, etc.

[0003] In the prior art, the method of constructing a three-dimensional model of an object surface by structured light scanning generally has the problem of feature degradation in practical applications, resulting in different degrees of shrinkage of curved surfaces such as spheres and cylinders in the three-dimensional model, rounding of right angles, increase of R angles, etc.; making the detailed features on the object surface in the three-dimensional model not three-dimensional enough and having a poor visual effect; at the same time, it affects the scanning accuracy and causes distortion of scanning data in some scenarios. In the same three-dimensional model, the more prominent the feature and the greater the surface curvature, that is, the greater the surface bending degree, the more serious the feature degradation phenomenon.

[0004] It can be seen that the prior art will have the problem of feature degradation when constructing a three-dimensional model of an object surface based on structured light scanning. Summary of the Invention

[0005] In view of this, it is necessary to provide a method, device, equipment and computer storage medium for compensating features of a three-dimensional model to solve the problem of feature degradation that occurs in the prior art when constructing a three-dimensional model of an object surface based on structured light scanning.

[0006] To solve the above problems, in the first aspect, the present invention provides a method for compensating features of a three-dimensional model, including: Performing tensor voting on the normal vector of a target point based on the neighborhood points of the target point in the three-dimensional point cloud data, and determining the type of the target point based on the tensor voting result; Denosing the normal vectors of the neighborhood points of the target point based on the type of the target point to obtain denoised neighborhood point normal vectors, and calculating the denoised curvature value of the target point based on the denoised neighborhood point normal vectors; Determining the feature compensation direction of the target point based on the denoised neighborhood point normal vectors of the target point, calculating the feature compensation value of the target point based on the denoised curvature value, and performing feature compensation on the target point based on the feature compensation direction and the feature compensation value.

[0007] In a possible implementation, based on the neighborhood points of the target point in the three-dimensional point cloud data, tensor voting is performed on the target point for the normal vector, and the type of the target point is determined based on the tensor voting result, including: Calculate the first covariance matrix of the normal vectors of the neighborhood points of the target point in the three-dimensional point cloud data, perform eigenvalue decomposition on the first covariance matrix, and obtain multiple eigenvalues; Determine the type of the target point according to the magnitude relationship of the multiple eigenvalues.

[0008] In a possible implementation, based on the type of the target point, the normal vectors of the neighborhood points of the target point are denoised to obtain the denoised neighborhood point normal vectors, including: Perform eigenvalue decomposition on the first covariance matrix to obtain multiple eigenvectors corresponding to the multiple eigenvalues; Determine the neighborhood point normal vector characteristics of the target point based on the type of the target point, update the multiple eigenvalues based on the neighborhood point normal vector characteristics to obtain multiple updated eigenvalues, and determine the second covariance matrix based on the updated eigenvalues and the multiple eigenvectors; Calculate the denoised normal vector of the target point and the denoised neighborhood point normal vectors of the neighborhood points based on the second covariance matrix.

[0009] In a possible implementation, calculate the denoised curvature value of the target point based on the denoised neighborhood point normal vectors, including: Calculate the third covariance matrix based on the denoised normal vector of the target point and the denoised neighborhood point normal vectors of the neighborhood points, and calculate the preliminary denoised curvature value of the target point and the preliminary denoised curvature values of the neighborhood points based on the third covariance matrix; Determine the weights of the neighborhood points based on the preliminary denoised curvature values of the neighborhood points and the preliminary denoised curvature value of the target point, and calculate the denoised curvature value of the target point by combining the preliminary denoised curvature values of the neighborhood points with the weights of the neighborhood points.

[0010] In a possible implementation, before determining the feature compensation direction of the target point based on the denoised neighborhood point normal vectors of the target point, including: Perform eigenvalue decomposition on the third covariance matrix to obtain multiple compensation judgment eigenvalues; When it is determined that the region where the target point is located is a non-planar region based on the multiple compensation judgment eigenvalues, perform feature compensation on the target point.

[0011] In a possible implementation, determine the feature compensation direction of the target point based on the denoised neighborhood point normal vectors of the target point, including: Determine the denoised neighborhood point plane based on the denoised neighborhood points and the denoised neighborhood point normal vectors, and calculate the common intersection point of the denoised neighborhood point planes; Determine the compensation direction vector based on the common intersection point and the target point, and determine the characteristic compensation direction of the target point based on the direction relationship between the compensation direction vector and the normal vector of the target point.

[0012] In a possible implementation manner, calculating the characteristic compensation value of the target point based on the denoised curvature value includes: Determine the fitting diameter of the fitting spherical surface based on the three-dimensional point cloud data of the standard spherical disk collected by the three-dimensional point cloud data acquisition device, and calculate the characteristic compensation value of each spherical point in the fitting spherical surface based on the fitting diameter and the standard diameter of the standard spherical disk; Perform least squares fitting on the characteristic compensation values to obtain the characteristic compensation parameters of the three-dimensional point cloud data acquisition device; Calculate the characteristic compensation value of the target point based on the characteristic compensation parameters and the denoised curvature value; Calculate the characteristic compensation value of the target point based on the characteristic compensation parameters and the denoised curvature value.

[0013] In a second aspect, the present invention further provides a three-dimensional model feature compensation device, including: A target point type determination module, configured to perform tensor voting on the normal vector of the target point based on the neighborhood points of the target point in the three-dimensional point cloud data, and determine the type of the target point based on the tensor voting result; A curvature value calculation module, configured to denoise the normal vectors of the neighborhood points of the target point based on the type of the target point to obtain the denoised neighborhood point normal vectors, and calculate the denoised curvature value of the target point based on the denoised neighborhood point normal vectors; A feature compensation module, configured to determine the characteristic compensation direction of the target point based on the denoised neighborhood point normal vectors of the target point, calculate the characteristic compensation value of the target point based on the denoised curvature value, and perform feature compensation on the target point based on the characteristic compensation direction and the characteristic compensation value.

[0014] In a third aspect, the present invention further provides a three-dimensional model scanning device, including a memory and a processor, wherein, The memory is used to store programs; The processor is coupled to the memory and is configured to execute the programs stored in the memory to implement the steps in the three-dimensional model feature compensation method in any of the above embodiments.

[0015] In a fourth aspect, the present invention further provides a computer-readable storage medium, configured to store computer-readable programs or instructions, and when the programs or instructions are executed by a processor, the steps in the three-dimensional model feature compensation method in any of the above embodiments can be implemented.

[0016] The beneficial effects of the present invention are as follows: The three-dimensional model feature compensation method provided by the present invention performs tensor voting on the normal vector of a target point through the neighborhood points of the target point in the three-dimensional point cloud data, determines the type of the target point based on the tensor voting result, denoises the normal vectors of the neighborhood points of the target point based on the type of the target point to obtain denoised neighborhood point normal vectors, which can effectively remove the noise points on the surface where the target point is located. Calculating the denoised curvature value of the target point based on the denoised neighborhood point normal vectors can further remove the noise features on the surface where the target point is located, retain the sharp features, determine the feature compensation direction of the target point through the denoised neighborhood point normal vectors of the target point, calculate the feature compensation value of the target point based on the denoised curvature value, and perform feature compensation on the target point based on the feature compensation direction and the feature compensation value, which can improve the feature compensation accuracy of the three-dimensional model surface and will not affect the planar features, ensuring that the three-dimensional model of the object surface constructed based on structured light scanning is not distorted. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those skilled in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0018] Figure 1 It is a schematic flowchart of a three-dimensional model feature compensation method provided by an embodiment of the present invention; Figure 2 It is a schematic flowchart of a method for determining the type of a target point provided by an embodiment of the present invention; Figure 3 It is a schematic flowchart of a method for denoising normal vectors provided by an embodiment of the present invention; Figure 4 It is a schematic flowchart of a method for calculating the denoised curvature value provided by an embodiment of the present invention; Figure 5 It is a schematic diagram of the surface curvature of a ball disk provided by an embodiment of the present invention; Figure 6 It is a histogram of the surface curvature provided by an embodiment of the present invention; Figure 7 It is a schematic flowchart of a method for determining feature compensation provided by an embodiment of the present invention; Figure 8 It is a schematic flowchart of a method for determining the feature compensation direction provided by an embodiment of the present invention; Figure 9 It is a schematic flowchart of a method for determining the feature compensation value provided by an embodiment of the present invention; Figure 10The structural schematic diagram of a three-dimensional model feature compensation device provided by an embodiment of the present invention; Figure 11 The structural schematic diagram of a three-dimensional model scanning device provided by an embodiment of the present invention. Specific embodiments

[0019] The preferred embodiments of the present invention will be specifically described below with reference to the accompanying drawings. The accompanying drawings form a part of this application and are used together with the embodiments of the present invention to explain the principles of the present invention, rather than to limit the scope of the present invention.

[0020] In the embodiments of the present invention, the descriptions such as "first" and "second" are only for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the quantity of the indicated technical features. Therefore, the technical features defined with "first" and "second" may explicitly or implicitly include at least one of such features.

[0021] Referring to "embodiments" herein means that the specific features, structures or characteristics described in connection with the embodiments may be included in at least one embodiment of the present invention. The phrase appears in various places in the specification and does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. Those skilled in the art will explicitly and implicitly understand that the embodiments described herein may be combined with other embodiments.

[0022] A specific embodiment of the present invention, as Figure 1 shown, discloses a three-dimensional model feature compensation method, including: S101, performing tensor voting on the normal vector of the target point based on the neighborhood points of the target point in the three-dimensional point cloud data, and determining the type of the target point based on the tensor voting result.

[0023] In the embodiments of the present invention, the three-dimensional point cloud data refers to the data obtained by a structured light three-dimensional scanner performing three-dimensional scanning on the surface of the object to be modeled, including the three-dimensional coordinates and laser reflection intensity of each point. The target point refers to the point in the three-dimensional point cloud data that needs to be feature-compensated, generally the points on the surface, such as the points on the edge, the points on the corner or other surface points; the neighborhood range of the target point can be set according to the actual situation, such as 3 times the point cloud resolution. A kd tree (K-Dimensional Tree) can be constructed based on the neighborhood of the target point, and tensor voting on the normal vector is performed within the neighborhood of the target point based on this kd tree, and the type of the target point is determined based on the tensor voting result. Among them, the types of the target point include plane points, plane intersection points, and corner points. The specific process of tensor voting and the specific process of determining the type of the target point will be described in detail later in the present invention.

[0024] S102. Denoise the normal vectors of the neighborhood points of the target point based on the type of the target point to obtain the denoised neighborhood point normal vectors, and calculate the denoised curvature value of the target point based on the denoised neighborhood point normal vectors.

[0025] In the embodiment of the present invention, after determining the type of the target point, denoise the normal vectors of the neighborhood points of the target point based on the type of the target point to obtain the denoised neighborhood point normal vectors. Specifically, if the target point is a planar point, most points in the neighborhood of the target point are coplanar, and the normal vectors of most points in the neighborhood are in the same direction as the normal vector of the target point. Remove the points whose normal vector directions are inconsistent with the normal vector of the target point, and the denoised neighborhood points can be obtained. Based on this, the denoised neighborhood point normal vectors can be obtained, and the denoised curvature value of the target point can be calculated. The denoised curvature value of the target point needs to be calculated in combination with the curvature values of the neighborhood points of the target point. The specific calculation method will be described in detail later in the present invention.

[0026] S103. Determine the feature compensation direction of the target point based on the denoised neighborhood point normal vectors of the target point, calculate the feature compensation value of the target point based on the denoised curvature value, and perform feature compensation on the target point based on the feature compensation direction and the feature compensation value.

[0027] In the embodiment of the present invention, when it is determined that the target point needs to be feature-compensated, the feature compensation direction of the target point can be determined according to the denoised neighborhood point normal vectors of the target point. For example, determine the feature compensation direction of the target point according to the direction of the denoised neighborhood point normal vectors of the target point. The specific determination of the feature compensation direction will be described in detail later in the present invention. For the feature compensation value of the target point, it can be determined according to the denoised curvature value of the target point in combination with the accuracy of the structured light 3D scanner. The specific determination process will be described in detail later in the present invention. After determining the feature compensation direction and the feature compensation value of the target point, perform feature compensation on the three-dimensional coordinates of the target point based on the feature compensation direction and the feature compensation value of the target point, and the feature compensation of the target point can be realized. For the surface three-dimensional scan point cloud data of a three-dimensional object, after performing feature compensation on all data points in the three-dimensional scan point cloud data, the feature compensation of the three-dimensional model can be realized.

[0028] The 3D model feature compensation method provided by the present invention performs tensor voting on the normal vector of a target point through the neighborhood points of the target point in the 3D point cloud data, determines the type of the target point based on the tensor voting result, denoises the normal vectors of the neighborhood points of the target point based on the type of the target point to obtain the denoised neighborhood point normal vectors, which can effectively remove the noise points on the surface where the target point is located. Calculating the denoised curvature value of the target point based on the denoised neighborhood point normal vectors can further remove the noise features on the surface where the target point is located, retain the sharp features, and determine the feature compensation direction of the target point through the denoised neighborhood point normal vectors of the target point, and calculate the feature compensation value of the target point based on the denoised curvature value. Feature compensation is performed on the target point based on the feature compensation direction and the feature compensation value, which can improve the feature compensation accuracy of the 3D model surface and will not affect the planar features, ensuring no distortion when constructing the 3D model of the object surface based on structured light scanning.

[0029] In some possible embodiments of the present invention, as Figure 2 shown, performing tensor voting on the normal vector of a target point through the neighborhood points of the target point in the 3D point cloud data and determining the type of the target point based on the tensor voting result includes: S201, calculating the first covariance matrix of the normal vectors of the neighborhood points of the target point in the 3D point cloud data, performing eigenvalue decomposition on the first covariance matrix to obtain multiple eigenvalues; S202, determining the type of the target point according to the magnitude relationship of the multiple eigenvalues.

[0030] In the embodiments of the present invention, during the process of performing tensor voting on the normal vector of a target point through the neighborhood points of the target point in the 3D point cloud data, it is necessary to first calculate the normal vectors of each point in the 3D point cloud data, and at the same time, outlier noise points can be detected and removed according to connectivity to achieve preliminary denoising. In the kd-tree of the constructed point cloud data, for the target point , cumulatively calculate the first covariance matrix of the normal vectors of all the neighborhood points of this target point , and the calculation formula is as follows:

[0031] Among them, is the first covariance matrix, is the normal vector of the neighborhood point of the target point , is the weight of the neighborhood point , is the neighborhood of the target point , and , indicating and The smaller the included angle is, the greater its weight. Adding the included angle weight can effectively suppress the influence of noise on the result. For the first covariance matrix Performing eigen decomposition can obtain three eigenvectors and three eigenvalues , , , where and ; then the first covariance matrix can be expressed as .

[0032] Furthermore, the magnitudes of the three eigenvalues , , reflect the intensities of the eigen decomposition of the normal vectors of each point in the neighborhood in the directions of the three eigenvectors. Based on this, the neighborhood where the point is located can be classified. When is approximately 1, and are approximately 0, the point is located on a plane, and the point is a plane point. When and are of comparable magnitudes, while is approximately 0, the point is located at the intersection of planes, and the point is an intersection point of planes. When , and are close in magnitude, the point is located at a corner, and the point is a corner point.

[0033] In the embodiments of the present invention, by performing tensor voting on the normal vector of the target point through the neighborhood points of the target point, the type of the target point and the neighborhood where the target point is located can be determined, which is convenient for subsequent judgment of feature compensation.

[0034] In some possible embodiments of the present invention, as Figure 3 shown, denoising the normal vectors of the neighborhood points of the target point based on the type of the target point to obtain denoised neighborhood point normal vectors includes: S301, performing eigen decomposition on the first covariance matrix to obtain a plurality of eigenvectors corresponding to a plurality of eigenvalues; S302, determining the normal vector features of the neighborhood points of the target point based on the type of the target point, updating the plurality of eigenvalues based on the neighborhood point normal vector features to obtain a plurality of updated eigenvalues, and determining a second covariance matrix based on the updated eigenvalues and the plurality of eigenvectors; S303, calculating the denoised normal vector of the target point and the denoised neighborhood point normal vectors of the neighborhood points based on the second covariance matrix.

[0035] In the embodiment of the present invention, as can be seen from the foregoing embodiment, when performing eigenvalue decomposition on the first covariance matrix, three eigenvectors are obtained and three eigenvalues , , , For the case where the above three types of target points belong to different types, the multiple eigenvalues can be updated based on the normal vector features of the neighborhood points to obtain multiple updated eigenvalues. Specifically: (1) When is approximately 1, and are approximately equal to 0, the point is located on a plane, and there is only one main orientation of the normal vector of the points in the neighborhood. The intensities in the other two directions are mainly caused by noise. At this time, update , ; (2) When and are of comparable magnitude, and is approximately equal to 0, the point is located at the intersection of planes, and there are two main orientations of the normal vector of the points in the neighborhood. The intensities in the remaining directions are mainly caused by noise. At this time, update , ; (3) When , and are close in magnitude, the point is located at a corner, and there are three main orientations of the normal vector of the points in the neighborhood. It is difficult to distinguish features from noise. At this time, update , .

[0036] For the updated eigenvalues obtained in the above three cases, the second covariance matrix can be constructed as:

[0037] At this time, the components caused by surface noise in the covariance matrix are basically removed, while the sharp features on the surface are retained.

[0038] According to the second covariance matrix, the normal vector of the target point can be updated to , where t is a proportionality coefficient, which can be set to 3 to prevent data distortion caused by excessive denoising.

[0039] Based on this, the normal vectors of all points in the 3D point cloud data can be updated, and the denoised neighborhood point normal vectors of the target points can be obtained.

[0040] In the embodiments of the present invention, the normal vectors of the target point and its neighboring points are updated by combining the tensor voting results with the normal vectors of the neighboring points of the target point, which can effectively remove surface noise, retain sharp features, and maximize the retention of surface features.

[0041] In some possible embodiments of the present invention, as Figure 4 shown, calculating the denoised curvature value of the target point based on the normal vectors of the denoised neighboring points includes: S401, calculating the third covariance matrix based on the denoised normal vector of the target point and the denoised normal vectors of the neighboring points, and calculating the preliminary denoised curvature value of the target point and the preliminary denoised curvature values of the neighboring points based on the third covariance matrix; S402, determining the weights of the neighboring points based on the preliminary denoised curvature values of the neighboring points and the preliminary denoised curvature value of the target point, and calculating the denoised curvature value of the target point by combining the preliminary denoised curvature values of the neighboring points with their weights.

[0042] In the embodiments of the present invention, after updating the normal vectors of the target point and its neighboring points, recalculating the covariance, eigenvectors, and eigenvalues according to the foregoing embodiments, the preliminary denoised curvature value of the target point can be determined as , and according to the analysis of the eigenvalues in the foregoing embodiments, it can be determined that when the target point is located on a plane, its preliminary denoised curvature value is relatively large; while when the target point is located at features such as chamfers, spherical surfaces, and cylindrical surfaces, is relatively small, and the sharper and more complex the feature, the smaller it is, and the smaller the value of is; and the value range of

[0043] is relatively wide, which can more clearly distinguish three-dimensional surfaces with different bending degrees and facilitate finding the corresponding compensation value rules.

[0044] where, , is the preliminary denoised curvature value of the neighboring point , is the preliminary denoised curvature value of the target point , is the neighboring point The weight value indicates that the greater the curvature difference between the neighborhood points and the target point, the smaller the contribution to the filtered result. Bilateral filtering can further suppress the influence of noise while preserving the original features, improving the generality of surface curvature. For example, the surface curvature of a spherical disk is as Figure 5 shown, and the histogram distribution of the surface curvature is as Figure 6 shown. Spheres with different diameters have different surface curvatures, while the surface curvatures of points on the surface of spheres with the same diameter are very close, meeting the requirements of feature compensation.

[0045] In the embodiment of the present invention, the curvature value of the target point is further denoised by the preliminary denoised curvature values of the target point and its neighborhood points, further reducing the influence of noise.

[0046] In some possible embodiments of the present invention, as Figure 7 shown, before determining the feature compensation direction of the target point based on the normal vector of the denoised neighborhood points of the target point, it includes: S701, performing eigen-decomposition on the third covariance matrix to obtain multiple compensation judgment eigenvalues; S702, when it is determined based on multiple compensation judgment eigenvalues that the region where the target point is located is a non-planar region, performing feature compensation on the target point.

[0047] In the embodiment of the present invention, before performing feature compensation on the target point, it is necessary to first determine whether the target point needs to be feature-compensated. Feature compensation needs to exclude planes and surfaces with insignificant curvature to prevent noise in these regions from being compensated, thereby reducing the accuracy of the 3D model. Specifically, eigen-decomposition can be performed on the third covariance matrix to obtain three eigenvalues , , . When any of the following three conditions is not satisfied, it can be determined that the target point is a non-planar region and feature compensation is required:

[0048]

[0049]

[0050] Among them, , and are the three eigenvalues of the third covariance matrix.

[0051] Based on this, it can be determined whether to perform feature compensation on the target point.

[0052] In the embodiment of the present invention, by determining whether the region where the feature point is located is a curved surface, it is prevented from performing feature compensation on planar points, resulting in distortion of the 3D model.

[0053] In some possible embodiments of the present invention, such as Figure 8 shown, determining the feature compensation direction of the target point based on the normal vector of the denoised neighborhood points of the target point includes: S801, determining the denoised neighborhood point plane based on the denoised neighborhood points and the normal vectors of the denoised neighborhood points, and calculating the common intersection point of each denoised neighborhood point plane; S802, determining the compensation direction vector based on the common intersection point and the target point, and determining the feature compensation direction of the target point based on the direction relationship between the compensation direction vector and the normal vector of the target point.

[0054] In the embodiments of the present invention, due to the difference between concave and convex surfaces in the three-dimensional scanned model, the normal vector of the convex surface is opposite to the feature degradation direction, while the point normal vector of the concave surface is the same as the feature degradation direction. If only expanding along the point cloud normal vector during compensation, it will cause more serious feature degradation of the concave surface. Therefore, it is necessary to first judge the concavity and convexity of the surface before compensation. To avoid the influence of locally fluctuating noise points, the denoised neighborhood point plane is determined based on the denoised neighborhood points and the normal vectors of the denoised neighborhood points, and the common intersection point of each denoised neighborhood point plane is calculated. Each neighborhood point and its normal vector can be used as a plane, and the covariance matrix of the neighborhood normal vectors is , and the intersection point of all planes in the neighborhood is calculated as , where , the intersection point and the target point the vector of is , if , then it is a convex surface, and the feature compensation direction is the direction of, if , then it is a concave surface, and the feature compensation direction is the opposite direction of.

[0055] The embodiments of the present invention determine whether the surface where the target point is located is a concave surface or a convex surface through the common intersection point of the planes of each point in the neighborhood of the target point, and then determine the feature compensation direction of the target point, ensuring the accuracy of feature compensation.

[0056] In some possible embodiments of the present invention, such as Figure 9 shown, calculating the feature compensation value of the target point based on the denoised curvature value includes: S901, determining the fitting diameter of the fitting spherical surface based on the three-dimensional point cloud data of the standard spherical disc collected by the three-dimensional point cloud data acquisition device, and calculating the feature compensation value of each spherical point in the fitting spherical surface based on the fitting diameter and the standard diameter of the standard spherical disc; S902, performing least squares fitting on the feature compensation values to obtain the feature compensation parameters of the three-dimensional point cloud data acquisition device; S903, calculating the feature compensation value of the target point based on the feature compensation parameters and the denoised curvature value.

[0057] In the embodiments of the present invention, for different structured light three-dimensional scanners, their camera parameters, laser stability, and system noise are all different. It is necessary to determine their characteristic compensation parameters. Specifically, there are two characteristic compensation parameters to be determined. The parameter a and the parameter b can be solved by first using a three-dimensional scanner to scan a standard spherical disk with a known diameter, calculating the surface curvature of all points in the three-dimensional point cloud data using the method in the foregoing embodiments, quickly screening the spherical surfaces using the curvature, and then for each screened spherical surface, removing noise points according to connectivity, fitting the spherical diameter parameter, comparing it with the true spherical diameter, and calculating the value to be compensated for each spherical surface point; finally, according to the surface curvature and the value to be compensated for each spherical surface point, performing a least squares fit uniformly to determine the parameters a and b values, so as to obtain an accurate compensation law. After determining the parameters a and b values, use the characteristic compensation value calculation formula to calculate the characteristic compensation value of the target point. Among them, is the characteristic compensation value of the target point, is the denoised curvature value of the target point.

[0058] Furthermore, according to the characteristic compensation value and the characteristic compensation direction, update the three-dimensional coordinates of the target point to achieve characteristic compensation. The updated three-dimensional coordinates of the target point are , where is the updated three-dimensional coordinates of the target point, is the original three-dimensional coordinates of the target point. For the surface three-dimensional scan point cloud data of a three-dimensional object, after performing characteristic compensation on all data points in the three-dimensional scan point cloud data, the characteristic compensation of the three-dimensional model can be achieved.

[0059] In order to better implement the three-dimensional model characteristic compensation method in the embodiments of the present invention, correspondingly, as Figure 10 shown, the embodiments of the present invention also provide a three-dimensional model characteristic compensation device. The three-dimensional model characteristic compensation device 1000 includes: A target point type determination module 1001, configured to perform tensor voting on the normal vector of the target point based on the neighborhood points of the target point in the three-dimensional point cloud data, and determine the type of the target point based on the tensor voting result; A curvature value calculation module 1002, configured to denoise the normal vector of the neighborhood points of the target point based on the type of the target point to obtain a denoised neighborhood point normal vector, and calculate the denoised curvature value of the target point based on the denoised neighborhood point normal vector; A feature compensation module 1003 is configured to determine a feature compensation direction of a target point based on the normal vector of the denoised neighborhood points of the target point, calculate a feature compensation value of the target point based on the denoised curvature value, and perform feature compensation on the target point based on the feature compensation direction and the feature compensation value.

[0060] The three-dimensional model feature compensation device 1000 provided in the above embodiment can implement the technical solutions described in the above embodiment of the three-dimensional model feature compensation method. The specific implementation principles of the above modules or units can be referred to the corresponding content in the above embodiment of the three-dimensional model feature compensation method, which will not be elaborated here.

[0061] As Figure 11 shown, the present invention also correspondingly provides a three-dimensional model scanning device 1100. The three-dimensional model scanning device 1100 includes a processor 1101, a memory 1102, and a display 1103. Figure 11 Only some components of the three-dimensional model scanning device 1100 are shown, but it should be understood that it is not required to implement all the shown components, and more or fewer components can be alternatively implemented.

[0062] In some embodiments, the processor 1101 may be a central processing unit (CPU), a microprocessor, or other data processing chips, and is configured to run program codes stored in the memory 1102 or process data, such as the three-dimensional model feature compensation method in the present invention.

[0063] In some embodiments, the processor 1101 may be a single server or a server group. The server group may be centralized or distributed. In some embodiments, the processor 1101 may be local or remote. In some embodiments, the processor 1101 may be implemented on a cloud platform. In some embodiments, the cloud platform may include a private cloud, a public cloud, a hybrid cloud, a community cloud, a distributed cloud, an internal cloud, a multi-cloud, etc., or any combination of the above.

[0064] In some embodiments, the memory 1102 may be an internal storage unit of the three-dimensional model scanning device 1100, such as a hard disk or a memory of the three-dimensional model scanning device 1100. In some other embodiments, the memory 1102 may also be an external storage device of the three-dimensional model scanning device 1100, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the three-dimensional model scanning device 1100.

[0065] Further, the memory 1102 may include both the internal storage unit of the three-dimensional model scanning device 1100 and an external storage device. The memory 1102 is used to store the application software and various types of data installed in the three-dimensional model scanning device 1100.

[0066] In some embodiments, the display 1103 may be an LED display, a liquid crystal display, a touch liquid crystal display, and an OLED (Organic Light-Emitting Diode) toucher, etc. The display 1103 is used to display the information of the three-dimensional model scanning device 1100 and to display a visual user interface. The components 1101-1103 of the three-dimensional model scanning device 1100 communicate with each other through a system bus.

[0067] In some embodiments, when the processor 1101 executes the three-dimensional model feature compensation program in the memory 1102, the following steps may be implemented: Perform tensor voting on the normal vector of the target point based on the neighborhood points of the target point in the three-dimensional point cloud data, and determine the type of the target point based on the tensor voting result; Denoise the normal vectors of the neighborhood points of the target point based on the type of the target point to obtain the denoised neighborhood point normal vectors, and calculate the denoised curvature value of the target point based on the denoised neighborhood point normal vectors; Determine the feature compensation direction of the target point based on the denoised neighborhood point normal vectors of the target point, calculate the feature compensation value of the target point based on the denoised curvature value, and perform feature compensation on the target point based on the feature compensation direction and the feature compensation value.

[0068] It should be understood that: when the processor 1101 executes the three-dimensional model feature compensation program in the memory 1102, in addition to the above functions, other functions may also be implemented. For details, reference may be made to the descriptions of the corresponding method embodiments above.

[0069] Correspondingly, an embodiment of the present application further provides a computer-readable storage medium. The computer-readable storage medium is used to store computer-readable programs or instructions. When the programs or instructions are executed by a processor, the steps or functions in the three-dimensional model feature compensation method provided by the above method embodiments can be implemented.

[0070] Those skilled in the art can understand that all or part of the processes of implementing the above method embodiments can be completed by instructing relevant hardware through a computer program. The program can be stored in a computer-readable storage medium. Among them, the computer-readable storage medium is a magnetic disk, an optical disk, a read-only memory, or a random access memory, etc.

[0071] As described above, it is only the preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.

Claims

1. A three-dimensional model feature compensation method, characterized in that: include: Performing tensor voting of normal vectors of the target point based on the neighborhood points of the target point in the three-dimensional point cloud data, and determining the type of the target point based on the tensor voting result; De-noising the normal vectors of the neighborhood points of the target point based on the type of the target point to obtain de-noised neighborhood point normal vectors, and calculating the de-noised curvature value of the target point based on the de-noised neighborhood point normal vectors; The feature compensation direction of the target point is determined based on the normal vector of the denoised neighborhood point of the target point, the feature compensation value of the target point is calculated based on the denoised curvature value, and feature compensation is performed on the target point based on the feature compensation direction and the feature compensation value.

2. The three-dimensional model feature compensation method according to claim 1, characterized in that: The step of performing tensor voting of a normal vector of the target point based on the neighborhood points of the target point in the three-dimensional point cloud data, and determining the type of the target point based on the tensor voting result, includes: Calculating a first covariance matrix of normal vectors of each neighborhood point of a target point in the three-dimensional point cloud data, and performing eigendecomposition on the first covariance matrix to obtain a plurality of eigenvalues; The type of the target point is determined according to the magnitude relationship of the multiple feature values.

3. The three-dimensional model feature compensation method according to claim 2, characterized in that: The denoising the normal vector of the neighborhood point of the target point based on the type of the target point to obtain the denoised neighborhood point normal vector includes: Performing eigendecomposition on the first covariance matrix to obtain a plurality of eigenvectors corresponding to the plurality of eigenvalues; Determine a neighborhood point normal vector feature of the target point based on the type of the target point, update the multiple eigenvalues ​​based on the neighborhood point normal vector feature to obtain multiple updated eigenvalues, and determine a second covariance matrix based on the updated eigenvalues ​​and the multiple eigenvectors; The denoised normal vector of the target point and the denoised neighborhood point normal vectors of the neighborhood points are calculated based on the second covariance matrix.

4. The three-dimensional model feature compensation method according to claim 3, characterized in that: The step of calculating the denoised curvature value of the target point based on the denoised neighbourhood point normal vector comprises: Calculating a third covariance matrix based on the denoised normal vector of the target point and the denoised normal vector of the neighbourhood point, and calculating preliminary denoised curvature values ​​of the target point and preliminary denoised curvature values ​​of the neighbourhood point based on the third covariance matrix; The weight of each neighborhood point is determined based on the preliminary denoised curvature value of each neighborhood point and the preliminary denoised curvature value of the target point, and the denoised curvature value of the target point is calculated based on the weight of each neighborhood point combined with the preliminary denoised curvature value of each neighborhood point.

5. The three-dimensional model feature compensation method according to claim 4, characterized in that: Before determining the feature compensation direction of the target point based on the denoised neighborhood point normal vector of the target point, the method includes: Performing eigendecomposition on the third covariance matrix to obtain a plurality of compensation judgment eigenvalues; When it is determined that the area where the target point is located is a non-flat area based on the multiple compensation judgment feature values, feature compensation is performed on the target point.

6. The three-dimensional model feature compensation method according to claim 1, characterized in that: The determining the feature compensation direction of the target point based on the normal vector of the denoised neighborhood point of the target point includes: Determine a denoised neighborhood point plane based on the denoised neighborhood points and the denoised neighborhood point normal vectors, and calculate a common intersection point of each denoised neighborhood point plane; A compensation direction vector is determined based on the common intersection point and the target point, and a characteristic compensation direction of the target point is determined based on a directional relationship between the compensation direction vector and a normal vector of the target point.

7. The three-dimensional model feature compensation method according to claim 1, characterized in that: The calculating the feature compensation value of the target point based on the denoised curvature value comprises: Determine a fitting diameter of a fitting sphere based on the three-dimensional point cloud data of a standard spherical disk collected by a three-dimensional point cloud data collection device, and calculate a characteristic compensation value of each spherical point in the fitting sphere based on the fitting diameter and the standard diameter of the standard spherical disk; Performing least square fitting on the characteristic compensation value to obtain characteristic compensation parameters of the three-dimensional point cloud data acquisition device; A feature compensation value of the target point is calculated based on the feature compensation parameter and the denoised curvature value.

8. A three-dimensional model feature compensation device, characterized in that: include: A target point type determination module, used to perform tensor voting of normal vectors of the target point based on the neighborhood points of the target point in the three-dimensional point cloud data, and determine the type of the target point based on the tensor voting result; A curvature value calculation module, used for denoising the normal vectors of the neighborhood points of the target point based on the type of the target point to obtain denoised neighborhood point normal vectors, and calculating the denoised curvature value of the target point based on the denoised neighborhood point normal vectors; A feature compensation module is used to determine the feature compensation direction of the target point based on the normal vector of the denoised neighborhood point of the target point, calculate the feature compensation value of the target point based on the denoised curvature value, and perform feature compensation on the target point based on the feature compensation direction and the feature compensation value.

9. A three-dimensional model scanning device, characterized in that: comprising a memory and a processor, wherein: The memory is used to store programs; The processor is coupled to the memory and is used to execute the program stored in the memory to implement the steps in the three-dimensional model feature compensation method described in any one of claims 1 to 7 above.

10. A computer-readable storage medium, characterized in that: Used to store computer-readable programs or instructions, which, when executed by a processor, can implement the steps in the three-dimensional model feature compensation method described in any one of claims 1 to 7.

Images