Three-dimensional thickness measurement method and medium

Abstract

The invention discloses a three-dimensional thickness measurement method and a medium, and the method comprises the following steps: S1, respectively scanning a first surface and a second surface, which are opposite to each other, of a measured object, and obtaining a first surface point cloud and a second surface point cloud of the measured object; s2, registering the first surface point cloud and the second surface point cloud of the measured object, and performing curved surface modeling to obtain a first curved surface model and a second curved surface model; and S3, calculating the thickness of the measured object based on the first curved surface model and the second curved surface model. According to the invention, by correcting the position attitude deviation of the scanned point clouds of the first surface and the second surface opposite to the measured object, the correct point cloud corresponding relation of the first surface and the second surface of the measured object is ensured, so that the accuracy of thickness calculation of the measured object is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of thickness measurement, more particularly to a three-dimensional thickness measurement method and medium. BACKGROUND

[0002] The existing technical means are as follows: ① contact measurement, using a micrometer, a thickness gauge and other manual contact measurement methods, the defects are: only discrete points can be obtained for thickness, the efficiency is low and it is difficult to reflect the thickness distribution of the whole plate, for thin and large measured objects, this method is time-consuming and labor-intensive, and due to the possible warping deformation of the measured object, it is difficult to ensure the constant measurement force of the contact measurement, affecting the accuracy; ② ultrasonic measurement, which can penetrate the measured object to measure the thickness, but needs to be in contact with the coupling agent, and the measured surface needs to be polished smooth, and manual point-by-point measurement still cannot quickly cover the whole measured object, the ultrasonic measurement method has limited sensitivity to local thinning and small thickness changes of the measured object, and the resolution of the obtained thickness distribution is low; ③ single-side optical scanning measurement, three-dimensional scanning technology based on laser or structured light can obtain the shape profile of the single side of the measured object, but only the single-side surface data cannot directly obtain the thickness information of the measured object, and the data of the opposite two surfaces must be obtained at the same time to calculate the thickness; ④ double-side optical scanning measurement, simultaneously scanning the opposite two surfaces of the measured object is a feasible scheme for obtaining the full-field thickness, but the problem of how to align the data of the opposite two surfaces of the measured object arises, thereby causing thickness calculation error, and the traditional method relies on strict mechanical installation alignment (for example, installing two sensors to face each other, and calibrating the distance through a standard block with known thickness), but the mechanical alignment is difficult to be absolutely accurate, and the equipment installation or vibration will introduce deviation. SUMMARY

[0003] The technical problem to be solved by the present application is to provide a three-dimensional thickness measurement method and medium.

[0004] The technical scheme adopted by the present application to solve the technical problem is: a three-dimensional thickness measurement method is constructed, the method comprises the following steps: S1: respectively scanning the opposite first surface and second surface of the measured object to obtain the first surface point cloud and the second surface point cloud of the measured object; S2: registering the first surface point cloud and the second surface point cloud of the measured object, and performing surface modeling to obtain the first surface model and the second surface model; S3: calculating the thickness of the measured object based on the first surface model and the second surface model.

[0005] In some embodiments, the step S1 comprises: obtaining the first surface target point cloud and the second surface target point cloud of the measured object based on the first surface target and the second surface target on the measured object; the first surface point cloud comprises the first surface target point cloud and the first surface surface point cloud; the second surface point cloud comprises the second surface target point cloud and the second surface surface point cloud; In step S2, the first surface point cloud and the second surface point cloud of the object under test are registered, including: registering the first surface point cloud and the second surface point cloud based on the first surface target point cloud and the second surface target point cloud.

[0006] In some embodiments, the registration of the first surface point cloud and the second surface point cloud based on the first surface target point cloud and the second surface target point cloud includes: S21: Identify the first target point cloud and the second target point cloud in the first point cloud and the second point cloud respectively; S22: Use the shape regularity features of the target to perform surface fitting on the point cloud of the first target and the point cloud of the second target respectively to obtain the center coordinates of the first target and the center coordinates of the second target. S23: Repeatedly acquire the center coordinates of a preset number of targets to form a first center coordinate set and a second center coordinate set; S24: Register the first center coordinate set and the second center coordinate set using a preset algorithm.

[0007] In some embodiments, step S24 includes: Calculate the rotation matrix and translation vector of corresponding points in the first and second center coordinate sets, and register the first and second face point clouds based on the constraint factors using the rotation matrix and translation vector.

[0008] In some embodiments, step S24 involves calculating the rotation matrix and translation vector of corresponding points in the first and second center coordinate sets, followed by: Determine if the determinant of the rotation matrix is ​​negative. If it is, adjust the rotation matrix by inverting the axes; otherwise, do nothing.

[0009] In some embodiments, step S24, which involves registering the first point cloud and the second point cloud based on the constraint factor using a rotation matrix and a translation vector, includes: registering the first point cloud and the second point cloud to the same coordinate system based on the constraint factor using a rotation matrix and a translation vector. In step S2, surface modeling is performed to obtain a first surface model and a second surface model, including: using a preset method to reconstruct the surface based on the first surface point cloud and the second surface point cloud in the same coordinate system to obtain the first surface model and the second surface model.

[0010] In some embodiments, the preset method for surface reconstruction is to interpolate and fit the first surface point cloud and the second surface point cloud respectively based on the Poisson surface reconstruction algorithm.

[0011] In some embodiments, step S3 includes: The sampling points of the first surface model and the triangular mesh of the second surface model are obtained to calculate the local thickness of the object under test.

[0012] In some embodiments, the step of obtaining sampling points of the first surface model and calculating the local thickness of the object under test using the triangular mesh of the second surface model includes: S31: Obtain the sampling points of the first surface model and the triangular mesh of the second surface model, and calculate the scalar parameters of the rays pointing from the sampling points along the normal direction into the interior of the object under test. S32: Obtain the ray pointing into the object under test in the direction of the normal of the sampling point based on scalar parameters; S33: Perform triangular processing on the ray based on the triangular mesh of the second surface model to obtain the distance between the sampling point and the second surface model; S34: Traverse all triangular meshes in the second surface model, and take the minimum distance between the sampling point and all the second surface models as the local thickness of the object at the sampling point.

[0013] In some embodiments, in step S31, the formula for calculating the scalar parameter is: , in, Approximate triangular mesh for the second surface model A vector in a triangular mesh j=1,2,3 Triangular mesh The plane normal vector, These are the sampling points selected on the first surface model. This is the unit normal vector at the sampling point.

[0014] In some embodiments, in step S32, the formula for calculating the ray is: .

[0015] In some embodiments, in step S32, the formula for calculating the distance between the ray and the second surface model is: .

[0016] In some embodiments, step S31 is followed by: The corresponding point of the sampling point in the normal direction of the second surface model is obtained based on the sampling point and the scalar parameter of the sampling point along the normal direction. Determine whether the corresponding point falls within the centroid region of the triangular mesh. If so, the scalar parameter of the sampling point along the normal direction is considered acceptable, and proceed to step S32.

[0017] In some embodiments, step S33 is followed by: Determine whether the distance between the sampling point and the second surface model is greater than 0. If yes, proceed to step S34; otherwise, calculate the scalar parameter of the ray pointing from the sampling point along the normal direction to the outside of the object in step S31, and re-execute steps S32 and S33.

[0018] In some embodiments, the method further includes: The thickness error of the measured object is made less than the sum of the surface modeling error and the discretization error of the rays pointing inward or outward in the direction of the sampling point normal. The formula for calculating the surface reconstruction error is: ;in, h is the point cloud density, and h is the reconstruction resolution. This refers to noise in the normal direction. The formula for calculating the discrete error of the ray pointing inward or outward from the sample point along the normal direction is: ;in, Triangular network for the second surface model The dimensions of the triangular face.

[0019] The present invention also provides a computer-readable storage medium, wherein the computer program, when executed by a processor, implements the three-dimensional thickness measurement method as described in any of the above embodiments.

[0020] The beneficial effects of this invention are that it involves first scanning the first and second opposing surfaces of an object to obtain point clouds of the first and second surfaces, then registering these point clouds and performing surface modeling to obtain a first surface model and a second surface model. Finally, the thickness of the object is calculated based on these two surface models. This invention improves the accuracy of thickness calculation by correcting the positional and orientation deviations of the point clouds scanned from the first and second opposing surfaces of the object, ensuring the correct correspondence between the point clouds of the first and second surfaces. Attached Figure Description

[0021] The present invention will be further described below with reference to the accompanying drawings and embodiments. In the accompanying drawings: Figure 1 This is a flowchart illustrating an embodiment of the three-dimensional thickness measurement method of the present invention; Figure 2 This is a schematic diagram of a linear laser three-dimensional sensor scanning the object being measured in one embodiment of the three-dimensional thickness measurement method of the present invention; Figure 3 This is a flowchart illustrating step S2 in one embodiment of the three-dimensional thickness measurement method of the present invention; Figure 4 This is a flowchart illustrating step S3 in one embodiment of the three-dimensional thickness measurement method of the present invention. Detailed Implementation

[0022] To provide a clearer understanding of the technical features, objectives, and effects of the present invention, specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0023] It should be noted that the flowcharts shown in the accompanying drawings are merely illustrative and do not necessarily include all content and operations / steps, nor do they necessarily have to be performed in the described order. For example, some operations / steps can be broken down, while others can be combined or partially combined; therefore, the actual execution order may change depending on the specific circumstances.

[0024] The block diagrams shown in the accompanying drawings are merely functional entities and do not necessarily correspond to physically independent entities. That is, these functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.

[0025] like Figure 1 As shown, the present invention provides a three-dimensional thickness measurement method, which includes the following steps: S1: Scan the first and second faces of the object to be measured respectively to obtain the point cloud of the first face and the point cloud of the second face of the object to be measured. S2: Register the point cloud of the first surface and the point cloud of the object under test, and perform surface modeling to obtain the first surface model and the second surface model; S3: Calculate the thickness of the object being measured based on the first surface model and the second surface model.

[0026] This invention improves the accuracy of thickness calculation by correcting the position and orientation deviation of the point cloud scans of the first and second surfaces of the object under test, ensuring the correct correspondence between the point clouds of the first and second surfaces of the object under test.

[0027] like Figure 2 As shown, in some embodiments, a line laser 3D sensor scans the first and second faces of the object under test to obtain the point cloud of the first face and the point cloud of the second face of the object under test.

[0028] Two sets of line laser 3D sensors are set up on the first and second faces of the object being measured, respectively. That is, the first and second line laser 3D sensors are arranged in a facing-to-facing structure, and the optical baselines of the two sets of line laser 3D sensors are parallel to each other and perpendicular to the surface of the object being measured, so that the laser lines projected by each sensor can cover the entire height direction of the object being measured. The line laser 3D sensor includes a laser projector and a high-resolution camera. The laser projector is used to generate a laser line and project it onto the surface of the object being measured to produce a laser point, while the high-resolution camera is used to acquire the laser point generated on the surface of the object being measured. The object being measured is fixed on the detection device to remain stable and not shake.

[0029] Before thickness measurement, the two sets of line laser 3D sensors undergo basic calibration, including intrinsic and extrinsic parameter calibration. This involves calibrating the internal parameters of the line laser 3D sensors and their relative positions in the external coordinate system to ensure the accuracy of the dimensions of the acquired point cloud and proper distortion correction. After calibration, the formal scanning phase begins. The two sets of line laser 3D sensors scan the object being measured synchronously. Specifically, the two sets of line laser 3D sensors move relative to each other along the axial direction of the object, and the laser line projected by the laser projector sweeps across the surface of the object perpendicular to the direction of movement. Here, the axial direction is merely an example and not intended to limit this application; other directions are also possible. The two sets of line laser 3D sensors can scan simultaneously to reduce measurement time.

[0030] in, Figure 2 In the diagram, l1 represents the path of the first laser projector projecting onto the first surface of the object under test, l1' represents the path of the laser points on the first surface of the object under test acquired by the first high-resolution camera, l2 represents the path of the second laser projector projecting onto the second surface of the object under test, l2' represents the path of the laser points on the second surface of the object under test acquired by the second high-resolution camera, and H represents the thickness of the object under test. The point clouds of the first and second surfaces of the object under test are calculated using the laser triangulation principle, i.e., using l1, l1', l2, and l2'.

[0031] In some embodiments, step S1 includes: acquiring a first-side target point cloud and a second-side target point cloud of the test object based on a first-side target and a second-side target on the test object; the first-side point cloud includes a first-side target point cloud and a first-side surface point cloud; the second-side point cloud includes a second-side target point cloud and a second-side surface point cloud. In step S2, the first surface point cloud and the second surface point cloud of the object under test are registered, including: registering the first surface point cloud and the second surface point cloud based on the first surface target point cloud and the second surface target point cloud.

[0032] During the scanning process, several targets are pre-arranged on the first and second faces of the object being measured to provide coordinate references. For example, the targets can be stainless steel spherical targets with a diameter of several millimeters, which are fixed to the edge area of ​​the object being measured using magnetism or clamps. The positions of the first and second spherical targets should be such that they can be observed by the two sets of laser 3D sensors during scanning, so that the target point cloud can be acquired during point cloud acquisition.

[0033] In some embodiments, as the scanning proceeds, two sets of line laser 3D sensors acquire point clouds of the target object and surface point clouds. To improve the quality of the point clouds, the acquired point clouds can be preprocessed, including filtering out outliers, using neighborhood statistics for noise reduction, and truncating irrelevant background points according to a depth threshold.

[0034] likeFigure 3 As shown, in some embodiments, registration of the first surface point cloud and the second surface point cloud based on the first surface target point cloud and the second surface target point cloud includes: S21: Identify the first target point cloud and the second target point cloud in the first point cloud and the second point cloud respectively; S22: Use the shape regularity features of the target to perform surface fitting on the point cloud of the first target and the point cloud of the second target respectively to obtain the center coordinates of the first target and the center coordinates of the second target. S23: Repeatedly acquire the center coordinates of a preset number of targets to form a first center coordinate set and a second center coordinate set; S24: Register the first center coordinate set and the second center coordinate set using a preset algorithm.

[0035] The first and second point clouds are processed separately to identify the target point clouds, i.e., the point cloud regions where the spherical targets are located. Utilizing the characteristic that the spherical targets have known diameters and regular shapes, for example, by filtering out a cluster of points with smaller curvature from the point cloud based on the known diameter of the spherical targets, spherical fitting is performed on the first and second target point clouds, i.e., spherical fitting is performed on the point cloud regions where the spherical targets are located, to calculate the center coordinates of each spherical target. The above steps are repeated until the first center coordinate set of n spherical targets is obtained from the first point cloud. ), and obtain the second center coordinate set of n spherical targets from the second point cloud ( ), ), and the center coordinates of the i-th spherical target in the first point cloud and the center coordinates of the i-th spherical target in the second point cloud ( ( ) as the corresponding points for registration, complete the registration of the first and second point clouds.

[0036] In some embodiments, step S24 includes: Calculate the rotation matrix and translation vector of corresponding points in the first and second center coordinate sets, and register the first and second face point clouds based on the constraint factors using the rotation matrix and translation vector.

[0037] The Umeyama least squares algorithm is used to calculate the rigid body transformation parameters that transform the coordinates of the second center coordinate set to the coordinates of the first center coordinate set, i.e., the points... Best alignment to point Or, place the point Best alignment to point The rigid body transformation parameters include the rotation matrix and the translation vector.

[0038] Specifically, the Umeyama least squares algorithm aims to minimize the mean square error between corresponding points to solve for the optimal rotation matrix. Translation vector , making , among which, for all Established.

[0039] In some embodiments, step S24 involves calculating the rotation matrix and translation vector of corresponding points in the first and second center coordinate sets, followed by: Determine if the determinant of the rotation matrix is ​​negative. If it is, adjust the rotation matrix by inverting the axes; otherwise, do nothing.

[0040] During the calculation of the rotation matrix and translation vector, the determinant of the rotation matrix is ​​used to check for mirror reversal. If mirror reversal is found, the first or second center coordinate set is transformed by axis inversion to eliminate mirror reversal, so that the registered first and second point clouds maintain the correct left-right relationship.

[0041] Specifically, the rotation matrix is ​​obtained through inspection. determinant To determine if a solution is negative, check if a mirror image has been found. This can be adjusted by reversing one of the axes. It is a valid rotation matrix, thus eliminating the mirror effect and ensuring that the point cloud of the first face and the point cloud of the second face match after transformation.

[0042] In some embodiments, step S24, registering the first surface point cloud and the second surface point cloud based on the constraint factor using a rotation matrix and a translation vector, includes: registering the first surface point cloud and the second surface point cloud to the same coordinate system based on the constraint factor using a rotation matrix and a translation vector. In step S2, surface modeling is performed to obtain a first surface model and a second surface model, including: using a preset method to reconstruct the surface based on the first surface point cloud and the second surface point cloud in the same coordinate system to obtain the first surface model and the second surface model.

[0043] The Umeyama least squares algorithm has a constraint scale factor of 1, which requires that the point cloud of the first facet and the point cloud of the second facet be in the same proportion to avoid unnecessary scaling.

[0044] The calculated rotation matrix and translation vector are applied to transform either the first or second point cloud, ultimately aligning them to the same 3D coordinate system, thus completing the registration. Specifically, assuming the second point cloud is aligned with the first, the rotation matrix R and translation vector t obtained from the Umeyama least squares algorithm are applied to the second point cloud, resulting in the coordinate transformation as follows: ,in, For all the second-face point clouds before the transformation, To align all the second-face point clouds, at this point, all the aligned second-face point clouds and the first-face point clouds are in the same coordinate system.

[0045] In the same coordinate system, if the point cloud of the first target surface coincides with the point cloud of the second target surface, and the error between the point cloud of the first surface and the corresponding point cloud of the second surface is within a preset range, the registration is considered successful. If the registration is unsuccessful, more targets can be added or the Iterative Closest Point (ICP) algorithm can be used to fine-tune the registration of the point cloud of the first surface and the point cloud of the second surface to further improve the registration accuracy.

[0046] In some embodiments, the preset method for surface reconstruction is to interpolate and fit the point cloud of the first surface and the point cloud of the second surface based on the Poisson surface reconstruction algorithm.

[0047] In the same three-dimensional coordinate system, surface reconstruction is performed on the first surface point cloud and the second surface point cloud respectively. The Poisson surface reconstruction algorithm is preferably used to interpolate and fit the discrete first surface point cloud and the second surface point cloud respectively to generate a three-dimensional first surface model and the second surface model.

[0048] Furthermore, during surface reconstruction, normal vector information is incorporated to ensure that the first and second surface models obtained from the surface reconstruction accurately fit the first and second point clouds. The first and second surface models obtained from the surface reconstruction are represented as triangular meshes or NURBS surfaces. These triangular meshes or NURBS surfaces are merely examples and are not intended to limit the scope of this application. The detail accuracy of the first and second surface models can be controlled by adjusting reconstruction parameters such as the octree depth.

[0049] In other embodiments, the surface reconstruction can also be combined with a polygon mesh optimization algorithm to eliminate small bumps or depressions caused by noise, thus ensuring the stability of the thickness calculation.

[0050] In some embodiments, step S3 includes: The sampling points of the first surface model and the triangular mesh of the second surface model are obtained to calculate the local thickness of the object under test.

[0051] Triangular meshes are discrete representations of a two-dimensional manifold or the surface of a three-dimensional object, composed of a large number of non-overlapping triangular facets. They are the "pixels of the 3D world," cutting complex surfaces into the simplest differentiable planar pieces, making the entire chain of geometry, vision, simulation, and manufacturing feasible.

[0052] In other embodiments, the local thickness of the object under test can be calculated by obtaining sampling points from the second surface model and the triangular mesh of the first surface model. The sampling points should be sufficiently dense, or the vertices of the triangulated mesh model can be directly used as sampling points.

[0053] like Figure 4 As shown, in some embodiments, obtaining sampling points of the first surface model and triangular meshes of the second surface model to calculate the local thickness of the object under test includes: S31: Obtain the sampling points of the first surface model and the triangular mesh of the second surface model, and calculate the scalar parameters of the rays pointing from the sampling points along the normal direction into the interior of the object under test. S32: Obtain the ray pointing into the object under test in the direction of the normal of the sampling point based on scalar parameters; S33: Perform triangular processing on the ray based on the triangular mesh of the second surface model to obtain the distance between the sampling point and the second surface model; S34: Traverse all triangular meshes in the second surface model, and take the minimum distance between the sampling point and all the second surface models as the local thickness of the object at the sampling point.

[0054] Specifically, sampling points are selected on the first surface model. Calculate the ray pointing in the direction of its normal towards the interior of the object being measured, i.e., towards the second surface of the object, and calculate the distance of this ray within the triangular mesh of the second surface model. Then, traverse all triangular meshes of the second surface model and take the minimum distance as the local thickness of the sampling point. Several sampling points in the first surface model can be selected, and the above process can be repeated to traverse all selected sampling points of the first surface model, ultimately forming a set of local thicknesses of the object being measured.

[0055] In some embodiments, in step S31, the formula for calculating the scalar parameter is: , in, Approximate triangular mesh for the second surface model A vector in a triangular mesh j=1,2,3 Triangular mesh The plane normal vector, These are the sampling points selected on the first surface model. This is the unit normal vector at the sampling point.

[0056] The formula for calculating the unit normal vector at the sampling point is: , This is the normal vector at the sampling point pointing in the direction of the internal normal to the object being measured. This is the length of the normal vector. In a triangular network... In, there exists Where x is a triangular network Any coordinate in the equation.

[0057] In some embodiments, in step S32, the formula for calculating the ray is: .

[0058] in, These are the sampling points selected on the first surface model. The unit normal vector at the sampling point. The scalar parameter is the ray pointing from the sampling point along the normal direction into the object being measured.

[0059] In some embodiments, in step S32, the formula for calculating the distance between the ray and the second surface model is: .

[0060] in, The ray pointing in the direction of the sampling point normal points into the interior of the object being measured, and ; It is a fixed format for mathematical expressions, indicating that the function... The value of the independent variable when the function reaches its minimum value, where 'a' is a function. The range of values ​​for the independent variable x. The meaning of this formula is: in all... And rays Falling into the triangular grid In the parameter t, take the function The value of the independent variable when it reaches its minimum value.

[0061] In some embodiments, in step S34, the formula for calculating the local thickness is: Where j represents the j triangular grids traversed. The meaning of this calculation formula is that when traversing j triangular meshes... The minimum value among the distances between the j rays obtained later and the second surface model is taken as the local thickness of the object at the sampling point.

[0062] In some embodiments, step S31 is followed by: The corresponding point of the sampling point in the normal direction of the second surface model is obtained based on the sampling point and the scalar parameter of the sampling point along the normal direction. Determine whether the corresponding point falls within the centroid region of the triangular mesh. If so, the scalar parameter of the sampling point along the normal direction is considered acceptable, and proceed to step S32.

[0063] The corresponding point is the intersection of the ray pointing from the sampling point into the interior of the object under test and the second surface model in the normal direction. If the sampling point of the first surface model cannot find a corresponding point on the second surface model, it may be in an edge region or due to error that the second surface model was missed. In this case, the average thickness of the neighboring points can be used as an approximation. The local thickness of the object under test at the sampling point can also be the distance between the sampling point and the corresponding point.

[0064] To determine whether a corresponding point falls within the centroid region of the triangular mesh, i.e., within the centroid coordinate threshold, the area method or the cross product sign-preserving calculation can be used.

[0065] In some embodiments, step S33 is followed by: Determine whether the distance between the sampling point and the second surface model is greater than 0. If yes, proceed to step S34; otherwise, calculate the scalar parameter of the ray pointing from the sampling point along the normal direction to the outside of the object in step S31, and re-execute steps S32 and S33.

[0066] After re-executing step S32 based on the scalar parameters of the ray pointing outward from the sample point along the normal direction, in step S33, the negative of the absolute value of the result of performing a triangle operation on the ray is taken as the distance between the sample point and the second surface model. This verification step takes into account the detection of a very small number of mirror mismatches or local penetrations: theoretically, the distance between the sample point and the second surface model should always be positive. If a negative value appears, it indicates that the local normal direction is incorrect or the first and second surface models have topological self-intersections, requiring re-unification of the normal direction or local densification of sampling correction during the surface reconstruction stage.

[0067] In some embodiments, the method further includes: The thickness error of the measured object is made less than the sum of the surface modeling error and the discretization error of the rays pointing inward or outward in the direction of the sampling point normal. The formula for calculating the surface reconstruction error is: ;in, h is the point cloud density, and h is the reconstruction resolution. This refers to noise in the normal direction. The formula for calculating the discrete error of the ray pointing inward or outward from the sample point along the normal direction is: ;in, Triangular network for the second surface model The dimensions of the triangular face.

[0068] The calculation error of local thickness mainly comes from surface reconstruction error and ray discretization error. A combined error is calculated based on these two errors, and the thickness error should be controlled within this combined error range. Experiments show that the thickness error should be better than ±0.03 mm. If the thickness error exceeds the preset range, it can be reduced by decreasing the resolution parameter in the Poisson surface reconstruction process, increasing the sampling density of sampling points, and improving the fineness of the triangular mesh faces.

[0069] In some embodiments, the method further includes: Traverse all sampling points of the first surface model in the object under test, calculate the local thickness corresponding to each sampling point, assign color mapping to sampling points of different thickness ranges, and obtain the thickness distribution map of the object under test.

[0070] Using preset colors, from blue representing thinness to red representing thickness, to represent the thickness of the object being measured, the colors are mapped to sampling points on the first surface model and their corresponding points on the second surface model to obtain a thickness distribution map, also known as a thickness heatmap. The thickness distribution map visualizes the thickness of the object being measured, making it easier to intuitively locate areas of abnormal thickness.

[0071] In some embodiments, the method further includes: The local thickness of all sampling points of the first surface model in the object under test is traversed, and the local thickness of all sampling points is normalized by the maximum local thickness and the minimum local thickness to obtain the thickness distribution in the normal direction of the object under test.

[0072] The formula for normalization is: ,in, The local thickness of the sampling point. For minimum local thickness, This represents the maximum local thickness. By normalizing the data, the thickness distribution along the normal direction of the measured object is obtained, enabling rapid identification of locally thinned areas.

[0073] In some embodiments, statistical analysis is performed based on the local thickness of all sampling points to draw a thickness distribution histogram.

[0074] Statistical analysis is performed on indicators such as maximum local thickness, minimum local thickness, average thickness, and standard deviation to create a thickness distribution histogram. For example, the thickness distribution histogram can show the main distribution range and deviation of local thicknesses.

[0075] The system can also visualize thickness distribution maps, thickness distribution along the normal direction, and thickness distribution histograms, and generate inspection reports. When there is a preset thickness tolerance range, the inspection report will give the judgment result of whether the thickness is qualified, and will mark and prompt areas that exceed the preset thickness tolerance range.

[0076] This invention achieves the following through the above process: 1. This method employs double-sided scanning to acquire the relative contours of the two sides of the entire object under test, and then uses an algorithm to calculate the local thickness in the normal direction of each sampling point, achieving a complete measurement of the thickness distribution of the object under test, rather than ordinary thickness calculation (i.e., thickness based on the Z-axis direction) which is limited to a few sampling points. This application can detect thickness anomalies in any area of ​​the entire object under test, such as local wear, corrosion thinning, etc., improving the defect detection rate.

[0077] 2. This invention utilizes an optical-based line laser 3D sensor for scanning, eliminating the need for contact with the object being measured, preventing damage to the object's surface, and requiring no coupling agent. The measurement process is fast, efficient, and safe. Compared to manual point-by-point inspection, this invention can complete the scanning and thickness calculation of the object in a shorter time, significantly improving thickness detection efficiency and making it suitable for rapid inspection of large-scale objects in production environments.

[0078] 3. By arranging spherical targets and using algorithmic registration, the dependence on mechanical installation accuracy is eliminated, achieving high-precision alignment of point clouds. Even if the measured object is deformed or the two sets of line laser 3D sensors are not perfectly aligned, the data correspondence between the first and second sides of the measured object can still be ensured by extracting the registration from the center coordinates of the spherical target. Based on the high extraction accuracy of the spherical target and the robustness of the registration algorithm, this invention can effectively correct the possible positional deviations of the measured object relative to the two sides of the scan, and improve the accuracy of the measured object thickness calculation.

[0079] 4. For test objects with uneven thickness, or test objects with ripples or curved surface structures, the thickness of the test object can be calculated by obtaining the ray in the normal direction of the sampling point. This method can accurately address situations where the two sides of the test object are not parallel or the surface has ripples. It can obtain the local thickness value at any point on the curved surface. The method of obtaining the ray in the normal direction of the sampling point to calculate the thickness of the test object has a wider applicability and coverage for test objects.

[0080] 5. Measurement results are output in the form of thickness distribution maps, thickness distribution along the normal direction, and thickness distribution histograms, allowing users to intuitively observe the spatial distribution and fluctuations of the thickness. The provided quantitative indicators such as maximum, minimum, and standard deviation facilitate comparison with thickness tolerance requirements, automatically determining whether the plate thickness is within the acceptable range, which is helpful for quality control and decision-making.

[0081] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the three-dimensional thickness measurement method as described in any of the above embodiments.

[0082] It is understood that the above embodiments only illustrate preferred embodiments of the present invention, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can freely combine the above technical features without departing from the concept of the present invention, and can also make several modifications and improvements, all of which fall within the protection scope of the present invention. Therefore, all equivalent transformations and modifications made with respect to the scope of the claims of the present invention should fall within the scope of the claims of the present invention.

Claims

1. A three-dimensional thickness measurement method, characterized in that, The method includes the following steps: S1: Scan the first and second faces of the object to be measured respectively to obtain the point cloud of the first face and the point cloud of the second face of the object to be measured. S2: Register the point cloud of the first surface and the point cloud of the object under test, and perform surface modeling to obtain the first surface model and the second surface model; S3: Calculate the thickness of the object being measured based on the first surface model and the second surface model.

2. The three-dimensional thickness measurement method according to claim 1, characterized in that, Step S1 includes: acquiring the first target point cloud and the second target point cloud of the test object based on the first target and the second target on the test object; the first target point cloud includes the first target point cloud and the first surface point cloud; the second target point cloud includes the second target point cloud and the second surface point cloud. In step S2, the first surface point cloud and the second surface point cloud of the object under test are registered, including: registering the first surface point cloud and the second surface point cloud based on the first surface target point cloud and the second surface target point cloud.

3. The three-dimensional thickness measurement method according to claim 2, characterized in that, The registration of the first surface point cloud and the second surface point cloud based on the first surface target point cloud includes: S21: Identify the first target point cloud and the second target point cloud in the first point cloud and the second point cloud respectively; S22: Use the shape regularity features of the target to perform surface fitting on the point cloud of the first target and the point cloud of the second target respectively to obtain the center coordinates of the first target and the center coordinates of the second target. S23: Repeatedly acquire the center coordinates of a preset number of targets to form a first center coordinate set and a second center coordinate set; S24: Register the first center coordinate set and the second center coordinate set using a preset algorithm.

4. The three-dimensional thickness measurement method according to claim 3, characterized in that, Step S24 includes: Calculate the rotation matrix and translation vector of corresponding points in the first and second center coordinate sets, and register the first and second face point clouds based on the constraint factors using the rotation matrix and translation vector.

5. The three-dimensional thickness measurement method according to claim 4, characterized in that, In step S24, the rotation matrix and translation vector of corresponding points in the first center coordinate set and the second center coordinate set are calculated, followed by: Determine if the determinant of the rotation matrix is ​​negative. If it is, adjust the rotation matrix by inverting the axes; otherwise, do nothing.

6. The three-dimensional thickness measurement method according to claim 5, characterized in that, In step S24, registering the first and second point clouds based on constraint factors using rotation matrices and translation vectors includes: registering the first and second point clouds to the same coordinate system based on constraint factors using rotation matrices and translation vectors. In step S2, surface modeling is performed to obtain a first surface model and a second surface model, including: using a preset method to reconstruct the surface based on the first surface point cloud and the second surface point cloud in the same coordinate system to obtain the first surface model and the second surface model.

7. The three-dimensional thickness measurement method according to claim 6, characterized in that, The preset method for surface reconstruction is as follows: interpolation and fitting of the first and second surface point clouds based on the Poisson surface reconstruction algorithm.

8. The three-dimensional thickness measurement method according to claim 1, characterized in that, Step S3 includes: The sampling points of the first surface model and the triangular mesh of the second surface model are obtained to calculate the local thickness of the object under test.

9. The three-dimensional thickness measurement method according to claim 8, characterized in that, The process of obtaining sampling points from the first surface model and calculating the local thickness of the object under test using the triangular mesh of the second surface model includes: S31: Obtain the sampling points of the first surface model and the triangular mesh of the second surface model, and calculate the scalar parameters of the rays pointing from the sampling points along the normal direction into the interior of the object under test. S32: Obtain the ray pointing into the object under test in the direction of the normal of the sampling point based on scalar parameters; S33: Perform triangular processing on the ray based on the triangular mesh of the second surface model to obtain the distance between the sampling point and the second surface model; S34: Traverse all triangular meshes in the second surface model, and take the minimum distance between the sampling point and all the second surface models as the local thickness of the object at the sampling point.

10. The three-dimensional thickness measurement method according to claim 9, characterized in that, In step S31, the formula for calculating the scalar parameter is: ， in, Let be a vector in the approximate triangular mesh △j of the second surface model, where the triangular mesh is... j=1,2,3 Triangular mesh The plane normal vector, where p is a sampling point selected on the first surface model. This is the unit normal vector at the sampling point.

11. The three-dimensional thickness measurement method according to claim 10, characterized in that, In step S32, the formula for calculating the ray is: .

12. The three-dimensional thickness measurement method according to claim 11, characterized in that, In step S32, the formula for calculating the distance between the ray and the second surface model is: .

13. The three-dimensional thickness measurement method according to claim 12, characterized in that, The process following step S31 also includes: The corresponding point of the sampling point in the normal direction of the second surface model is obtained based on the sampling point and the scalar parameter of the sampling point along the normal direction. Determine whether the corresponding point falls within the centroid region of the triangular mesh. If so, the scalar parameter of the sampling point along the normal direction is considered acceptable, and proceed to step S32.

14. The three-dimensional thickness measurement method according to claim 10, characterized in that, The step S33 is followed by: Determine whether the distance between the sampling point and the second surface model is greater than 0. If yes, proceed to step S34; otherwise, calculate the scalar parameter of the ray pointing from the sampling point along the normal direction to the outside of the object in step S31, and re-execute steps S32 and S33.

15. The three-dimensional thickness measurement method according to claim 14, characterized in that, The method further includes: The thickness error of the measured object is made less than the sum of the surface modeling error and the discretization error of the rays pointing inward or outward in the direction of the sampling point normal. The formula for calculating the surface reconstruction error is: Where ρ is the point cloud density and h is the reconstruction resolution. This refers to noise in the normal direction. The formula for calculating the discrete error of the ray pointing inward or outward from the sample point along the normal direction is: Where Δ is the triangular network of the second surface model. The dimensions of the triangular face.

16. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the three-dimensional thickness measurement method as described in any one of claims 1-15.

Images