A method for measuring micro-deformation of a physical model test structure

Abstract

The application provides a physical model test structure micro-deformation measurement method and relates to the technical field of structure deformation measurement. In view of the problems of low measurement efficiency and poor measurement accuracy of the structure micro-deformation measurement method in the prior art, the application obtains original point cloud data before and after deformation based on a binocular camera, rotates the original point cloud data before and after deformation by 180 degrees around a horizontal coordinate axis to obtain converted point cloud data, determines a left side region first effective measurement region and a right side region second effective measurement region in the converted point cloud data according to camera parameters, processes the first effective measurement region to obtain a first continuous surface before deformation and a second continuous surface after deformation, processes the second effective measurement region to obtain a third continuous surface before deformation and a fourth continuous surface after deformation, respectively splices the surfaces before and after deformation to form original surfaces and deformed surfaces, and determines micro-deformation. The application has high measurement efficiency and measurement accuracy.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of structural deformation measurement, in particular to a physical model test structural micro-deformation measurement method. BACKGROUND

[0002] The structural micro-deformation in the physical model test refers to the small and difficult-to-directly-observe displacement or deformation of the model structure in the test, which needs to be captured by high-precision measurement means.

[0003] At present, the structural micro-deformation measurement mainly adopts the contact sensor method and the laser interference and scanning method. The contact sensor method converts the deformation into an electrical signal after arranging physical sensors on the structure surface to calculate the deformation, but this method is easily disturbed by environmental factors such as temperature drift, and has poor adaptability to complex curved surface structures. The laser interference and scanning method uses a laser beam to project onto the structure surface, and analyzes the deformation through reflected light interference fringes or point cloud displacement to achieve micron-level precision, but the full-field scanning takes too long.

[0004] The patent with publication number CN119245531A discloses a structural deformation monitoring method based on binocular vision and coordinate conversion, which comprises calibrating the left and right binocular stereo vision cameras based on a calibration board to obtain multiple calibration board images of the calibration board under different postures; collecting target structure images based on the binocular stereo vision camera; calculating calibration parameters based on the multiple calibration board images, and obtaining a stiffness conversion matrix when the calibration board and the target structure surface coincide; and performing coordinate conversion on the measurement points of the target structure image based on the stiffness conversion matrix. The coordinate conversion depends on the stiffness conversion matrix, needs to be realized through multiple complex operations, and the effective area is not considered for screening, resulting in incomplete coverage area.

[0005] In summary, the measurement efficiency and the measurement accuracy of the structural micro-deformation measurement method in the prior art are low.

[0006] Therefore, it is of great significance to develop a physical model test structural micro-deformation measurement method for improving the measurement efficiency and the measurement accuracy. SUMMARY

[0007] In view of the problems of low measurement efficiency and poor measurement accuracy of the structural micro-deformation measurement method in the prior art, the present application proposes a physical model test structural micro-deformation measurement method, which specifically comprises the following steps:

[0008] S1, obtaining the original point cloud data of the to-be-measured structure before and after deformation based on a binocular camera, wherein the to-be-measured structure is divided into a left side area and a right side area, a left camera in the binocular camera photographs the left side area, and a right camera in the binocular camera photographs the right side area;

[0009] S2. Rotate the original point cloud data before and after deformation by 180° around the horizontal coordinate axis to obtain the transformed point cloud data before and after deformation.

[0010] S3. Based on the camera parameters of the left camera, determine the first effective measurement area of ​​the left region in the converted point cloud data; based on the camera parameters of the right camera, determine the second effective measurement area of ​​the right region in the converted point cloud data.

[0011] S4. Perform meshing and interpolation processing on the transformed point cloud data before and after deformation in the first effective measurement area to obtain the first continuous surface before deformation and the second continuous surface after deformation; perform meshing and interpolation processing on the transformed point cloud data before and after deformation in the second effective measurement area to obtain the third continuous surface before deformation and the fourth continuous surface after deformation.

[0012] S5. Join the first continuous surface and the third continuous surface to form the original surface, and join the second continuous surface and the fourth continuous surface to form the deformed surface;

[0013] S6. Determine the micro-deformation of the structure under test based on the difference between the original surface and the deformed surface.

[0014] Furthermore, in step S3, determining the first effective measurement area of ​​the left region in the converted point cloud data based on the camera parameters of the left camera includes: determining the planar field of view, target depth, and tolerance value of the left region based on the camera parameters of the left camera; and determining the first effective measurement area in the converted point cloud data based on the planar field of view, target depth, and tolerance value of the left region.

[0015] Furthermore, the left camera parameters include the field of view (FAR) parameters of the left camera and the vertical distance from the left camera to the left region of the structure under test. Based on the left camera parameters, the planar field of view, target depth, and tolerance value of the left region are determined, including: calculating the horizontal field of view width of the left region based on the FAR parameters and the vertical distance from the left camera to the left region of the structure under test; calculating the vertical field of view height of the left region based on the FAR parameters and the vertical distance from the left camera to the left region of the structure under test; determining the planar field of view range of the left region based on the horizontal field of view width and the vertical field of view height of the left region; and determining the target depth and tolerance value of the left region based on the vertical distance from the left camera to the left region of the structure under test.

[0016] Furthermore, in step S3, determining the second effective measurement area of ​​the right region in the converted point cloud data based on the camera parameters of the right camera includes: determining the planar field of view, target depth, and tolerance value of the right region based on the camera parameters of the right camera; and determining the second effective measurement area in the converted point cloud data based on the planar field of view, target depth, and tolerance value of the right region.

[0017] Furthermore, the right-side camera parameters include the field of view (FAR) parameters of the right-side camera and the vertical distance from the right-side camera to the right-side region of the structure under test. Based on the right-side camera parameters, the planar field of view, target depth, and tolerance value of the right-side region are determined, including: calculating the horizontal field of view width of the right-side region based on the FAR parameters of the right-side camera and the vertical distance from the right-side camera to the right-side region of the structure under test; calculating the vertical field of view height of the right-side region based on the FAR parameters of the right-side camera and the vertical distance from the right-side camera to the right-side region of the structure under test; determining the planar field of view range of the right-side region based on the horizontal field of view width and the vertical field of view height of the right-side region; and determining the target depth and tolerance value of the right-side region based on the vertical distance from the right-side camera to the right-side region of the structure under test.

[0018] Furthermore, in step S4, the transformed point cloud data before and after deformation in the first effective measurement area are respectively subjected to meshing and interpolation processing to obtain the first continuous surface before deformation and the second continuous surface after deformation. This includes: according to the planar field of view of the left region, the transformed point cloud data before and after deformation in the first effective measurement area are respectively subjected to planar meshing to obtain the first mesh model before deformation and the second mesh model after deformation; the mesh nodes in the first mesh model are subjected to local weighted interpolation to obtain the first continuous surface, and the mesh nodes in the second mesh model are subjected to local weighted interpolation to obtain the second continuous surface.

[0019] Furthermore, a first continuous surface is obtained by performing local weighted interpolation on the grid nodes in the first grid model, and a second continuous surface is obtained by performing local weighted interpolation on the grid nodes in the second grid model. This includes: using the Thiessen polygon method to generate Thiessen polygons at each transformed point cloud data point in the first and second grid models, with each transformed point cloud data point corresponding to a Thiessen polygon; determining the target Thiessen polygon where the grid node is located, with the transformed point cloud data in the target Thiessen polygon serving as the natural neighbor points of the grid node; determining the weight of the natural neighbor points based on the area of ​​the target Thiessen polygon where the natural neighbor points are located; calculating the depth data of the corresponding grid node based on the weight of the natural neighbor points, and determining the first and second continuous surfaces.

[0020] Furthermore, in step S4, the transformed point cloud data before and after deformation in the second effective measurement area are respectively subjected to meshing and interpolation processing to obtain the third continuous surface before deformation and the fourth continuous surface after deformation. This includes: according to the planar field of view of the right region, the transformed point cloud data before and after deformation in the second effective measurement area are respectively subjected to planar meshing to obtain the third mesh model before deformation and the fourth mesh model after deformation; the mesh nodes in the third mesh model are subjected to local weighted interpolation to obtain the third continuous surface, and the mesh nodes in the fourth mesh model are subjected to local weighted interpolation to obtain the fourth continuous surface.

[0021] Furthermore, a third continuous surface is obtained by performing local weighted interpolation on the grid nodes in the third grid model, and a fourth continuous surface is obtained by performing local weighted interpolation on the grid nodes in the fourth grid model. This includes: using the Thiessen polygon method to generate Thiessen polygons at each transformed point cloud data point in the third and fourth grid models, with each transformed point cloud data point corresponding to one Thiessen polygon; determining the target Thiessen polygon where the grid node is located, with the transformed point cloud data in the target Thiessen polygon serving as the natural neighbor points of the grid node; determining the weight of the natural neighbor points based on the area of ​​the target Thiessen polygon where the natural neighbor points are located; calculating the depth data of the corresponding grid node based on the weight of the natural neighbor points, and determining the third and fourth continuous surfaces.

[0022] Furthermore, in step S5, splicing the first continuous surface and the third continuous surface to form an original surface, and splicing the second continuous surface and the fourth continuous surface to form a deformable surface, includes: determining the maximum value of the first continuous surface on the horizontal coordinate axis; determining the minimum value of the third continuous surface on the horizontal coordinate axis; determining a horizontal displacement based on the maximum value of the first continuous surface on the horizontal coordinate axis and the minimum value of the third continuous surface on the horizontal coordinate axis; splicing the first continuous surface and the third continuous surface based on the horizontal displacement to form an original surface, and splicing the second continuous surface and the fourth continuous surface to form a deformable surface.

[0023] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0024] This invention utilizes binocular cameras to acquire raw point cloud data before and after deformation. The raw point cloud data is rotated 180° around the horizontal coordinate axis to obtain transformed point cloud data. Based on the camera parameters of the left camera, a first effective measurement region on the left side is determined from the transformed point cloud data. Based on the camera parameters of the right camera, a second effective measurement region on the right side is determined from the transformed point cloud data. The first effective measurement region is meshed and interpolated to obtain a first continuous surface before deformation and a second continuous surface after deformation. The second effective measurement region is meshed and interpolated to obtain a third continuous surface before deformation and a fourth continuous surface after deformation. The surfaces before and after deformation are then stitched together to form the original surface and the deformed surface, determining the micro-deformation. Notably, rotating the raw point cloud data 180° around the horizontal coordinate axis solves the problem of coordinate system inconsistency caused by binocular camera installation. Automatic coordinate rotation replaces traditional manual calibration matrix transformation, eliminating the need for manual adjustment of the coordinate reference and significantly improving computational efficiency. In addition, determining the effective measurement area based on camera parameters can eliminate interference points such as edge anomalies, reduce the interference of invalid data on deformation calculation, ensure that micro-deformation results only reflect the true deformation of the structure, and improve the accuracy of micro-deformation measurement. Attached Figure Description

[0025] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0026] Figure 1 This is a flowchart of a method for measuring micro-deformation of a physical model test structure provided in an embodiment of the present invention. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0028] The specific embodiments of the present invention will be described below.

[0029] To address the issues of low measurement efficiency and poor accuracy in existing methods for measuring structural micro-deformation, this invention utilizes a binocular camera to acquire raw point cloud data before and after deformation. This raw point cloud data is then rotated 180° around the horizontal coordinate axis to obtain transformed point cloud data. Based on camera parameters, a first effective measurement region on the left and a second effective measurement region on the right are determined from the transformed point cloud data. The first effective measurement region is processed to obtain a first continuous surface before deformation and a second continuous surface after deformation. The second effective measurement region is processed to obtain a third continuous surface before deformation and a fourth continuous surface after deformation. These surfaces are then stitched together to form the original surface and the deformed surface, thus determining the amount of micro-deformation. This invention offers high measurement efficiency and accuracy.

[0030] This invention provides a method for measuring the micro-deformation of a physical model experimental structure. Figure 1 This is a flowchart of a method for measuring micro-deformation of a physical model test structure provided in an embodiment of the present invention, as shown below. Figure 1 As shown, the specific steps include the following:

[0031] S1. Acquire the original point cloud data of the structure under test before and after deformation based on a binocular camera. The structure under test is divided into a left region and a right region. The left camera of the binocular camera captures the left region, and the right camera of the binocular camera captures the right region.

[0032] A binocular camera is an imaging device consisting of a left camera and a right camera. It achieves non-contact measurement by acquiring three-dimensional point cloud data of the target. The structure under test can include physical models or engineering components that require micro-deformation monitoring. The raw point cloud data is a three-dimensional coordinate dataset directly acquired by the binocular camera, recording the X, Y, and Z three-dimensional coordinate information of the surface of the structure under test.

[0033] The test area is divided into a left region and a right region. A stereo camera is activated; the left camera captures the left region of the structure under test, generating the original point cloud data of the left region before deformation. The right camera captures the right region of the structure under test, generating the original point cloud data of the right region before deformation. Deformation is applied to the structure under test, and the above shooting operation is repeated. The left camera generates the original point cloud data of the left region after deformation, and the right camera generates the original point cloud data of the right region after deformation. These four sets of original point cloud data are saved as the basis for subsequent coordinate calibration, effective region extraction, and micro-deformation calculation. By using left and right partitioned shooting, the limitation of the single camera's field of view is solved, allowing coverage of larger-sized test structures. Simultaneously, independent acquisition from the left and right sides improves data processing efficiency.

[0034] S2. Rotate the original point cloud data before and after deformation by 180° around the horizontal coordinate axis to obtain the transformed point cloud data before and after deformation.

[0035] The horizontal coordinate axis, also known as the X-axis in the 3D coordinate system, serves as the reference axis for point cloud data rotation operations. Transformed point cloud data refers to the point cloud data obtained after rotating the original point cloud data 180° around the X-axis. Specifically, the Y-axis and Z-axis coordinates of the original point cloud data are negativeed, achieving a 180° flip around the X-axis, outputting transformed point cloud data that conforms to a unified coordinate system, thus correcting coordinate direction deviations during binocular camera shooting. This solves the coordinate system mirroring problem caused by the left and right viewing angles of the binocular camera. The transformed data is aligned under the same X-axis reference, avoiding subsequent stitching misalignments due to inconsistent coordinate references, and providing an accurate spatial basis for calculating full-field deformation. Compared to traditional methods that require manual calculation of calibration matrices for coordinate transformation, which typically takes more than 10 minutes, the rotation operation in this embodiment is automatically executed by a corresponding function, taking less than 1 second, significantly shortening the data processing cycle and greatly improving measurement efficiency.

[0036] S3. Based on the camera parameters of the left camera, determine the first effective measurement area of ​​the left region in the converted point cloud data, and based on the camera parameters of the right camera, determine the second effective measurement area of ​​the right region in the converted point cloud data.

[0037] Camera parameters are core metrics for evaluating camera performance and applicable scenarios. These parameters include, for example, the field of view. The first effective measurement region is a front-view, interference-free area selected from the converted point cloud data in the left region based on the left-side camera parameters. The second effective measurement region is a front-view, interference-free area selected from the converted point cloud data in the right region based on the right-side camera parameters. Using the camera parameters as the basis for calculation, the physical extent of the effective region is derived. Effective point clouds are extracted from the converted point cloud data, invalid data is eliminated, and the core measurement area is locked, removing interference for subsequent high-precision deformation calculations.

[0038] Specifically, based on the camera parameters of the left camera, the first effective measurement area of ​​the left region is determined in the converted point cloud data, including: determining the planar field of view, target depth, and tolerance value of the left region based on the camera parameters of the left camera; and determining the first effective measurement area in the converted point cloud data based on the planar field of view, target depth, and tolerance value of the left region.

[0039] The left-side camera parameters include the field of view (FAR) parameters of the left-side camera and the vertical distance from the left-side camera to the left-side region of the structure under test. Based on the left-side camera parameters, the planar field of view, target depth, and tolerance value of the left-side region are determined, including: calculating the horizontal field of view width of the left-side region based on the FAR parameters of the left-side camera and the vertical distance from the left-side camera to the left-side region of the structure under test; calculating the vertical field of view height of the left-side region based on the FAR parameters of the left-side camera and the vertical distance from the left-side camera to the left-side region of the structure under test; determining the planar field of view range of the left-side region based on the horizontal field of view width and the vertical field of view height of the left-side region; and determining the target depth and tolerance value of the left-side region based on the vertical distance from the left-side camera to the left-side region of the structure under test.

[0040] The field of view parameters of the left camera include the horizontal field of view a1 and the vertical field of view b1. The vertical distance from the left camera to the left side of the structure under test is d1. The horizontal width of the left area's planar field of view is 2 × d1 × tan(a1 / 2), and the vertical height is 2 × d1 × tan(b1 / 2). The target depth is the theoretical Z-axis coordinate of the left camera facing the plane of the structure under test, and the target depth is -d1. The tolerance value refers to the allowable fluctuation range of depth set to compensate for point cloud noise. For example, the tolerance value is d1 × 0.15, which means the tolerance value is 15% of the camera distance. The tolerance value is proportional to the shooting distance, rather than a fixed value, to ensure that the tolerance value is dynamically adjusted with the camera shooting distance at different shooting distances, always conforming to the true distribution of point cloud noise in different scenes, and avoiding the problem of missing effective points or selecting too many interference points caused by a fixed tolerance.

[0041] In this embodiment, by dynamically calculating the boundary based on camera parameters, it can adapt to different shooting distances and structural shapes, increasing the effective measurement coverage of the left region to ≥95% and avoiding the loss of effective data due to fixed ranges. Through precise planar field of view and depth constraints, invalid data such as reflection points and edge distortion points can be effectively eliminated, providing high-quality point clouds for subsequent mesh generation and natural neighbor interpolation, and reducing the interference of invalid data on deformation calculation.

[0042] Specifically, based on the camera parameters of the right-side camera, the second effective measurement area of ​​the right-side region is determined in the converted point cloud data, including: determining the planar field of view, target depth, and tolerance value of the right-side region based on the camera parameters of the right-side camera; and determining the second effective measurement area in the converted point cloud data based on the planar field of view, target depth, and tolerance value of the right-side region.

[0043] The parameters of the right-side camera include the field of view (FAR) of the right-side camera and the vertical distance from the right-side camera to the right region of the structure under test. Based on the camera parameters of the right-side camera, the planar field of view, target depth, and tolerance value of the right-side region are determined, including: calculating the horizontal field of view width of the right-side region based on the FAR of the right-side camera and the vertical distance from the right-side camera to the right region of the structure under test; calculating the vertical field of view height of the right-side region based on the FAR of the right-side camera and the vertical distance from the right-side camera to the right region of the structure under test; determining the planar field of view range of the right-side region based on the horizontal field of view width and the vertical field of view height of the right-side region; and determining the target depth and tolerance value of the right-side region based on the vertical distance from the right-side camera to the right region of the structure under test.

[0044] The field of view parameters of the right camera include the horizontal field of view a2 and the vertical field of view b2. The vertical distance from the left camera to the left side of the structure under test is d2. The horizontal width of the right area's planar field of view is 2 × d2 × tan(a2 / 2), and the vertical height is 2 × d2 × tan(b2 / 2). The target depth is the theoretical Z-axis coordinate of the plane of the structure under test facing the right camera, and the target depth is -d2. The tolerance value refers to the allowable fluctuation range of depth set to compensate for point cloud noise. For example, the tolerance value is d2 × 0.15, which means the tolerance value is 15% of the camera distance. The tolerance value is proportional to the shooting distance, rather than a fixed value, to ensure that the tolerance value is dynamically adjusted with the camera shooting distance at different shooting distances, always conforming to the true distribution of point cloud noise in different scenes, and avoiding the problem of missing effective points or selecting too many interference points caused by a fixed tolerance.

[0045] In this embodiment, by dynamically calculating the boundary based on camera parameters, it can adapt to different shooting distances and structural shapes, increasing the effective measurement coverage of the right region to ≥95% and avoiding the loss of effective data due to fixed ranges. Through precise planar field of view and depth constraints, invalid data such as reflection points and edge distortion points can be effectively eliminated, providing high-quality point clouds for subsequent mesh generation and natural neighbor interpolation, and reducing the interference of invalid data on deformation calculation.

[0046] S4. Perform meshing and interpolation processing on the transformed point cloud data before and after deformation in the first effective measurement area to obtain the first continuous surface before deformation and the second continuous surface after deformation; perform meshing and interpolation processing on the transformed point cloud data before and after deformation in the second effective measurement area to obtain the third continuous surface before deformation and the fourth continuous surface after deformation.

[0047] Specifically, the transformed point cloud data before and after deformation in the first effective measurement area are respectively meshed and interpolated to obtain the first continuous surface before deformation and the second continuous surface after deformation. This includes: according to the planar field of view of the left area, the transformed point cloud data before and after deformation in the first effective measurement area are respectively meshed to obtain the first mesh model before deformation and the second mesh model after deformation; the mesh nodes in the first mesh model are locally weighted and interpolated to obtain the first continuous surface, and the mesh nodes in the second mesh model are locally weighted and interpolated to obtain the second continuous surface.

[0048] Meshing involves dividing the effective region's XY plane into a regular grid with a fixed step size. The grid nodes serve as the reference points for constructing the continuous surface. The grid step size is dynamically calculated based on the effective region's dimensions. For example, if the grid step size is 1% of the average width and height of the effective region, and the effective region's width is 945mm and its height is 546mm, then the grid step size would be approximately 7.45mm. Interpolation involves calculating the weights of the point cloud data surrounding each grid node and using a weighted average to complete the Z-axis coordinate value of each grid node, thereby converting discrete points into a continuous surface.

[0049] Based on the X-axis and Y-axis coordinates of the point cloud of the first effective measurement area, the X-axis and Y-axis ranges are determined, thus obtaining the boundary of the planar field of view of the left region, ensuring that the mesh covers the entire first effective measurement area. The mesh step size is calculated based on the width and height of the first effective measurement area. Regular mesh coordinates are generated based on the range of the first effective measurement area and the mesh step size. Since only the Z-axis height changes before and after deformation, the XY-axis mesh does not need to be generated repeatedly. This mesh coordinate matrix serves as both the first and second mesh models.

[0050] Based on the transformed point cloud data before deformation, local weighted interpolation is performed on the grid nodes in the first mesh model to calculate the Z-axis height of each grid node before deformation, generating the first continuous surface. Based on the transformed point cloud data after deformation, local weighted interpolation is performed on the grid nodes in the second mesh model to calculate the Z-axis height of each grid node after deformation, generating the second continuous surface. The same XY mesh is used before and after deformation, with only the Z-axis height differing, avoiding calculation errors in deformation caused by mesh position deviations. Discrete point clouds cannot be directly compared for deformation differences; converting discrete points into a continuous surface facilitates the calculation of micro-deformations.

[0051] Based on the above embodiments, a first continuous surface is obtained by performing local weighted interpolation on the grid nodes in the first grid model, and a second continuous surface is obtained by performing local weighted interpolation on the grid nodes in the second grid model. This includes: using the Thiessen polygon method, generating Thiessen polygons at each transformed point cloud data point in the first and second grid models, with each transformed point cloud data point corresponding to one Thiessen polygon; determining the target Thiessen polygon where the grid node is located, with the transformed point cloud data in the target Thiessen polygon serving as the natural neighbor points of the grid node; determining the weight of the natural neighbor points based on the area of ​​the target Thiessen polygon where the natural neighbor points are located; calculating the depth data of the corresponding grid node based on the weight of the natural neighbor points, and determining the first and second continuous surfaces.

[0052] The Voronoi diagram method, also known as the Thiessen polygon method, is a spatial partitioning algorithm that uses discrete transformed point cloud data as generation points to divide the plane into non-overlapping polygons. Each polygon contains only one generation point, and the distance from any point within the polygon to that generation point is less than the distance to any other generation point.

[0053] For example, within the first effective measurement area, there are 100 transformed point cloud data points before transformation. 100 non-overlapping polygons are generated using the Thiessen polygon method, with each polygon uniquely corresponding to one transformed point cloud data point. In the first mesh model, the target Thiessen polygon containing a given mesh node is determined, and the transformed point cloud data corresponding to the target Thiessen polygon is the natural neighbor point. There can be multiple target Thiessen polygons, corresponding to multiple natural neighbor points. The area of ​​the Thiessen polygon containing each natural neighbor point is measured, and the weight of each neighbor point is determined. The depth data (i.e., Z-axis coordinate) of the mesh node is obtained by multiplying the Z-axis coordinate of each natural neighbor point by its corresponding weight and then summing the results. The above steps are repeated to calculate the depth data of all mesh nodes in the first mesh model. These depth data points are arranged according to their mesh positions, forming the first continuous surface before deformation. Referring to the above embodiment, the same operation is performed on the second mesh model using the Thiessen polygon method to obtain the second continuous surface.

[0054] Specifically, the transformed point cloud data before and after deformation in the second effective measurement area are meshed and interpolated to obtain the third continuous surface before deformation and the fourth continuous surface after deformation. This includes: based on the planar field of view of the right region, the transformed point cloud data before and after deformation in the second effective measurement area are meshed to obtain the third mesh model before deformation and the fourth mesh model after deformation; the mesh nodes in the third mesh model are locally weighted and interpolated to obtain the third continuous surface, and the mesh nodes in the fourth mesh model are locally weighted and interpolated to obtain the fourth continuous surface.

[0055] Based on the X-axis and Y-axis coordinates of the point cloud of the second effective measurement area, the X-axis and Y-axis ranges are determined, thus obtaining the boundary of the planar field of view of the right region, ensuring that the mesh covers the entire second effective measurement area. The mesh step size is calculated based on the width and height of the second effective measurement area. Regular mesh coordinates are generated based on the range of the second effective measurement area and the mesh step size. Since only the Z-axis height changes before and after deformation, the XY-axis mesh does not need to be generated repeatedly. This mesh coordinate matrix serves as both the third and fourth mesh model.

[0056] Based on the transformed point cloud data before deformation, local weighted interpolation is performed on the grid nodes in the third mesh model to calculate the Z-axis height of each grid node before deformation, generating the third continuous surface. Based on the transformed point cloud data after deformation, local weighted interpolation is performed on the grid nodes in the fourth mesh model to calculate the Z-axis height of each grid node after deformation, generating the fourth continuous surface. The same XY mesh is used before and after deformation, with only the Z-axis height differing, avoiding calculation errors in deformation caused by mesh position deviations. Discrete point clouds cannot directly compare deformation differences; converting discrete points into a continuous surface facilitates the calculation of micro-deformations.

[0057] Based on the above embodiments, a third continuous surface is obtained by performing local weighted interpolation on the grid nodes in the third grid model, and a fourth continuous surface is obtained by performing local weighted interpolation on the grid nodes in the fourth grid model. This includes: using the Thiessen polygon method to generate Thiessen polygons at each transformed point cloud data location in the third and fourth grid models, with each transformed point cloud data corresponding to one Thiessen polygon; determining the target Thiessen polygon where the grid node is located, with the transformed point cloud data in the target Thiessen polygon serving as the natural neighbor points of that grid node; determining the weight of the natural neighbor points based on the area of ​​the target Thiessen polygon where the natural neighbor points are located; calculating the depth data of the corresponding grid node based on the weight of the natural neighbor points, and determining the third and fourth continuous surfaces.

[0058] For example, within the second effective measurement area, there are 100 transformed point cloud data points before transformation. 100 non-overlapping polygons are generated using the Thiessen polygon method, with each polygon uniquely corresponding to one transformed point cloud data point. In the third mesh model, the target Thiessen polygon containing a given mesh node is determined, and the transformed point cloud data corresponding to the target Thiessen polygon is the natural neighbor point. There can be multiple target Thiessen polygons, corresponding to multiple natural neighbor points. The area of ​​the Thiessen polygon containing each natural neighbor point is measured, and the weight of each neighbor point is determined. The depth data (i.e., Z-axis coordinate) of the mesh node is obtained by multiplying the Z-axis coordinate of each natural neighbor point by its corresponding weight and then summing the results. The above steps are repeated to calculate the depth data of all mesh nodes in the third mesh model. These depth data points are arranged according to their mesh positions, forming the third continuous surface before deformation. Referring to the above embodiment, the same operation is performed on the fourth mesh model using the Thiessen polygon method to obtain the fourth continuous surface.

[0059] When performing local weighted interpolation on grid nodes in the first and second effective measurement regions, the set of natural neighborhood points of each grid node is dynamically adjusted according to the density of the transformed point cloud around that node. For example, grid nodes in sparse point cloud regions automatically expand their neighborhood to include more transformed point cloud as references, while grid nodes in dense point cloud regions automatically shrink their neighborhood to reduce redundant reference points. Ultimately, this ensures that the interpolation calculation for each grid node can adapt to the point cloud distribution characteristics of its location. Compared to the stepped surface of traditional linear interpolation, local weighted interpolation, through dynamic neighborhood and weighted averaging, makes the gradient of continuous surfaces smoother and avoids stepped jagged edges.

[0060] S5. Join the first continuous surface and the third continuous surface to form the original surface, and join the second continuous surface and the fourth continuous surface to form the deformed surface.

[0061] Specifically, the maximum value of the first continuous surface on the horizontal coordinate axis is determined; the minimum value of the third continuous surface on the horizontal coordinate axis is determined; the horizontal displacement is determined based on the maximum value of the first continuous surface on the horizontal coordinate axis and the minimum value of the third continuous surface on the horizontal coordinate axis; based on the horizontal displacement, the first continuous surface and the third continuous surface are spliced ​​together to form the original surface, and the second continuous surface and the fourth continuous surface are spliced ​​together to form the deformed surface.

[0062] The maximum value of the first continuous surface on the horizontal coordinate axis refers to the maximum value of the X-coordinate of all grid nodes on the first continuous surface, which is the right boundary of the left surface. The minimum value of the third continuous surface on the horizontal coordinate axis refers to the minimum value of the X-coordinate of all grid nodes on the third continuous surface, which is the left boundary of the right surface. The horizontal displacement is determined based on the maximum and minimum values ​​of the first and third continuous surfaces on the horizontal coordinate axis. The horizontal displacement is then added to the X-coordinate of all grid nodes on the third continuous surface to ensure a seamless connection between the left boundary of the right surface and the right boundary of the left surface. This allows the translated third continuous surface to merge with the first continuous surface in the same XY coordinate system, forming the original surface covering the entire field.

[0063] Since the X-axis boundaries of the left and right surfaces change only in depth along the Z-axis before and after deformation, while the X-axis range remains unchanged, the horizontal displacement after deformation is the same as the horizontal displacement in the above embodiment. Based on the horizontal displacement calculated above, the horizontal displacement is added to the X-coordinates of all mesh nodes on the fourth continuous surface to align it with the right boundary of the second continuous surface, and then merged to form the deformed surface.

[0064] S6. Determine the micro-deformation of the structure under test based on the difference between the original surface and the deformed surface.

[0065] Because the splicing logic of the original and deformed surfaces is consistent, the XY axis ranges, the number of mesh nodes, and their positions on the original and deformed surfaces correspond perfectly. For example, if the coordinates of a mesh node on the original surface are (X=100mm, Y=200mm), there must be a mesh node on the deformed surface with the exact same coordinates, providing a basis for comparing the same location for depth difference calculation. For all mesh nodes in the entire field, the Z-axis depth of the deformed surface and the Z-axis depth of the original surface are calculated one by one to obtain the local micro-deformation of each node. The micro-deformation of all nodes is then integrated to form a full-field micro-deformation dataset.

[0066] Building upon the above embodiments, the full-field micro-deformation dataset can also be displayed through an engineering-grade visualization and interactive analysis system. For example, integrating industrial software interaction logic supports one-click deformation analysis, replacing the manual data export process. Micro-deformation quantities are converted into color deformation cloud maps, with different colors indicating the degree of deformation, and contour lines are overlaid to visually display the deformation gradient, meeting the needs of engineering reports for result presentation.

[0067] This embodiment acquires original point cloud data before and after deformation using a binocular camera. The original point cloud data is rotated 180° around the horizontal coordinate axis to obtain transformed point cloud data. Based on the camera parameters of the left camera, a first effective measurement region on the left side is determined from the transformed point cloud data. Based on the camera parameters of the right camera, a second effective measurement region on the right side is determined from the transformed point cloud data. The first effective measurement region is meshed and interpolated to obtain a first continuous surface before deformation and a second continuous surface after deformation. The second effective measurement region is meshed and interpolated to obtain a third continuous surface before deformation and a fourth continuous surface after deformation. The surfaces before and after deformation are then stitched together to form the original surface and the deformed surface, determining the micro-deformation. Rotating the original point cloud data 180° around the horizontal coordinate axis solves the problem of coordinate system inconsistency caused by binocular camera installation. Automatic coordinate rotation replaces traditional manual calibration matrix transformation, eliminating the need for manual adjustment of the coordinate reference and significantly improving computational efficiency. In addition, determining the effective measurement area based on camera parameters can eliminate interference points such as edge anomalies, reduce the interference of invalid data on deformation calculation, ensure that micro-deformation results only reflect the true deformation of the structure, and improve the accuracy of micro-deformation measurement.

[0068] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the technical solutions of the embodiments of the present invention.

Claims

1. A method for measuring the micro-deformation of a physical model test structure, characterized in that, include: S1. Obtain the original point cloud data of the structure under test before and after deformation based on a binocular camera. The structure under test is divided into a left region and a right region. The left camera of the binocular camera captures the left region, and the right camera of the binocular camera captures the right region. S2. Rotate the original point cloud data before and after deformation by 180° around the horizontal coordinate axis to obtain the transformed point cloud data before and after deformation. S3. Based on the camera parameters of the left camera, determine the first effective measurement area of ​​the left region in the converted point cloud data; based on the camera parameters of the right camera, determine the second effective measurement area of ​​the right region in the converted point cloud data. S4. Perform meshing and interpolation processing on the transformed point cloud data before and after deformation in the first effective measurement area to obtain the first continuous surface before deformation and the second continuous surface after deformation; perform meshing and interpolation processing on the transformed point cloud data before and after deformation in the second effective measurement area to obtain the third continuous surface before deformation and the fourth continuous surface after deformation. S5. Join the first continuous surface and the third continuous surface to form the original surface, and join the second continuous surface and the fourth continuous surface to form the deformed surface; specifically including: Determine the maximum value of the first continuous surface on the horizontal coordinate axis; Determine the minimum value of the third continuous surface on the horizontal coordinate axis; The horizontal displacement is determined based on the maximum value of the first continuous surface on the horizontal axis and the minimum value of the third continuous surface on the horizontal axis. Based on the horizontal displacement, the first continuous surface and the third continuous surface are spliced ​​together to form the original surface, and the second continuous surface and the fourth continuous surface are spliced ​​together to form the deformed surface; S6. Determine the micro-deformation of the structure under test based on the difference between the original surface and the deformed surface.

2. The method for measuring micro-deformation of a physical model test structure according to claim 1, characterized in that, In step S3, based on the camera parameters of the left-side camera, the first effective measurement area of ​​the left-side region is determined from the converted point cloud data, including: Based on the camera parameters of the left-side camera, determine the planar field of view, target depth, and tolerance value of the left-side region; Based on the planar field of view, target depth, and tolerance value of the left-side region, a first effective measurement area is determined in the converted point cloud data.

3. The method for measuring micro-deformation of a physical model test structure according to claim 2, characterized in that, The left camera parameters include the field of view angle of the left camera and the vertical distance from the left camera to the left side of the structure under test. Based on the left camera parameters, the planar field of view, target depth, and tolerance value of the left side are determined, including: The horizontal field of view of the left region is calculated based on the field of view angle parameters of the left camera and the vertical distance from the left camera to the left region of the structure under test. The vertical field of view height of the left region is calculated based on the field of view angle parameters of the left camera and the vertical distance from the left camera to the left region of the structure under test. The planar field of view range of the left-side region is determined based on the horizontal field of view width and the vertical field of view height of the left-side region. The target depth and tolerance value of the left region are determined based on the vertical distance from the left camera to the left region of the structure under test.

4. The method for measuring micro-deformation of a physical model test structure according to claim 1, characterized in that, In step S3, based on the camera parameters of the right-side camera, a second effective measurement region of the right-side region is determined from the converted point cloud data, including: Based on the camera parameters of the right-side camera, determine the planar field of view, target depth, and tolerance value of the right-side region; Based on the planar field of view, target depth, and tolerance value of the right-hand region, a second effective measurement area is determined in the converted point cloud data.

5. The method for measuring micro-deformation of a physical model test structure according to claim 4, characterized in that, The parameters of the right-side camera include the field of view angle of the right-side camera and the vertical distance from the right-side camera to the right region of the structure under test. Based on the camera parameters of the right-side camera, the planar field of view, target depth, and tolerance value of the right-side region are determined, including: The horizontal field of view of the right region is calculated based on the field of view angle parameters of the right camera and the vertical distance from the right camera to the right region of the structure under test. The vertical field of view height of the right area is calculated based on the field of view angle parameters of the right camera and the vertical distance from the right camera to the right area of ​​the structure under test. The planar field of view range of the right-side region is determined based on the horizontal field of view width and the vertical field of view height of the right-side region. The target depth and tolerance value of the right-side region are determined based on the vertical distance from the right-side camera to the right-side region of the structure under test.

6. The method for measuring micro-deformation of a physical model test structure according to claim 2, characterized in that, In step S4, the transformed point cloud data before and after deformation in the first effective measurement area are respectively subjected to meshing and interpolation processing to obtain the first continuous surface before deformation and the second continuous surface after deformation, including: Based on the planar field of view of the left region, the transformed point cloud data before and after deformation in the first effective measurement region are divided into planar grids to obtain the first grid model before deformation and the second grid model after deformation. A first continuous surface is obtained by performing local weighted interpolation on the mesh nodes in the first mesh model, and a second continuous surface is obtained by performing local weighted interpolation on the mesh nodes in the second mesh model.

7. The method for measuring micro-deformation of a physical model test structure according to claim 6, characterized in that, A first continuous surface is obtained by performing local weighted interpolation on the mesh nodes in the first mesh model, and a second continuous surface is obtained by performing local weighted interpolation on the mesh nodes in the second mesh model, including: Using the Thiessen polygon method, Thiessen polygons are generated at each transformed point cloud data point in the first and second mesh models, with each transformed point cloud data point corresponding to one Thiessen polygon. The target Thiessen polygon where the grid node is located is determined, and the transformed point cloud data in the target Thiessen polygon is used as the natural neighbor points of the grid node. The weight of the natural neighbor point is determined based on the area of ​​the target Thiessen polygon where the natural neighbor point is located. The depth data of the corresponding grid nodes are calculated based on the weights of the natural neighboring points to determine the first and second continuous surfaces.

8. The method for measuring micro-deformation of a physical model test structure according to claim 4, characterized in that, In step S4, the transformed point cloud data before and after deformation in the second effective measurement area are respectively subjected to meshing and interpolation processing to obtain the third continuous surface before deformation and the fourth continuous surface after deformation, including: Based on the planar field of view of the right region, the transformed point cloud data before and after deformation in the second effective measurement region are divided into planar grids to obtain the third grid model before deformation and the fourth grid model after deformation. The third continuous surface is obtained by performing local weighted interpolation on the grid nodes in the third grid model, and the fourth continuous surface is obtained by performing local weighted interpolation on the grid nodes in the fourth grid model.

9. The method for measuring micro-deformation of a physical model test structure according to claim 8, characterized in that, To obtain a third continuous surface, local weighted interpolation is performed on the mesh nodes in the third mesh model; to obtain a fourth continuous surface, local weighted interpolation is performed on the mesh nodes in the fourth mesh model, including: Using the Thiessen polygon method, Thiessen polygons are generated at each transformed point cloud data point in the third and fourth grid models, with each transformed point cloud data point corresponding to one Thiessen polygon. The target Thiessen polygon where the grid node is located is determined, and the transformed point cloud data in the target Thiessen polygon is used as the natural neighbor points of the grid node. The weight of the natural neighbor point is determined based on the area of ​​the target Thiessen polygon where the natural neighbor point is located. The depth data of the corresponding grid nodes are calculated based on the weights of the natural neighboring points, and the third and fourth continuous surfaces are determined.

Images