Flexible plate molded surface dynamic measurement method and system based on three-dimensional vision method

Abstract

The invention belongs to the technical field of wind tunnel tests, and discloses a flexible plate profile dynamic measurement method and system based on a three-dimensional vision method. The annular coding points with the unique identification codes are arranged on the outer wall of the flexible plate, unambiguous matching is achieved by means of the global unique identification codes of the coding points, the camera self-calibration method based on polar geometric constraint and light beam adjustment optimization is combined, and the calibration accuracy can be greatly improved under the condition that a checkerboard calibration plate does not need to be placed again. And external parameter changes caused by environmental disturbance (such as vibration) are estimated and corrected in real time. The method is suitable for a large-amplitude high-frequency vibration environment (through a real-time dynamic correction mode, calibration parameters are updated in each frame) and is also suitable for a low-frequency low-amplitude vibration environment (through an intermittent correction mode, calibration is triggered according to a preset interval), the bottleneck problem of precision attenuation of traditional static calibration in a dynamic environment is effectively solved, and the calibration accuracy is improved. And the long-term maintenance-free operation of the measurement system is realized.

Description

Technical Field

[0001] This invention relates to the field of wind tunnel testing technology, and in particular to a method and system for dynamic measurement of flexible plate profiles based on three-dimensional vision. Background Technology

[0002] Wind tunnels simulate real flight environments by generating controllable airflow, providing crucial data support for aircraft aerodynamic design. Flexible-walled nozzles, by varying their profiles, can create flow fields at different Mach numbers within a single wind tunnel. Developing efficient and accurate dynamic measurement technology for the entire profile of flexible-walled nozzles is of great significance in ensuring flow field quality.

[0003] A flexible-wall nozzle is a long rectangular plate-shell structure whose profile adjustment function is achieved through the elastic deformation of a flexible wall panel (hereinafter referred to as the flexible panel), providing aerodynamic profiles at different Mach numbers without replacing the nozzle. The inner wall is a smooth air passage, and the outer wall is hinged to an actuator rod. The profile of the nozzle's inner wall is mainly determined by measuring the deformation of the outer wall, combined with mechanical principles for indirect measurement. For the deformation measurement of the outer wall, contact measurement methods are currently mainly used, such as resistance strain gauges, fiber optic grating sensors, or displacement sensors. Although high-precision measurements can be achieved, the sensors must be tightly attached to the structural surface. To obtain high-precision deformation information across the entire range, multiple sensors need to be attached to the outer wall, inevitably increasing the number of signal acquisition channels and the difficulty of real-time data processing. Since the flexible-wall structure is a moving part, sensors are prone to detachment during long-term monitoring, leading to measurement failure; moreover, strain sensors output strain, which requires integration to obtain displacement, a process that introduces significant additional errors. The deployment of a large number of displacement sensors also significantly increases the complexity of the internal structure of the flexible-wall nozzle, posing challenges to subsequent equipment maintenance.

[0004] Using non-contact vision measurement technology to measure the shape and deformation of the nozzle outer wall has significant advantages: the vision measurement system has full-field three-dimensional displacement measurement, large range, high precision, and real-time capability, making it particularly suitable for dynamic monitoring of complex structures such as flexible plates; the vision measurement camera can be mounted on a fixed bracket inside the wind tunnel, eliminating the need for any active sensors on the flexible plate, avoiding measurement instability caused by connecting sensor circuits to moving parts, and greatly improving the maintainability of the measurement system. However, stereo vision measurement faces serious challenges in practical engineering applications: First, camera calibration is a key technology for achieving high-precision outer wall shape measurement, especially in vibration environments where camera calibration parameters change. Developing a stable and reliable self-calibration method is crucial for achieving long-term maintenance-free vision measurement; second, considering the actual length of the nozzle (meter-level), achieving full-domain deformation measurement requires multiple stereo vision measurement systems to work together. Coordinate transformation and synchronous measurement between the cameras are also issues that must be addressed in achieving dynamic shape measurement of the nozzle outer wall. Traditional feature point-based stitching methods are prone to mismatches, leading to cumulative errors.

[0005] In view of this, the present invention is proposed. Summary of the Invention

[0006] The present invention aims to solve at least one of the above-mentioned technical problems, and provides a method and system for dynamic measurement of flexible plate surface based on three-dimensional vision method.

[0007] To achieve the above objectives, the first aspect of the present invention provides: A dynamic measurement method for flexible plate profiles based on 3D vision includes the following steps: Multiple ring-shaped coding points are arranged on the outer wall surface of the flexible plate, and each ring-shaped coding point has a unique identification code; Images of the flexible plate outer wall containing the ring-shaped coding points were acquired from different perspectives. Using the unique identifier of the ring-shaped coding point, the camera that acquires images is self-calibrated to determine the camera's internal and external parameters; The three-dimensional coordinates of the ring-shaped coding point are calculated using the internal and external parameters. Based on the aforementioned three-dimensional coordinates, coordinate systems from different perspectives are unified into the same coordinate system to obtain three-dimensional point cloud data of the entire outer wall of the flexible plate. The three-dimensional point cloud data is fitted to obtain the global surface data of the flexible plate outer wall.

[0008] Preferably, the annular coding point includes a central positioning circle and an outer coding ring. The coding ring is divided into two or more sector-shaped regions along the circumference, and the unique identification code is formed by the concave-convex shape or color contrast of the regions.

[0009] Preferably, the coding ring is divided into 15 sector-shaped regions along the circumferential direction.

[0010] Preferably, the self-calibration of the camera acquiring images using the unique identifier of the ring-shaped coding point includes: Extract the pixel coordinates and corresponding unique identifiers of the circular coding points in each image; Based on the unique identifier, a matching relationship of coding points between different images is established to eliminate matching ambiguity; Using the matching relationship and polar geometry constraints, the internal and external parameters of the camera are calculated and optimized.

[0011] Preferably, unifying coordinate systems from different perspectives into the same coordinate system includes: Using the common ring-shaped coding points in the overlapping areas of adjacent viewpoints, an unambiguous correspondence of feature points is established based on the unique identifier; The transformation matrix between adjacent coordinate systems is calculated using the iterative nearest point algorithm. By minimizing the reprojection error of all point clouds using a global optimization algorithm, the transformation matrix from each viewpoint coordinate system to a unified coordinate system is determined.

[0012] Preferably, the surface fitting of the three-dimensional point cloud data includes: Discrete three-dimensional coordinates are used as nodes, and a spatial interpolation method is used to reconstruct the continuous flexible plate outer wall surface. The spatial interpolation method includes at least one of finite element shape function interpolation, spline interpolation, or radial basis function interpolation.

[0013] Preferably, after obtaining the global profile data of the flexible plate's outer wall, the method further includes: By repeatedly performing the acquisition, self-calibration, calculation, unification, and fitting steps at different times, time-series global surface data can be obtained. By comparing the global surface data at different times, dynamic deformation information of the flexible plate outer wall is obtained.

[0014] Preferably, the step of acquiring images of the flexible plate outer wall containing the annular coding points from different perspectives includes: achieving synchronous acquisition of multi-view images through an external synchronization trigger signal, with synchronization accuracy reaching sub-millisecond level.

[0015] Preferably, the ring-shaped coding points are prepared on the outer surface of the flexible board by laser engraving.

[0016] The second aspect of the present invention provides: A dynamic measurement system for flexible plate surfaces based on 3D vision methods includes: The coding unit is used to arrange multiple annular coding points on the outer wall surface of the flexible board, and each annular coding point has a unique identification code. The acquisition unit is used to acquire images of the flexible plate outer wall containing the annular coding points from different perspectives; The self-calibration unit is used to perform self-calibration on the camera that acquires images using the unique identifier of the ring coding point, and to determine the camera's internal and external parameters. A calculation unit is used to calculate the three-dimensional coordinates of the ring-shaped coding point using the internal and external parameters; A unified unit is used to unify the coordinate systems of different perspectives into the same coordinate system based on the three-dimensional coordinates, so as to obtain three-dimensional point cloud data of the entire outer wall of the flexible plate. The fitting unit is used to perform surface fitting on the three-dimensional point cloud data to obtain the global surface data of the flexible plate outer wall.

[0017] Compared with the prior art, the present invention has the following beneficial effects: This invention utilizes a ring of uniquely coded points arranged on the outer wall of a flexible plate to achieve unambiguous matching. Combined with a camera self-calibration method based on polar geometry constraints and bundle adjustment optimization, it can estimate and correct for changes in external parameters caused by environmental disturbances (such as vibration) in real time without needing to reposition the checkerboard calibration plate. This method is applicable to both high-amplitude, high-vibration environments (updating calibration parameters every frame through a real-time dynamic correction mode) and low-frequency, low-amplitude vibration environments (triggered at predetermined intervals through an intermittent correction mode). It effectively overcomes the bottleneck problem of accuracy degradation in traditional static calibration under dynamic environments, enabling long-term maintenance-free operation of the measurement system.

[0018] This invention acquires images of the flexible slab's outer wall containing the annular coded points from different perspectives. Utilizing common annular coded points in the overlapping fields of view between adjacent groups, it establishes unambiguous feature point correspondences based on unique identifiers. Combining iterative nearest-point algorithms and global optimization algorithms, it unifies multiple coordinate systems into a single coordinate system. This method fundamentally eliminates the risk of mismatches in traditional feature point matching, minimizes reprojection errors through global bundle adjustment, effectively suppresses error accumulation during multi-system stitching, and achieves high-precision global surface measurement of the flexible slab's outer wall at the meter-scale. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating the dynamic measurement method for flexible plate surfaces based on 3D vision. Figure 2 A schematic diagram of 15 equally divided circular coding points; Figure 3 This is a schematic diagram of a dynamic measurement device for flexible plate profiles based on a 3D vision method. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be described in detail below with reference to specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be arbitrarily combined with each other. Those skilled in the art should understand that the following descriptions are merely exemplary means of implementing this invention and not limiting conditions; any technical means employed to achieve the same or similar technical effects as this invention should fall within the protection scope of this invention.

[0021] Figure 1 This is a flowchart illustrating a dynamic measurement method for flexible plate surfaces based on 3D vision. (Reference) Figure 1 The first embodiment of the present invention provides a dynamic measurement method for flexible plate surfaces based on three-dimensional vision methods, including the following steps: S101, Multiple ring-shaped coding points are arranged on the outer wall surface of the flexible plate, and each ring-shaped coding point has a unique identification code.

[0022] This step aims to establish a benchmark point system for full-field measurement. The key is to replace the traditional unmarked dots with ring-shaped coded points that have unique identification, so as to provide a data foundation for subsequent multi-camera stitching and self-calibration under vibration environment.

[0023] Distributed measurement reference points are established on the surface of a large flexible panel, requiring each point to have a globally unique identity to resolve the ambiguity problem of feature point matching during multi-view stitching. This invention employs a ring-shaped coding point structure consisting of a central positioning circle and an outer coding ring, ensuring no duplication of coding points across the entire field through a specific coding mechanism. Figure 2 The typical structure of the ring coding point is shown. The ring coding point includes: (1) the central positioning circle: the black solid circle in the center of the figure is used to provide a high-precision geometric center positioning reference. The principle is to use ellipse fitting and other methods to accurately extract the center pixel coordinates of the central circle to achieve sub-pixel level positioning of the coding point; (2) the outer coding ring: the ring area surrounding the central positioning circle is fixedly divided into n sector areas along the circumference (n is an integer ≥2, such as 12, 15, 16, etc., the specific value is determined according to the required coding capacity), which is the core feature of the ring coding point with global uniqueness.

[0024] Each sector has an independent coded state, distinguished by color contrast (e.g., black and white) or embossed form (e.g., grooves and protrusions created by laser engraving), corresponding to "0" and "1" in binary code. The sector regions are arranged clockwise (or counter-clockwise), and their state combinations form an n-bit binary code, which can generate 2^n... n A unique identifier. For example, for a flexible printed circuit board (FPCB) in the 2050mm × 580mm range, a 15-bit equally divided coding ring can generate 32768 (i.e., 2... 15) unique identifier (reference) Figure 2 This is sufficient to cover thousands of coding points across the entire field without repetition. Figure 2 The specific black and white distribution pattern displayed in the image (such as black sector blocks corresponding to "1" and white gaps corresponding to "0") corresponds to a specific unique identifier.

[0025] Ring-shaped coding dots can be prepared on the outer surface of flexible panels using various methods, including laser engraving, mechanical etching, screen printing, and inkjet printing followed by adhesion. Laser engraving has been widely used due to its significant advantages: laser ablation can create high-contrast, permanent black-and-white marks on metal surfaces, resulting in high-contrast, sharp edges that facilitate image recognition; laser engraving is a localized modification of the material surface without any added mass (no adhesive layer, no sticker thickness), and does not affect the aerodynamic performance or structural dynamics of the flexible panel; laser marking is a physically permanent mark that can be used for a long time without fading, making it particularly suitable for harsh engineering environments such as wind tunnels with high temperatures, high pressures, and strong airflows. Compared to organic dyes or adhesive labels, it exhibits better thermal stability under high-temperature heating conditions, ensuring long-term stability and maintainability of visual measurements.

[0026] Regarding the arrangement of the coding points, conventional strategies in this field, such as regular grid arrays or pseudo-random distribution, can be used to arrange the coding points on the outer surface of the flexible plate. The arrangement density should be determined according to the measurement accuracy requirements and camera resolution, typically ensuring that each camera's field of view contains a sufficient number of coding points (e.g., no less than 20) to ensure calibration stability. Given the advantages of laser engraving technology, such as high temperature resistance, no added mass, and good long-term stability, it is particularly suitable for harsh environments such as wind tunnels.

[0027] S102, Acquire images of the flexible plate outer wall containing the annular coding points from different perspectives.

[0028] This step aims to obtain two-dimensional projection information of the encoded points through multi-view images. The key is to achieve synchronous triggering of multiple cameras to ensure the temporal consistency of dynamic measurements.

[0029] Multi-view image acquisition can be achieved using single-camera multi-angle shooting or multi-camera stereo vision arrays to acquire images from different perspectives. For large flexible panels (e.g., meter-scale), multiple sets of stereo vision measurement systems (e.g., 5 sets with a total of 10 cameras) are typically arranged along the length of the flexible panel, with adjacent camera sets maintaining a certain overlap in the field of view (e.g., no less than 20%) to ensure the continuity of subsequent coordinate transformations. Camera selection, lens focal length, shooting distance, and other parameters can be determined according to the measurement field of view and accuracy requirements using conventional methods in this field.

[0030] To achieve dynamic measurement, image acquisition from each viewpoint should employ an externally triggered synchronization mode. Specifically, a periodic square wave signal can be generated using a signal generator, and the rising edge of the square wave can be used to trigger synchronized exposure of all cameras, achieving sub-millisecond (e.g., microsecond) synchronization accuracy. Of course, other synchronization methods known in the art can also be used, such as software triggering or GPS clock synchronization, as long as the timing consistency requirements of dynamic measurement are met.

[0031] S103, using the unique identifier of the ring coding point, the camera that acquires the image is self-calibrated to determine the camera's internal and external parameters.

[0032] This step aims to solve the problem of camera calibration parameter drift under vibration. The key is to use the globally unique identifier of the coding point to achieve unambiguous matching, so that calibration correction can be completed without repositioning the checkerboard calibration plate.

[0033] For code point identification and matching, firstly, the images from each camera are preprocessed (such as Gaussian filtering, OTSU binarization, and edge detection) to extract the coordinates of the center circle pixel of the ring code point and the encoding state of the outer ring code, and to determine the unique identifier of the code point.

[0034] Based on the unique identifier, coded point matching relationships between different images (or different cameras) can be directly established. Due to the global uniqueness of the identifier, unambiguous feature point correspondence can be achieved even under conditions of large viewpoint changes or partial occlusion, fundamentally eliminating the common mismatch problems in traditional feature point matching (such as SIFT, SURF, etc.).

[0035] For camera parameter calculation, by utilizing the established matching relationship and combining it with polar geometry constraints, a system of equations concerning the camera's extrinsic parameters (rotation matrix and translation vector) can be constructed. Specifically, conventional nonlinear optimization methods in this field, such as Singular Value Decomposition (SVD), Bundle Adjustment, or the Levenberg-Marquardt optimization algorithm, can be used to calculate and optimize the camera's intrinsic parameters (such as focal length, principal point, and distortion coefficients) and extrinsic parameters.

[0036] As an example of optimization techniques, camera parameters can be calculated and optimized using methods based on epipolar geometry and bundle adjustment: First, equations are established using epipolar geometry constraints. The spatial point P satisfies the epipolar constraint at the imaging points p and p′ of the left and right cameras: p′ T Fp=0; where F is the basic matrix, containing the internal and external parameters of the left and right cameras: F=K′ -T tˆRK -1 Where K and K′ are the intrinsic parameter matrices of the left and right cameras, respectively. -TLet K' be the transpose of the inverse matrix, t and R be the translation vector and rotation matrix of the right camera relative to the left camera, respectively, and tˆ be the antisymmetric matrix of t.

[0037] Once the camera's intrinsic parameters are determined, the common viewpoints of each spatial point in the left and right images can be used to construct an equation regarding the extrinsic parameters of the vision system. In this measurement system, the pixel coordinates of the ring-shaped coded points on the flexible wall in the left and right cameras can be precisely determined, and the identification codes of the ring-shaped coded points ensure unique pairing of each coded point in the left and right cameras. Using the above method, no fewer than 60 equations can be established. After solving for the essential matrix, the initial estimates of the rotation matrix R and translation vector t are obtained using the Singular Value Decomposition (SVD) method.

[0038] Furthermore, the camera calibration parameters can be optimized using the inverse depth bundle adjustment method. Assume the pixel coordinates of the i-th matching point pair are... And the corresponding inverse depth parameter is Then the nonlinear least squares model in the bundle adjustment method is: ; Here, argmin indicates that the algorithm automatically finds the most accurate camera calibration parameters through iterative optimization, L represents the cost function, P represents the projection transformation, and D and D′ are the distortion coefficients of the left and right cameras, respectively. The above equation defines a nonlinear optimization problem with 6+m optimization variables for m pairs of matching points, which can be stably solved using the Levenberg-Marquardt method.

[0039] It should be noted that the above mathematical model is only a preferred embodiment of this patent. Those skilled in the art can achieve the same technical effect by using other equivalent optimization algorithms (such as the Gauss-Newton method, the conjugate gradient method, etc.) or equivalent mathematical expressions, depending on the actual accuracy requirements and computing resources.

[0040] Given the fixed camera mounting method, changes in external parameters caused by external vibrations and other environmental factors will only occur around the initial external parameters. Therefore, the self-calibrated external parameters are strictly limited to a certain range of variation. This complete matching-initial estimation-robust optimization process can dynamically track and correct external parameter drift, effectively overcoming the bottleneck problem of accuracy degradation in traditional static calibration under dynamic environments.

[0041] Based on the actual vibration environment characteristics, the self-calibration can be configured into two operating modes: Real-time dynamic correction mode: Suitable for environments with large-amplitude, high-frequency vibrations (such as wind tunnel aerodynamic excitation). In this mode, the calibration parameter update frequency is synchronized with the camera frame rate (e.g., multiple times per frame or per second), continuously using the encoded point data of the current frame to correct for instantaneous changes in external parameters caused by vibration, ensuring dynamic measurement accuracy.

[0042] Intermittent Correction Mode: Suitable for low-frequency, low-amplitude vibration environments of 20-200Hz isolated by vibration damping devices. In this mode, vibration damping devices (such as rubber vibration isolation pads, wire rope vibration isolators, etc.) first suppress large-amplitude vibrations, and then self-calibrate at a lower frequency (such as every predetermined time interval s seconds, or when a displacement change is detected to exceed a set threshold) to correct long-term slow drift, significantly reducing computational resource consumption.

[0043] S104, using the internal and external parameters, calculate the three-dimensional coordinates of the ring-shaped coding point.

[0044] This step aims to recover the three-dimensional spatial coordinates of the encoded point from the two-dimensional image coordinates using the principle of stereo vision triangulation.

[0045] Based on the calibrated camera parameters and the pixel coordinates of the coded points in the image, the three-dimensional spatial position of the coded points is reconstructed through triangulation. Specifically, after self-calibration, the same coded point is successfully matched on the left and right cameras using triangulation (the match is confirmed using a unique identifier). By utilizing the relative pose of the cameras and combining epipolar geometric constraints to solve for the scaling factor, the precise position of the center of the coded point in the camera coordinate system is calculated, thus achieving high-precision three-dimensional reconstruction of the coded point.

[0046] The three-dimensional coordinates can also be calculated using direct linear transformation (DLT) or other conventional stereo vision reconstruction methods in the field.

[0047] S105, Based on the three-dimensional coordinates, the coordinate systems of different perspectives are unified into the same coordinate system to obtain the three-dimensional point cloud data of the entire outer wall of the flexible plate.

[0048] This step aims to solve the problem of inconsistent coordinate systems in multi-camera systems and achieve the stitching of global point clouds. The key lies in using the unambiguity of the coded point identifier to achieve accurate registration.

[0049] Since each group of stereo vision measurement systems uses a different coordinate system, it is necessary to transform all the coordinate systems into the same coordinate system. By utilizing the common ring-shaped coded points in the overlapping areas between adjacent groups, unambiguous and accurate feature point matching is established based on the unique identifier, fundamentally eliminating mismatches. At the same time, its circular contour and positioning ring features also provide sub-pixel-level high-precision 3D point cloud data, ensuring the accuracy of the paired data in the registration algorithm.

[0050] For adjacent viewpoints (or adjacent camera groups) with overlapping fields of view, the common ring-coded points in the overlapping region (identifying the same point based on a unique identifier) ​​can be used to calculate the rotation matrix and translation vector (i.e., transformation matrix) between adjacent coordinate systems using conventional point cloud registration algorithms in this field, such as the Iterative Closest Point (ICP) algorithm or Singular Value Decomposition (SVD).

[0051] As an exemplary registration method, the Iterative Closest Point (ICP) algorithm can be used to calculate the transformation matrix between adjacent coordinate systems. Assuming P1 and P2 are two point clouds to be registered, the transformation matrix between adjacent stereo vision coordinate systems can be obtained by minimizing the following objective function: ; Where R is the rotation matrix, t is the translation vector, and p 1a and p 2a These are the corresponding points established in P1 and P2, respectively, and N is the number of corresponding points.

[0052] After the co-view point cloud registration of each adjacent stereo vision system is completed, cumulative errors may exist between multiple point clouds, affecting the accuracy of the final stitching. As an exemplary optimization method, a global optimization algorithm can be used for error correction. By minimizing the reprojection error between all point clouds, the relative pose of each point cloud is optimized. Assume P c For the c-th point cloud, T c Let c be the pose transformation matrix of the c-th point cloud, and minimize the objective function of global bundle adjustment, denoted as: ; where p cd Let d be the feature point in the c-th point cloud. These are the feature points after pose transformation.

[0053] Using the aforementioned ICP registration algorithm and optimization, transformation matrices between five sets of stereo vision coordinate systems were determined, transforming the measured point cloud of the entire flexographic surface into the same coordinate system, thus realizing the output of the flexographic surface's global deformation. To match the actual flexographic surface's coordinate system, coordinate transformation can be performed to convert the overall measured coordinates into a coordinate system with the fixed end of the flexographic surface as the origin and the length direction as the X-axis.

[0054] In addition to ICP based on coded point identifiers, conventional point cloud registration methods in this field, such as conventional ICP algorithms and Procrustes analysis, can also be used. However, conventional methods are easily affected by mismatched point pairs, while methods based on coded point identifiers can fundamentally eliminate mismatches.

[0055] S106, perform surface fitting on the three-dimensional point cloud data to obtain the global surface data of the flexible plate outer wall.

[0056] This step aims to reconstruct the discrete three-dimensional coordinates into a continuous surface, which is a post-processing step in measurement and can be achieved using conventional interpolation methods in this field.

[0057] As an exemplary reconstruction method, a spatial interpolation method based on finite element shape functions can be used to reliably reconstruct continuous deformation information from a finite number of measurement points. For an arbitrary quadrilateral element with known four vertices, the displacement of any point inside it can be expressed as: ; ; ; Where u, v, and w are the displacements of any point along the X, Y, and Z directions, respectively, and α1 to α... 12 All are coefficients in the expression.

[0058] Twelve equations can be established using the four vertices, allowing us to solve for the twelve coefficients of the shape function: ; Among them, u n v n w n (n=1,2,3,4) are the displacement components of the nth vertex along the X, Y, and Z directions, respectively; x1 to x4 and y1 to y4 are the x and y coordinates of the four vertices.

[0059] In this way, the three-dimensional displacement of any point inside the quadrilateral can be obtained, thus realizing the calculation of the global displacement field of the flexible plate. This method ensures the continuity and smoothness of the displacement field of the flexible plate, can capture local deformation features to achieve high-precision reconstruction, and has high computational stability.

[0060] Continuous surfaces can also be reconstructed using conventional spatial interpolation methods in this field, such as bicubic spline interpolation, radial basis function (RBF) interpolation, and Kriging interpolation.

[0061] Those skilled in the art can choose an appropriate interpolation method based on the requirements for surface smoothness and computational efficiency. For example, interpolation methods based on finite element shape functions can be naturally integrated with subsequent finite element analyses (such as flexible plate deformation analysis), while spline interpolation methods have better overall smoothness.

[0062] After obtaining the global profile data of the flexible plate's outer wall, in order to achieve dynamic measurement, the above steps S102 to S106 can be repeated at different times to obtain a time-series global profile data sequence. By comparing the global profile data at different times (such as the difference with the initial time data, or the difference between adjacent time data), the dynamic deformation information of the flexible plate's outer wall (such as displacement field, velocity field, etc.) can be obtained.

[0063] Figure 3 This is a schematic diagram of a dynamic measurement system for flexible plate surfaces based on a 3D vision method. (Reference) Figure 3The second embodiment of the present invention provides a dynamic measurement system 300 for flexible plate surface based on a three-dimensional vision method, including: an encoding unit 301, an acquisition unit 302, a self-calibration unit 303, a calculation unit 304, a unification unit 305, and a fitting unit 306. The specific functions of each unit are described in detail below.

[0064] The coding unit 301 is used to arrange multiple annular coding points on the outer wall surface of the flexible plate, wherein each annular coding point has a unique identifier. Acquisition unit 302 is used to acquire images of the flexible plate outer wall containing the annular coding points from different perspectives; Self-calibration unit 303 is used to perform self-calibration on the camera that acquires images using the unique identifier of the ring coding point, and to determine the camera's internal and external parameters. The calculation unit 304 is used to calculate the three-dimensional coordinates of the ring-shaped coding point using the internal parameters and external parameters; The unified unit 305 is used to unify the coordinate systems of different perspectives into the same coordinate system based on the three-dimensional coordinates to obtain three-dimensional point cloud data of the entire outer wall of the flexible plate. Fitting unit 306 is used to perform surface fitting on the three-dimensional point cloud data to obtain the global surface data of the flexible plate outer wall.

[0065] To further illustrate the specific application of the present invention, an exemplary embodiment is given below. This embodiment should not be construed as limiting the scope of protection of the present invention.

[0066] Example The flexible plate is made of special structural steel with a yield strength greater than 1000MPa. Its overall dimensions are 2050 mm × 580 mm × 6 mm. One end is screwed to a rigid reinforcing support, and five reinforcing ribs are machined at equal intervals along the axial direction on the outer wall surface. A 15-bit annular coded dot array (divided into 15 equal fan-shaped areas, generating 32,768 unique identifiers) is laser-engraved on its outer wall surface, with the coded dots distributed in a regular grid pattern.

[0067] Five sets of stereo vision measurement systems (a total of 10 high-resolution industrial cameras, 2448×2048 pixels, 60fps) are arranged along the length of the flexible board. Each set consists of two cameras forming a stereo measurement unit, with a 20% overlap of the field of view between adjacent sets. Multi-camera synchronous exposure is controlled by an external synchronization trigger signal, enabling synchronous acquisition of multi-view images with sub-millisecond synchronization accuracy.

[0068] 1. Accuracy assessment experiment The length scale was measured, and the distance between the centers of the two coding points was 141.99 mm. The scale was placed in the fields of view of five sets of binocular stereo vision systems (A, B, C, D, and E). Each set of vision measurement systems performed coding point identification, 3D reconstruction, and calculation of the distance between the two coding points. The results are shown in Table 1.

[0069] Table 1 Measurement results of scale coding point length .

[0070] Experimental results show that the average measurement error of the five sets of cameras is less than 0.1 mm, which verifies the absolute measurement accuracy of the system.

[0071] 2. Repeatability and accuracy test of outer wall profile In the flexible plate actuation test platform, five push rods (i.e., actuating rods) were pushed at 20%, 60%, and 100% strokes, with stroke parameters shown in Table 2. The flexible plate underwent elastic deformation under the drive of the push rods. Three full-area surface measurements of the outer wall were performed on the flexible plate under three different loading states, and the root mean square error of the three measurement results was compared, with the results shown in Table 3.

[0072] Table 2 Push rod stroke parameters

[0073] Table 3 Repeatability accuracy results .

[0074] Experimental results show that the system can maintain extremely high repeatability accuracy under three different strokes, with root mean square errors of less than 0.05 mm (maximum 0.048 mm in the Z direction), verifying the stability of the system under vibration-free conditions.

[0075] 3. Verification of self-calibration effectiveness under vibration environment A vibration system was used to excite the flexible plate actuation test platform at 200 Hz for approximately 1 minute. A self-calibration method (real-time dynamic correction mode) was employed to correct and read the new camera parameters for calibration. The surface measurement experiment was repeated, and the full-area surface profile of the flexible plate's outer wall under three different pusher strokes was measured three times. The results are shown in Table 4.

[0076] Table 4. Repeatability accuracy results after introducing vibration .

[0077] Experimental results show that after the flexible plate actuation test platform is subjected to external excitations such as 200Hz vibration, the camera self-calibration technology can effectively ensure the effectiveness of external parameter correction and still guarantee extremely high repeatability measurement accuracy (maximum 0.041mm in the Z direction), verifying the long-term maintenance-free measurement capability of the present invention under large-amplitude high-frequency vibration environment.

[0078] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application 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 of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A dynamic measurement method for flexible plate profiles based on three-dimensional vision, characterized in that, Includes the following steps: Multiple ring-shaped coding points are arranged on the outer wall surface of the flexible plate, and each ring-shaped coding point has a unique identification code; Images of the flexible plate outer wall containing the ring-shaped coding points were acquired from different perspectives. Using the unique identifier of the ring-shaped coding point, the camera that acquires images is self-calibrated to determine the camera's internal and external parameters; The three-dimensional coordinates of the ring-shaped coding point are calculated using the internal and external parameters. Based on the aforementioned three-dimensional coordinates, coordinate systems from different perspectives are unified into the same coordinate system to obtain three-dimensional point cloud data of the entire outer wall of the flexible plate. The three-dimensional point cloud data is fitted to obtain the global surface data of the flexible plate outer wall.

2. The method for dynamic measurement of flexible plate surface based on three-dimensional vision as described in claim 1, characterized in that, The annular coding point includes a central positioning circle and an outer coding ring. The coding ring is divided into two or more sector-shaped regions along the circumference. The unique identification code is formed by the concave and convex shape or color contrast of the regions.

3. The method for dynamic measurement of flexible plate surface based on three-dimensional vision as described in claim 2, characterized in that, The coding ring is divided into 15 sector-shaped regions along the circumference.

4. The method for dynamic measurement of flexible plate surface based on three-dimensional vision as described in claim 1, characterized in that, Using the unique identifier of the ring-shaped coding point, the camera that acquires images performs self-calibration, including: Extract the pixel coordinates and corresponding unique identifiers of the circular coding points in each image; Based on the unique identifier, a matching relationship of coding points between different images is established to eliminate matching ambiguity; Using the matching relationship and polar geometry constraints, the internal and external parameters of the camera are calculated and optimized.

5. The method for dynamic measurement of flexible plate surface based on three-dimensional vision as described in claim 1, characterized in that, Unifying coordinate systems from different perspectives into the same coordinate system includes: Using the common ring-shaped coding points in the overlapping areas of adjacent viewpoints, an unambiguous correspondence of feature points is established based on the unique identifier; The transformation matrix between adjacent coordinate systems is calculated using the iterative nearest point algorithm. By minimizing the reprojection error of all point clouds using a global optimization algorithm, the transformation matrix from each viewpoint coordinate system to a unified coordinate system is determined.

6. The method for dynamic measurement of flexible plate surface based on three-dimensional vision as described in claim 1, characterized in that, The surface fitting of the three-dimensional point cloud data includes: Discrete three-dimensional coordinates are used as nodes, and a spatial interpolation method is used to reconstruct the continuous flexible plate outer wall surface. The spatial interpolation method includes at least one of finite element shape function interpolation, spline interpolation, or radial basis function interpolation.

7. The method for dynamic measurement of flexible plate surface based on three-dimensional vision as described in claim 1, characterized in that, After obtaining the global profile data of the flexible plate's outer wall, the following is also included: By repeatedly performing the acquisition, self-calibration, calculation, unification, and fitting steps at different times, time-series global surface data can be obtained. By comparing the global surface data at different times, dynamic deformation information of the flexible plate outer wall is obtained.

8. The method for dynamic measurement of flexible plate surface based on three-dimensional vision as described in claim 1, characterized in that, The step of acquiring images of the flexible plate outer wall containing the annular coding points from different perspectives includes: achieving synchronous acquisition of multi-view images through an external synchronization trigger signal, with synchronization accuracy reaching the sub-millisecond level.

9. The method for dynamic measurement of flexible plate surface based on three-dimensional vision as described in claim 1, characterized in that, The ring-shaped coding points are prepared on the outer surface of the flexible board by laser engraving.

10. A dynamic measurement system for flexible plate surfaces based on three-dimensional vision methods, characterized in that, include: The coding unit is used to arrange multiple annular coding points on the outer wall surface of the flexible board, and each annular coding point has a unique identification code. The acquisition unit is used to acquire images of the flexible plate outer wall containing the annular coding points from different perspectives; The self-calibration unit is used to perform self-calibration on the camera that acquires images using the unique identifier of the ring coding point, and to determine the camera's internal and external parameters. A calculation unit is used to calculate the three-dimensional coordinates of the ring-shaped coding point using the internal and external parameters; A unified unit is used to unify the coordinate systems of different perspectives into the same coordinate system based on the three-dimensional coordinates, so as to obtain three-dimensional point cloud data of the entire outer wall of the flexible plate. The fitting unit is used to perform surface fitting on the three-dimensional point cloud data to obtain the global surface data of the flexible plate outer wall.

Images