A method and system for full-field strain measurement during the tensioning process of precast piles based on digital image correlation.
By using a mobile robot and vision sensors to capture images during the tensioning process of precast piles, decomposing pixel displacements, filtering and smoothing them, generating reference speckle images, and allocating adaptive search windows, the problem of unstable image matching in the production line environment was solved, and stable measurement of the strain of precast piles across the entire field was achieved.
Patent Information
- Application Number
- CN202611311691.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-27
- Publication Date
- 2026-11-17
- Estimated Expiration
- 2046-08-27
AI Technical Summary
In existing technologies, the image matching in precast pile tensioning operations is unstable in a production line environment, resulting in unreliable strain results across the entire field and making it difficult to achieve stable measurements.
By controlling a mobile robot to follow a precast pile, the system captures initial and deformed speckle images using a vision sensor, decomposes pixel displacement vectors, performs filtering and smoothing, generates motion-compensated images, establishes deformation gradients, allocates adaptive variable-scale search windows, iteratively solves for target matching coordinates, and outputs the full-field strain distribution.
Stable measurement of the full-field strain on the surface of precast piles was achieved under assembly line tensioning conditions, reducing the impact of shaking and movement, and improving the stability and computational efficiency of image matching.
Smart Images

Figure CN122813696B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of precast pile testing technology, and in particular to a method and system for measuring full-field strain during the tensioning process of precast piles based on digital image correlation. Background Technology
[0002] The full-field strain measurement method during the precast pile tensioning process is a detection method that obtains deformation information of a continuous area on the surface of the pile body during the precast pile tensioning operation, and thereby judges the state of tension force transmission. This method can be used to detect local abnormal deformation, stress concentration in the end anchorage area, and uneven tensioning, and has promising applications in precast pile production quality control, automated testing, and production line manufacturing process monitoring.
[0003] In the prior art, a system and method for measuring non-uniform strain on concrete surfaces discloses a scheme that acquires speckle images of the concrete surface before and after deformation using a binocular high-resolution industrial camera, and performs matching calculations on the images to obtain the non-uniform strain distribution on the concrete surface. This scheme illustrates that digital image correlation can be used for measuring strain on concrete surfaces.
[0004] However, the above methods are mainly for concrete samples with relatively stable positions. During measurement, the camera and the sample usually need to maintain a stable correspondence. The precast pile tensioning operation is in a production line environment, where the pile body will continuously move, and the tensioning equipment and production line will cause vibration. The surface of the precast pile also has the characteristics of a cylindrical curved surface. When the tension force is transmitted along the pile body, the degree of deformation at different positions is not consistent, which can easily lead to unstable image matching and discontinuous identification of local deformation, thus affecting the reliability of the overall strain results. Therefore, the existing technology has technical problems such as insufficient coverage of single-point measurement and difficulty in obtaining the overall strain distribution of the precast pile surface stably under production line tensioning conditions. Summary of the Invention
[0005] The purpose of this application is to provide a method and system for measuring the full-field strain of precast piles during the tensioning process based on digital image correlation, so as to solve the problems in the prior art where single-point measurement coverage is insufficient and fixed visual measurement is difficult to stably obtain the full-field strain distribution on the surface of precast piles under assembly line tensioning conditions.
[0006] To address the aforementioned technical problems, in a first aspect, this application provides a method for measuring the full-field strain during the tensioning process of precast piles based on digital image correlation, the method comprising:
[0007] The mobile robot is controlled to follow the precast pile undergoing tensioning along the production line, and uses a vision sensor to capture the initial speckle image of the surface of the precast pile, and captures the deformation speckle image of the surface of the precast pile during the tensioning process.
[0008] Extract the pixel displacement vector of the tension end plate between adjacent image frames in the initial speckle image, and decompose the pixel displacement vector into global motion trajectory and mechanical jitter;
[0009] The initial speckle image is filtered and smoothed according to the mechanical vibration. The filtered initial speckle image is then transformed according to the global motion trajectory to generate a motion-compensated image. The surface geometric curvature and deflection rotation angle of the precast pile are obtained. The motion-compensated image is then flattened based on the deflection rotation angle and the surface geometric curvature to generate a reference speckle image.
[0010] The physical tension displacement of the precast pile is obtained, and a deformation gradient with decreasing axial distribution distance of the physical tension displacement is established. An adaptive variable-scale correlation search window is allocated in the reference speckle image according to the deformation gradient.
[0011] Using the reference speckle image as a reference and the deformed speckle image as a target, the target matching coordinates are iteratively solved within the relevant search window. Abnormal matching points are separated based on the spatial continuity of the target matching coordinates. Interpolation calculations are performed on valid matching points that are in the same topological structure as the abnormal matching points to generate corrected coordinate points.
[0012] The corrected coordinate points are converted into a three-dimensional displacement field in reverse. Strain calculation is performed along the stress transmission path of the three-dimensional displacement field to output the full-field strain distribution of the precast pile.
[0013] Optionally, the step of allocating an adaptively scaled correlation search window in the reference speckle image based on the deformation gradient includes:
[0014] Obtain the parameters of the prestressed tendons pre-installed inside the precast pile, determine the principal stress direction based on the arrangement direction or tensioning direction of the prestressed tendons, and divide the feature sampling area in the reference speckle image based on the principal stress direction.
[0015] Image sampling grids with different resolutions are assigned to the feature sampling region and the peripheral region independent of the feature sampling region, and the measurement reference axis of the image sampling grid is adjusted to remain parallel to the principal stress direction;
[0016] Based on the deformation gradient, the window scaling factor and lateral displacement of each grid node in the image sampling grid are calculated, and an adaptive variable-scale correlation search window is allocated in the image sampling grid according to the window scaling factor and the lateral displacement.
[0017] Optionally, the step of calculating the window scaling factor and lateral displacement of each grid node within the image sampling grid based on the deformation gradient, and allocating an adaptive variable-scale correlation search window within the image sampling grid based on the window scaling factor and the lateral displacement, includes:
[0018] The longitudinal deformation along the measurement reference axis and the lateral deformation perpendicular to the measurement reference axis are analyzed.
[0019] The vertical deformation is compared with the preset standard deformation value to generate a window scaling factor, and the horizontal deformation is used as the lateral displacement.
[0020] Extract the initial window size of the image sampling grid, and perform a product operation on the initial window size according to the window scaling factor to obtain the target window size;
[0021] Based on the lateral displacement, the node center coordinates of the mesh node are spatially translated, and the translated coordinates and the target window size are used to establish the relevant search window.
[0022] Optionally, the step of separating abnormal matching points based on the spatial continuity of the target matching coordinates, and calling valid matching points in the same topological structure as the abnormal matching points to perform interpolation calculations and generate corrected coordinate points includes:
[0023] Obtain the neighborhood coordinate points corresponding to each target matching coordinate, and calculate the relative displacement difference between the target matching coordinates and the coordinates corresponding to the neighborhood coordinate points;
[0024] When the relative displacement difference exceeds the preset elastic deformation threshold, the coordinate point corresponding to the target matching coordinate is determined to be an abnormal matching point;
[0025] Design a polygonal topology that surrounds the abnormal matching points, and filter the valid matching points located at the vertices of the polygonal topology;
[0026] A surface fitting function is constructed using the coordinates of the valid matching points, and the target matching coordinates corresponding to the abnormal matching points are corrected using the surface fitting function to obtain the corrected coordinate points.
[0027] Optionally, the step of constructing a surface fitting function using the coordinates of the valid matching points, and using the surface fitting function to correct the target matching coordinates corresponding to the abnormal matching points to obtain corrected coordinate points, includes:
[0028] Obtain the actual displacement corresponding to the valid matching point;
[0029] Using the coordinates of the effective matching points as input variables and the actual displacement as output variables, a surface fitting function is constructed, wherein the surface fitting function has a polynomial and polynomial coefficients.
[0030] Substitute the target matching coordinates corresponding to the abnormal matching points into the surface fitting function to perform interpolation calculations and obtain the fitted displacement.
[0031] The fitted displacement is superimposed on the original coordinates of the abnormal matching point in the reference speckle image to obtain the corrected coordinate point.
[0032] Optionally, the step of performing strain calculation along the stress transmission path of the three-dimensional displacement field and outputting the full-field strain distribution of the precast pile includes:
[0033] In the three-dimensional displacement field, an axial computational grid is constructed along the direction parallel to the central axis of the precast pile, and a circumferential computational grid is constructed along the tangential direction perpendicular to the central axis.
[0034] Solve for the axial displacement difference of the three-dimensional displacement field on the axial computational grid and the circumferential displacement difference on the circumferential computational grid, respectively.
[0035] Based on the axial displacement difference, the circumferential displacement difference, and the grid spacing of the axial and circumferential calculation grids, the axial strain components, circumferential strain components, and shear strain components of each grid node are analyzed.
[0036] The axial strain component, the circumferential strain component, and the shear strain component are spatially distributed and integrated to output the full-field strain distribution of the precast pile.
[0037] Optionally, the step of extracting the pixel displacement vector of the tensioned end plate in the initial speckle image between adjacent image frames, and decomposing the pixel displacement vector into a global motion trajectory and mechanical jitter, includes:
[0038] In the initial speckle image, the tensioned end plate that has not undergone stress deformation is identified, and the tensioned end plate is taken as the target area;
[0039] The pixel displacement vector of the target region between adjacent image frames is calculated using an image matching algorithm;
[0040] Perform a frequency domain transformation on the pixel displacement vector to obtain a spectral sequence in the frequency domain space;
[0041] Based on a preset cutoff frequency, low-frequency and high-frequency components are separated from the spectral sequence.
[0042] The low-frequency component and the high-frequency component are respectively subjected to inverse frequency domain transformation to reconstruct the global motion trajectory and mechanical jitter.
[0043] Optionally, the step of obtaining the surface geometric curvature and deflection rotation angle of the precast pile, and applying surface flattening processing to the motion compensation image based on the deflection rotation angle and the surface geometric curvature to generate a reference speckle image includes:
[0044] Extract the side profile of the precast pile in the motion-compensated image, calculate the inclination angle of the side profile relative to the axial baseline of the image coordinate system, and determine the inclination angle as the deflection rotation angle.
[0045] Perform image rotation correction on the motion-compensated image according to the stated deflection rotation angle to generate a pose-aligned image;
[0046] Obtain the cross-sectional diameter of the precast pile, determine the cross-sectional radius based on the cross-sectional diameter, and use the reciprocal of the cross-sectional radius as the surface geometric curvature;
[0047] A cylindrical projection model is established based on the surface geometric curvature. The pose-aligned image is then equidistantly unfolded onto a two-dimensional plane using the cylindrical projection model to generate a reference speckle image.
[0048] Optionally, the step of inversely converting the corrected coordinate points into a three-dimensional displacement field includes:
[0049] The central angles of each corrected coordinate point and the reference coordinate points that are not considered as abnormal matching points in the original three-dimensional space are determined according to the preset conversion relationship between pixels and physical scale.
[0050] Based on the central angle, perform three-dimensional coordinate mapping on the corrected coordinate point and the reference coordinate point that is not used as an abnormal matching point, and generate a three-dimensional displacement field in cylindrical coordinate system based on the difference of the mapped three-dimensional coordinates.
[0051] Secondly, this application provides a full-field strain measurement system for the tensioning process of precast piles based on digital image correlation, the system comprising:
[0052] The capture module is used to control the mobile robot to follow the precast pile undergoing tensioning along the production line, and to capture the initial speckle image of the surface of the precast pile using a vision sensor, and capture the deformation speckle image of the surface of the precast pile during the tensioning process.
[0053] The decomposition module is used to extract the pixel displacement vector of the tension end plate in the initial speckle image between adjacent image frames, and decompose the pixel displacement vector into the global motion trajectory and mechanical jitter.
[0054] The generation module is used to perform filtering and smoothing processing on the initial speckle image according to the mechanical vibration, perform coordinate transformation on the filtered initial speckle image according to the global motion trajectory to generate a motion compensation image, obtain the surface geometric curvature and deflection rotation angle of the precast pile, and apply surface flattening processing to the motion compensation image based on the deflection rotation angle and the surface geometric curvature to generate a reference speckle image.
[0055] The allocation module is used to obtain the physical tension displacement of the precast pile, establish a deformation gradient with decreasing axial distribution distance of the physical tension displacement, and allocate an adaptive variable-scale correlation search window in the reference speckle image according to the deformation gradient.
[0056] The calling module is used to iteratively solve the target matching coordinates within the relevant search window, taking the reference speckle image as a reference and the deformed speckle image as the target, to separate abnormal matching points based on the spatial continuity of the target matching coordinates, and to call the valid matching points that are in the same topological structure as the abnormal matching points to perform interpolation calculations to generate corrected coordinate points.
[0057] The output module is used to inversely convert the corrected coordinate points into a three-dimensional displacement field, perform strain calculation along the stress transmission path of the three-dimensional displacement field, and output the full-field strain distribution of the precast pile.
[0058] The full-field strain measurement method for precast pile tensioning process provided in this application has the following advantages: By controlling a mobile robot to follow the precast pile undergoing tensioning along the production line and acquiring initial speckle images and deformed speckle images, continuous image data of the same pile surface before and after tensioning can be obtained; by extracting the pixel displacement vector of the tensioning end plate and decomposing it into global motion trajectory and mechanical jitter, the overall following movement and short-term vibration interference can be distinguished; by generating a reference speckle image through filtering, smoothing, coordinate transformation, and surface flattening, the influence of jitter, movement, and cylindrical curved surfaces on image matching can be reduced; by establishing the deformation gradient through physical tensioning displacement and allocating an adaptive variable-scale correlation search window, a more suitable matching range can be obtained for different deformation regions; by iteratively solving the target matching coordinates and correcting abnormal matching points, the influence of local mismatches on the displacement results can be reduced; by converting the corrected coordinate points into a three-dimensional displacement field and performing strain calculation, the full-field strain distribution of the precast pile surface can be output.
[0059] Furthermore, by obtaining the prestressed tendon parameters to determine the principal stress direction, the feature sampling area and the peripheral area are divided in the reference speckle image, and image sampling grids with different resolutions are set for different areas. Then, the window scaling factor and lateral displacement are calculated according to the deformation gradient to allocate the adaptive variable-scale related search window. This enables the key stress area to obtain more detailed sampling and a more suitable search range, while reducing redundant calculations in the peripheral area, thereby improving the stability and computational efficiency of image matching. Attached Figure Description
[0060] To more clearly illustrate the technical solutions of the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0061] Figure 1 This is a flowchart illustrating a full-field strain measurement method for precast pile tensioning based on digital image correlation disclosed in this application.
[0062] Figure 2 This is a schematic diagram of the mobile robot following and data collection structure disclosed in this application;
[0063] Figure 3 This is a schematic diagram of the abnormal matching point topology interpolation correction disclosed in this application;
[0064] Figure 4 This is a schematic diagram of the three-dimensional displacement field inversion and strain calculation disclosed in this application;
[0065] Figure 5 This is a schematic diagram of the structure of a full-field strain measurement system for the tensioning process of precast piles based on digital image correlation method disclosed in this application. Detailed Implementation
[0066] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some embodiments of the present application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0067] The core of this application is to provide a method for measuring the full-field strain during the tensioning process of precast piles based on digital image correlation. A flowchart of one specific implementation is shown below. Figure 1 As shown, the method includes:
[0068] S101. Control the mobile robot to follow the precast pile undergoing tensioning along the production line, use a vision sensor to capture the initial speckle image of the surface of the precast pile, and capture the deformation speckle image of the surface of the precast pile during the tensioning process.
[0069] like Figure 2 As shown, the mobile robot 21, vision sensor 22, precast pile 23, surface speckle layer 24, tensioning end plate 25, tensioning equipment 26, assembly line track 27, and computing unit 28 together constitute a follow-and-acquire relationship. The surface speckle layer includes a speckle texture layer disposed on the surface of the precast pile, used to form initial speckle images and deformed speckle images by the vision sensor before and after tensioning, and to provide the vision sensor with continuously recognizable image textures. The tensioning equipment includes on-site operating equipment that applies tension force to the tensioning end plate, used to bring the precast pile into the tensioned state. The assembly line track includes a guide path for the mobile robot to move along the production line direction, used to maintain the follow-and-acquire relationship between the mobile robot and the precast pile. The computing unit includes processing components that perform follow-and-acquire control, image reception, and data processing, used to coordinate the mobile robot and vision sensor to complete image acquisition.
[0070] The initial speckle image includes image data of the speckle texture on the surface of the precast pile before tensioning, which is used as a reference for subsequent displacement comparison; the deformation speckle image includes image data of the speckle texture on the surface of the precast pile after the stress changes during tensioning, which is used to reflect the deformation of the surface of the precast pile.
[0071] In one specific implementation, the computing unit controls a mobile robot to move along a production line track, ensuring that the vision sensor maintains a relatively stable imaging position relative to the precast pile being tensioned. The tensioning equipment applies tension to the precast pile through a tensioning end plate. The vision sensor first acquires an initial speckle image covering the surface speckle layer, and then continuously or at preset time intervals acquires deformed speckle images of the surface speckle layer during the tensioning process. Thus, the computing unit can obtain image changes of the same precast pile surface area before and after tensioning, providing an image data basis for subsequent displacement calculations and strain measurements.
[0072] S102. Extract the pixel displacement vector of the tension end plate between adjacent image frames in the initial speckle image, and decompose the pixel displacement vector into global motion trajectory and mechanical jitter.
[0073] The tensioning end plate includes an end plate that is set at the tensioning end of the precast pile and cooperates with the tensioning equipment. In this application scenario, the tensioning end plate mainly serves as a rigid reference area in the image to observe the relative movement between the mobile robot, the vision sensor, and the precast pile. The pixel displacement vector includes the image position change of the target area between adjacent image frames, which usually includes the direction of movement and the distance of movement. The global motion trajectory is used to represent the overall relative movement between the precast pile and the vision sensor with a change frequency less than a preset cutoff frequency. The mechanical vibration is used to represent the short-term vibration caused by the mobile robot, the production line, or the tensioning equipment.
[0074] In this embodiment of the application, S102 specifically includes:
[0075] S1021. Identify the tensioned end plate that has not undergone stress deformation in the initial speckle image, and take the tensioned end plate as the target area.
[0076] The target region includes the local image region used for tracking in the initial speckle image. This region can cover the edges, corners, or locations where the grayscale fluctuation is less than a preset grayscale fluctuation threshold of the tensioned end plate.
[0077] In this embodiment, the computing unit can use edge detection, contour recognition or template matching to determine the position of the tension end plate in the initial speckle image. Since the tension end plate can be regarded as a rigid component in a short time, its image position change mainly reflects the overall movement and jitter. Therefore, the computing unit delineates it as the target area for subsequent image matching.
[0078] S1022. Calculate the pixel displacement vector of the target region between adjacent image frames using an image matching algorithm.
[0079] Among them, the image matching algorithm is used to find the position in the subsequent frame that is most similar to the target region in the previous frame. This algorithm can use a normalized cross-correlation algorithm or a feature point matching algorithm.
[0080] In this embodiment, the calculation unit uses the target region in the previous frame as a reference template, searches for similar image blocks in the vicinity of the next frame, and compares the grayscale similarity between each candidate image block and the reference template; the position with the highest similarity is determined as the position of the target region in the next frame; and the pixel displacement vector is obtained by using the image coordinate difference between the position in the next frame and the position in the previous frame.
[0081] S1023. Perform frequency domain transformation on the pixel displacement vector to obtain the spectral sequence in the frequency domain space.
[0082] Among them, frequency domain transformation is used to convert the pixel displacement vector that changes with time into a combination of different frequency components; the spectral sequence includes the intensity of the change corresponding to each frequency component, which is used to determine whether the displacement change mainly comes from steady motion or short-term vibration.
[0083] In this embodiment, the computing unit arranges multiple pixel displacement vectors in the order of image frames to form a displacement time series; then, the displacement time series is processed by fast Fourier transform to obtain a spectral sequence. Through the spectral sequence, the originally mixed overall movement and vibration changes are converted into distinguishable frequency components.
[0084] S1024. Based on the preset cutoff frequency, separate the low-frequency component and the high-frequency component from the spectral sequence.
[0085] The preset cutoff frequency is used to divide the low-frequency component and the high-frequency component; the part with a frequency lower than the preset cutoff frequency is regarded as the low-frequency component, and the low-frequency component corresponds to the overall movement with a frequency change lower than the cutoff frequency; the part with a frequency greater than or equal to the preset cutoff frequency is regarded as the high-frequency component, and the high-frequency component corresponds to short-term jitter.
[0086] In this embodiment, the portion of the spectral sequence below a preset cutoff frequency is classified as a low-frequency component, and the portion above the preset cutoff frequency is classified as a high-frequency component. The preset cutoff frequency can be set according to the mobile robot's operating speed, the visual sensor's sampling frequency, and the vibration of the production line.
[0087] S1025. Perform inverse frequency domain transformation on the low-frequency component and the high-frequency component respectively to reconstruct the global motion trajectory and mechanical jitter.
[0088] Among them, the inverse frequency domain transform is used to convert the separated frequency components back into displacement curves that vary with time.
[0089] In this embodiment, the computing unit performs inverse frequency domain transformation on the low-frequency component and the high-frequency component respectively: the low-frequency component is transformed into a global motion trajectory, which represents the overall relative motion generated by the visual sensor as the mobile robot follows the precast pile; the high-frequency component is transformed into mechanical jitter, which represents the rapid vibration superimposed during the shooting process.
[0090] As an example, the tensioning end plate is first identified in the initial speckle image, and the region with clear edges and stable grayscale in the tensioning end plate is taken as the target region. Next, a normalized cross-correlation algorithm is used in adjacent image frames to find the new position of this target region, thus obtaining the pixel displacement vector. Then, the pixel displacement vectors obtained from multiple consecutive adjacent image frames are arranged in chronological order and a spectral sequence is obtained through a fast Fourier transform. Then, the spectral sequence is divided into low-frequency and high-frequency components according to a preset cutoff frequency. Finally, inverse frequency domain transforms are performed on the low-frequency and high-frequency components respectively to obtain the global motion trajectory and mechanical jitter. The above example is only one example of this application. In practical applications, settings can be made according to the visual sensor sampling conditions and the operating status of the production line; this application does not limit this.
[0091] Through the above steps, this application can distinguish between overall following motion and short-term vibration interference before subsequent image compensation, so that subsequent speckle matching is more focused on the actual deformation of the precast pile surface.
[0092] S103. Perform filtering and smoothing processing on the initial speckle image according to the mechanical vibration, perform coordinate transformation on the filtered initial speckle image according to the global motion trajectory to generate a motion compensation image, obtain the surface geometric curvature and deflection rotation angle of the precast pile, apply surface flattening processing to the motion compensation image based on the deflection rotation angle and the surface geometric curvature to generate a reference speckle image.
[0093] The filtering and smoothing process includes weakening short-term jitter, local misalignment, or grayscale disturbance caused by mechanical vibration in the initial speckle image, which is used to make the speckle texture in the image more stable; coordinate transformation is used to adjust the image position according to the global motion trajectory, so that the same precast pile surface area in adjacent image frames is aligned as much as possible; motion compensation image includes speckle image after jitter weakening and overall position correction, which is used as the image basis for subsequent surface flattening processing; surface geometric curvature is used to represent the degree of curvature of the outer surface of the precast pile. In the scenario of this application, the precast pile can be approximated as a cylindrical component, so the surface geometric curvature can be determined by the cross-sectional radius of the precast pile; deflection rotation angle is used to represent the degree of inclination of the precast pile relative to the axial baseline in the image.
[0094] In this embodiment of the application, S103 specifically includes:
[0095] S1031. Extract the side profile of the precast pile in the motion-compensated image, calculate the inclination angle of the side profile relative to the axial reference line of the image coordinate system, and determine the inclination angle as the deflection rotation angle.
[0096] The side contour line includes the line corresponding to the outer edge of the precast pile in the motion-compensated image; the axial reference line includes the reference straight line in the image coordinate system used to represent the length direction of the precast pile.
[0097] In this embodiment, an edge detection algorithm can be used to identify the outer edge of the precast pile, and then a straight line fitting method can be used to extract the side contour line from the outer edge. Then, the tilt angle between the side contour line and the axial baseline is compared, and the tilt angle is used as the deflection rotation angle. The deflection rotation angle is used to indicate whether the precast pile is tilted in the image and the tilt direction.
[0098] S1032. Perform image rotation correction on the motion compensation image according to the deflection rotation angle to generate a pose-aligned image.
[0099] Among them, the pose-aligned image includes an image in which the axial direction of the precast pile is basically consistent with the image coordinate direction after rotation correction.
[0100] In this embodiment, performing reverse rotation processing on the motion compensation image according to the deflection rotation angle ensures that the length direction of the precast pile in the image is consistent with the axial baseline. This processing step corrects the originally tilted precast pile image to a regularly placed state, thereby generating a pose-aligned image. The pose-aligned image facilitates subsequent surface unfolding in a unified direction.
[0101] S1033. Obtain the cross-sectional diameter of the precast pile, determine the cross-sectional radius based on the cross-sectional diameter, and use the reciprocal of the cross-sectional radius as the surface geometric curvature.
[0102] The cross-sectional diameter includes the width of the precast pile perpendicular to its length, which can be obtained from the precast pile specification documents, production parameters, or on-site calibration data; the cross-sectional radius is half of the cross-sectional diameter.
[0103] In this embodiment, the cross-sectional diameter of the precast pile is read, and the cross-sectional radius is obtained based on the cross-sectional diameter. After obtaining the cross-sectional radius of the precast pile, the surface geometric curvature is determined as the reciprocal of the cross-sectional radius. That is, the larger the cross-sectional radius, the smaller the curvature of the precast pile surface; the smaller the cross-sectional radius, the greater the curvature of the precast pile surface. Here, the surface geometric curvature describes the degree of curvature of the cylindrical surface of the precast pile, and the cross-sectional radius represents the distance from the center of the precast pile cross-section to the surface.
[0104] As an example, if the cross-sectional diameter is 600 mm, then the cross-sectional radius is 300 mm, and the surface geometric curvature is 1 divided by 300, resulting in one three-hundredth of a millimeter. The above example is merely one illustration of this application; in practical applications, it can be set according to the specifications of the precast piles or the results of on-site calibration, and this application does not limit it in this regard.
[0105] S1034. Establish a cylindrical projection model based on the surface geometric curvature, and use the cylindrical projection model to equidistantly unfold the pose alignment image into a two-dimensional plane to generate a reference speckle image.
[0106] The cylindrical projection model includes geometric mapping rules for describing the correspondence between the cylindrical surface and the two-dimensional plane. This cylindrical projection model is not a model obtained through sample training, but is constructed based on the cross-sectional radius of the precast pile, the surface geometric curvature, and the conversion relationship between pixels and physical scale.
[0107] In this embodiment, the correspondence between the surface of the precast pile and the planar state is determined according to the surface geometric curvature; then, the image points located on the visible surface of the cylinder in the pose alignment image are transformed to the corresponding positions in the two-dimensional plane according to the cylindrical projection model; after equidistant unfolding, the precast pile surface, which was originally a cylindrical curved surface, is unfolded into a planar image that is easy for image matching, thereby generating a reference speckle image.
[0108] As an example, the initial speckle image is first filtered and smoothed using mechanical vibration, and then coordinate transformed according to the global motion trajectory to obtain a motion-compensated image. Next, the side contour lines of the precast pile are identified in the motion-compensated image, and the inclination angle of the side contour lines relative to the axial baseline is calculated to obtain the deflection rotation angle. Then, the motion-compensated image is rotated and corrected according to this deflection rotation angle to obtain a pose-aligned image. Next, the diameter of the precast pile's cross-section is read, and the cross-sectional radius and surface geometric curvature are obtained from the cross-sectional diameter. Finally, a cylindrical projection model is established using the surface geometric curvature, and the pose-aligned image is isometrically unfolded into a two-dimensional planar image using the cylindrical projection model to obtain a reference speckle image. The above example is merely one example of this application; in practical applications, adjustments can be made according to the shape of the precast pile and the image acquisition conditions, and this application does not limit this.
[0109] The above processing steps can reduce the impact of mechanical vibration, overall movement, and cylindrical surface projection on speckle matching, enabling subsequent displacement calculations to be performed on a more stable and uniform reference speckle image.
[0110] S104. Obtain the physical tension displacement of the precast pile, establish a deformation gradient with decreasing axial distribution distance of the physical tension displacement, and allocate an adaptive variable-scale correlation search window in the reference speckle image according to the deformation gradient.
[0111] Among them, physical tension displacement includes the actual displacement data output by the tensioning equipment during the tensioning process, which is used to represent the overall displacement of the precast pile after being pulled at the tensioning end; axial distribution distance includes the actual distance from the tensioning end plate or tensioning end along the central axis of the precast pile to the corresponding measurement position of each image sampling grid node, which is used to determine the distance order of different measurement positions in the deformation gradient; deformation gradient includes the distribution relationship of physical tension displacement gradually transmitted along the axial direction of the precast pile and decreasing with increasing distance, which is used to illustrate the difference in the degree of deformation at different axial positions; adaptive variable-scale correlation search window includes image matching areas whose size and center position can change with the degree of deformation, which is used to improve the positioning stability of subsequent speckle matching.
[0112] In this embodiment of the application, S104 specifically includes:
[0113] S1041. Obtain the prestressed tendon parameters inside the precast pile, determine the principal stress direction according to the arrangement direction or tensioning direction in the prestressed tendon parameters, and divide the feature sampling area in the reference speckle image according to the principal stress direction.
[0114] Among them, the prestressing tendon parameters include the arrangement position, arrangement direction, tensioning direction and corresponding tensioning control information of the prestressing tendons inside the precast pile, which are used to represent the main transmission path of tension force inside the precast pile; the principal stress direction includes the main direction of stress transmission during tensioning, which is used to determine the direction that needs to be sampled and matched in the reference speckle image; the feature sampling area includes the image area in the reference speckle image corresponding to the principal stress direction, which is used to centrally describe the location on the surface of the precast pile where strain changes are more likely to occur.
[0115] In this embodiment, prestressing tendon parameters are first read from the tensioning equipment control record, precast pile production parameter table, or precast pile design data. Then, the arrangement direction or tensioning direction in the prestressing tendon parameters is identified, and the direction in which the tension force is mainly transmitted is determined as the principal stress direction. When multiple groups of prestressing tendons exist, the direction in which the tension force value reaches a preset principal stress threshold can be determined as the principal stress direction. Alternatively, the tension force ratio of each group of prestressing tendons can be compared with a preset ratio threshold, and the principal stress direction can be determined comprehensively according to the direction that reaches the preset ratio threshold. Then, the principal stress direction is mapped to a reference speckle image, and a feature sampling area is divided near the tensioning end and within the image band area through which the prestressing tendon projection passes. Thus, the principal stress direction and the feature sampling area can be used in subsequent steps.
[0116] S1042. Assign image sampling grids with different resolutions to the feature sampling region and the peripheral region independent of the feature sampling region, and adjust the measurement reference axis of the image sampling grid to keep it parallel to the principal stress direction.
[0117] The image sampling grid includes multiple grid nodes arranged on the reference speckle image to determine the location for subsequent image matching; the measurement reference axis includes the axis in the image sampling grid used to represent the main measurement direction, so as to keep the displacement calculation direction consistent with the tension force transmission direction; the peripheral area includes the area in the reference speckle image other than the feature sampling area, used to supplement the overall deformation of the precast pile surface.
[0118] In this embodiment, a high-resolution image sampling grid is first allocated within the feature sampling region, ensuring that the distance between adjacent grid nodes is less than a preset feature node spacing threshold, in order to record image changes at the concentrated tension transmission location. Next, a low-resolution image sampling grid is allocated within the outer region, ensuring that the distance between adjacent grid nodes is greater than or equal to a preset outer node spacing threshold, thereby reducing redundant calculations in areas where the deformation is less than a preset outer deformation threshold. Then, using the principal stress direction as a reference, the orientation of the image sampling grid is rotated or rearranged so that the measurement reference axis of the image sampling grid remains parallel to the principal stress direction. Thus, the image sampling grid can both cover the entire area of the reference speckle image and highlight the main tension transmission area, while also providing grid nodes and initial window sizes for subsequent steps.
[0119] S1043. Based on the deformation gradient, calculate the window scaling factor and lateral displacement of each grid node in the image sampling grid, and allocate an adaptive variable-scale related search window in the image sampling grid according to the window scaling factor and the lateral displacement.
[0120] In this embodiment of the application, S1043 may specifically include the following steps:
[0121] A1. Analyze the longitudinal deformation along the measurement reference axis and the lateral deformation perpendicular to the measurement reference axis; A2. Compare the longitudinal deformation with a preset standard deformation value to generate a window scaling factor, and use the lateral deformation as a lateral displacement; A3. Extract the initial window size of the image sampling grid, and perform a product operation on the initial window size according to the window scaling factor to obtain the target window size; A4. Perform a spatial translation on the node center coordinates of the grid nodes according to the lateral displacement, and establish a relevant search window with the translated coordinates and the target window size.
[0122] The window scaling factor indicates the degree to which the relevant search window needs to be enlarged or reduced relative to the initial window size; the lateral displacement indicates the distance that the center position of the relevant search window needs to be moved laterally relative to the node center coordinates of the grid node; the preset standard deformation value includes a reference deformation degree preset according to the precast pile specifications, tensioning process and visual sensor resolution, used to determine whether the relevant search window needs to be enlarged at a certain grid node; the initial window size includes the window size preset by the image sampling grid before adaptive adjustment; and the target window size includes the window size adjusted according to the window scaling factor.
[0123] In this embodiment, the deformation gradient is first decomposed into deformation information in two directions using the measurement reference axis as a directional reference: longitudinal deformation along the measurement reference axis, which represents the main elongation change caused by the transmission of tension force along the axial direction of the precast pile; and lateral deformation perpendicular to the measurement reference axis, which represents the positional offset of the precast pile surface caused by deflection, local rotation, or lateral deformation. The longitudinal and lateral deformations are used in subsequent steps to calculate the window scaling factor and lateral displacement, respectively.
[0124] Next, the longitudinal deformation at each grid node is compared with the preset standard deformation value: when the longitudinal deformation is greater than or equal to the preset standard deformation value, it indicates that the speckle position change at the grid node has reached the condition for expanding the search range, and the window scaling factor is increased so that the subsequent related search windows cover the target window size; when the longitudinal deformation is less than the preset standard deformation value, it indicates that the speckle position change at the grid node has not reached the condition for expanding the search range, and the window scaling factor is reduced or maintained so that the subsequent matching process reduces invalid searches; at the same time, the lateral deformation is used as a lateral displacement to indicate that the center position of the related search window should move in a direction perpendicular to the measurement reference axis.
[0125] Then, the initial window size preset for each grid node is read, and the initial window size is multiplied by the window scaling factor to obtain the target window size. This product operation is only used to adjust the window size proportionally. Both the initial window size and the target window size are in image pixels, and the window scaling factor is a dimensionless proportional value. Therefore, the calculation result is consistent in terms of dimensions.
[0126] Finally, the lateral displacement is converted into pixel offsets in the reference speckle image, and the center coordinates of the grid nodes are moved along a direction perpendicular to the measurement reference axis to obtain the translated coordinates. Then, using the translated coordinates as the center of the correlation search window and the target window size as the size of the correlation search window, an adaptively scaled correlation search window is established within the image sampling grid. Thus, the correlation search window can adjust its size not only with the degree of longitudinal deformation but also with the lateral position offset.
[0127] After the above steps, the vertical deformation, horizontal deformation, window scaling factor, lateral displacement, and target window size corresponding to each grid node have been obtained. To more intuitively illustrate how the deformation gradient affects the allocation of the relevant search window, Table 1 shows the correspondence between different image regions and the relevant search window adjustment methods:
[0128] Table 1. Correspondence between different image regions and related search window adjustment methods.
[0129]
[0130] Table 1 above is merely an example of this application, used to illustrate the allocation rules of the relevant search window in different image regions. In practical applications, the density of the image sampling grid, the initial window size, the preset standard deformation value, and the window adjustment method can be set according to the precast pile specifications, tensioning process, and image acquisition conditions; this application does not impose any limitations on these.
[0131] As an example, firstly, the prestressing tendon parameters of a precast pile are read to determine that the tension force is mainly transmitted along the axial direction of the precast pile. Then, the area near this transmission path is divided into a feature sampling region in the reference speckle image. Next, an image sampling grid with adjacent node spacing less than a preset feature node spacing threshold is set in the feature sampling region, and an image sampling grid with adjacent node spacing greater than or equal to a preset peripheral node spacing threshold is set in the outer region, ensuring that the measurement reference axis is parallel to the principal stress direction. Then, the deformation gradient is analyzed into longitudinal and transverse deformation variables. The size of the correlation search window is adjusted using the longitudinal deformation variable, and the center position of the correlation search window is moved using the transverse deformation variable. Finally, an adaptive variable-scale correlation search window adapted to different deformation degrees is obtained in the reference speckle image. The above example is only one example of this application; in practical applications, it can be set according to requirements, and this application does not limit it.
[0132] This application establishes a deformation gradient through physical tension displacement and adjusts the size and position of the relevant search window according to the deformation gradient, so that the subsequent image matching process can take into account both the matching reliability of key deformation areas and the computational efficiency of the overall image area.
[0133] S105. Using the reference speckle image as a reference and the deformed speckle image as the target, iteratively solve the target matching coordinates within the relevant search window, separate abnormal matching points according to the spatial continuity of the target matching coordinates, call the valid matching points that are in the same topological structure as the abnormal matching points to perform interpolation calculations, and generate corrected coordinate points.
[0134] like Figure 3As shown, the target matching coordinate 31 is represented by the center coordinate point to be determined, and the neighboring coordinate points 32 are represented by a number of coordinate points distributed around the center coordinate point; when the relative displacement difference between the target matching coordinate 31 and the neighboring coordinate points 32 exceeds a preset elastic deformation threshold, the coordinate point corresponding to the target matching coordinate 31 is determined to be an abnormal matching point 33. Figure 3 In the diagram, the judgment circle outside the central coordinate point represents the abnormal matching point 33; the polygonal topology 34 is set around the abnormal matching point 33, and the coordinate points located at the vertices of the polygonal topology 34 that are not judged as abnormal are considered valid matching points 35. The surface fitting function and the corrected coordinate points are part of the subsequent calculation results. Figure 3 No separate reference numerals are set for the attached figures.
[0135] The target matching coordinates include the position coordinates of the speckle region in the reference speckle image and the corresponding position in the deformed speckle image, which are used to represent the movement result of the speckle region during the tensioning process; spatial continuity includes the characteristic that the displacement change of adjacent surface positions should be smooth and continuous, which is used to determine whether a matching result deviates significantly from the surrounding position; abnormal matching points include coordinate points whose matching results are unreliable due to speckle blurring, local occlusion, illumination changes or vibration interference; valid matching points include matching points that have not been judged as abnormal matching points and can represent the real displacement changes of the surrounding area.
[0136] In this embodiment of the application, S105 specifically includes:
[0137] S1051. Obtain the neighborhood coordinate points corresponding to each target matching coordinate, and calculate the relative displacement difference between the target matching coordinates and the coordinates corresponding to the neighborhood coordinate points.
[0138] Among them, the neighborhood coordinate points include the matching coordinates of adjacent grid nodes located around a target matching coordinate, which are used to reflect the displacement change trend near the target matching coordinate; the relative displacement difference includes the difference between the displacement result corresponding to the target matching coordinate and the displacement result corresponding to the neighborhood coordinate points, which are used to determine whether the target matching coordinate is consistent with the surrounding position.
[0139] In this embodiment, firstly, starting from the grid nodes in the reference speckle image, a gray-level correlation matching algorithm is used to compare the speckle texture within the corresponding relevant search window. Then, the position with the highest texture similarity among the candidate positions in the deformed speckle image that reaches a preset similarity threshold is selected as the initial matching position. Secondly, the search range is narrowed near the initial matching position and the comparison is repeated to make the target matching coordinates gradually approach the real position. Next, according to the adjacency relationship of the image sampling grid, neighborhood coordinate points are extracted for each target matching coordinate. Then, the displacements of the target matching coordinates and neighborhood coordinate points relative to their original positions in the reference speckle image are calculated respectively. Finally, the differences in the displacement direction and displacement magnitude are compared to obtain the relative displacement difference.
[0140] S1052. When the relative displacement difference exceeds the preset elastic deformation threshold, the coordinate point corresponding to the target matching coordinate is determined as an abnormal matching point.
[0141] Among them, the preset elastic deformation threshold includes a displacement difference limit set in advance according to the material characteristics of the precast pile, the tensioning process requirements and the image sampling grid spacing, which is used to represent the range of displacement changes that are allowed to occur between adjacent positions under normal elastic deformation.
[0142] In this embodiment, the relative displacement difference is compared with a preset elastic deformation threshold: when the relative displacement difference used for judgment does not exceed the preset elastic deformation threshold, it indicates that the displacement change of the target matching coordinate and the surrounding neighboring coordinate points remains continuous, and the target matching coordinate is retained; when the relative displacement difference used for judgment exceeds the preset elastic deformation threshold, it indicates that the target matching coordinate has a sudden change relative to the surrounding neighboring coordinate points, the corresponding coordinate point is judged as an abnormal matching point, and the abnormal matching point is handed over to subsequent steps for correction preparation.
[0143] S1053. Design a polygonal topology that surrounds the abnormal matching points, and filter the valid matching points located at the vertices of the polygonal topology.
[0144] The polygon topology includes the connection relationships between adjacent points established around the anomalous matching points, which are used to determine which valid matching points can jointly describe the displacement changes near the anomalous matching points; the valid matching points at the vertices include matching points located at the boundary of the polygon topology that have not been identified as anomalous matching points, which are used to provide a reliable basis for subsequent interpolation calculations.
[0145] In this embodiment, with the abnormal matching point as the center, adjacent normal matching points are selected in the image sampling grid, and these normal matching points are connected according to their spatial positions in the image to form a polygonal topology. This polygonal topology can be a triangle, quadrilateral, or other polygons that can enclose the abnormal matching point. Then, it is checked whether the matching points at each vertex of the polygonal topology meet the spatial continuity requirement, and points that have been determined to be abnormal matching points or whose relative displacement difference exceeds a preset elastic deformation threshold are removed, while the valid matching points at the vertices are retained. The valid matching points obtained in this way can reflect the local deformation trend around the abnormal matching point that meets the spatial continuity requirement.
[0146] S1054. Construct a surface fitting function using the coordinates of the effective matching points, and use the surface fitting function to correct the target matching coordinates corresponding to the abnormal matching points to obtain the corrected coordinate points.
[0147] In this embodiment of the application, S1054 may specifically include the following steps:
[0148] B1. Obtain the actual displacement corresponding to the valid matching point; B2. Construct a surface fitting function using the coordinates of the valid matching point as input variables and the actual displacement as output variables, wherein the surface fitting function has a polynomial and polynomial coefficients; B3. Substitute the target matching coordinates corresponding to the abnormal matching point into the surface fitting function to perform interpolation calculation and obtain the fitted displacement; B4. Superimpose the fitted displacement onto the original coordinates of the abnormal matching point in the reference speckle image to obtain the corrected coordinate point.
[0149] The surface fitting function can be a local displacement change function constructed on the spot based on the coordinates of the effective matching point and the actual displacement. It is used to describe the relationship between the displacement around the abnormal matching point and the change in spatial position. This surface fitting function is not a pre-trained model, but is temporarily constructed based on the effective matching points within the same topology each time the abnormal matching point needs to be corrected. The actual displacement includes the positional difference between the original coordinates of the effective matching point in the reference speckle image and the target matching coordinates in the deformed speckle image, which is used to represent the real movement result of the effective matching point during the tensioning process. The fitted displacement includes the displacement of the abnormal matching point calculated using the local displacement change trend of the effective matching point, which is used to replace the original unreliable displacement result of the abnormal matching point.
[0150] In this embodiment, the effective matching points obtained through filtering are first read, and the original coordinates of each effective matching point in the reference speckle image and the target matching coordinates in the deformed speckle image are obtained respectively. The actual displacement corresponding to each effective matching point is determined by the coordinate difference, and the actual displacement is used as the output basis for constructing the surface fitting function in the subsequent steps.
[0151] Next, for each displacement component in the actual displacement, a surface fitting function is constructed using the least squares surface fitting method. Specifically: the coordinates of the effective matching points are converted into local coordinates, and the local coordinates are normalized using the spacing of the image sampling grid so that the input variables do not have physical dimensions; when the actual displacement includes axial displacement and circumferential displacement, the polynomial coefficients are solved using the corresponding displacement components as output variables, thereby obtaining the surface fitting function. It can be expressed using the polynomial relationship in formula (1):
[0152] (1)
[0153] in, This represents the displacement obtained by fitting at the normalized local coordinates; This represents the normalized horizontal coordinate of the valid matching point in the local coordinate system. This represents the normalized longitudinal coordinate of the valid matching point in the local coordinate system; Indicates the order of a polynomial; Indicates and index and index The corresponding polynomial coefficients; express The exponent in the current polynomial term, express The exponent in the current polynomial term, and and The sum does not exceed .because and The input variables have been normalized using the spacing of the image sampling grid, therefore they are dimensionless parameters. The dimensions of the polynomial are consistent with the actual displacement, and both sides of the polynomial maintain dimensional consistency. The polynomial order can be determined based on the number and distribution of effective matching points: a low-order polynomial can be used when the number of effective matching points is less than a preset threshold; a high-order polynomial can be used when the number of effective matching points is greater than or equal to the preset threshold, and the difference in spacing between effective matching points is less than a preset distribution uniformity threshold. These thresholds can be preset based on the image sampling grid density and the relevant search window size.
[0154] Then, the target matching coordinates corresponding to the abnormal matching points are converted into normalized local coordinates, and these normalized local coordinates are substituted into the surface fitting function. The surface fitting function performs interpolation calculations on the location of the abnormal matching points according to the local displacement trend formed by the surrounding valid matching points, and obtains the fitted displacement. This fitted displacement does not come directly from the original erroneous matching results of the abnormal matching points, but from the continuous change trend of valid matching points within the same topology.
[0155] Finally, using the original coordinates of the anomalous matching points in the reference speckle image as the starting point, the fitted displacement is superimposed onto these original coordinates to obtain the corrected coordinate points. These corrected coordinate points replace the previously unreliable target matching coordinates of the anomalous matching points and participate in the subsequent three-dimensional displacement field transformation along with the target matching coordinates of points not identified as anomalous.
[0156] After completing the above processing, it is possible to distinguish between normal matching points, abnormal matching points, and valid matching points used for correction. To more intuitively illustrate the judgment criteria and processing methods for different matching points in this application scenario, Table 2 provides an explanation of the matching point judgment and correction process:
[0157] Table 2. Explanation of Matching Point Determination and Correction Processing
[0158]
[0159] Table 2 above is merely an example of this application, used to illustrate the determination and correction methods for different matching points in this application scenario. In practical applications, the selection method of neighborhood coordinate points, the shape of the polygon topology, the preset elastic deformation threshold, and the order of the surface fitting function can be set according to the precast pile specifications, image sampling grid density, and on-site image quality; this application does not impose any limitations on these.
[0160] As an example, firstly, grayscale correlation matching is performed on the reference speckle image and the deformed speckle image within the relevant search window to obtain the target matching coordinates, and then the neighboring coordinate points around the target matching coordinates are extracted; next, the relative displacement difference between the target matching coordinates and the neighboring coordinate points is compared, and when the relative displacement difference exceeds a preset elastic deformation threshold, the coordinate point is determined to be an abnormal matching point; then, a polygonal topology is established around the abnormal matching point, and valid matching points located at the vertices of the polygonal topology are selected; finally, a surface fitting function is constructed using the coordinates of the valid matching points and the actual displacement, and the target matching coordinates corresponding to the abnormal matching points are interpolated and corrected to obtain the corrected coordinate points. The above example is only one example of this application, and in practical applications, it can be set according to requirements, which is not limited in this application.
[0161] This application separates anomalous matching points by spatial continuity and generates corrected coordinate points using effective matching points within the same topology, which can improve the continuity and reliability of speckle matching results and provide a stable data foundation for subsequent three-dimensional displacement field conversion and full-field strain distribution output.
[0162] S106. Convert the corrected coordinate points into a three-dimensional displacement field in reverse, perform strain calculation along the stress transmission path of the three-dimensional displacement field, and output the full-field strain distribution of the precast pile.
[0163] like Figure 4 As shown, the calculation unit first uses the matched and corrected two-dimensional position in the reference speckle image as the flattened coordinates 41. Figure 4 The coordinates of the unfolded coordinates are represented by coordinates on a two-dimensional unfolded plane. Then, based on the conversion relationship between pixels and physical scale, the lateral position of the unfolded coordinates 41 is converted into the arc length of a circle, and the central angle 42 is determined in combination with the radius of the precast pile cross-section. Figure 4 The angle between two radii within a circular cross-section is used to represent the stress. Then, the flattened coordinates 41 are restored to the three-dimensional position of the precast pile's cylindrical surface, and the coordinate differences between the corresponding three-dimensional positions before and after tensioning are compared. Strain calculations are then performed along the axial calculation mesh 43, the circumferential calculation mesh 44, and the stress transmission path 45. The axial calculation mesh 43 is represented by mesh lines extending along the central axis of the cylindrical surface, the circumferential calculation mesh 44 is represented by circumferential mesh lines extending along the circumference of the cylindrical surface, and the stress transmission path 45 is represented by arrows that propagate along the central axis and diffuse circumferentially. The cylindrical mapping model, the three-dimensional displacement field, and the full-field strain distribution are calculation relationships or output results. Figure 4 No separate reference numerals are set for the attached figures.
[0164] The three-dimensional displacement field includes displacement data of each measured position on the surface of the precast pile in three-dimensional space, which is used to represent the actual spatial deformation of the precast pile during the tensioning process; the stress transmission path includes the path of the tension force transmitted from the tensioning end along the central axis of the precast pile and diffused in the circumferential direction on the cylindrical surface, which is used to determine the main calculation direction when solving the strain; the overall strain distribution includes the axial strain components, circumferential strain components and shear strain components at different positions on the surface of the precast pile, which is used to reflect the overall stress and deformation state of the surface of the precast pile during the tensioning process.
[0165] In this embodiment of the application, S106 specifically includes:
[0166] S1061. Determine the central angle of each corrected coordinate point and the reference coordinate point that is not considered an abnormal matching point in the original three-dimensional space according to the preset conversion relationship between pixels and physical scale.
[0167] The conversion relationship between pixels and physical scales includes the correspondence between image pixel length and the actual length of the precast pile surface, which can be determined by visual sensor calibration, precast pile cross-sectional diameter, or on-site calibration ruler. The original three-dimensional space includes the physical coordinate space of the precast pile surface in the actual cylindrical structure before the reference speckle image is generated. The original three-dimensional space uses the central axis of the precast pile as the axial reference, the center of the precast pile cross-section as the reference center of the circumferential position, and the radius of the precast pile cross-section as the distance from the surface point to the central axis. The central angle includes the angular position of a certain surface position on the precast pile cylindrical cross-section relative to the center of the cross-section, which is used to restore the lateral position in the two-dimensional unfolded image to the cylindrical surface in the original three-dimensional space. The reference coordinate points include coordinate points that have not been identified as abnormal matching points and can be used directly without interpolation correction.
[0168] In this embodiment, firstly, a preset conversion relationship between pixels and physical scales is read, converting the pixel distance in the reference speckle image into the actual distance on the surface of the precast pile. Next, the lateral positions of each corrected coordinate point and the reference coordinate points not considered as abnormal matching points are extracted from the reference speckle image, and these lateral positions are converted into the circumferential arc length on the cylindrical surface of the precast pile. Then, the ratio of the circumferential arc length to the cross-sectional radius is processed to obtain the central angle of the corresponding position in the original three-dimensional space. Since both the circumferential arc length and the cross-sectional radius are length quantities, the central angle obtained after the ratio process no longer carries a length unit and is only used to represent the angular position on the cylindrical cross-section. The central angle in this process can be used as input for subsequent three-dimensional coordinate mapping.
[0169] S1062. Perform three-dimensional coordinate mapping on the corrected coordinate point and the reference coordinate point that is not used as an abnormal matching point according to the central angle, and generate a three-dimensional displacement field in cylindrical coordinate system according to the difference of the mapped three-dimensional coordinates.
[0170] The cylindrical coordinate system includes radial position, angular position in the circumferential direction, and axial position, which are used to describe the position of points on the surface of cylindrical components such as precast piles; the three-dimensional coordinate mapping includes the process of converting two-dimensional image coordinates into spatial coordinates on the cylindrical surface.
[0171] In this embodiment, the radial position of the surface point is first determined by the cross-sectional radius of the precast pile, the angular position of the surface point in the circumferential direction is determined by the central angle, and the axial position of the surface point is determined by the coordinate conversion result along the length of the precast pile in the image. Then, three-dimensional coordinate mapping is performed on the corrected coordinate points and the reference coordinate points that are not considered abnormal matching points, so that the points in the two-dimensional image are re-corresponded to the cylindrical surface of the precast pile. Then, the three-dimensional coordinates of the same measurement position before and after tensioning are compared, the difference in the three-dimensional coordinates after mapping is obtained, and the difference in the three-dimensional coordinates of each measurement position is arranged according to the spatial position to generate a three-dimensional displacement field in the cylindrical coordinate system.
[0172] S1063. In the three-dimensional displacement field, an axial computational grid is constructed along the direction parallel to the central axis of the precast pile, and a circumferential computational grid is constructed along the tangential direction perpendicular to the central axis.
[0173] The axial computational grid includes computational nodes arranged along the length of the precast pile, used to calculate the expansion and contraction changes caused by the transmission of tension force along the pile body; the circumferential computational grid includes computational nodes arranged along the tangent direction of the precast pile circumference, used to calculate the expansion, contraction or local displacement of the precast pile surface along the circumferential direction.
[0174] In this embodiment, the direction of the central axis of the precast pile is first determined in a three-dimensional displacement field, and the direction parallel to the central axis is taken as the axial calculation direction. Then, adjacent measurement positions are connected along this axial calculation direction to construct an axial calculation grid. Next, adjacent measurement positions are connected along the tangent direction of the cylindrical surface perpendicular to the central axis to construct a circumferential calculation grid. Thus, the axial calculation grid and the circumferential calculation grid correspond to the main elongation direction and the circumferential deformation direction of the surface during the tensioning process of the precast pile, respectively.
[0175] S1064. Solve for the axial displacement difference of the three-dimensional displacement field on the axial calculation grid and the circumferential displacement difference on the circumferential calculation grid, respectively.
[0176] Among them, the axial displacement difference includes the displacement difference between adjacent grid nodes in the axial calculation grid in the axial direction, which is used to represent the local expansion and contraction along the length of the precast pile; the circumferential displacement difference includes the displacement difference between adjacent grid nodes in the circumferential calculation grid in the circumferential direction, which is used to represent the local deformation of the cylindrical surface along the circumferential direction.
[0177] In this embodiment, the calculation unit reads the three-dimensional displacement data of each grid node from the three-dimensional displacement field: first, it calculates the axial displacement of adjacent grid nodes one by one along the axial direction to obtain the axial displacement difference; then, it calculates the circumferential displacement of adjacent grid nodes one by one along the circumferential direction to obtain the circumferential displacement difference. Through this processing, the displacement result at each position in the three-dimensional displacement field is converted into the displacement change result between adjacent positions, providing a data basis for subsequent steps to analyze strain components.
[0178] S1065. Based on the axial displacement difference, the circumferential displacement difference, and the grid spacing of the axial calculation grid and the circumferential calculation grid, the axial strain component, circumferential strain component, and shear strain component of each grid node are analyzed.
[0179] Among them, the axial strain component is used to represent the degree of tension or compression of the precast pile along the central axis; the circumferential strain component is used to represent the degree of expansion or contraction of the cylindrical surface of the precast pile along the circumferential direction; the shear strain component is used to represent the degree of relative displacement between the axial and circumferential directions; the grid spacing includes the actual physical distance between adjacent grid nodes, which is used to convert the displacement difference into the amount of deformation per unit length.
[0180] In this embodiment, the finite difference method is used to perform strain calculation: First, the axial displacement difference is divided by the grid spacing of the axial calculation grid to obtain the axial strain component, which represents the unit length expansion and contraction of adjacent grid nodes in the direction of the precast pile's central axis; then, the circumferential displacement difference is divided by the grid spacing of the circumferential calculation grid to obtain the circumferential strain component, which represents the unit length expansion or contraction of adjacent grid nodes in the direction of the circumference tangent of the precast pile; then, the circumferential displacement difference of adjacent nodes on the axial calculation grid is obtained, and the circumferential displacement difference is divided by the grid spacing of the axial calculation grid to obtain the first displacement term; at the same time, the axial displacement difference of adjacent nodes on the circumferential calculation grid is obtained, and the axial displacement difference is divided by the grid spacing of the circumferential calculation grid to obtain the second displacement term; finally, the first displacement term and the second displacement term are added together to obtain the shear strain component. It can be expressed by formulas (2), (3) and (4):
[0181] (2)
[0182] (3)
[0183] (4)
[0184] in, Indicates the axial strain component; Indicates the circumferential strain component; Represents the shear strain component; This represents the difference in axial displacement along the circumferential computational mesh; This represents the difference in circumferential displacement of the calculated mesh along the axial direction; Indicates the grid spacing of the axial computational mesh; This represents the grid spacing of the circumferential computational mesh. The grid spacing of the circumferential computational mesh is the arc length distance between adjacent nodes on the cylindrical surface along the circumferential computational mesh.
[0185] In the above formula, the axial displacement difference, the circumferential displacement difference, the grid spacing of the axial calculation grid, and the grid spacing of the circumferential calculation grid are all length quantities; the displacement difference divided by the corresponding grid spacing yields a dimensionless result, so the axial strain components, circumferential strain components, and shear strain components are consistent in dimensions.
[0186] S1066. The axial strain component, the circumferential strain component and the shear strain component are spatially distributed and integrated to output the full-field strain distribution of the precast pile.
[0187] Spatial distribution integration includes the process of arranging and combining different strain components according to the three-dimensional position of each grid node on the surface of the precast pile cylinder, in order to form a viewable overall strain result.
[0188] In this embodiment, the axial strain component, circumferential strain component, and shear strain component are first bound to the corresponding grid node positions. Then, the grid nodes are arranged according to the axial position and central angle position in the cylindrical coordinate system, and the three types of strain components are combined into full-field strain data. Finally, the full-field strain distribution of the precast pile can be output in the form of strain cloud map, node data table, or test report, so that the testers can view the deformation state at different positions on the surface of the precast pile during the tensioning process.
[0189] After completing the above processing, the two-dimensional image matching results have been restored to three-dimensional spatial deformation results, and the strain components in different directions have been further analyzed. To more intuitively illustrate the correspondence between the three-dimensional displacement field and strain calculation, Table 3 provides a description of the three-dimensional displacement field strain calculation object:
[0190] Table 3 Description of the objects for three-dimensional displacement field strain calculation
[0191]
[0192] Table 3 above is merely an example of this application, used to illustrate the relationship between various processed objects in the three-dimensional displacement field strain calculation process. In practical applications, the method for determining the central angle, the three-dimensional coordinate mapping method, the grid spacing of the axial calculation grid and the circumferential calculation grid, and the output form of the full-field strain distribution can be set according to the precast pile specifications, the visual sensor calibration results, and the detection requirements. This application does not impose any limitations on these aspects.
[0193] As an example, firstly, the lateral positions of the corrected coordinate points and reference coordinate points (not considered as anomalous matching points) in the reference speckle image are converted into central angles in the original three-dimensional space using the conversion relationship between pixels and physical scales. Secondly, three-dimensional coordinate mapping is performed according to the central angle, cross-sectional radius, and axial position, and a three-dimensional displacement field in cylindrical coordinates is generated using the difference between the mapped three-dimensional coordinates. Next, axial and circumferential computational grids are constructed on the three-dimensional displacement field. Then, the axial and circumferential displacement differences are solved, and the ratio of the displacement differences to the corresponding grid spacing is processed to analyze the axial strain components, circumferential strain components, and shear strain components. Finally, the strain components are integrated according to the spatial position of the grid nodes to output the full-field strain distribution of the precast pile. The above example is only one example of this application, and it can be set according to the requirements in practical applications. This application does not limit it.
[0194] This application converts the modified coordinate points into a three-dimensional displacement field and performs strain calculations along the stress transmission path, thereby converting the two-dimensional speckle matching results into spatial strain results on the surface of the precast pile, so that the local deformation and overall deformation during the tensioning process can be continuously presented.
[0195] This application, through S101-S106, can continuously acquire speckle images during the precast pile tensioning operation and compensate for the effects of overall movement, mechanical vibration, and cylindrical surface projection caused by follow-up shooting; at the same time, it adaptively sets the relevant search window according to the tensioning displacement distribution and corrects abnormal matching points to reduce the impact of local mismatches on displacement calculation; furthermore, it converts the corrected two-dimensional matching results into a three-dimensional displacement field and performs strain calculation, thereby obtaining the full-field strain distribution during the precast pile tensioning process and improving the continuity, stability, and reliability of tensioning deformation measurement.
[0196] Figure 5 This application provides a schematic diagram of a specific implementation of a full-field strain measurement system for the tensioning process of precast piles based on digital image correlation, as shown in the embodiments of this application. Figure 5 The system may include:
[0197] The capture module 51 is used to control the mobile robot to follow the precast pile undergoing tensioning along the production line, capture the initial speckle image of the surface of the precast pile using a vision sensor, and capture the deformation speckle image of the surface of the precast pile during the tensioning process.
[0198] Decomposition module 52 is used to extract the pixel displacement vector of the tension end plate between adjacent image frames in the initial speckle image, and decompose the pixel displacement vector into global motion trajectory and mechanical jitter;
[0199] The generation module 53 is used to perform filtering and smoothing processing on the initial speckle image according to the mechanical vibration, perform coordinate transformation on the filtered initial speckle image according to the global motion trajectory to generate a motion compensation image, obtain the surface geometric curvature and deflection rotation angle of the precast pile, and apply surface flattening processing to the motion compensation image based on the deflection rotation angle and the surface geometric curvature to generate a reference speckle image.
[0200] The allocation module 54 is used to obtain the physical tension displacement of the precast pile, establish a deformation gradient with decreasing axial distribution distance of the physical tension displacement, and allocate an adaptive variable-scale correlation search window in the reference speckle image according to the deformation gradient.
[0201] The module 55 is invoked to iteratively solve the target matching coordinates within the relevant search window, using the reference speckle image as a reference and the deformed speckle image as the target, separate abnormal matching points according to the spatial continuity of the target matching coordinates, call the valid matching points that are in the same topology as the abnormal matching points to perform interpolation calculations, and generate corrected coordinate points.
[0202] Output module 56 is used to inversely convert the corrected coordinate points into a three-dimensional displacement field, perform strain calculation along the stress transmission path of the three-dimensional displacement field, and output the full-field strain distribution of the precast pile.
[0203] The full-field strain measurement system for the precast pile tensioning process based on digital image correlation method in this application embodiment is used to implement the aforementioned full-field strain measurement method for the precast pile tensioning process based on digital image correlation method. Therefore, the specific implementation of the full-field strain measurement system for the precast pile tensioning process based on digital image correlation method can be found in the embodiment section of the precast pile tensioning process full-field strain measurement method based on digital image correlation method above. The specific implementation can be referred to the description of the corresponding embodiment, and will not be repeated here.
[0204] The foregoing has provided a detailed description of the full-field strain measurement method and system for the tensioning process of precast piles based on digital image correlation. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are merely for the purpose of helping to understand the method and its core ideas. It should be noted that those skilled in the art can make various improvements and modifications to this application without departing from its principles, and these improvements and modifications also fall within the protection scope of this application.
Claims
1. A method for measuring the full-field strain during the tensioning process of precast piles based on digital image correlation, characterized in that, include: The mobile robot is controlled to follow the precast pile undergoing tensioning along the production line, and uses a vision sensor to capture the initial speckle image of the surface of the precast pile, and captures the deformation speckle image of the surface of the precast pile during the tensioning process. Extract the pixel displacement vector of the tension end plate between adjacent image frames in the initial speckle image, and decompose the pixel displacement vector into global motion trajectory and mechanical jitter; The initial speckle image is filtered and smoothed according to the mechanical vibration. The filtered initial speckle image is then transformed according to the global motion trajectory to generate a motion-compensated image. The surface geometric curvature and deflection rotation angle of the precast pile are obtained. The motion-compensated image is then flattened based on the deflection rotation angle and the surface geometric curvature to generate a reference speckle image. The physical tension displacement of the precast pile is obtained, and a deformation gradient with decreasing axial distribution distance of the physical tension displacement is established. An adaptive variable-scale correlation search window is allocated in the reference speckle image according to the deformation gradient. Using the reference speckle image as a reference and the deformed speckle image as a target, the target matching coordinates are iteratively solved within the relevant search window. Abnormal matching points are separated based on the spatial continuity of the target matching coordinates. Interpolation calculations are performed on valid matching points that are in the same topological structure as the abnormal matching points to generate corrected coordinate points. The corrected coordinate points are converted into a three-dimensional displacement field in reverse. Strain calculation is performed along the stress transmission path of the three-dimensional displacement field to output the full-field strain distribution of the precast pile.
2. The method according to claim 1, characterized in that, The step of allocating an adaptively scaled correlation search window in the reference speckle image based on the deformation gradient includes: Obtain the parameters of the prestressed tendons pre-installed inside the precast pile, determine the principal stress direction based on the arrangement direction or tensioning direction of the prestressed tendons, and divide the feature sampling area in the reference speckle image based on the principal stress direction. Image sampling grids with different resolutions are assigned to the feature sampling region and the peripheral region independent of the feature sampling region, and the measurement reference axis of the image sampling grid is adjusted to remain parallel to the principal stress direction; Based on the deformation gradient, the window scaling factor and lateral displacement of each grid node in the image sampling grid are calculated, and an adaptive variable-scale correlation search window is allocated in the image sampling grid according to the window scaling factor and the lateral displacement.
3. The method according to claim 2, characterized in that, The step of calculating the window scaling factor and lateral displacement of each grid node within the image sampling grid based on the deformation gradient, and allocating an adaptive variable-scale correlation search window within the image sampling grid based on the window scaling factor and the lateral displacement, includes: The longitudinal deformation along the measurement reference axis and the lateral deformation perpendicular to the measurement reference axis are analyzed. The vertical deformation is compared with the preset standard deformation value to generate a window scaling factor, and the horizontal deformation is used as the lateral displacement. Extract the initial window size of the image sampling grid, and perform a product operation on the initial window size according to the window scaling factor to obtain the target window size; Based on the lateral displacement, the node center coordinates of the mesh node are spatially translated, and the translated coordinates and the target window size are used to establish the relevant search window.
4. The method according to claim 1, characterized in that, The step of separating abnormal matching points based on the spatial continuity of the target matching coordinates, and then calling valid matching points in the same topological structure as the abnormal matching points to perform interpolation calculations and generate corrected coordinate points includes: Obtain the neighborhood coordinate points corresponding to each target matching coordinate, and calculate the relative displacement difference between the target matching coordinates and the coordinates corresponding to the neighborhood coordinate points; When the relative displacement difference exceeds the preset elastic deformation threshold, the coordinate point corresponding to the target matching coordinate is determined to be an abnormal matching point; Design a polygonal topology that surrounds the abnormal matching points, and filter the valid matching points located at the vertices of the polygonal topology; A surface fitting function is constructed using the coordinates of the valid matching points, and the target matching coordinates corresponding to the abnormal matching points are corrected using the surface fitting function to obtain the corrected coordinate points.
5. The method according to claim 4, characterized in that, The step of constructing a surface fitting function using the coordinates of the valid matching points, and then using the surface fitting function to correct the target matching coordinates corresponding to the abnormal matching points to obtain corrected coordinate points, includes: Obtain the actual displacement corresponding to the valid matching point; Using the coordinates of the effective matching points as input variables and the actual displacement as output variables, a surface fitting function is constructed, wherein the surface fitting function has a polynomial and polynomial coefficients. Substitute the target matching coordinates corresponding to the abnormal matching points into the surface fitting function to perform interpolation calculations and obtain the fitted displacement. The fitted displacement is superimposed on the original coordinates of the abnormal matching point in the reference speckle image to obtain the corrected coordinate point.
6. The method according to claim 1, characterized in that, The strain calculation performed along the stress transmission path of the three-dimensional displacement field outputs the full-field strain distribution of the precast pile, including: In the three-dimensional displacement field, an axial computational grid is constructed along the direction parallel to the central axis of the precast pile, and a circumferential computational grid is constructed along the tangential direction perpendicular to the central axis. Solve for the axial displacement difference of the three-dimensional displacement field on the axial computational grid and the circumferential displacement difference on the circumferential computational grid, respectively. Based on the axial displacement difference, the circumferential displacement difference, and the grid spacing of the axial and circumferential calculation grids, the axial strain components, circumferential strain components, and shear strain components of each grid node are analyzed. The axial strain component, the circumferential strain component, and the shear strain component are spatially distributed and integrated to output the full-field strain distribution of the precast pile.
7. The method according to claim 1, characterized in that, The step of extracting the pixel displacement vector of the tensioned end plate in the initial speckle image between adjacent image frames, and decomposing the pixel displacement vector into a global motion trajectory and mechanical jitter, includes: In the initial speckle image, the tensioned end plate that has not undergone stress deformation is identified, and the tensioned end plate is taken as the target area; The pixel displacement vector of the target region between adjacent image frames is calculated using an image matching algorithm; Perform a frequency domain transformation on the pixel displacement vector to obtain a spectral sequence in the frequency domain space; Based on a preset cutoff frequency, low-frequency and high-frequency components are separated from the spectral sequence. The low-frequency component and the high-frequency component are respectively subjected to inverse frequency domain transformation to reconstruct the global motion trajectory and mechanical jitter.
8. The method according to claim 1, characterized in that, The process of acquiring the surface geometric curvature and deflection rotation angle of the precast pile, and applying surface flattening processing to the motion compensation image based on the deflection rotation angle and the surface geometric curvature to generate a reference speckle image includes: Extract the side profile of the precast pile in the motion-compensated image, calculate the inclination angle of the side profile relative to the axial baseline of the image coordinate system, and determine the inclination angle as the deflection rotation angle. Perform image rotation correction on the motion-compensated image according to the stated deflection rotation angle to generate a pose-aligned image; Obtain the cross-sectional diameter of the precast pile, determine the cross-sectional radius based on the cross-sectional diameter, and use the reciprocal of the cross-sectional radius as the surface geometric curvature; A cylindrical projection model is established based on the surface geometric curvature. The pose-aligned image is then equidistantly unfolded onto a two-dimensional plane using the cylindrical projection model to generate a reference speckle image.
9. The method according to claim 1, characterized in that, The step of inversely converting the corrected coordinate points into a three-dimensional displacement field includes: The central angles of each corrected coordinate point and the reference coordinate points that are not considered as abnormal matching points in the original three-dimensional space are determined according to the preset conversion relationship between pixels and physical scale. Based on the central angle, perform three-dimensional coordinate mapping on the corrected coordinate point and the reference coordinate point that is not used as an abnormal matching point, and generate a three-dimensional displacement field in cylindrical coordinate system based on the difference of the mapped three-dimensional coordinates.
10. A full-field strain measurement system for the tensioning process of precast piles based on digital image correlation, characterized in that, include: The capture module is used to control the mobile robot to follow the precast pile undergoing tensioning along the production line, and to capture the initial speckle image of the surface of the precast pile using a vision sensor, and capture the deformation speckle image of the surface of the precast pile during the tensioning process. The decomposition module is used to extract the pixel displacement vector of the tension end plate in the initial speckle image between adjacent image frames, and decompose the pixel displacement vector into the global motion trajectory and mechanical jitter. The generation module is used to perform filtering and smoothing processing on the initial speckle image according to the mechanical vibration, perform coordinate transformation on the filtered initial speckle image according to the global motion trajectory to generate a motion compensation image, obtain the surface geometric curvature and deflection rotation angle of the precast pile, and apply surface flattening processing to the motion compensation image based on the deflection rotation angle and the surface geometric curvature to generate a reference speckle image. The allocation module is used to obtain the physical tension displacement of the precast pile, establish a deformation gradient with decreasing axial distribution distance of the physical tension displacement, and allocate an adaptive variable-scale correlation search window in the reference speckle image according to the deformation gradient. The calling module is used to iteratively solve the target matching coordinates within the relevant search window, taking the reference speckle image as a reference and the deformed speckle image as the target, to separate abnormal matching points based on the spatial continuity of the target matching coordinates, and to call the valid matching points that are in the same topological structure as the abnormal matching points to perform interpolation calculations to generate corrected coordinate points. The output module is used to inversely convert the corrected coordinate points into a three-dimensional displacement field, perform strain calculation along the stress transmission path of the three-dimensional displacement field, and output the full-field strain distribution of the precast pile.
Citation Information
Patent Citations
Tubular pile concrete crack detection method and device
CN110530714A
Steel trestle deformation monitoring method based on non-contact full-field measurement technology DIC
CN118706015A
