A high-precision global quasi-static strain vision measurement method

Abstract

The present invention discloses a high-precision global quasi-static strain vision measurement method, which relates to the technical field of strain vision measurement. The specific steps include: Step 1, obtaining high-resolution continuous image deformation data and correcting the geometric distortion of the deformed image through an orthorectification method, that is, obtaining high-resolution and continuous image deformation data through a high-precision image acquisition device and precisely correcting the existing geometric distortion in the deformed image by using the orthorectification method; Step 2, performing noise reduction processing on the image through singular value decomposition technology and conducting image displacement analysis on the image after the noise reduction processing is completed, that is, using advanced singular value decomposition technology to deeply perform noise reduction processing on the image and effectively removing various noise interferences in the image; Step 3, further processing the data obtained after strain calculation and visually displaying the strain result through visual output. The present invention provides an efficient and reliable solution for global strain measurement of complex structures.

Description

Technical Field

[0001] The present invention belongs to the technical field of strain vision measurement, and particularly relates to a high-precision global quasi-static strain vision measurement method. Background Art

[0002] In the fields of engineering and materials science, accurately measuring the global strain distribution of an object under stress is crucial for evaluating structural safety, predicting failure modes, and optimizing designs. Although traditional strain measurement techniques (such as resistance strain gauges and fiber optic sensors) have relatively high local measurement accuracy, they have the following significant drawbacks: they can only obtain discrete point data and cannot achieve global strain distribution analysis; contact installation easily interferes with the surface state of the object being measured and is not suitable for sensitive materials such as thin films and biological tissues; comprehensive measurement of complex geometric structures or large-sized objects is costly and difficult to implement.

[0003] In recent years, non-contact vision measurement techniques (such as digital image correlation method, DIC) have gradually become a research hotspot. DIC calculates the global strain distribution by tracking the displacement changes of feature points on the object surface, but its accuracy and reliability are limited by image quality, noise interference, and geometric distortion correction ability. In the prior art, the image acquisition system often causes cumulative strain calculation errors due to lens distortion, camera viewing angle deviation, or image blurring under dynamic loading; in addition, image denoising algorithms (such as mean filtering and wavelet transform) are difficult to balance between retaining details and suppressing noise and are prone to losing tiny deformation information.

[0004] Therefore, there is an urgent need for a high-precision global quasi-static strain vision measurement method to overcome the limitations of traditional methods in terms of accuracy, efficiency, and applicability. Summary of the Invention

[0005] The purpose of the present invention is to provide a high-precision global quasi-static strain vision measurement method to solve the technical problems existing in the prior art in terms of accuracy, efficiency, and applicability.

[0006] To achieve the above purpose, the present invention adopts the following technical solutions:

[0007] A high-precision global quasi-static strain vision measurement method includes:

[0008] Step 1: Obtain high-resolution continuous image deformation data and correct the geometric distortion of the deformed image through an orthorectification method;

[0009] That is, obtain high-resolution and continuous image deformation data through a high-precision image acquisition device, and accurately correct the geometric distortion existing in the deformed image by using the orthorectification method;

[0010] Step 2: Denoise the image using singular value decomposition technology and perform image displacement analysis on the denoised image;

[0011] Step 3: Further process the data obtained after strain calculation and visually display the strain results through visualization output.

[0012] Furthermore, obtain high-resolution continuous image deformation data. The specific method is as follows:

[0013] Select a high-resolution CMOS camera with a resolution of n×m pixels and a frame rate equal to f frames per second to adapt to the dynamic deformation of the monitored object during the quasi-static loading process. By equipping the camera with a high-precision lens, such as a telecentric lens or an ultra-wide-angle lens, geometric distortion can be reduced and the field of view can be expanded. The camera is fixed in front of the object to be measured through a high-precision three-dimensional adjustment bracket, and the internal and external parameters of the camera are calibrated through a high-precision calibration board to establish the mapping relationship between pixel coordinates and actual physical dimensions, completing the configuration of the image acquisition system. During the loading process, the system continuously acquires the image sequence of the object to be measured to ensure continuous and comparable deformation data.

[0014] Furthermore, correct the geometric distortion of the deformed image through the orthorectification method. The specific method is as follows:

[0015] Using the configured image acquisition system, during the quasi-static loading process, continuously capture the surface images of the object at a preset frame rate to generate time series data. Among them, geometric correction requires using multiple frames in the continuously captured image sequence for distortion correction. Distortion correction requires defining the source points and target points. Select four feature points on the surface of the object to be measured or the calibration board as the source points, and the coordinates of the source points are represented by where i = 1, 2, 3, 4. The source points are located at the four corners or specific areas of the image. Define the target points corresponding to each source point based on the actual physical dimensions, and the coordinates of the target points are represented by The target points form a rectangular area, and its coordinates are strictly aligned with the actual physical coordinate system. Solve the perspective transformation matrix according to the origin and the target points. By comparing the length of the known-sized line segment on the calibration board with the corresponding length of the corrected image, calculate the geometric distortion residual. If the residual is greater than the preset tolerance, iteratively adjust the transformation matrix H until the accuracy requirement is met to obtain the corrected image. Denote the corrected image sequence matrix as M.

[0016] Furthermore, solve the perspective transformation matrix according to the origin and the target points. The specific method is as follows:

[0017] Use to represent the formula for solving the perspective transformation matrix, where represents the scaling factor of the i-th target point, and Use the formula represents a perspective transformation matrix, where, represents the transformation parameters to be solved for the perspective transformation matrix. For each pair of source points and target points, the transformation parameters are solved through the perspective transformation matrix solution formula, and the corresponding linear equations can be obtained. Solve the linear equations corresponding to all source points and target points, apply the calculated perspective transformation matrix H to each pixel point of the original image, calculate the gray value of each pixel in the target image through bilinear interpolation, and generate an orthorectified image. During the continuous image acquisition process, the offset of the source point is detected in real time. If the offset exceeds the preset threshold a, the perspective transformation matrix H is recalculated until a perspective transformation matrix that meets the threshold requirements is obtained.

[0018] Furthermore, the image is denoised by singular value decomposition technology. The specific method is as follows:

[0019] After obtaining the corrected image, perform singular value decomposition on the image sequence matrix M by singular value decomposition technology to obtain singular values and singular vectors. Preset the singular value energy threshold Q. With the sum of singular value energies being greater than or equal to the threshold Q as the condition, select the singular modes that meet this condition and have the smallest number as the signal modes to form a mode set , construct a weight matrix according to the mode set, and use the formula to perform weighted average filtering on the original gray values, where, represents the filtered image sequence matrix, W represents the weight matrix constructed from the mode set is represented by the formula represents, represents the mode set is the transpose of. After performing weighted average filtering on the original gray values using the weight matrix, then replace the gray value in the original image with the filtered gray value, and reconstruct the image using the retained singular values and singular vectors to complete image denoising and obtain the filtered image sequence.

[0020] Furthermore, perform image displacement analysis on the image after denoising. The specific method is as follows:

[0021] After completing image denoising, perform displacement analysis on the image using the digital image correlation method. The image is a two-dimensional matrix composed of pixels, and each pixel point has a corresponding coordinate position. Calculate the displacement vector between the current frame and the reference frame through the digital image correlation method, perform interpolation on each pixel point of the image using numerical interpolation methods to obtain a smooth and continuous displacement field, and use the formula represents the displacement field, where, (x, y) represents the coordinates of any pixel point in the image, represents the displacement of the pixel point with coordinates (x, y) in the x direction, It represents the displacement of the pixel point with coordinates (x, y) in the y direction. Strain is calculated through the gradient of the displacement field. The strain of each point is calculated to obtain the strain distribution on the surface of the entire photographed object.

[0022] Furthermore, the strain is calculated through the gradient of the displacement field. The specific method is as follows:

[0023] Using It represents the formula for calculating strain through the gradient of the displacement field. Here, x represents the abscissa of any pixel point in the image, and y represents the ordinate of any pixel point in the image. It represents the displacement of each pixel point in the x direction. It represents the displacement of each pixel point in the y direction. It represents the normal strain of each pixel point in the x direction. It represents the normal strain of each pixel point in the y direction. It represents the shear strain of each pixel point.

[0024] Furthermore, the data obtained after the strain calculation is further processed, and the strain result is visually displayed through visualization output. The specific method is as follows:

[0025] The high-frequency noise in the data after the strain calculation is removed through low-pass filtering to smooth the strain data, eliminate the abnormal fluctuations caused by measurement errors and noise, and obtain the filtered strain field data. Then, the out-of-plane correction is used to solve the geometric distortion problem that appears on the surface of the object in the experiment. By adopting a geometric correction algorithm, the out-of-plane correction of the strain data is carried out to ensure that the strain data accurately reflects the true deformation of the object surface. The strain result is displayed graphically through a visualization tool.

[0026] To sum up, due to the adoption of the above technical solutions, the beneficial effects of the present invention are as follows:

[0027] 1. The present invention realizes the accurate monitoring and analysis of the deformation of the monitored object during the quasi-static loading process by acquiring high-resolution continuous image deformation data and using the orthorectification method to correct geometric distortion, which helps to improve the accuracy of deformation analysis. By real-time detecting the offset of the source point and recalculating the perspective transformation matrix when necessary, it helps to improve the accuracy of the corrected image.

[0028] 2. Through singular value decomposition, the present invention identifies and distinguishes the signal components and noise components in the image, effectively filters out the noise mode, realizes the noise reduction processing of the image, uses the weight matrix to perform weighted average filtering on the original gray value, further enhances the noise reduction effect, makes the filtered image sequence clearer, retains better details, and after completing the noise reduction processing, uses the digital image correlation method to perform displacement analysis on the image, and can accurately calculate the displacement vector between the current frame and the reference frame, which helps to improve the data analysis efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0030] Figure 1 Shows a step diagram of a high-precision global quasi-static strain vision measurement method;

[0031] Figure 2 Shows a flowchart of a high-precision global quasi-static strain vision measurement method. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0032] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0033] Embodiment 1. As Figure 1 、 Figure 2 shown, a high-precision global quasi-static strain vision measurement method specifically includes the following steps:

[0034] Step 1. Obtain high-resolution continuous image deformation data, and correct the geometric distortion of the deformed image through an orthorectification method.

[0035] Select a high-resolution CMOS camera with a resolution of n×m pixels and a frame rate equal to f frames per second to adapt to the dynamic deformation of the monitored object during the quasi-static loading process. By equipping the camera with a high-precision lens, such as a telecentric lens or an ultra-wide-angle lens, geometric distortion is reduced and the field of view is expanded. The camera is fixed in front of the object to be measured through a high-precision three-dimensional adjustment bracket, and the internal and external parameters of the camera are calibrated through a high-precision calibration board to establish the mapping relationship between pixel coordinates and actual physical dimensions, completing the configuration of the image acquisition system. During the loading process, the system continuously acquires the image sequence of the object to be measured to ensure continuous and comparable deformation data. In this embodiment, n×m is set to 1000×2000 and f is set to 50;

[0036] In the image acquisition and geometric correction stage, using the configured image acquisition system, during the quasi-static loading process, continuously capture the surface images of the object at a preset frame rate to generate time series data. Among them, geometric correction requires using multiple frames of images in the continuously captured image sequence for distortion correction. Distortion correction requires defining source points and target points, including selecting four feature points on the surface of the object to be measured or on the calibration board as source points, and the coordinates of the source points are represented by where i = 1, 2, 3, 4. The source points are located at the four corners or specific regions in the image. Define the target points corresponding to each source point based on the actual physical dimensions, and the coordinates of the target points are represented by The target points form a rectangular area, and its coordinates are strictly aligned with the actual physical coordinate system. Solve the perspective transformation matrix according to the origin and the target points. The specific formula is as follows:

[0037] ;

[0038] where, represents the scaling factor of the i-th target point, and H represents the perspective transformation matrix;

[0039] Furthermore, the specific expression of the perspective transformation matrix is as follows;

[0040] ;

[0041] where, represents the transformation parameters to be solved of the perspective transformation matrix. In this embodiment, , for each pair of source points and target points, the transformation parameters are solved through the perspective transformation matrix solution formula, and the corresponding linear equations can be obtained. Solve the linear equations corresponding to all source points and target points, apply the calculated perspective transformation matrix H to each pixel point of the original image, calculate the gray value of each pixel in the target image through bilinear interpolation, and generate the orthorectified image. During the continuous image acquisition process, the offset of the source point is detected in real time. If the offset exceeds the preset threshold a, the perspective transformation matrix H is recalculated to ensure that the proportional error between the corrected image and the actual physical size is less than the actual error acceptance range. By comparing the length of the line segment with known size on the calibration plate and the corresponding length of the corrected image, the geometric distortion residual is calculated. If the residual is greater than the preset tolerance, the transformation matrix H is iteratively adjusted until the accuracy requirement is met, and the corrected image is obtained. Denote the matrix of the corrected image sequence as M.

[0042] Step 2: Denoise the image through singular value decomposition technology and perform image displacement analysis on the denoised image.

[0043] After obtaining the corrected image, perform singular value decomposition on the image sequence matrix M through singular value decomposition technology to obtain singular values and singular vectors. Preset the singular value energy threshold Q. Taking the sum of singular value energies being greater than or equal to the threshold Q as the condition, select the singular modes that meet this condition and have the least number as the signal modes to form a mode set. , construct a weight matrix according to the mode set, and use this weight matrix to perform weighted average filtering on the original gray values. The specific formula is as follows:

[0044] ;

[0045] where, represents the filtered image sequence matrix, W represents the weight matrix constructed from the mode set , which is represented by the formula , represents the transpose of the mode set . After performing weighted average filtering on the original gray values using the weight matrix, then replace the gray values in the original image with the filtered gray values, and reconstruct the image using the reserved singular values and singular vectors to complete image denoising and obtain the filtered image sequence.

[0046] After completing image denoising, perform displacement analysis on the image using digital image correlation method. The image is a two-dimensional matrix composed of pixels, and each pixel point has a corresponding coordinate position. Calculate the displacement vector between the current frame and the reference frame through digital image correlation method, and perform interpolation on each pixel point of the image using numerical interpolation method to obtain a smooth and continuous displacement field. The specific expression of the displacement field is as follows:

[0047] ;

[0048] Among them, (x, y) represents the coordinates of any pixel point in the image. represents the displacement of the pixel point with coordinates (x, y) in the x direction. represents the displacement of the pixel point with coordinates (x, y) in the y direction.

[0049] The strain is calculated through the gradient of the displacement field, and the specific formula is as follows:

[0050] ;

[0051] Among them, x represents the abscissa of any pixel point in the image, and y represents the ordinate of any pixel point in the image. represents the displacement of each pixel point in the x direction. represents the displacement of each pixel point in the y direction. represents the normal strain of each pixel point in the x direction. represents the normal strain of each pixel point in the y direction. represents the shear strain of each pixel point.

[0052] By calculating the above strain values for each point in the image, the strain distribution on the surface of the entire photographed object can be obtained.

[0053] Step 3: Further process the data obtained after the strain calculation and visually display the strain results through visualization.

[0054] There may be high-frequency fluctuations caused by noise in the data obtained after the strain calculation. These high-frequency noises need to be removed through low-pass filtering. The rule is that low-frequency signals can pass through normally, while high-frequency signals exceeding the set critical value are blocked and attenuated. However, the degree of blocking and attenuation will change according to different frequencies and different filtering purposes. This frequency is the cut-off frequency. When the frequency domain is higher than this cut-off frequency, the signal is completely blocked or assigned a value of 0. Since all low-frequency signals are allowed to pass through in this process, in this embodiment, signals above 1 HZ are filtered by applying Gaussian filtering to smooth the strain data and eliminate abnormal fluctuations caused by measurement errors and noises, obtaining the filtered strain field data. Then, the geometric distortion problem that appears on the surface of the object in the experiment is solved through out-of-plane correction. By adopting a geometric correction algorithm, such as the least squares method, the strain data is corrected out-of-plane to ensure that the strain data accurately reflects the true deformation of the object surface. The strain result after the above processing is recorded as the output strain result.

[0055] The strain results are presented in a graphical manner. The specific methods include heat maps, 3D strain maps, and strain animations. Through these visualization tools, it is convenient to visually observe the strain distribution on the surface of the object. Finally, the strain data is exported in common formats such as CSV and Excel for further analysis and recording.

[0056] As described above, only the preferred specific embodiments of the present invention are provided, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, making equivalent substitutions or changes should be covered within the protection scope of the present invention.

[0057] The preferred embodiments of the present invention disclosed above are only used to help illustrate the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the present invention to only the specific embodiments. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the present invention, so that those skilled in the art can well understand and utilize the present invention. The present invention is only limited by the claims and their full scope and equivalents.

Claims

1. A high-precision global quasi-static strain visual measurement method, characterized in that: include: Step 1: Obtain high-resolution continuous image deformation data and correct the geometric distortion of the deformed image by using the orthorectification method; That is, high-resolution and continuous image deformation data is obtained through high-precision image acquisition equipment, and the geometric distortion existing in the deformed image is accurately corrected using the orthorectification method; Step 2: De-noise the image by using singular value decomposition technology, and perform image displacement analysis on the image after de-noising; Step 3: further process the data obtained after strain calculation, and intuitively display the strain results through visual output; The data obtained after strain calculation is further processed, and the strain results are intuitively displayed through visual output. The specific method is as follows: Low-pass filtering is used to remove high-frequency noise that appears in the data after strain calculation, smooth the strain data, eliminate abnormal fluctuations caused by measurement errors and noise, and obtain filtered strain field data. Then, out-of-plane correction is used to solve the geometric distortion problem on the surface of the object in the experiment. By adopting a geometric correction algorithm, the strain data is corrected out of the plane to ensure that the strain data accurately reflects the actual deformation of the object surface, and the strain results are displayed in a graphical way through visualization tools.

2. A high-precision global quasi-static strain visual measurement method according to claim 1, characterized in that: Obtain high-resolution continuous image deformation data. The specific method is: A high-resolution CMOS camera with a resolution of n×m pixels and a frame rate equal to f frames per second is selected to adapt to the dynamic deformation of the object under quasi-static loading. The camera is equipped with a high-precision lens, fixed in front of the object under test through a high-precision three-dimensional adjustment bracket, and the internal and external parameters of the camera are calibrated through a high-precision calibration plate. The mapping relationship between pixel coordinates and actual physical dimensions is established to complete the configuration of the image acquisition system. During the loading process, the system continuously collects image sequences of the object under test to ensure the acquisition of continuous and comparable deformation data.

3. A high-precision global quasi-static strain visual measurement method according to claim 1, characterized in that: The geometric distortion of the deformed image is corrected by the orthorectification method. The specific method is as follows: Using the configured image acquisition system, during the quasi-static loading process, the surface images of the object are continuously captured at a preset frame rate to generate time series data. The geometric correction requires the use of multiple frames of images in the continuously captured image sequence for distortion correction. The distortion correction requires the definition of source points and target points. Four feature points are selected on the surface of the object to be measured or on the calibration plate as the source points. The coordinates of the source points are expressed as Indicated by, where i=1, 2, 3, 4, the source points are located in the four corners or specific areas of the image, and the target point corresponding to each source point is defined based on the actual physical size. The coordinates of the target point are expressed as It means that the target points constitute a rectangular area, whose coordinates are strictly aligned with the actual physical coordinate system. The perspective transformation matrix is ​​solved according to the origin and the target point. The geometric distortion residual is calculated by comparing the length of the line segment of known size on the calibration plate with the corresponding length of the corrected image. If the residual is greater than the preset tolerance, the transformation matrix H is iteratively adjusted until the accuracy requirement is met to obtain the corrected image. The corrected image sequence matrix is ​​recorded as M.

4. A high-precision global quasi-static strain visual measurement method according to claim 3, characterized in that: Solve the perspective transformation matrix based on the origin and target point. The specific method is: use Represents the formula for solving the perspective transformation matrix, where represents the scaling factor of the i-th target point, and , using the formula represents the perspective transformation matrix, where The transformation parameters to be solved of the perspective transformation matrix are represented. For each pair of source points and target points, the transformation parameters are solved by the perspective transformation matrix solution formula to obtain the corresponding linear equations. The linear equations corresponding to all source points and target points are solved, and the calculated perspective transformation matrix H is applied to each pixel point of the original image. The grayscale value of each pixel in the target image is calculated by bilinear interpolation to generate an image corrected by orthographic projection. During continuous image acquisition, the offset of the source point is detected in real time. If the offset exceeds the preset threshold a, the perspective transformation matrix H is recalculated until a perspective transformation matrix that meets the threshold requirement is obtained.

5. A high-precision global quasi-static strain visual measurement method according to claim 1, characterized in that: The image is denoised using singular value decomposition technology. The specific method is as follows: After obtaining the corrected image, the singular value decomposition technique is used to decompose the image sequence matrix M to obtain singular values ​​and singular vectors. The singular value energy threshold Q is preset. The sum of the singular value energy is greater than or equal to the threshold Q. The singular modes that meet this condition and have the least number are selected as signal modes to form a mode set. , construct the weight matrix according to the modal set, and use the formula Perform weighted average filtering on the original gray value, where represents the filtered image sequence matrix, W represents the modality set The constructed weight matrix is ​​given by the formula express, Represents a modal set The weight matrix is ​​used to perform weighted average filtering on the original grayscale values, and then the filtered grayscale values ​​are used to replace the grayscale values ​​in the original image. The image is reconstructed using the retained singular values ​​and singular vectors to complete image denoising and obtain a filtered image sequence.

6. A high-precision global quasi-static strain visual measurement method according to claim 1, characterized in that: Perform image displacement analysis on the image that has completed noise reduction processing. The specific method is as follows: After image denoising, the digital image correlation method is used to perform displacement analysis on the image. The image is a two-dimensional matrix composed of pixels, and each pixel has a corresponding coordinate position. The displacement vector between the current frame and the reference frame is calculated by the digital image correlation method. The numerical interpolation method is used to interpolate each pixel of the image to obtain a smooth and continuous displacement field. The formula represents the displacement field, where (x, y) represents the coordinates of any pixel in the image. Indicates the displacement of the pixel with coordinates (x, y) in the x direction. It represents the displacement of a pixel with coordinates (x, y) in the y direction. The strain is calculated by the gradient of the displacement field. The strain amount is calculated for each point to obtain the strain distribution of the entire surface of the photographed object.

7. A high-precision global quasi-static strain visual measurement method according to claim 6, characterized in that: The strain is calculated by the gradient of the displacement field. The specific method is: The formula for calculating strain through the gradient of the displacement field is represented, where x represents the horizontal coordinate of any pixel in the image, and y represents the vertical coordinate of any pixel in the image. Represents the displacement of each pixel in the x direction, Represents the displacement of each pixel in the y direction, represents the positive strain of each pixel in the x direction, represents the positive strain of each pixel in the y direction, Represents the shear strain at each pixel.

Images