Random grid and special-shaped sub-area division method for cracks in digital image correlation

Abstract

The present invention discloses a random grid and special-shaped sub-area division method for cracks in digital image correlation, which relates to experimental mechanics, non-contact full-field crack displacement measurement, and digital image correlation fields. The method is a method that can non-uniformly grid a deformed image, and intelligently identify discontinuous areas such as cracks and mark the contours through a deep learning intelligent algorithm, automatically or manually add nodes outside the crack contour, and define special-shaped sub-areas of the nodes that exclude crack areas. According to the defined nodes and sub-areas, correlation matching is performed with a reference image, and then the displacements of all nodes are calculated, and then the full-field displacement is interpolated. This method can effectively solve the problem in traditional digital image correlation methods that the existence of discontinuous areas such as cracks leads to the breakage of sub-areas, which in turn causes the missing of calculation results in certain areas, resulting in erroneous calculation results.

Description

Technical Field

[0001] The invention relates to the fields of experimental mechanics, non-contact full-field crack displacement measurement and digital image correlation, and in particular to a random grid and special-shaped sub-area division method for cracks in digital image correlation. Background Art

[0002] In many fields such as aerospace, building bridges, etc., ensuring structural reliability is always the top priority. With the development of science and technology, being able to control and repair structures before irreversible damage occurs has become an important goal for many scientists. The measurement results of material response load characterization provided by experimental mechanics have rich guiding significance for improving design, improving structure and monitoring damage. Displacement and strain, as important parameters of experimental mechanics, play a vital role in analyzing the mechanical properties of models, verifying basic assumptions and feature identification. Faced with increasingly complex measurement needs, traditional displacement and strain measurement methods are stretched. At the same time, the displacement and strain measurement results at limited locations are far from being competent for the analysis of complex models. In addition, in response to the constraints of various measurement conditions, emerging measurement technologies that can achieve full-field strain and non-contact measurement have emerged.

[0003] Different methods that can perform full-field measurement each have their own advantages in accuracy and applicability. Digital Image Correlation (DIC), which can also meet non-contact measurement conditions, has been widely used in experimental mechanics due to its advantages such as low experimental cost and relatively simple data acquisition process. It has low requirements for measurement environment and vibration isolation, strong universality, and is widely used in various engineering and disciplines, such as aerospace, biomedicine, etc.

[0004] In the traditional digital image correlation method (DIC), when encountering the displacement measurement problem of discontinuous areas such as cracks, the crack area cannot be avoided when dividing the nodes because the uniform grid division method is used. As a result, when calculating the node displacement of the crack area, the pixel grayscale distribution of the sub-area is discontinuous because the node sub-area contains crack information. When performing correlation matching calculations between node sub-areas, the discontinuity of the sub-area grayscale leads to errors in the calculation of node displacement or no displacement information is obtained at all. Summary of the invention

[0005] The present invention aims at the problems existing in the prior art and discloses a random grid and shaped sub-area division method for cracks in digital image correlation. The method is capable of gridding a deformed image non-uniformly, and at the same time, for discontinuous areas such as cracks, automatically or manually adding nodes outside the crack contour, and defining shaped sub-areas of the nodes to exclude the crack area. According to the defined nodes and sub-areas, correlation matching is performed with the reference image, and then the displacement of all nodes is calculated, and then the full-field displacement is obtained by interpolation. The method solves the technical defects of the prior art that the node displacement calculation is wrong or the displacement information cannot be obtained at all.

[0006] The present invention is achieved in that:

[0007] A random grid and shaped sub-area division method for cracks in digital image correlation, the method comprising:

[0008] Step 1: collect reference images and deformed images with cracks; Step 2: identify crack features on the deformed image with cracks, define the number, position and contour of cracks, and represent each crack contour with a function; Step 3: preset the crack extension range in the deformed image, and expand each crack contour along the normal direction outside the contour to obtain the extended contour area; Step 4: arrange a certain number of randomly distributed nodes outside the extended contour area of ​​the deformed image as the initial nodes for mesh division; Step 5: define the matching sub-area of ​​each initial node on the deformed image according to the set matching sub-area size and the relative position of each initial node and the crack; Step 6: select several nodes required to be added to the encrypted grid in the extended crack contour and other areas where the grid needs to be encrypted by intelligent point arrangement or manual point selection, and define the special-shaped matching sub-area of ​​each added node on the deformed image; Step 7: the node group of the deformed image is composed of the initial nodes and the added nodes, and the displacement of each node is calculated by the digital image correlation method according to the matching sub-area information of each node; Step 8: the displacement of each node can be interpolated to obtain the full-field displacement.

[0009] Furthermore, the step 1 is specifically as follows: building an experimental platform, using a digital camera to collect speckle images of the object before and after deformation, collecting the digital speckle image of the object before the experiment and recording it as a reference image; collecting the digital speckle image of the object with cracks after the experiment and recording it as a deformation image.

[0010] Furthermore, the step 2 is specifically as follows: through an image recognition or feature recognition method, wherein the method is feature recognition of a deep learning neural network, the number of cracks in the deformed image and the corresponding position and contour range are detected, and its contour is represented by a linear or nonlinear function.

[0011] Furthermore, the step 5 is specifically as follows: define the position of each initial node matching sub-area:

[0012] The radius M of the node matching sub-area is preset. Usually, the value range of the node sub-area is: a square area with the node as the center and a side length of 2M+1, with a size of (2M+1)×(2M+1); when some areas of the node sub-area contain cracks, the node position remains unchanged, and the sub-area is translated in the direction away from the crack so that the boundary of the sub-area is outside the crack contour. At this time, it is necessary to ensure that the node and the sub-area boundary are kept at a distance of more than two pixels.

[0013] Furthermore, in step six, the method of adding nodes is specifically as follows: within the range of the extended crack contour or the area where the mesh needs to be encrypted, nodes are automatically selected by an intelligent algorithm, or nodes are manually added by human-computer interaction.

[0014] Furthermore, in step 6, the method for defining the special-shaped matching sub-area is specifically as follows:

[0015] The radius M of the node sub-region is preset, and a square area with the node as the center and a side length of 2M+1 is taken as the sub-region; when some areas of the sub-region contain cracks, the node position remains unchanged, and the sub-region is translated in the direction away from the crack to reduce the area of ​​the crack region in the sub-region, but it is necessary to ensure that the node and the sub-region boundary are kept at a distance of more than two pixels; when the sub-region after translation still contains a crack area, the grayscale information of the crack area in the sub-region is removed, and the square sub-region becomes a polygonal sub-region, which is called an irregular sub-region.

[0016] Furthermore, in step seven, the digital image correlation method is specifically as follows: according to the calculated sub-area range of all nodes of the deformed image, a correlation criterion is selected, and correlation matching is performed with the reference image before deformation to obtain the displacement of each node.

[0017] Furthermore, the correlation criterion includes all correlation functions in the digital image correlation method, such as a zero-mean normalized covariance cross-correlation function or a normalized least square distance correlation function.

[0018] Furthermore, the specific expression of the zero-mean normalized cross-correlation function (ZNCC) is as follows:

[0019]

[0020] Among them, x, y are the horizontal and vertical coordinates, 2M+1 is the side length of the feature description domain, f(x,y) represents the grayscale distribution of the reference node feature description domain, and f m is the grayscale mean of the region, g(x',y') represents the grayscale distribution of the deformation node feature description domain, and g m is the mean grayscale value of the area.

[0021] The beneficial effects of the present invention and the prior art are:

[0022] The present invention proposes a random grid and irregular sub-area division method for cracks in digital image correlation, which can effectively solve the problem of inaccurate displacement measurement results for discontinuous areas such as cracks. The present invention randomly arranges nodes in the non-crack area of ​​the deformed image to divide the non-uniform grid. At the same time, nodes are automatically or manually arranged outside the crack area and irregular sub-areas of the nodes are divided. The irregular sub-areas can solve the problem of sub-area fracture in traditional DIC. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 The present invention provides a random grid and an offset sub-area division method for cracks in digital image correlation;

[0024] Figure 2 A method for dividing a random grid and a method for dividing a special-shaped sub-region for cracks in digital image correlation according to the present invention;

[0025] Figure 3 It is a simulated speckle image of a random grid and a special-shaped sub-area division method for cracks in a digital image correlation of the present invention;

[0026] Figure 4 A displacement true solution cloud diagram of a random grid and a special-shaped sub-area division method for cracks in digital image correlation according to the present invention;

[0027] Figure 5 The initial grid and the additional grid of the random grid and the special-shaped sub-area division method for cracks in the digital image correlation of the present invention;

[0028] Figure 6 This is a displacement calculation solution cloud diagram of a random grid and special-shaped sub-area division method for cracks in digital image correlation in the present invention. DETAILED DESCRIPTION

[0029] In order to make the purpose, technical solution and effect of the present invention clearer and more specific, the present invention is further described in detail by enumerating examples below. It should be noted that the specific implementation described here is only used to explain the present invention and is not used to limit the present invention.

[0030] To achieve the purpose of the present invention, the method of the present invention comprises the following specific steps:

[0031] Step 1: Collect the digital speckle image of the object before the experiment, which is recorded as the reference image; collect the digital speckle image of the object with cracks after the experiment, which is recorded as the deformation image;

[0032] Step 2: Perform crack feature recognition on the deformation image, identify the number, location and contour range of the cracks, and represent the contour with a function;

[0033] Step 3: Set the number of initial nodes to n; arrange n randomly distributed initial nodes in the non-crack area of ​​the deformed image;

[0034] Step 4: Set the crack contour extension range to 3 pixels and expand the crack contour range outward.

[0035] Step 5: Assign a square area with the node as the center and a side length of 2M+1 to each node as the matching sub-area of ​​the node; if part of the sub-area of ​​some nodes is located in the crack area, the node position remains unchanged, and the sub-area is translated in the direction away from the crack so that the sub-area boundary is outside the crack area; at this time, it is necessary to ensure that the node is still more than 2 pixels away from the sub-area boundary, such as Figure 1 , 2 As shown;

[0036] Step 6: Automatically select nodes within the extended outline using an intelligent algorithm, or manually add nodes through human-computer interaction; the specific method of adding nodes is: automatically select nodes within the extended outline using an intelligent algorithm, or manually add nodes through human-computer interaction, such as Figure 1 , 2 As shown in the figure, the definition method of the special-shaped matching sub-area of ​​the added node is as follows: similar to the translation sub-area method of the initial node, the sub-area boundary of the added node is set outside the crack contour, and the sub-area is arranged in the direction away from the crack, such as Figure 1 , 2 If the division range of some added node sub-areas is limited by the crack area, the crack area can be removed from the sub-area to form a special-shaped sub-area distributed along the crack. The specific division method is as follows: Figure 2 As shown in the figure, the filling areas marked by ① and ② are partially located in the crack area, so the crack area is removed from the atomic area to form a special-shaped sub-area. It is necessary to ensure that the node is more than 2 pixels away from the sub-area boundary;

[0037] Step 7: Similar to the method of translating sub-regions of the initial node, set the sub-region boundary of the added node to be located on the crack contour, and arrange the sub-regions in the direction away from the crack. If the sub-region division range of some added nodes is limited by the crack area, the crack area can be removed from the sub-region to form a special-shaped sub-region distributed along the crack.

[0038] Step 8: The initial nodes and the added nodes form a node group of the deformed image, and the displacement of each node is calculated by a digital image correlation method according to the matching sub-area information of each node;

[0039] Step 9: The displacement of each node can be interpolated to obtain the displacement of the entire field.

[0040] The method of the present invention is described below with specific examples:

[0041] The simulated speckle image is generated by a computer Gaussian speckle algorithm with a pixel size of 500 × 500 and a speckle number of 1000. A reference image is generated, and then a simulated crack is set to generate a deformation image, such as Figure 3 As shown in the figure, the image is divided into four quadrants, and the fourth quadrant is uniformly compressed from 90° to 87° to simulate the crack and generate a deformation image. The displacement cloud map is shown in Figure 4 The displacement field is calculated by the above algorithm, and then the calculated result is compared with the true solution.

[0042] The specific implementation steps are as follows:

[0043] Step 1: Mark the crack area of ​​the deformed image;

[0044] Step 2: Set 4000 randomly distributed initial nodes in the non-crack area of ​​the deformed image;

[0045] Step 3: According to the relative position of the node and the crack, a matching sub-area with a size of 41×41 and different positions is allocated to each node;

[0046] Step 4: Expand the crack contour outward by 5 pixels, manually add nodes within the expanded contour, and assign a matching sub-area of ​​size 41×41 and different positions to each node; Figure 5 The manually added node position can be seen (the encrypted part in the ellipse is the added mesh);

[0047] Step 5: According to the grayscale information of the matching sub-areas of all nodes of the deformed image, correlation matching is performed with the reference image to obtain the matching displacement of each node;

[0048] Step 6: Interpolate the displacement of each node to get the full-field displacement. The displacement cloud diagram is as follows: Figure 6 As shown. Figure 6 Compute the solution and Figure 4 By comparing the calculated solution with the real solution, it is found that the calculated solution matches the real solution well, and there is no displacement loss caused by sub-area fracture, which shows that the proposed method is real, effective and accurate.

[0049] Finally, it should be noted that the above implementation scheme is only used to illustrate the implementation method of the present invention rather than to limit it; people should understand that modifications to the implementation process of the invention or equivalent replacement of part of the algorithm process will not deviate from the spirit of the technical solution of the present invention, and should be included in the scope of the technical solution for which protection is requested by the present invention.

Claims

1. A random grid and shaped sub-area division method for cracks in digital image correlation, characterized in that: The method described is: Step 1: Collect a reference image and a deformed image with cracks; Step 2: Perform crack feature recognition on the deformation image containing cracks, define the number, position and contour of the cracks, and represent each crack contour with a function; Step 3: In the deformation image, preset the crack extension range, and expand each crack contour along the outer normal direction of the contour to obtain the extended contour area; Step 4: Arrange a certain number of randomly distributed nodes outside the extended contour area of ​​the deformed image as the initial nodes for mesh division; Step 5: According to the set matching sub-region size and the relative position of each initial node and the crack, the matching sub-region of each initial node is defined on the deformation image; Step 6: Select several nodes to be added to the mesh in the extended crack contour and other areas where mesh encryption is required by means of intelligent point placement or manual point selection, and define the special-shaped matching sub-area of ​​each added node on the deformed image; the definition method of the special-shaped matching sub-area is as follows: preset the radius of the node sub-area M , take the node as the center, 2 M The square area with a side length of +1 is the sub-area; when some areas of the sub-area contain cracks, the node position remains unchanged, the node is kept at a distance of more than two pixels from the sub-area boundary, and the sub-area is translated in the direction away from the crack to reduce the area of ​​the crack area in the sub-area; when the sub-area after translation still contains crack areas, the grayscale information of the crack area in the sub-area is removed, and the sub-area becomes a polygonal sub-area, which is called a special-shaped sub-area; Step 7: The initial nodes and the added nodes form a node group of the deformed image, and the displacement of each node is calculated by a digital image correlation method according to the matching sub-region information of each node; Step 8: Interpolate the displacement of each node to get the displacement of the entire field.

2. The random grid and shaped sub-area division method for cracks in digital image correlation according to claim 1, characterized in that: The step 1 is specifically as follows: building an experimental platform, using a digital camera to collect speckle images of the object before and after deformation, collecting the digital speckle image of the object before the experiment and recording it as a reference image; collecting the digital speckle image of the object with cracks after the experiment and recording it as a deformation image.

3. The random grid and shaped sub-area division method for cracks in digital image correlation according to claim 1, characterized in that: The step 2 is specifically as follows: through image recognition or feature recognition methods, wherein the method is feature recognition of a deep learning neural network, the number of cracks in the deformed image and the corresponding positions and contour ranges are detected, and their contours are represented by linear or nonlinear functions.

4. The random grid and shaped sub-area division method for cracks in digital image correlation according to claim 1, characterized in that: The step five is specifically as follows: Define the position of each initial node matching sub-area: preset the radius of the node matching sub-area M , the value range of the node sub-area is: centered on the node, 2 M +1 is a square area with a side length of (2 M +1)×(2 M +1); when some areas of the node sub-region contain cracks, the node position remains unchanged, and the sub-region is translated in the direction away from the crack so that the sub-region boundary is outside the crack contour and the node is kept at least two pixels away from the sub-region boundary.

5. The random grid and shaped sub-area division method for cracks in digital image correlation according to claim 1, characterized in that: In the step six, the method of adding nodes is specifically: within the range of the extended crack contour or the area where the mesh needs to be encrypted, nodes are automatically selected through an intelligent algorithm, or nodes are manually added through human-computer interaction.

6. The random grid and shaped sub-area division method for cracks in digital image correlation according to claim 1, characterized in that: In the step 7, the digital image correlation method is specifically as follows: According to the calculated sub-area range of all nodes of the deformed image, a correlation criterion is selected, and correlation matching is performed with the reference image before deformation, so as to obtain the displacement of each node.

7. The random grid and shaped sub-area division method for cracks in digital image correlation according to claim 6, characterized in that: The correlation criterion includes a zero-mean normalized covariance cross-correlation function or a normalized least square distance correlation function.

Images